Focus tracking in dialogs

[0001] The present application relates to U.S. patent application entitled APPLICATION ABSTRACTION WITH DIALOG PURPOSE having Ser. No. 10/087,608, filed Oct. 21, 2001, and published as U.S. 2003/0130854 and U.S. patent application entitled APPLICATION ABSTRACTION WITH DIALOG PURPOSE having Ser. No. 10/426,053, filed Apr. 28, 2003, the contents of which are hereby incorporated be reference in their entirety.

[0002] The present invention relates to access of information over a wide area network such as the Internet. More particularly, the present invention relates to web enabled recognition allowing information and control on a client side device to be entered using a variety of methods.

[0003] Small computing devices such as personal information managers (PIM), devices and portable phones are used with ever increasing frequency by people in their day-to-day activities. With the increase in processing power now available for microprocessors used to run these devices, the functionality of these devices are increasing, and in some cases, merging. For instance, many portable phones now can be used to access and browse the Internet as well as can be used to store personal information such as addresses, phone numbers and the like.

[0004] In view that these computing devices are being used for browsing the Internet, or are used in other server/client architectures, it is therefore necessary to enter information into the computing device. Unfortunately, due to the desire to keep these devices as small as possible in order that they are easily carried, conventional keyboards having all the letters of the alphabet as isolated buttons are usually not possible due to the limited surface area available on the housings of the computing devices.

[0005] To address this problem, there has been increased interest and adoption of using voice or speech to provide and access such information, particularly over a wide area network such as the Internet. Published U.S. patent application, U.S. 2003/0130854, entitled APPLICATION ABSTRACTION WITH DIALOG PURPOSE and U.S. patent application entitled APPLICATION ABSTRACTION WITH DIALOG PURPOSE having Ser. No. 10/426,053 and filed Apr. 28, 2003 describe a method and system defining controls for a web server to generate client side markups that include recognition and/or audible prompting.

[0006] Each of the controls perform a role in the dialog. For instance, controls can include prompt object used to generate corresponding markup for the client device to present information to the user, or generate markups for the client device to ask a question. An answer control or object generates markup for the client device so that a grammar used for recognition is associated with an input field related to a question that has been asked. If it is unclear whether or not a recognized result is correct, a confirmation mechanism can be activated and generate markup to confirm a recognized result. A command control generates markup that allows the user to provide commands, which are other than the expected answers to a specific question, and thus, allows the user to navigate through the web server application, for example. A module, when executed such as on a client, creates a dialog to solicit and provide information as a function of the controls.

[0007] The module can use a control mechanism that identifies an order for the dialog, for example, an order for asking questions. The controls include activation logic that may activate other controls based on the answer given by the user. In many cases, the controls specify and allow the user to provide extra answers, which are commonly answers to questions yet to be asked, and thereby, cause the system to skip such questions since such answers have already been provided. This type of dialog is referred to as “mixed-initiative” since the system and the user have some control of dialog flow.

[0008] However, when users are allowed to provide many pieces of information in one sentence, it becomes difficult to ensure that the system will respond appropriately. For example, suppose a system asks a user for a phone number. In this example, the phone number includes an area code, a local number and an extension. In a mixed-initiative dialog, the user could provide the full number or just a portion of it. The system may need to confirm portions of the number that have been given and would need to ask for the remaining portions of the number. If the user denies or corrects a portion that the system misunderstood, the system would need to ask it again. Ideally, the system would make sure to always confirm or ask a question about the portions of the number that the user just provided. In contrast, if the system were to confirm or ask a question about another portion of the number, the dialog would seem confusing and hard to follow. Given the large number of possible dialog flows, which can be based on the number of permutations due to the number of extra answers that can be provided, a logical dialog flow is difficult to achieve. In some cases, the system may follow a hard-coded path through the dialog and appears from the user's point-of-view, to ignore the information it was given. However, it is usually processed later, which can further add to the confusion.

[0009] There is thus an ongoing need to improve upon the methods used to provide speech recognition in an application such as server/client architecture such as the Internet. In particular, a method, system or authoring tool that addresses one, several or all of the foregoing disadvantages and thus provides generation of speech-enabled recognition and/or speech-enabled prompting in an application is needed.

[0010] Controls are provided for a web server to generate client side markups that include recognition and/or audible prompting. The controls comprise elements of a dialog such as a question, answer, confirmation, command or statement. A module forms a dialog by making use of the information carried in the controls.

[0011] Each of the controls perform a role in the dialog. For instance, controls can include prompt object used to generate corresponding markup for the client device to present information to the user, or generate markups for the client device to ask a question. An answer control or object generates markup for the client device so that a grammar used for recognition is associated with an input field related to a question that has been asked. If it is unclear whether or not a recognized result is correct, a confirmation mechanism can be activated and generate markup to confirm a recognized result. A module, when executed such as on a client, creates a dialog to solicit and provide information as a function of the controls.

[0012] An aspect of the present invention is to allow the system to automatically adapt the dialogue flow so that it stays focused on the user's most recent inputs. Generally, whenever recognition results are received, this information is retained in a manner so as to provide an order indicating the relative order it was received. In this manner, the most recently recognition results can be identified. In one embodiment, memory is used in the form of a “stack”. The stack comprises identifiers related to recognition results received. When the dialog is created, it looks for controls related to the recognition results identified at the top of the stack, for example, whether this recognition result needs to be confirmed. Although the controls typically include means such as an attribute to indicate a selected order for execution, this means that controls later in the selected order than others can be “promoted” and run before them, provided that they are related to the top-most of the stack whereas the others are not.

[0029] Before describing architecture of web based recognition and methods for implementing the same, it may be useful to describe generally computing devices that can function in the architecture. Referring now to FIG. 1, an exemplary form of a data management device (PIM, PDA or the like) is illustrated at 30. However, it is contemplated that the present invention can also be practiced using other computing devices discussed below, and in particular, those computing devices having limited surface areas for input buttons or the like. For example, phones and/or data management devices will also benefit from the present invention. Such devices will have an enhanced utility compared to existing portable personal information management devices and other portable electronic devices, and the functions and compact size of such devices will more likely encourage the user to carry the device at all times. Accordingly, it is not intended that the scope of the architecture herein described be limited by the disclosure of an exemplary data management or PIM device, phone or computer herein illustrated.

[0030] An exemplary form of a data management mobile device 30 is illustrated in FIG. 1. The mobile device 30 includes a housing 32 and has an user interface including a display 34, which uses a contact sensitive display screen in conjunction with a stylus 33. The stylus 33 is used to press or contact the display 34 at designated coordinates to select a field, to selectively move a starting position of a cursor, or to otherwise provide command information such as through gestures or handwriting. Alternatively, or in addition, one or more buttons 35 can be included on the device 30 for navigation. In addition, other input mechanisms such as rotatable wheels, rollers or the like can also be provided. However, it should be noted that the invention is not intended to be limited by these forms of input mechanisms. For instance, another form of input can include a visual input such as through computer vision.

[0031] Referring now to FIG. 2, a block diagram illustrates the functional components comprising the mobile device 30. A central processing unit (CPU) 50 implements the software control functions. CPU 50 is coupled to display 34 so that text and graphic icons generated in accordance with the controlling software appear on the display 34. A speaker 43 can be coupled to CPU 50 typically with a digital-to-analog converter 59 to provide an audible output. Data that is downloaded or entered by the user into the mobile device 30 is stored in a non-volatile read/write random access memory store 54 bi-directionally coupled to the CPU 50. Random access memory (RAM) 54 provides volatile storage for instructions that are executed by CPU 50, and storage for temporary data, such as register values. Default values for configuration options and other variables are stored in a read only memory (ROM) 58. ROM 58 can also be used to store the operating system software for the device that controls the basic functionality of the mobile 30 and other operating system kernel functions (e.g., the loading of software components into RAM 54).

[0032] RAM 54 also serves as a storage for the code in the manner analogous to the function of a hard drive on a PC that is used to store application programs. It should be noted that although non-volatile memory is used for storing the code, it alternatively can be stored in volatile memory that is not used for execution of the code.

[0033] Wireless signals can be transmitted/received by the mobile device through a wireless transceiver 52, which is coupled to CPU 50. An optional communication interface 60 can also be provided for downloading data directly from a computer (e.g., desktop computer), or from a wired network, if desired. Accordingly, interface 60 can comprise various forms of communication devices, for example, an infrared link, modem, a network card, or the like.

[0034] Mobile device 30 includes a microphone 29, and analog-to-digital (A/D) converter 37, and an optional recognition program (speech, DTMF, handwriting, gesture or computer vision) stored in store 54. By way of example, in response to audible information, instructions or commands from a user of device 30, microphone 29 provides speech signals, which are digitized by A/D converter 37. The speech recognition program can perform normalization and/or feature extraction functions on the digitized speech signals to obtain intermediate speech recognition results. Using wireless transceiver 52 or communication interface 60, speech data is transmitted to a remote recognition server 204 discussed below and illustrated in the architecture of FIG. 4. Recognition results are then returned to mobile device 30 for rendering (e.g. visual and/or audible) thereon, and eventual transmission to a web server 202 (FIG. 5), wherein the web server 202 and mobile device 30 operate in a client/server relationship. Similar processing can be used for other forms of input. For example, handwriting input can be digitized with or without pre-processing on device 30. Like the speech data, this form of input can be transmitted to the recognition server 204 for recognition wherein the recognition results are returned to at least one of the device 30 and/or web server 202. Likewise, DTMF data, gesture data and visual data can be processed similarly. Depending on the form of input, device 30 (and the other forms of clients discussed below) would include necessary hardware such as a camera for visual input.

[0035] In addition to the portable or mobile computing devices described above, it should also be understood that the present invention can be used with numerous other computing devices such as a general desktop computer. For instance, the present invention will allow a user with limited physical abilities to input or enter text into a computer or other computing device when other conventional input devices, such as a full alpha-numeric keyboard, are too difficult to operate.

[0036] The invention is also operational with numerous other general purpose or special purpose computing systems, environments or configurations. Examples of well known computing systems, environments, and/or configurations that may be suitable for use with the invention include, but are not limited to, wireless or cellular telephones, regular telephones (without any screen), personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.

[0037] The following is a brief description of a general purpose computer 120 illustrated in FIG. 3. However, the computer 120 is again only one example of a suitable computing environment and is not intended to suggest any limitation as to the scope of use or functionality of the invention. Neither should the computer 120 be interpreted as having any dependency or requirement relating to any one or combination of components illustrated therein.

[0038] The invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices. Tasks performed by the programs and modules are described below and with the aid of figures. Those skilled in the art can implement the description and figures as processor executable instructions, which can be written on any form of a computer readable medium.

[0039] With reference to FIG. 3, components of computer 120 may include, but are not limited to, a processing unit 140, a system memory 150, and a system bus 141 that couples various system components including the system memory to the processing unit 140. The system bus 141 may be any of several types of bus structures including a memory bus or memory controller, a peripheral bus, and a local bus using any of a variety of bus architectures. By way of example, and not limitation, such architectures include Industry Standard Architecture (ISA) bus, Universal Serial Bus (USB), Micro Channel Architecture (MCA) bus, Enhanced ISA (EISA) bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus also known as Mezzanine bus. Computer 120 typically includes a variety of computer readable mediums. Computer readable mediums can be any available media that can be accessed by computer 120 and includes both volatile and nonvolatile media, removable and non-removable media. By way of example, and not limitation, computer readable mediums may comprise computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by computer 120.

[0040] Communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. The term “modulated data signal” means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, FR, infrared and other wireless media. Combinations of any of the above should also be included within the scope of computer readable media.

[0041] The system memory 150 includes computer storage media in the form of volatile and/or nonvolatile memory such as read only memory (ROM) 151 and random access memory (RAM) 152. A basic input/output system 153 (BIOS), containing the basic routines that help to transfer information between elements within computer 120, such as during start-up, is typically stored in ROM 151. RAM 152 typically contains data and/or program modules that are immediately accessible to and/or presently being operated on by processing unit 140. By way of example, and not limitation, FIG. 3 illustrates operating system 54, application programs 155, other program modules 156, and program data 157.

[0042] The computer 120 may also include other removable/non-removable volatile/nonvolatile computer storage media. By way of example only, FIG. 3 illustrates a hard disk drive 161 that reads from or writes to non-removable, nonvolatile magnetic media, a magnetic disk drive 171 that reads from or writes to a removable, nonvolatile magnetic disk 172, and an optical disk drive 175 that reads from or writes to a removable, nonvolatile optical disk 176 such as a CD ROM or other optical media. Other removable/non-removable, volatile/nonvolatile computer storage media that can be used in the exemplary operating environment include, but are not limited to, magnetic tape cassettes, flash memory cards, digital versatile disks, digital video tape, solid state RAM, solid state ROM, and the like. The hard disk drive 161 is typically connected to the system bus 141 through a non-removable memory interface such as interface 160, and magnetic disk drive 171 and optical disk drive 175 are typically connected to the system bus 141 by a removable memory interface, such as interface 170.

[0043] The drives and their associated computer storage media discussed above and illustrated in FIG. 3, provide storage of computer readable instructions, data structures, program modules and other data for the computer 120. In FIG. 3, for example, hard disk drive 161 is illustrated as storing operating system 164, application programs 165, other program modules 166, and program data 167. Note that these components can either be the same as or different from operating system 154, application programs 155, other program modules 156, and program data 157. Operating system 164, application programs 165, other program modules 166, and program data 167 are given different numbers here to illustrate that, at a minimum, they are different copies.

[0044] A user may enter commands and information into the computer 120 through input devices such as a keyboard 182, a microphone 183, and a pointing device 181, such as a mouse, trackball or touch pad. Other input devices (not shown) may include a joystick, game pad, satellite dish, scanner, or the like. These and other input devices are often connected to the processing unit 140 through a user input interface 180 that is coupled to the system bus, but may be connected by other interface and bus structures, such as a parallel port, game port or a universal serial bus (USB). A monitor 184 or other type of display device is also connected to the system bus 141 via an interface, such as a video interface 185. In addition to the monitor, computers may also include other peripheral output devices such as speakers 187 and printer 186, which may be connected through an output peripheral interface 188.

[0045] The computer 120 may operate in a networked environment using logical connections to one or more remote computers, such as a remote computer 194. The remote computer 194 may be a personal computer, a hand-held device, a server, a router, a network PC, a peer device or other common network node, and typically includes many or all of the elements described above relative to the computer 120. The logical connections depicted in FIG. 3 include a local area network (LAN) 191 and a wide area network (WAN) 193, but may also include other networks. Such networking environments are commonplace in offices, enterprise-wide computer networks, intranets and the Internet.

[0046] When used in a LAN networking environment, the computer 120 is connected to the LAN 191 through a network interface or adapter 190. When used in a WAN networking environment, the computer 120 typically includes a modem 192 or other means for establishing communications over the WAN 193, such as the Internet. The modem 192, which may be internal or external, may be connected to the system bus 141 via the user input interface 180, or other appropriate mechanism. In a networked environment, program modules depicted relative to the computer 120, or portions thereof, may be stored in the remote memory storage device. By way of example, and not limitation, FIG. 3 illustrates remote application programs 195 as residing on remote computer 194. It will be appreciated that the network connections shown are exemplary and other means of establishing a communications link between the computers may be used.

[0047] FIG. 4 illustrates architecture 200 for web based recognition as can be used with the present invention. Generally, information stored in a web server 202 can be accessed through mobile device 30 (which herein also represents other forms of computing devices having a display screen, a microphone, a camera, a touch sensitive panel, etc., as required based on the form of input), or through phone 80 wherein information is requested audibly or through tones generated by phone 80 in response to keys depressed and wherein information from web server 202 is provided only audibly back to the user.

[0048] In this exemplary embodiment, architecture 200 is unified in that whether information is obtained through device 30 or phone 80 using speech recognition, a single recognition server 204 can support either mode of operation. In addition, architecture 200 operates using an extension of well-known markup languages (e.g. HTML, XHTML, cHTML, XML, WML, and the like). Thus, information stored on web server 202 can also be accessed using well-known GUI methods found in these markup languages. By using an extension of well-known markup languages, authoring on the web server 202 is easier, and legacy applications currently existing can be also easily modified to include voice or other forms of recognition.

[0049] Generally, device 30 executes HTML+ scripts, or the like, provided by web server 202. When voice recognition is required, by way of example, speech data, which can be digitized audio signals or speech features wherein the audio signals have been preprocessed by device 30 as discussed above, are provided to recognition server 204 with an indication of a grammar or language model to use during speech recognition. The implementation of the recognition server 204 can take many forms, one of which is illustrated, but generally includes a recognizer 211. The results of recognition are provided back to device 30 for local rendering if desired or appropriate. Upon compilation of information through recognition and any graphical user interface if used, device 30 sends the information to web server 202 for further processing and receipt of further HTML scripts, if necessary.

[0050] As illustrated in FIG. 4, device 30, web server 202 and recognition server 204 are commonly connected, and separately addressable, through a network 205, herein a wide area network such as the Internet. It therefore is not necessary that any of these devices be physically located adjacent to each other. In particular, it is not necessary that web server 202 includes recognition server 204. In this manner, authoring at web server 202 can be focused on the application to which it is intended without the authors needing to know the intricacies of recognition server 204. Rather, recognition server 204 can be independently designed and connected to the network 205, and thereby, be updated and improved without further changes required at web server 202. As discussed below, web server 202 can also include an authoring mechanism that can dynamically generate client-side markups and scripts. In a further embodiment, the web server 202, recognition server 204 and client 30 may be combined depending on the capabilities of the implementing machines. For instance, if the client comprises a general purpose computer, e.g. a personal computer, the client may include the recognition server 204. Likewise, if desired, the web server 202 and recognition server 204 can be incorporated into a single machine.

[0051] Access to web server 202 through phone 80 includes connection of phone 80 to a wired or wireless telephone network 208, that in turn, connects phone 80 to a third party gateway 210. Gateway 210 connects phone 80 to a telephony voice browser 212. Telephone voice browser 212 includes a media server 214 that provides a telephony interface and a voice browser 216. Like device 30, telephony voice browser 212 receives HTML scripts or the like from web server 202. In one embodiment, the HTML scripts are of the form similar to HTML scripts provided to device 30. In this manner, web server 202 need not support device 30 and phone 80 separately, or even support standard GUI clients separately. Rather, a common markup language can be used. In addition, like device 30, voice recognition from audible signals transmitted by phone 80 are provided from voice browser 216 to recognition server 204, either through the network 205, or through a dedicated line 207, for example, using TCP/IP. Web server 202, recognition server 204 and telephone voice browser 212 can be embodied in any suitable computing environment such as the general purpose desktop computer illustrated in FIG. 3.

[0052] However, it should be noted that if DTMF recognition is employed, this form of recognition would generally be performed at the media server 214, rather than at the recognition server 204. In other words, the DTMF grammar would be used by the media server 214.

[0053] Referring back to FIG. 4, web server 202 can include a server side plug-in authoring tool or module 209 (e.g. ASP, ASP+, ASP.Net by Microsoft Corporation, JSP, Javabeans, or the like). Server side plug-in module 209 can dynamically generate client-side markups and even a specific form of markup for the type of client accessing the web server 202. The client information can be provided to the web server 202 upon initial establishment of the client/server relationship, or the web server 202 can include modules or routines to detect the capabilities of the client device. In this manner, server side plug-in module 209 can generate a client side markup for each of the voice recognition scenarios, i.e. voice only through phone 80 or multimodal for device 30. By using a consistent client side model, application authoring for many different clients is significantly easier.

[0054] In addition to dynamically generating client side markups, high-level dialog modules, discussed below, can be implemented as a server-side control stored in store 211 for use by developers in application authoring. In general, the high-level dialog modules 211 would generate dynamically client-side markup and script in both voice-only and multimodal scenarios based on parameters specified by developers. The high-level dialog modules 211 can include parameters to generate client-side markups to fit the developers' needs.

[0055] Before describing further aspect of the present invention, it may be helpful to first discuss an exemplary form of extensions to the markup language for use in web based recognition.

[0056] As indicated above, the markup languages such as HTML, XHTML cHTML, XML, WML or any other SGML-derived markup, which are used for interaction between the web server 202 and the client device 30 and phone 80, are extended to include controls and/or objects that provide recognition in a client/server architecture. Generally, controls and/or objects can include one or more of the following functions: recognizer controls and/or objects for recognizer configuration, recognizer execution and/or post-processing; synthesizer controls and/or objects for synthesizer configuration and prompt playing; grammar controls and/or objects for specifying input grammar resources; and/or binding controls and/or objects for processing recognition results. The extensions are designed to be a lightweight markup layer, which adds the power of an audible, visual, handwriting, etc. interface to existing markup languages. As such, the extensions can remain independent of: the high-level page in which they are contained, e.g. HTML; the low-level formats which the extensions used to refer to linguistic resources, e.g. the text-to-speech and grammar formats; and the individual properties of the recognition and speech synthesis platforms used in the recognition server 204.

[0057] It should be noted, a markup language extension such as speech application language tags (SALT) can be used. SALT is a developing standard for enabling access to information, applications and web services from personal computers, telephones, tablet PCs and wireless mobile devices, for example. SALT extends existing markup languages such as HTML, XHTML and XML. An example of the SALT specification can be found in Published U.S. patent application, U.S. 2003/0130854, entitled APPLICATION ABSTRACTION WITH DIALOG PURPOSE, which is herein incorporated by reference in its entirety. The SALT specification may be found online at http://www.SALTforum.org. Further details regarding the extensions are not necessary for understanding the present invention.

[0058] Although speech recognition will be discussed below, it should be understood that the techniques, tags and server side controls described hereinafter can be similarly applied in handwriting recognition, gesture recognition and image recognition.

[0059] At this point though, a particular mode of entry should be discussed. In particular, use of speech recognition in conjunction with at least a display and, in a further embodiment, a pointing device as well which enables the coordination of multiple modes of input, e.g. to indicate the fields for data entry, is particularly useful. Specifically, in this mode of data entry, the user is generally able to coordinate the actions of the pointing device with the speech input, so for example the user is under control of when to select a field and provide corresponding information relevant to the field. For instance, a credit card submission graphical user interface (GUI) is illustrated in FIG. 5, a user could first decide to enter the credit card number in field 252 and then enter the type of credit card in field 250 followed by the expiration date in field 254. Likewise, the user could return back to field 252 and correct an errant entry, if desired. When combined with speech recognition, an easy and natural form of navigation is provided. As used herein, this form of entry using both a screen display allowing free form actions of the pointing device on the screen, e.g. the selection of fields and recognition is called “multimodal”. When rendered using the phone 80 in a voice-only application, the user would be prompted to provide the information illustrated in FIG. 5.

[0060] As indicated above, server side plug-in module 209 outputs client side markups when a request has been made from the client device 30 or telephony voice browser 212. Although possibly described below with respect to the client device, it should be understood that the telephony voice browser 212 is inferred as an example device for voice-only applications. In short, the server side plug-in module 209 allows the website, and thus, the application and services provided by the application to be defined or constructed. The instructions in the server side plug-in module 209 are made of a complied code. The code is run when a web request reaches the web server 202. The server side plug-in module 209 then outputs a new client side markup page that is sent to the client device 30 or telephony voice browser 212. As is well known, this process is commonly referred to as rendering. The server side plug-in module 209 operates on “controls” that abstract and encapsulate the markup language, and thus, the code of the client side markup page. Such controls that abstract and encapsulate the markup language and operate on the webserver 202 include or are equivalent to “Servlets” or “Server-side plug ins” to name a few.

[0061] As is known, server side plug-in modules of the prior art can generate client side markup for visual rendering and interaction with the client device 30. Three different approaches are provided herein for extending the server side plug-in module 209 to include recognition and audible prompting extensions such as the exemplary client side extensions discussed above. In a first approach illustrated schematically in FIG. 6, the current, visual, server side controls (which include parameters for visual display such as location for rendering, font, foreground color, background color, etc.) are extended to include parameters or attributes for recognition and audibly prompting for related recognition. Using speech recognition and associated audible prompting by way of example, the attributes generally pertain to audible prompting parameters such as whether the prompt comprises inline text for text-to-speech conversion, playing of a prerecorded audio file (e.g. a wave file), the location of the data (text for text-to-speech conversion or a prerecorded audio file) for audible rendering, etc. For recognition, the parameters or attributes can include the location of the grammar to be used during recognition, confidence level thresholds, etc. Since the server side plug-in module 209 generates client side markup, the parameters and attributes for the controls for the server side plug-in module 209 relate to the extensions provided in the client side markup for recognition and/or audible prompting.

[0062] The controls indicated at 300A in FIG. 6 are controls, which are well-known in website application development or authoring tools such as ASP, ASP+, ASP.Net, JSP, Javabeans, or the like. Such controls are commonly formed in a library and used by controls 302 to perform a particular visual task. Library 300A includes methods for generating the desired client markup, event handlers, etc. Examples of visual controls 302 include a “Label” control that provides a selected text label on a visual display such as the label “Credit Card Submission” 304 in FIG. 5. Another example of a higher level visual control 302 is a “Textbox”, which allows data to be entered in a data field such as is indicated at 250 in FIG. 5. The existing visual controls 302 are also well-known. In the first approach for extending server side plug-in module controls to include recognition and/or audible prompting, each of the visual controls 302 would include further parameters or attributes related to recognition or audible prompting. In the case of the “label” control, which otherwise provides selected text on a visual display, further attributes may include whether an audio data file will be rendered or text-to-speech conversion will be employed as well as the location of this data file. A library 300B, similar to library 300A, includes further markup information for performing recognition and/or audible prompting. Each of the visual controls 302 is coded so as to provide this information to the controls 300B as appropriate to perform the particular task related to recognition or audible prompting.

[0063] As another example, the “Textbox” control, which generates an input field on a visual display and allows the user of the client device 30 to enter information, would also include appropriate recognition or audible prompting parameters or attributes such as the grammar to be used for recognition. It should be noted that the recognition or audible prompting parameters are optional and need not be used if recognition or audible prompting is not otherwise desired.

[0064] In general, if a control at level 302 includes parameters that pertain to visual aspects, the control will access and use the library 300A. Likewise, if the control includes parameters pertaining to recognition and/or audible prompting the control will access or use the library 300B. It should be noted that libraries 300A and 300B have been illustrated separately in order to emphasize the additional information present in library 300B and that a single library having the information of libraries 300A and 300B can be implemented.

[0065] In this approach, each of the current or prior art visual controls 302 are extended to include appropriate recognition/audible prompting attributes. The controls 302 can be formed in a library. The server side plug-in module 209 accesses the library for markup information. Execution of the controls generates a client side markup page, or a portion thereof, with the provided parameters.

[0066] In a second approach illustrated in FIG. 7, new visual, recognition/audible prompting controls 304 are provided such that the controls 304 are a subclass relative to visual controls 302, wherein recognition/audible prompting functionality or markup information is provided at controls 304. In other words, a new set of controls 304 are provided for recognition/audible prompting and include appropriate parameters or attributes to perform the desired recognition or an audible prompting related to a recognition task on the client device 30. The controls 304 use the existing visual controls 302 to the extent that visual information is rendered or obtained through a display. For instance, a control “SpeechLabel” at level 304 uses the “Label” control at level 302 to provide an audible rendering and/or visual text rendering. Likewise, a “SpeechTextbox” control would associate a grammar and related recognition resources and processing with an input field. Like the first approach, the attributes for controls 304 include where the grammar is located for recognition, the inline text for text-to-speech conversion, or the location of a prerecorded audio data file that will be rendered directly or a text file through text-to-speech conversion. The second approach is advantageous in that interactions of the recognition controls 304 with the visual controls 302 are through parameters or attributes, and thus, changes in the visual controls 302 may not require any changes in the recognition controls 304 provided the parameters or attributes interfacing between the controls 304 and 302 are still appropriate. However, with the creation of further visual controls 302, a corresponding recognition/audible prompting control at level 304 may also have to be written.

[0067] A third approach is illustrated in FIG. 8. Generally, controls 306 of the third approach are separate from the visual controls 302, but are associated selectively therewith as discussed below. In this manner, the controls 306 do not directly build upon the visual controls 302, but rather provide recognition/audible prompting enablement without having to rewrite the visual controls 302. The controls 306, like the controls 302, use a library 300. In this embodiment, library 300 includes both visual and recognition/audible prompting markup information and as such is a combination of libraries 300A and 300B of FIG. 6.

[0068] There are significant advantages to this third approach. Firstly, the visual controls 302 do not need to be changed in content. Secondly, the controls 306 can form a single module which is consistent and does not need to change according to the nature of the speech-enabled control 302. Thirdly, the process of speech enablement, that is, the explicit association of the controls 306 with the visual controls 302 is fully under the developer's control at design time, since it is an explicit and selective process. This also makes it possible for the markup language of the visual controls to receive input values from multiple sources such as through recognition provided by the markup language generated by controls 306, or through a conventional input device such as a keyboard. In short, the controls 306 can be added to an existing application authoring page of a visual authoring page of the server side plug-in module 209. The controls 306 provide a new modality of interaction (i.e. recognition and/or audible prompting) for the user of the client device 30, while reusing the visual controls' application logic and visual input/output capabilities. In view that the controls 306 can be associated with the visual controls 302 whereat the application logic can be coded, controls 306 may be hereinafter referred to as “companion controls 306” and the visual controls 302 be referred to as “primary controls 302”. It should be noted that these references are provided for purposes of distinguishing controls 302 and 306 and are not intended to be limiting. For instance, the companion controls 306 could be used to develop or author a website that does not include visual renderings such as a voice-only website. In such a case, certain application logic could be embodied in the companion control logic.

[0069] A first exemplary set of companion controls 306 are further illustrated in FIG. 9. The set of companion controls 306 can be grouped as output controls 308 and input controls 310. Output controls 308 provide “prompting” client side markups, which typically involves the playing of a prerecorded audio file, or text for text-to-speech conversion, the data included in the markup directly or referenced via a URL. Although a single output control can be defined with parameters to handle all audible prompting, in the exemplary embodiment, the forms or types of audible prompting in a human dialog are formed as separate controls. In particular, the output controls 308 can include a “Question” control 308A, a “Confirmation” control 308B and a “Statement” control 308C, which will be discussed in detail below. Likewise, the input controls 310 can also form or follow human dialog and include a “Answer” control 310A and a “Command” control 310B. The input controls 310 are discussed below, but generally the input controls 310 associate a grammar with expected or possible input from the user of the client device 30.

[0070] Although the question control 308A, confirmation control 308B, statement control 308C, answer control 310A, command control 310B, other controls as well as the general structure of these controls, the parameters and event handlers, are specifically discussed with respect to use as companion controls 306, it should be understood that these controls, the general structure, parameters and event handlers can be adapted to provide recognition and/or audible prompting in the other two approaches discussed above with respect to FIGS. 6 and 7. For instance, the parameter “ClientToSpeechEnable”, which comprises one exemplary mechanism to form the association between a companion control and a visual control, would not be needed when embodied in the approaches of FIGS. 6 and 7.

[0071] In a multimodal application, at least one of the output controls 308 or one of the input controls 310 is associated with a primary or visual control 302. In the embodiment illustrated, the output controls 308 and input controls 310 are arranged or organized under a “Question/Answer” (hereinafter also “QA”) control 320. QA control 320 is executed on the web server 202, which means it is defined on the application development web page held on the web server using the server-side markup formalism (ASP, JSP or the like), but is output as a different form of markup to the client device 30 or telephony voice browser 212. Although illustrated in FIG. 9 where the QA control appears to be formed of all of the output controls 308 and the input controls 310, it should be understood that these are merely options wherein one or more may be included for a QA control.

[0072] At this point it may be helpful to explain use of the controls 308 and 310 in terms of application scenarios. Referring to FIG. 10 and in a voice-only application QA control 320 could comprise a single question control 308A and an answer control 310A. The question control 308A contains one or more prompt objects or controls 322, while the answer control 310A can define a grammar through grammar object or control 324 for recognition of the input data and related processing on that input. Line 326 represents the association of the QA control 320 with the corresponding primary control 302, if used. In a multimodal scenario, where the user of the client device 30 may touch on the visual textbox, for example with a “TapEvent”, an audible prompt may not be necessary. For example, for a primary control comprising a textbox having visual text forming an indication of what the user of client device should enter in the corresponding field, a corresponding QA control 320 may or may not have a corresponding prompt such as an audio playback or a text-to-speech conversion, but would have a grammar corresponding to the expected value for recognition, and event handlers 328 to process the input, or process other recognizer events such as no speech detected, speech not recognized, or events fired on timeouts (as illustrated in “Eventing” below).

[0073] In general, the QA control through the output controls 308 and input controls 310 and additional logic can perform one or more of the following: provide output audible prompting, collect input data, perform confidence validation of the input result, allow additional types of input such as “help” commands, or commands that allow the user of the client device to navigate to other selected areas of the website, allow confirmation of input data and control of dialog flow at the website, to name a few. In short, the QA control 320 contains all the controls related to a specific topic. In this manner, a dialog is created through use of the controls with respect to the topic in order to inform to obtain information, to confirm validity, or to repair a dialog or change the topic of conversation.

[0074] In one method of development, the application developer can define the visual layout of the application using the visual controls 302. The application developer can then define the spoken interface of the application using companion controls 306 (embodied as QA control 320, or output controls 308 and input control 310). As illustrated in FIGS. 9 and 10, each of the companion controls 306 are then linked or otherwise associated with the corresponding primary or visual control 302 to provide recognition and audible prompting. Of course if desired, the application developer can define or encode the application by switching between visual controls 302 and companion controls 306, forming the links therebetween, until the application is completely defined or encoded.

[0075] At this point, it may be helpful to provide a short description of each of the output controls 308 and input controls 310. Detailed descriptions are provided below for this embodiment in Appendix A.

Questions, Answers and Commands

[0076] Generally, as indicated above, the question controls 308A and answer controls 310A in a QA control 320 hold the prompt and grammar resources relevant to the primary control 302, and related binding (associating recognition results with input fields of the client-side markup page) and processing logic. The presence, or not, of question controls 308A and answer controls 310A determines whether speech output or recognition input is enabled on activation. Command controls 310B and user initiative answers are activated by specification of the Scope property on the answer controls 310A and command controls 310B.

[0077] In simple voice-only applications, a QA control 320 will typically hold one question control or object 308A and one answer control or object 310A. Although not shown in the example below, command controls 310B may also be specified, e.g. Help, Repeat, Cancel, etc., to enable user input which does not directly relate to the answering of a particular question.

[0078] A typical ‘regular’ QA control for voice-only dialog is as follows:

[0079]

	<Speech:QA
	id=“QA_WhichOne”
	ControlsToSpeechEnable=“textBox1”
	runat=“server” >
	<Question >
	<prompt> Which one do you want?
	</prompt>
	</Question>
	<Answer >
	<grammar src=“whichOne.gram” />
	</Answer>
	</Speech:QA>

[0080] (The examples provided herein are written in the ASP.Net framework by example only and should not be considered as limiting the present invention.)

[0081] In this example, the QA control can be identified by its “id”, while the association of the QA control with the desired primary or visual control is obtained through the parameter “ControlsToSpeechEnable”, which identifies one or more primary controls by their respective identifiers. If desired, other well-known techniques can be used to form the association. For instance, direct, implicit associations are available through the first and second approaches described above, or separate tables can be created used to maintain the associations. The parameter “runat” instructs the web server that this code should be executed at the webserver 202 to generate the correct markup.

[0082] A QA control might also hold only a statement control 308C, in which case it is a prompt-only control without active grammars (e.g. for a welcome prompt). Similarly a QA control might hold only an answer control 310A, in which case it may be a multimodal control, whose answer control 310A activates its grammars directly as the result of an event from the GUI, or a scoped mechanism (discussed below) for user initiative.

[0083] It should also be noted that a QA control 320 may also hold multiple output controls 308 and input controls 310 such as multiple question controls 308A and multiple answers controls 310A. This allows an author to describe interactional flow about the same entity within the same QA control. This is particularly useful for more complex voice-only dialogs. So a mini-dialog which may involve different kinds of question and answer (e.g. asking, confirming, giving help, etc.), can be specified within the wrapper of the QA control associated with the visual control which represents the dialog entity. A complex QA control is illustrated in FIG. 10.

[0084] The foregoing represent the main features of the QA control. Each feature is described from a functional perspective below.

Answer Control

[0085] The answer control 310A abstracts the notion of grammars, binding and other recognition processing into a single object or control. Answer controls 310A can be used to specify a set of possible grammars relevant to a question, along with binding declarations and relevant scripts. Answer controls for multimodal applications such as “Tap-and-Talk” are activated and deactivated by GUI browser events. The following example illustrates an answer control 310A used in a multimodal application to select a departure city on the “mouseDown” event of the textbox “txtDepCity”, and write its value into the primary textbox control:

[0086]

	<Speech:QA
	controlsToSpeechEnable=“txtDepCity”
	runat=“server”>
	<Answer	id=“AnsDepCity”
	StartEvent=“onMouseDown”
	StopEvent=“onMouseUp”
	/>
	<grammar src=“/grammars/depCities.gram”/>
	<bind value=“//sml/DepCity”
	targetElement=“txtCity” />
	</Answer>
	</Speech:QA>

[0087] Typical answer controls 310A in voice-only applications are activated directly by question controls 308A as described below.

[0088] The answer control further includes a mechanism to associate a received result with the primary controls. Herein, binding places the values in the primary controls; however, in another embodiment the association mechanism may allow the primary control to look at or otherwise access the recognized results.

Question Control

[0089] Question controls 308A abstracts the notion of the prompt tags into an object which contains a selection of possible prompts and the answer controls 310A which are considered responses to the question. Each question control 308A is able to specify which answer control 310A it activates on its execution. This permits appropriate response grammars to be bundled into answer controls 310A, which reflect relevant question controls 308A.

[0090] The following question control 308A might be used in a voice-only application to ask for a Departure City:

[0091]

	<Speech:QA id=“QADepCity”
	controlsToSpeechEnable=“txtDepCity”
	runat=“server” >
	<Question id=“Q1” Answers=“AnsDepCity” >
	<prompt>
	Please give me the departure
	city.
	</prompt>
	</Question>
	<Answer id=“AnsDepCity” ... />
	</Speech:QA>

[0092] In the example below, different prompts can be called depending on an internal condition of the question control 308A. The ability to specify conditional tests on the prompts inside a question control 308A means that changes in wording can be accommodated within the same functional unit of the question control 308A.

[0093]

	<Speech:QA	id=“QADepCity”
	controlsToSpeechEnable=“txtDepCity”
	runat=“server” >
	<Question id=“Q1” Answers=“AnsDepCity” >
	<prompt count=“1”>
	Now I need to get the departure city.
	Where would you like to fly from?
	</prompt>
	<prompt count=“2”>
	Which departure city?
	</prompt>
	</Question>
	<Answer id=“AnsDepCity” ... />
	</Speech:QA>

Conditional QA Control

[0094] The following example illustrates how to determine whether or not to activate a QA control based upon information known to the application. The example is a portion of a survey application. The survey is gathering information from employees regarding the mode of transportation they use to get to work.

[0095] The portion of the survey first asks whether or not the user rides the bus to work. If the answer is:

• • [0096] Yes, the next question asks how many days last week the users rode the bus.
  • [0097] No, the “number of days rode the bus” question is bypassed.

[0098]

	<asp:Label id=“lblDisplay1”
	text=“Do you ride the bus to work?”
	runat=“server”/>
	<asp:DropDownList id=“lstRodeBusYN” runat=“server”>
	<asp:ListItem
	selected=“true”>No</asp:ListItem>
	<asp:ListItem>Yes</asp:ListItem>
	</asp:DropDownList>
	<Speech:QA id=“QA_RideBus
	ControlsToSpeechEnable=“lstRodeBusYN”
	runat=“server” >
	<SDN:Question id=“Q_RideBus” >
	<prompt bargeIn=“False”>
	Do you ride the bus to work?
	</prompt>
	</SDN:Question>
	<SDN:Answer id=“A_RideBus”
	autobind=“False”
	StartEvent=“onMouseDown”
	StopEvent=“onMouseUp”
	runat=“server”
	onClientReco=“ProcessRideBusAnswer”
	<grammar src=“ . . . ” /> <--! “yes/no”
	grammar -->
	</SDN:Answer>
	</Speech:QA>
	<asp:Label id=“lblDisplay2”
	enabled=“False”
	text=“How many days last week did you ride
	the bus to work?”
	runat=“server”/>
	<asp:DropDownList id=“lstDaysRodeBus” enabled=“False”
	runat=“server”>
	<asp:ListItem selected=“true”
	>0</asp:ListItem>
	<asp:ListItem>1</asp:ListItem>
	<asp:ListItem>2</asp:ListItem>
	<asp:ListItem>3</asp:ListItem>
	<asp:ListItem>4</asp:ListItem>
	<asp:ListItem>5</asp:ListItem>
	<asp:ListItem>6</asp:ListItem>
	<asp:ListItem>7</asp:ListItem>
	</asp:DropDownList>
	<Speech:QA id=“QA_DaysRodeBus”
	ControlsToSpeechEnable=“lstDaysRodeBus”
	ClientTest=“RideBusCheck”
	runat=“server” >
	<Question id=“Q_DaysRodeBus” >
	<prompt bargeIn=“False”>
	How many days last week did you ride the
	bus to work?
	</prompt>
	</SDN:Question>
	<SDN:Answer id=“A_DaysRodeBus”
	autobind=“False”
	StartEvent=“onMouseDown”
	StopEvent=“onMouseUp”
	runat=“server”
	onClientReco=“ProcessDaysRodeBusAnswer”
	<grammar src=“ . . . ” /> <--! “numbers”
	grammar -->
	</SDN:Answer>
	</Speech:QA>
	<script language=“jscript”>
	function ProcessRideBusAnswer( ) {
	<--! using SML attribute of the Event object,
	determine yes or no answer -->
	<--! then select the appropriate item in the
	dropdown listbox	-->
	<--! and enable the next label and dropdown
	listbox if answer is “yes”	-->
	if <--! Answer is “yes” --> {
	lstRodeBusYN.selectedIndex=2
	lblDisplay2.enabled=“true”
	lstDaysRodeBus.enabled=“true” }
	}
	function RideBusCheck( ) {
	if lstRodeBusYN.selectedIndex=“1” <--!
	this is no -->
	then return “False”
	endif
	}
	function ProcessDaysRodeBusAnswer( ) {
	<--! case statement to select proper
	dropdown item -->
	}
	</script>

[0099] In the example provided above, the QA control “QA_DaysRodeBus” is executed based on a boolean parameter “ClientTest”, which in this example, is set based on the function RideBusCheck( ). If the function returns a false condition, the QA control is not activated, whereas if a true condition is returned the QA control is activated. The use of an activation mechanism allows increased flexibility and improved dialog flow in the client side markup page produced. As indicated in Appendix A many of the controls and objects include an activation mechanism.

Command Control

[0100] Command controls 310B are user utterances common in voice-only dialogs which typically have little semantic import in terms of the question asked, but rather seek assistance or effect navigation, e.g. help, cancel, repeat, etc. The Command control 310B within a QA control 306 can be used to specify not only the grammar and associated processing on recognition (rather like an answer control 310A without binding of the result to an input field), but also a ‘scope’ of context and a type. This allows for the authoring of both global and context-sensitive behavior on the client side markup.

[0101] As appreciated by those skilled in the art from the foregoing description, controls 306 can be organized in a tree structure similar to that used in visual controls 302. Since each of the controls 306 are also associated with selected visual controls 302, the organization of the controls 306 can be related to the structure of the controls 302.

[0102] The QA controls 302 may be used to speech-enable both atomic controls (textbox, label, etc.) and container controls (form, panel, etc.) This provides a way of scoping behaviour and of obtaining modularity of subdialog controls. For example, the scope will allow the user of the client device to navigate to other portions of the client side markup page without completing a dialog.

[0103] In one embodiment, “Scope” is determined as a node of the primary controls tree. The following is an example “help” command, scoped at the level of the “Pnl1” container control, which contains two textboxes.

[0104]

	<asp:panel id=“Pnl1” ...>
	<asp:textbox id=“tb1” ... />
	<asp:textbox id=“tb2” ... />
	</asp:panel>
	<Speech:QA ... >
	<Command
	id=“HelpCmd1”
	scope=“Pnl1”
	type=“help”
	onClientReco=“GlobalGiveHelp( )” >
	<Grammar src=“grammars/help.gram”/>
	</Command>
	</Speech:QA>
	<script>
	function GlobalGiveHelp( ) {
	...
	}
	</script>

[0105] As specified, the “help” grammar will be active in every QA control relating to “Pnl1” and its contents. The GlobalGiveHelp subroutine will execute every time “help” is recognized. To override this and achieve context-sensitive behavior, the same typed command can be scoped to the required level of context:

[0106]

	<Speech:QA ... >
	<Command
	id=“HelpCmd2”
	scope=“Tb2”
	type=“help”
	onClientReco=“SpecialGiveHelp( )” >
	<Grammar src=“grammars/help.gram”/>
	</Command>
	</Speech:QA>
	<script>
	function SpecialGiveHelp( ) {
	...
	}
	</script>

Confirmation Control

[0107] The QA control 320 can also include a method for simplifying the authoring of common confirmation subdialogs. The following QA control exemplifies a typical subdialog which asks and then confirms a value:

[0108]

	<Speech:QA
	id=“qaDepCity”
	controlsToSpeechEnable=“txtDepCity”
	runat=“server” >
	<!-- asking for a value -->
	<Question	id=“AskDepCity”
	type=“ask”
	Answers=“AnsDepCity” >
	<prompt> Which city? </prompt>
	</Question>
	<Answer	id=“AnsDepCity”
		confirmThreshold=“60” >
	<grammar src=“grammars/depCity.gram”
	/>
	</Answer>
	<!-- confirming the value -->
	<Confirm	id=“ConfirmDepCity”
		Answers=“AnsConfDepCity” >
	<prompt>
	Did you say <value
	targetElement=“txtDepCity/Text”>?
	</prompt>
	</Confirm>
	<Answer	id=“AnsConfDepCity” >
	<grammar
	src=“grammars/YesNoDepCity.gram” />
	</Answer>
	</Speech:QA>

[0109] In this example, a user response to ‘which city?’ which matches the AnsDepCity grammar but whose confidence level does not exceed the confirmThreshold value will trigger the confirm control 308. More flexible methods of confirmation available to the author include mechanisms using multiple question controls and multiple answer controls.

[0110] In a further embodiment, additional input controls related to the confirmation control include an accept control, a deny control and a correct control. Each of these controls could be activated (in a manner similar to the other controls) by the corresponding confirmation control and include grammars to accept, deny or correct results, respectively. For instance, users are likely to deny be saying “no”, to accept by saying “yes” or “yes+current value” (e.g., “Do you want to go to Seattle?” “Yes, to Seattle”), to correct by saying “no”+new value (e.g., “Do you want to go to Seattle” “No, Pittsburgh”).

Statement Control

[0111] The statement control allows the application developer to provide an output upon execution of the client side markup when a response is not required from the user of the client device 30. An example could be a “Welcome” prompt played at the beginning of execution of a client side markup page.

[0112] An attribute can be provided in the statement control to distinguish different types of information to be provided to the user of the client device. For instance, attributes can be provided to denote a warning message or a help message. These types could have different built-in properties such as different voices. If desired, different forms of statement controls can be provided, i.e. a help control, warning control, etc. Whether provided as separate controls or attributes of the statement control, the different types of statements have different roles in the dialog created, but share the fundamental role of providing information to the user of the client device without expecting an answer back.

Eventing

[0113] Event handlers as indicated in FIG. 10 are provided in the QA control 320, the output controls 308 and the input controls 310 for actions/inactions of the user of the client device 30 and for operation of the recognition server 204 to name a few, other events are specified in Appendix A. For instance, mumbling, where the speech recognizer detects that the user has spoken but is unable to recognize the words and silence, where speech is not detected at all, are specified in the QA control 320. These events reference client-side script functions defined by the author. In a multimodal application specified earlier, a simple mumble handler that puts an error message in the textbox could be written as follows:

[0114]

	<Speech:QA
	controlsToSpeechEnable=“txtDepCit
	y”  onClientNoReco=“OnMumble( )”
	runat=“server”>
	<Answer id=“AnsDepCity”
	StartEvent=“onMouseDown”
	StopEvent=“onMouseUp”
	>
	<grammar src=“/grammars/depCities.gram”/>
	<bind value=“//sml/DepCity”
	targetElement=“txtCity” />
	</Answer>
	</Speech:QA>
	<script>
	function OnMumble( ) {
	txtDepCity.value=“ . . .recognition
	error. . . ”;
	}
	</script>

Control Execution Algorithm

[0115] In one embodiment, a client-side script or module (herein referred to as “RunSpeech”) is provided to the client device. The purpose of this script is to execute dialog flow via logic, which is specified in the script when executed on the client device 30, i.e. when the markup pertaining to the controls is activated for execution on the client due to values contained therein. The script allows multiple dialog turns between page requests, and therefore, is particularly helpful for control of voice-only dialogs such as through telephony browser 216. The client-side script RunSpeech is executed in a loop manner on the client device 30 until a completed form in submitted, or a new page is otherwise requested from the client device 30.

[0116] It should be noted that in one embodiment, the controls can activate each other (e.g. question control activating a selected answer control) due to values when executed on the client. However, in a further embodiment, the controls can “activate” each other in order to generate appropriate markup, in which case server-side processing may be implemented.

[0117] Generally, in one embodiment, the algorithm generates a dialog turn by outputting speech and recognizing user input. The overall logic of the algorithm is as follows for a voice-only scenario:

• • [0118] 1. Find next active output companion control;
  • [0119] 2. If it is a statement, play the statement and go back to 1; If it is a question or a confirm go to 3;
  • [0120] 3. Collect expected answers;
  • [0121] 4. Collect commands;
  • [0122] 5. Play output control and listen in for input;
  • [0123] 6. Activate recognized Answer or Command object or, issue an event if none is recognized;
  • [0124] 7. Go back to 1.
    In the multimodal case, the logic is simplified to the following algorithm:
  • [0125] 1. Wait for triggering event—i.e., user tapping on a control;
  • [0126] 2. Collect expected answers;
  • [0127] 3. Listen in for input;
  • [0128] 4. Activate recognized Answer object or, if none, throw event;
  • [0129] 5. Go back to 1.

[0130] The algorithm is relatively simple because, as noted above, controls contain built-in information about when they can be activated. The algorithm also makes use of the role of the controls in the dialogue. For example statements are played immediately, while questions and confirmations are only played once the expected answers have been collected.

[0131] In a further embodiment, implicit confirmation can be provided whereby the system confirms a piece of information and asks a question at the same time. For example the system could confirm the arrival city of a flight and ask for the travel date in one utterance: “When do you want to go to Seattle?” (i.e. asking ‘when’ and implicitly confirming ‘destination: Seattle’). If the user gives a date then the city is considered implicitly accepted since, if the city was wrong, users would have immediately challenged it. In this scenario, it becomes clear that the knowledge of what a user is trying to achieve is vitally important: are they answering the question, or are they correcting the value, or are they asking for help? By using the role of the user input in the dialogue the system can know when to implicitly accept a value.

[0132] In summary, a dialog is created due to the role of the control in the dialog and the relationship with other controls, wherein the algorithm executes the controls and thus manages the dialog. Each control contains information based on its type which is used by the execution algorithm to select (i.e. make active for execution) a given control according to whether or not it serves a useful purpose at that point in the dialog on the client. For example, confirmation controls are only active when there is a value to confirm and the system does not have sufficient confidence in that value to proceed. In a further implementation, most of these built-in pieces of information can be overridden or otherwise adapted by application developers.

[0133] The following table summarizes the controls, their corresponding role in the dialog and the relationship with other controls.

[0134]

			Relationship with other
	Control	Role in dialogue	controls
	Statement	output: present	(none)
		information to
		users
	Question	output: ask	selects expected Answers
		question	as a response
	Confirmation	output: confirm a	selects potential input
		value obtained from	controls as a response,
		the user	typically Accept, Deny,
			Correct
	Answer	input: provide an	selected by
		answer to a	Question/Confirmation
		question
	Command	input: seek to	scoped to other controls
		repair a dialog, or
		change the topic of
		conversation
	Accept	input: confirm a	selected by a
		value in response	confirmation
		to a confirmation
	Deny	input: deny a value	selected by a
		in response to a	confirmation
		confirmation
	Correct	input: correct a	selected by a
		value in response	confirmation
		to a confirmation
	QA		(wrapper: contains all
			the controls related to
			a specific topic)

The use of these controls may be explained with an illustration of a simple human/computer dialog. In the dialog below, each dialog turn on the part of the System or the User is characterized according to the control (indicated in parentheses) which reflects its purpose in the dialog.

• [0135] 1. System (Statement): “Welcome to the travel booking service”.
• [0136] 2. System (Question): “Where would you like to go?”
• [0137] 3. User (Answer): “San Francisco.”
• [0138] 4. System (Confirmation): “Did you say Seattle?”
• [0139] 5. User (Deny): “No.”
• [0140] 6. System (Question): “Where would you like to go?”
• [0141] 7. User (Answer): “San Francisco.”
• [0142] 8. System (Confirmation): “Did you say Seattle?”
• [0143] 9. User (Correct): “I said San Francisco.”
• [0144] 10. System (Confirmation): “Did you say San Francisco?”
• [0145] 11. User (Correct): “Yes.”
• [0146] 12. System (Question): “When would you like to leave?”
• [0147] 13. User (Command): “Help.”

[0148] Turn 1 is a statement on the part of the System. Since a statement control activates no answer controls in response, the system does not expect input. The system goes on to activate a question control at turn 2. This in turn activates a set of possible answer controls, including one which holds a grammar containing the cities available through the service, including “San Francisco”, “Seattle”, etc., which permits the user to provide such a city in turn 3.

[0149] The user's turn 3 is misrecognized by the system. Although the system believes it has a value from an answer control for the city, its confidence in that value is low (rightly so, since it has recognized incorrectly). This low confidence value in a just-received answer control is sufficient information for RunSpeech to trigger a confirmation control on the part of the system, as generated at turn 4. The confirmation control in turn activates a deny control, a correct control and an accept control and makes their respective grammars available to recognize the user's next turn. User turns 5, 9 and 11 illustrate example responses for these controls. Turn 5 of the user simply denies the value “no”. This has the effect of removing the value from the system, so the next action of RunSpeech is to ask the question again to re-obtain the value (turn 6).

[0150] Turns 7 and 8 return us to a confirmation control as with 3 and 4.

[0151] User turn 9 is a correct control, which has again been activated as a possible response to the confirmation control. A correct control not only denies the value undergoing confirmation, it also provides a new value. So user turn 9 is recognized by the system as a correct control with a new value which, correctly this time, is recognized as “San Francisco”.

[0152] The system's confidence in the new value is low, however, and yet another confirmation control is generated at turn 10. This in turn activates accept, deny and correct controls in response, and user turn 11 (“Yes” matches an accept control grammar. The recognition of the accept control has the effect of ‘grounding’ the system's belief in the value which it is trying to obtain, and so RunSpeech is now able to select other empty values to obtain. In turn 12, a new question control is output which asks for a date value. The user's response this time (turn 13) is a command: “help”. Command controls are typically activated in global fashion, that is, independently of the different question controls and confirmation controls on the part of the system. In this way the user is able to ask for help at any time, as he does in turn 13. Command controls may also be more sensitively enabled by a mechanism that scopes their activation according to which part of the primary control structure is being talked about.

[0153] Referring back to the algorithm, in one exemplary embodiment, the client-side script RunSpeech examines the values inside each of the primary controls and an attribute of the QA control, and any selection test of the QA controls on the current page, and selects a single QA control for execution. For example, within the selected QA control, a single question and its corresponding prompt are selected for output, and then a grammar is activated related to typical answers to the corresponding question. Additional grammars may also be activated, in parallel, allowing other commands (or other answers), which are indicated as being allowable. Assuming recognition has been made and any further processing on the input data is complete, the client-side script RunSpeech will begin again to ascertain which QA control should be executed next. An exemplary implementation and algorithm of RunSpeech is provided in Appendix A.

[0154] It should be noted that the use of the controls and the RunSpeech algorithm or module is not limited to the client/server application described above, but rather can be adapted for use with other application abstractions. For instance, an application such as VoiceXML, which runs only on the client device 30 or telephony voice browser 212, could conceivably include further elements or controls such as question and answer provided above as part of the VoiceXML browser and operating in the same manner. In this case the mechanisms of the RunSpeech algorithm described above could be executed by default by the browser without the necessity for extra script. Similarly, other platforms such as finite state machines can be adapted to include the controls and RunSpeech algorithm or module herein described.

Synchronization

[0155] As noted above, the companion controls 306 are associated with the primary controls 302 (the existing controls on the page). As such the companion controls 306 can re-use the business logic and presentation capabilities of the primary controls 302. This is done in two ways: storing values in the primary controls 302 and notifying the primary controls of the changes 302.

[0156] The companion controls 306 synchronize or associates their values with the primary controls 302 via the mechanism called binding. Binding puts values retrieved from recognizer into the primary controls 302, for example putting text into a textbox, herein exemplified with the answer control. Since primary controls 302 are responsible for visual presentation, this provides visual feedback to the users in multimodal scenarios.

[0157] The companion controls 306 also offer a mechanism to notify the primary controls 302 that they have received an input via the recognizer. This allows the primary controls 302 to take actions, such as invoking the business logic. (Since the notification amounts to a commitment of the companion controls 306 to the values which they write into the primary controls 302, the implementation provides a mechanism to control this notification with a fine degree of control. This control is provided by the RejectThreshold and ConfirmThreshold properties on the answer control, which specify numerical acoustic confidence values below which the system should respectively reject or attempt to confirm a value.)

[0158] A second exemplary set of companion controls 400 is illustrated in FIG. 11. In this embodiment, the companion controls 400 generally include a QA control 402, a Command control 404, a CompareValidator control 406, a Custom Validator control 408 and a semantic map 410. The semantic map 410 is schematically illustrated and includes SemanticItemSemanticItems 412 that form a layer between the visual domain primary controls 402 (e.g. HTML and a non-visual recognition domain of the companion controls 400.

[0159] At this point, it should be emphasized that that although the organization of the companion controls QA and Command is different than that of the first set of companion controls discussed above, the functionality remains the same. In particular, the QA control 402 includes a Prompt property that references Prompt objects to perform the functions of output controls, i.e. that provide “prompting” client side markups for human dialog, which typically involves the playing of a prerecorded audio file, or text for text-to-speech conversion, the data included in the markup directly or referenced via a URL. Likewise, the input controls are embodied as the QA control 402 and Command Control 404 and also follow human dialog and include the Prompt property (referencing a Prompt object) and an Answer property that references at least one Answer object. Both the QA control 402 and the Command control 404 associate a grammar with expected or possible input from the user of the client device 30. The QA control 402 in this embodiment can thus be considered a question control, an answer control as well as a confirm control and a statement control since it includes properties necessary for performing these functions.

[0160] Although the QA control 402, Command control 404, Compare Validator control 406 and Custom Validator control 408 and other controls as well as the general structure of these controls, the parameters and event handlers, are specifically discussed with respect to use as companion controls 400, it should be understood that these controls, the general structure, parameters and event handlers can be adapted to provide recognition and/or audible prompting in the other two approaches discussed above with respect to FIGS. 6 and 7. For instance, the Semantic Map 410, which comprises another exemplary mechanism to form the association between the companion controls and visual control 302, would not be needed when embodied in the approaches of FIGS. 6 and 7.

[0161] At this point, it may be helpful to provide a short description of each of the controls. Detailed descriptions are provided below in Appendix B.

QA Control

[0162] In general, the QA control 402 through the properties illustrated can perform one or more of the following: provide output audible prompting, collect input data, perform confidence validation of the input result, allow confirmation of input data and aid in control of dialog flow at the website, to name a few. In other words, the QA control 402 contains properties that function as controls for a specific topic.

[0163] The QA control 402, like the other controls, is executed on the web server 202, which means it is defined on the application development web page held on the web server using the server-side markup formalism (ASP, JSP or the like), but is output as a different form of markup to the client device 30. Although illustrated in FIG. 11 where the QA control appears to be formed of all of the properties Prompt, Reco, Answers, ExtraAnswers and Confirms, it should be understood that these are merely options wherein one or more may be included for a QA control.

[0164] At this point it may be helpful to explain use of the QA controls 402 in terms of application scenarios. Referring to FIG. 11 and in a voice-only application QA control 402 could function as a question and an answer in a dialog. The question would be provided by a Prompt object, while a grammar is defined through grammar object for recognition of the input data and related processing on that input. An Answers property associates the recognized result with a SemanticItem 412 in the Semantic Map 410 using an Answer object, which contains information on how to process recognition results. Line 414 represents the association of the QA control 402 with the Semantic Map 410, and to a SemanticItem 412 therein. Many SemanticItems 412 are individually associated with a visual or primary control 302 as represented by line 418, although one or more SemanticItems 412 may not be associated with a visual control and used only internally. In a multimodal scenario, where the user of the client device 30 may touch on the visual textbox, for example with a “TapEvent”, an audible prompt may not be necessary. For example, for a primary control comprising a textbox having visual text forming an indication of what the user of client device should enter in the corresponding field, a corresponding QA control 402 may or may not have a corresponding prompt such as an audio playback or a text-to-speech conversion, but would have a grammar corresponding to the expected value for recognition, and event handlers to process the input, or process other recognizer events such as no speech detected, speech not recognized, or events fired on timeouts.

[0165] In a further embodiment, the recognition result includes a confidence level measure indicating the level of confidence that the recognized result was correct. A confirmation threshold can also be specified in the Answer object, for example, as ConfirmThreshold equals 0.7. If the confirmation level exceeds the associated threshold, the result can be considered confirmed.

[0166] It should also be noted that in addition, or in the alternative, to specifying a grammar for speech recognition, QA controls and/or Command controls can specify Dtmf (dual tone modulated frequency) grammars to recognize telephone key activations in response to prompts or questions. Appendix B provides details of a Dtmf object that applies a different modality of grammar (a keypad input grammar rather than, for example, a speech input grammar) to the same question. Some of the properties of the Dtmf object include Preflush, which is a flag indicating if “type-ahead” functionality is allowed in order that the user can provide answers to questions before they are asked. Other properties include the number of milliseconds to wait for receiving the first key press, InitialTimeOut, and the number of milliseconds to wait before adjacent key presses, InterdigitTimeOut. Client-side script functions can be specified for execution through other properties, for example, when no key press is received, OnClientSilence, or when the input is not recognized, OnClientNoReco, or when an error is detected OnClientError.

[0167] At this point it should be noted that when a Semanticitem 412 of the Semantic map 410 is filled, through recognition for example, speech or Dtmf, several actions can be taken. First, an event can be issued or fired indicating that the value has been “changed”. Depending on if the confirmation level was met, another event that can be issued or fired includes a “confirm” event that indicates that the corresponding SemanticItemSemanticItem has been confirmed. These events are used for controlling dialog.

[0168] The Confirms property can also include answer objects having the structure similar to that described above with respect to the Answers property in that it is associated with a SemanticItem 412 and can include a ConfirmThreshold if desired. The Confirms property is not intended to obtain a recognition result per se, but rather, to confirm a result already obtained and ascertain from the user whether the result obtained is correct. The Confirms property is a collection of Answer objects used to assert whether the value of a previously obtained result was correct. The containing QA's Prompt object will inquire about these items, and obtains the recognition result from the associated SemanticItem 412 and forms it in a question such as “Did you say Seattle?” If the user responds with affirmation such as “Yes”, the confirmed event is then fired. If the user responds in the negative such as “No”, the associated SemanticItem 412 is cleared.

[0169] It should be noted in a further embodiment, the Confirms property can also accept corrections after a confirmation prompt has been provided to the user. For instance, in response to a confirmation prompt “Did you say Seattle?” the user may respond “San Francisco” or “No, San Francisco”, in which case, the QA control has received a correction. Having information as to which SemanticItem is being confirmed through the Answer object, the value in the SemanticItem can be replaced with the corrected value. It should also be noted that if desired, confirmation can be included in a further prompt for information such as “When did you want to go to Seattle?”, where the prompt by the system includes a confirmation for “Seattle” and a further prompt for the day of departure. A response by the user providing a correction to the place of destination would activate the Confirms property to correct the associated SemanticItem, while a response with only a day of departure would provide implicit confirmation of the destination.

[0170] The ExtraAnswers property allows the application author to specify Answer objects that a user may provide in addition to a prompt or query that has been made. For instance, if a travel oriented system prompts a user for a destination city, but the user responds by indicating “Seattle tomorrow”, the Answers property that initially prompted the user will retrieve and therefore bind the destination city “Seattle” to the appropriate SemanticItem, while the ExtraAnswers property can process “Tomorrow” as the next succeeding day (assuming that the system knows the current day), and thereby, bind this result to the appropriate SemanticItem in the Semantic Map. The ExtraAnswers property includes one or more Answer objects defined for possible extra information the user may also state. In the example provided above, having also retrieved information as to the day of departure, the system would then not need to reprompt the user for this information, assuming that the confirmation level exceeded the corresponding ConfirmThreshold. If the confirmation level did not exceed the corresponding threshold, the appropriate Confirms property would be activated.

Command Control

[0171] Command controls 404 are user utterances common in voice-only dialogs which typically have little semantic import in terms of the question asked, but rather seek assistance or effect navigation, e.g. help, cancel, repeat, etc. The Command control 404 can include a Prompt property to specify a prompt object. In addition, the Command control 404 can be used to specify not only the grammar (through a Grammar property) and associated processing on recognition (rather like an Answer object without binding of the result to an SemanticItem), but also a ‘scope’ of context and a type. This allows for the authoring of both global and context-sensitive behavior on the client side markup. The Command control 404 allows additional types of input such as “help” commands, or commands that allow the user of the client device to navigate to other selected areas of the website.

CompareValidator Control

[0172] The CompareValidator control compares two values according to an operator and takes an appropriate action. The values to be compared can be of any form such as integers, strings of text, etc. The CompareValidator includes a property SematicItemtoValidate that indicates the SemanticItem that will be validated. The SemanticItem to be validated can be compared to a constant or another SemanticItem, where the constant or other SemanticItem is provided by properties ValuetoCompare and SematicItemtoCompare, respectively. Other parameters or properties associated with the CompareValidator include Operator, which defines the comparison to be made and Type, which defines the type of value, for example, integer or string of the SemanticItems.

[0173] If the validation associated with the CompareValidator control fails, a Prompt property can specify a Prompt object that can be played instructing the user that the result obtained was incorrect. If upon comparison the validation fails, the associated SemanticItem defined by SematicItemtoValidate is indicated as being empty, in order that the system will reprompt the user for a correct value. However, it may be helpful to not clear the incorrect value of the associated SemanticItem in the Semantic Map in the event that the incorrect value will be used in a prompt to the user reiterating the incorrect value. The CompareValidator control can be triggered either when the value of the associated SemanticItem changes value or when the value has been confirmed, depending on the desires of the application author.

CustomValidator Control

[0174] The CustomValidator control is similar to the CompareValidator control. A property SematicItemtoValidate indicates the SemanticItem that will be validated, while a property ClientValidationFunction specifies a custom validation routine through an associated function or script. The function would provide a Boolean value “yes” or “no” or an equivalent thereof whether or not the validation failed. A Prompt property can specify a Prompt object to provide indications of errors or failure of the validation. The CustomValidator control can be triggered either when the value of the associated SemanticItem changes value or when the value has been confirmed, depending on the desires of the application author.

Call Control

[0175] In a further embodiment, controls are provided that enable application authors to create speech applications that handle telephony transactions. In general, the controls implement or invoke well-known telephony transactions such as ECMA (European Computer Manufactures Associated) CSTA (Computer Supported Telecommunication Application) messages, eventing and services. As is known, CSTA specifies application interfaces and protocols for monitoring and controlling calls and devices in a communication network. These calls and devices may support various media and can reside in various network environments such as IP, Switched Circuit Networks and mobile networks.

[0176] In the illustrated embodiment, the controls available to the application author include a SmexMessage control (SMEX-Simple Message Exchange), a TransferCall control, a MakeCall control, a DisconnectCall control and an AnswerCall control. Like the controls described above, these controls can be executed on the server so as to generate client-side markup that when executed on the client device perform the desired telephony transaction.

[0177] Referring to FIG. 4, the client-side markup generated by server 202 can be executed by voice browser 216, which in turn provides telephony transactions instructions (e.g. CSTA service calls) to the media server 214 and gateway 210 as necessary to perform the desired telephony transaction. Appendix B provides detailed information regarding each of the properties available in the controls. The controls are commonly used in a voice-only mode such as by voice browser 216 in FIG. 4; however, it should be understood that applications can be written also to be executed in an multi-modal client device.

Control Execution Algorithm

[0178] As in the previous set of controls, a client-side script or module (herein referred to as “RunSpeech”) is provided to the client device for the controls of FIG. 11. Again, the purpose of this script is to execute dialog flow via logic, which is specified in the script when executed on the client device 30, i.e. when the markup pertaining to the controls is activated for execution on the client due to values contained therein. The script allows multiple dialog turns between page requests, and therefore, is particularly helpful for control of voice-only dialogs such as through telephony browser 216. The client-side script RunSpeech is executed in a loop manner on the client device 30 until a completed form is submitted, or a new page is otherwise requested from the client device 30.

[0179] Generally, in one embodiment, the algorithm generates a dialog turn by outputting speech and recognizing user input. The overall logic of the algorithm is as follows for a voice-only scenario (reference is made to Appendix B for properties or parameters not otherwise discussed above):

• • [0180] 1. Find the first active (as defined below) QA, CompareValidator or CustomValidator control in speech index order.
  • [0181] 2. If there is no active control, submit the page.
  • [0182] 3. Otherwise, run the control.
    A QA is considered active if and only if:
  • [0183] 1. The QA's clientActivationFunction either is not present or returns true, AND
  • [0184] 2. If the Answers property collection is non empty, the State of all of the SemanticItems pointed to by the set of Answers is Empty OR
  • [0185] 3. If the Answers property collection is empty, the State at least one SemanticItem in the Confirm array is NeedsConfirmation.
    However, if the QA has PlayOnce true and its Prompt has been run successfully (reached OnComplete) the QA will not be a candidate for activation.
    A QA is run as follows:
  • [0186] 1. If this is a different control than the previous active control, reset the prompt Count value.
  • [0187] 2. Increment the Prompt count value
  • [0188] 3. If PromptSelectFunction is specified, call the function and set the Prompt's inlineprompt to the returned string.
  • [0189] 4. If a Reco object is present, start it. This Reco should already include any active command grammar.
    A Validator (either a CompareValidator or a CustomValidator) is active if:
  • [0190] 1. The SemanticItemToValidate has not been validated by this validator and its value has changed.
    A CompareValidator is run as follows:
  • [0191] 1. Compare the values of the SemanticItemToCompare or ValueToCompare and SemanticItemToValidate according to the validator's Operator.
  • [0192] 2. If the test returns false, empty the text field of the SemanticItemToValidate and play the prompt.
  • [0193] 3. If the test returns true, mark the SemanticItemToValidate as validated by this validator.
    A CustomValidator is run as follows:
  • [0194] 1. The ClientValidationFunction is called with the value of the SemanticItemToValidate.
  • [0195] 2. If the function returns false, the semanticItem cleared and the prompt is played, otherwise as validated by this validator.
    A Command is considered active if and only if:
  • [0196] 1. It is in Scope, AND
  • [0197] 2. There is not another Command of the same Type lower in the scope tree.

[0198] In the multi-modal case, the logic is simplified to the following algorithm:

• • [0199] 1. Wait for triggering event—i.e., user tapping on a control;
  • [0200] 2. Collect expected answers;
  • [0201] 3. Listen in for input;
  • [0202] 4. Bind result to SemanticItem, or if none, throw event;
  • [0203] 5. Go back to 1.
    Focus Tracking

[0204] The following discussion regarding focus tracking will be described with respect to use of SemanticItems as described above. However, it should be understood that this is but one embodiment and the techniques described below with regard to processing recognition results and maintaining focus on information recently recognized or provided from the user can be applied to the other embodiments described above.

[0205] The foregoing algorithm for the voice-only scenario uses QA (Question-Answer) controls and the SemanticItem to formulate the dialogs. As described above each SemanticItem contains a recognition result, the confidence that the system has in it, and its current state. QA controls contain information, including prompts and grammars, that are used to ask questions, recognize answers and update the SemanticItems. QA controls also contain answer and extra-answers objects that are used to specify the QA activation logic and the processing to be done with the results. Both answers and extra-answers take the recognition results returned by the speech recognizer and update SemanticItems with the values extracted from the recognition results. The difference between answers and extra-answers lies in the activation logic used by the system: if a SemanticItem already contains a value, the system will not process answers related to it. On the other hand, extra-answers can be activated irrespective of whether their related SemanticItem already contains a value or not.

[0206] Although the foregoing algorithm works well in many applications, problems can arise as discussed in the Background Section when, for example, extra answers are being processed in a mixed-initiative dialogue with the user. An aspect of the present invention is to allow the system to automatically adapt the dialogue flow so that it stays focused on the user's most recent input. In the algorithm discussed above, the dialogue flow is constrained by two main sources: the activation logic of the QA controls (based on the SemanticItem state and the QAs' answer/confirm/extra-answer specification) and the speech index of the QA control. Generally, this aspect of the invention adds a third constraint, herein referred to as “focus”. Whenever a SemanticItem is modified, which herein represents recently received recognition results, this information is retained in a manner so as to provide an order indicating when SemanticItems have been changed relative to each other. In this manner, the most recently changed SemanticItem can be identified. In one form, memory is used in the form of a “stack”, which is pictorially illustrated in FIG. 13 at 450. The stack 450 comprises identifiers such as 451, 452, 453 and 454 of SemanticItems related to recognition results received. As a SemanticItem is changed through the receipt of recognition results, it is added to the stack 450. Then, when the RunSpeech algorithm looks for a suitable QA control to execute next, it will only consider QA controls that are related to the SemanticItem at the top of the stack 450. This means that QAs that are later in speech index order than other QAs can be “promoted” and run before them, provided that they are related to the top-most SemanticItems whereas the others are not, and it is active. If no suitable QA can be found, the stack 450 is “popped” or decremented and the RunSpeech algorithm searches for suitable QAs again. This process may repeat until the stack 450 is empty, in which case the RunSpeech algorithm acts with the usual (non-focused) behavior as described above.

[0207] A system wide “focussing” value can be used to identify if focussing is to be performed or not. With focussing, the general algorithm described above can be represented as follows, where other portions are the same.

• • [0208] 1. If focussing is desired and the stack is not null (indicating focused SemanticItems are present), find the first active (as defined above) QA corresponding to the SemaniticItem at the top of the stack.
    • [0209] Otherwise, if focussing is not desired or the stack is null (indicating no SemanticItems are present), find the first active (as defined above) QA, CompareValidator or CustomValidator control in speech index order.
  • [0210] 2. If there is no active control, submit the page.
  • [0211] 3. Otherwise, run the control.

[0212] Another way to describe this technique is that all QA controls that are not related to the SemanticItem at the top of the stack are removed and the active QA is then selected as usual.

[0213] As appreciated by those skilled in the art, the use of a stack is but one form to accomplish focusing as described above. Other forms of focussing or prioritizing which QA control will be executed in order to maintain focus include, but are not limited to, counters, pointers, indices, time stamps, etc. as well known to those skilled in the art.

[0214] It should also be noted in a further embodiment, that the mechanism or manner in which the most recent SemanticItems have been saved need not ensure that they all be saved indefinitely. Rather, referring to FIG. 13 where the stack 450, illustrated by way of example for storing such information, is some selected or finite length such that when the stack 450 is full and a further SemanticItem placed thereon, it causes the lowest or oldest SemanticItem to be pushed off. This technique may be convenient in that the dialogue created thereby will not return to some item of information that the user spoke long ago.

[0215] The stack or other form of memory is also accessible to the application author through program logic such that it can be erased or reset for example if the user moves on to a different subject where the information related to the semantic information will not be used.

[0216] In one exemplary embodiment, the use of focus is controlled by a Boolean property on each QA control herein referred to as “focusing”. If the focus property is set to true, the SemanticItems modified by that QA are put on the stack 450 and the focusing mechanism operates until the stack 450 is emptied as described above, if a system wide focussing parameter is not present or is set for focussing. If the focus property set to false, the SemanticItems are not put on the stack 450 and focusing does not take place, at least with respect to that SemanticItem(s).

[0217] An exemplary embodiment for activation logic with respect to determining if any Qas are related to recently received recognition results for the QA is illustrated in FIG. 14. Generally, at least one answer or confirm must be related to the focused SemanticItem.

[0218] At step 460 the most recent SemanticItem is identified from the stack 450. At step 462, each QA control's corresponding answers are compared for a relation to the most recent SemanticItem. If a related answer is found, the QA control considered for execution. This is subject to the QA being active under the usual activation conditions (semantic item should be empty, etc.)

[0219] If at step 462 no QA control is found based on related answers, the method continues at step 464 whereat each QA control's corresponding confirms are compared for a relation to the most recent SemanticItem. If a related confirm is found the QA control is considered for execution.

[0220] If at step 464 no QA control is found based on related confirms, the method continues at step 466 whereat the Semantic Item is removed from the stack. If the stack is not empty or null, the method returns to step 460.

[0221] In the embodiment described above, one SemanticItem at a time is added to the stack 450 when each recognition result is obtained. In another embodiment, each element 451, 452, 453 and 454 in the stack 450 can represent an array each array contains one or more SemanticItems. The method of activating a QA illustrated in FIG. 14 is essentially the same however, each SemanticItem in the array for each layer of the stack 450 is examined before the array is removed from the stack 450.

[0222] In some cases a QA control may need to be played in speech index order, irrespective of being identified by the method of FIG. 14. In a further embodiment, a property is provided for each QA control specifying whether or not it should be included in the focusing mechanism, if identified.

[0223] The following provides an example of how the focussing mechanism retains order or focus in a mixed-initiative travel dialogue requiring a departure city and a destination city. Suppose the question or prompt is rendered “What are your travel plans?”. A user may answer “I'd like to go to Seattle.” or “I'd like to leave from Paris”. In the first case, the dialogue should go on with the confirmation: “Did you say Seattle?” and then ask for the departure city since this information was not given.

[0224] In the second case, the system should confirm the departure city first and then ask about the destination city second. If the user provided both the departure city and the destination city, the application author may want to confirm the departure city first and then the arrival city. Writing such a dialogue flow without using automated focusing is time consuming and error prone. Using the focusing mechanism, the application author does not have to worry about getting the ordering right. (However, the application author will still need to worry about some of the ordering (e.g., which city to confirm when both are given at the same time) although not as much as before.) If the user provides the departure city, the associated SemanticItem is pushed on the stack. When the control algorithm RunSpeech looks for a suitable QA control to execute, only QA controls related to that SemanticItem will be considered. In this case the QA confirming the departure city will be executed, even if the QA asking for the destination city comes earlier in speech index order.

[0225] The focusing problem becomes quickly intractable if three or more pieces of information can be provided. In the example above, assume the airline to be used is also desired besides the departure and destination cities. A user may provide input to the initial question of travel plans by stating, “On United Airlines, I want to depart from Seattle”, or “I want to fly to Minneapolis on American Airlines”, or “I want to fly from Seattle to Chicago.” Each of these examples provides a different set of information. Trying to predict all possible dialogue flows and making sure that the system appropriately confirms (and asks again in case the user denies that the recognition result is correct) and asks for remaining required information in a logical manner is a very difficult task. However, this becomes straightforward using the focusing mechanism.

Assisted Multi-Modal Dialogue

[0226] The foregoing has provided separate algorithms for controlling dialogue in a voice-only scenario and in a multi-modal scenario. Described below is a single algorithm that can be used to control the dialog in either the voice-only scenario or the multi-modal scenario, which, among other benefits, would allow the user to easily switch between modes of operation for any page loaded on the client device.

[0227] Currently, in the voice-only scenario, the algorithm starts once the page has been loaded on the client device, for example, the voice browser 216, and stops once all the information has been obtained. The page is then sent from the client device to the server 202.

[0228] In a modified form discussed below, pressing on a textbox causes the algorithm to run a dialogue associated with that textbox, rather than starting at the beginning of the page. The dialogue can include focusing as described above. This gives the user the ability to enter information as he/she desires. It should be noted in the multi-modal algorithm discussed earlier such a dialog was not run.

[0229] FIG. 15 pictorially illustrates information to be gathered organized as “topics” in order to execute portions of the dialog. In the illustrative example, the topics pertain to a travel site that allows users to input information related to a departure city 500, a departure date and time 502, an arrival city 504 and an arrival date and time 506. Each topic 500, 502, 504 and 506 comprises a collection of one or more questions, answers, commands or validators such as illustrated in FIG. 11 to form a dialog for each corresponding topic. Each of the collections 500, 502, 504 and 506 includes a label or identifier 500 A, 502 A, 504 A and 506 A, respectively. In addition, the collections can be grouped in two or more sets also identified by a label or identifier. In the illustrated example, a larger collection is identified as 508 A (representative of the complete page), which comprises the collections 500, 502, 504 and 506 as a group or hierarchy.

[0230] Although the each of the collections 500, 502, 504 and 506 can be constructed by using a combination of the various controls illustrated in FIG. 11 and discussed above, if desired, the controls can be further grouped together in a larger control forming a template. In this manner, an application author can select the larger control and modify the individual controls therein as necessary depending upon the topic under development. Graphical user interfaces can be employed where the application author can select the desired controls and “drag and drop” on a panel to construct a topic. Such techniques are well-known in website design and other application development environments.

[0231] Organization of the controls in this manner allows convenient execution of the dialog in both the voice-only scenario and the multi-modal scenario. In particular, by using the collection identifiers 500A, 502A, 504A, 506A and 508A the control algorithm can be instructed to execute the corresponding dialog for each collection 500, 502, 504 and 506, individually, or as a group 508. For example, a simple J-script command such as of the form:

• [0232] Runspeech.ActiveQA=(Collection Identifier)
  can be used to identify collection and thus the corresponding dialog to execute.

[0233] In the voice-only mode scenario, the “Collection Identifier” can be set to identifier 508 A, in which case the control algorithm will execute in the manner discussed above for the voice-only mode of operation to execute the complete dialog. However, it should be noted that in an alternative embodiment, a separate manager algorithm or module 512 can be used to individually and sequentially activate each of the collections 500, 502, 504 and 506. In the example illustrated in FIG. 15, the manager algorithm 512 would issue a command identifying that collection 502 is active, whereupon after control algorithm executes the dialog of collection 502, control returns back to the manager algorithm 512. The manager algorithm 512 would then issue a command identifying that collection 504 is now active. This process is repeated until each of the collections 500, 502, 504 and 506 have been activated as prescribed by the manager algorithm 512.

[0234] For the multi-modal application, the control algorithm maintains a list of those collections 500, 502, 504 and 506 that are considered active, for instance, a textbox on a page that needs input data. It should be noted in the illustrated embodiment, the control algorithm can ascertain the SemanticItems associated with each collection that needs to be filled. Whereupon, the data in the SemanticItem is used to fill in the corresponding textbox.

[0235] FIG. 16 illustrates an exemplary display rendering 520 for the travel page. As data is entered, more collections may become active or may not be relevant and thus deactivated. As described above, each collection 500, 502, 504 and 506 has an associated identifier or label 500 A, 502 A, 504 A and 506 A in the dialog. Each textbox in the page provided to the client also has an associated identifier or label (not shown). When a textbox, or other button such as button 522 associated with each textbox, is activated by the user, a simple function can then be called using the label or identifier of the textbox or other button 522 as an input parameter to identify the corresponding collection identifier that the selected textbox or button 522 is associated with. With the collection identifier determined, the control algorithm can be executed upon the dialog of the corresponding collection, for example, by issuing the J-script command provided above.

[0236] Organization of the page into topics with corresponding dialogs identified by collection identifiers provides additional benefits not previously available in the multi-modal scenario. As stated above, the control algorithm in the multi-modal scenario begins execution of the dialog associated with the textbox that has been selected. Once the dialogue has been completed for the selected textbox or a group of textboxes, having obtained the speech input and performed any required validation or confirmation, the control algorithm stops, and possibly waits for further direction from the user. Alternatively, depending on the author of the application, an additional “stop” button 526 can be provided such that when a user initiates the stop button 526 the dialogue for the previously activated textbox would stop. In yet another embodiment, the user can provide an audible command such as “cancel” via a command control, which would also stop the control algorithm with respect to the textbox that has been selected. Each of the afore-mentioned techniques, enable the application author to allow the user to selectively stop the dialog in the multi-modal scenario. Suitable mechanisms or methods can be provided in the control algorithm to selectively stop execution of the control algorithm based on user action. For example, an application program interface can be used to stop an executing topic and direct the control algorithm to begin execution of another topic.

[0237] Another situation which may require the control algorithm to halt execution occurs when a confirmation prompt or other confirmation output is provided in the multi-modal scenario. By contrast, in a voice-only browser once input speech is obtained from the user, the control algorithm may execute a confirmation control (automatically or based on a confidence threshold not being exceeded), where the confirmation control audibly returns back to the user what has been recognized. The confirmation control then asks if what has been recognized is correct. The user answers whether or not it has been recognized as correct and the control algorithm proceeds to the next control such as the next question or QA control.

[0238] However, in the multi-modal scenario, the input speech that has been recognized is generally rendered in the textbox that has been selected. Again, based upon automatic confirmation or whether or not a confidence threshold has been exceeded, a confirmation prompt can be displayed asking whether or not the input speech has been properly recognized. In FIG. 16, a confirmation prompt is illustrated at 530 in response to a confidence threshold not being exceeded. In this example, the prompt 530 includes a “yes” button 532 and a “no” button 534. The confirmation prompt could be rendered audibly and/or the user's response could be provided by buttons 532 and 534 or via speech input.

[0239] At this point, it should be noted that displaying a prompt rather than audibly rendering the prompt when available through a multi-modal capable device may be preferred depending on the data being rendered or the current operating environment. For example, audible rendering of sensitive data such as confirmation of a social security number may not be preferred in view that such information could possibly be overheard by others. Thus, parameters can be established and associated with information being gathered allowing the application author to designate that if said information is rendered it should be rendered as a default as a prompt if a multi-modal client is being used. Likewise, given a noisy operating environment, audible rendering of prompts may not be practical. Methods can be employed by the application author to enable the user to change the settings of these parameters, if desired.

[0240] Also, audible rendering of prompts typically includes active monitoring of the length of silence that may result after rendering the audible prompt. If a sufficient amount of silence occurs after playing the prompt, the user may not have heard the prompt and some methods are then employed to continue the dialog, for example, by replaying the prompt. However, when a prompt is displayed, such as the confirmation 530 of FIG. 16, the notion of time may be irrelevant. Thus, if a prompt is being displayed rather than being rendered audibly, it may be appropriate to disable or otherwise modify silence measuring methods occurring after rendering. In effect, the control algorithm could suspend the dialog after displaying a prompt, such as confirmation prompt 530, and begin again with the user selecting one of the buttons “yes” or “no” in the prompt 530. Depending on if any measurement of inactivity is being monitored by the control algorithm, activation of the buttons “yes” or “no” could employ a method to resume processing of the dialog by the control algorithm.

[0241] Referring back to FIG. 16, if the user indicates that the recognition was correct, the recognized input can be rendered in the textbox, whereas if the user indicates that the recognition was incorrect, the recognized input would be discarded. However, in the multi-modal scenario, it is also desirable to allow the user to ignore answering the displayed or otherwise rendered confirmation question, and proceed to another textbox or operation with respect to the form provided on the client device. Thus, it is desirable to allow the control algorithm to stop execution of the portion of the control algorithm with respect to the textbox selected if the user does not respond to the confirmation response. Depending on the author of the application, the author may choose to ignore the input speech provided to the textbox, or accept it with the intent of reconfirming its accuracy at a later time.

[0242] In many applications, it is common to include a “submit” button 532 when the user believes that he/she has entered all required information. As indicated above, the user may not have chosen to confirm that a particular value was properly recognized for a given textbox; however, upon activation of the submit button 532 the author of the application may at that time force confirmation of the previously ignored confirmation prompt, or choose to accept the recognized input as correct and return the form to the server for processing.

[0243] Another improvement for the multi-modal scenario allows the user to change the value for the textbox either through voice recognition or through standard graphical user interface methods such as with a keyboard, a handwriting recognition, etc. Using voice recognition, the control algorithm initially receives input speech and ascertains a recognized input for the input speech. This recognized input is retained by the control algorithm in memory, for example in the corresponding SemanticItem discussed above along with status information such as whether or not the recognized input has been confirmed. At sometime prior to submission of the page to the web server, the recognized input is associated with the primary control. However, in the multi-modal scenario, the user also has the option of entering the value through a graphical user interface rather than using voice recognition. If the user chooses to input a value using a graphical user interface, the value entered by the user is also replicated in the corresponding SemanticItem where the status information, for example, can be considered as confirmed because the user manually entered the assumed correct value. The SemanticItems and the status information is exposed through suitable methods to allow the author when writing the application to update these values if the user of the client device chooses to use a graphical user interface to input values. Likewise, if the user has provided input speech for a particular entry which was recognized and displayed back to the user, using the graphical user interface, the user can correct the value rather than using the confirmation control. Again, by exposing the SemanticItems and status information maintained by the control algorithm, the control algorithm and status information can be maintained in correspondence.

[0244] Exposing the SemanticItems and the status information, also allows the application author to reset selected portions of the dialog information as desired during execution. For instance, a reset button can be provided for the each portion of the dialog separately based on the collections 500, 502, 504 and 506. These buttons are indicated at 550, 552, 554 and 556, respectively. If the user, activates any of the buttons 550, 552, 554 or 556, the control algorithm can ascertain which collection has been identified for resetting using a simple function (in a manner similar to that discussed above with respect to identifying textboxes) and then remove the corresponding values in the associated SemanticItems for that collection and/or change the status information to signify that new data is required. Although in FIG. 16 use of the reset buttons 550, 552, 554 and 556 may appear simplistic, it should be noted the embodiment illustrated is merely exemplary, and that the concept can be applied to more complex data entry forms where resetting would involve more textboxes or complexity.

[0245] In the multi-modal scenario where the user has the ability to select a textbox for entry of data, typically the user will know initially what the form is asking for due to the fact that there is a visual prompt associated with the textbox. For example, it is clear in FIG. 16 that textbox 536 is for the “Departure City”. Thus, although the control algorithm is running the dialogue associated with the textbox in the multi-modal scenario, there is probably no need to ask the initial prompt in the collection 500, as would be necessary in the voice-only scenario, since the user probably is provided with a visual prompt. However, it is also desirable that the page be renderable in either a multi-modal scenario or a voice-only scenario. Therefore, the control algorithm can maintain in memory status information as to whether or not the user is operating in a multi-modal scenario, or whether the page is being rendered in a voice-only scenario such as through voice browser 216. In one embodiment, information can be easily ascertained by the client device upon receipt of the form to determine the appropriate mode of entry. For example, if the client device is a voice browser 216, the mode of entry needs to be operation of the control algorithm in a voice-only mode. The voice-only mode is particularly suitable for focusing as described above. However, if the client device is a multi-modal device such as a PDA, then the mode of entry can be optionally defaulted to a preselected mode of entry as well as be changed such as through activation of a button 540.

[0246] Being able to switch the mode of entry for a page rendered on the client device can be very advantageous. For instance, if the user is entering data in a multi-modal scenario by selecting textboxes and providing voice or speech input which is recognized and displayed in the corresponding textboxes, the user may want to switch to a voice-only mode of entry, for instance, if the user needs to perform another activity which prevents him/her from selecting textboxes for entry such a case when the user may be driving. Button 540 or other suitable indicating device allows the mode of entry to be changed. If the user has been operating in a multi-modal for of entry, the control algorithm can then begin processing the page, for example, at the beginning and execute the remaining dialog with voice prompts until the form is completed. Switching to the situation using voice prompts for questions and confirmations (although the system would otherwise display the results) would be another form in which focusing may be desired. If, on the other hand, the user has been operating in voice-only operation, upon activation of a button on the display or other form of user input such as a keyword spoken by the user, or merely selecting a textbox, the control algorithm can then switch to a multi-modal for of entry allowing the user to navigate through the page as desired. In such a case, the stack or other memory mechanism retaining focussing information can then be reset since the user will be generally identifying textboxes to which spoken input will be directed. It should be noted that switching from a voice-only mode to a multi-modal mode, or vice-versa, can be for the whole page as described above, or for just a portion of the dialog associated with one or more textboxes or other required input.

[0247] As indicated above, for a multi-modal form, the control algorithm can automatically determine that the initial prompt associated with any textbox not be played in view of the fact that a visual prompt is probably already associated with the textbox selected by the user. However, if upon selection, the user does not provide any voice input, the client device can then play the initial prompt. However, depending upon the application, there may be other audible prompts that should be played depending upon, for example, entry of incorrect information. For instance, if the user has entered a credit card number of insufficient length, the control algorithm operating in the multi-modal operation, would then play or otherwise render a prompt stating that the credit card number was of insufficient length and should contain a given amount of digits. Thus, although the initial prompt may not be played in a multi-modal scenario, other prompts associated with the dialog for the selected textbox may need to be rendered. Flexibility is provided for the application author to enable the subsequent prompts to be rendered audibly, visually or both audibly and visually, as desired and depending on the capabilities of the device.

[0248] From the foregoing, a method and system are provided for generating mark-up for client side devices for speech-enabled applications including telephony applications that further provides focussing. The same set of controls can be used in three different forms of interaction including Voice-only, Tap-and-talk (multi-modal) and Hands-free (multi-modal). In Voice-only dialogs are provided on a GUI-less browser such as for telephony applications. This kind of application is driven by a dialog-flow manager that runs on the client (RunSpeech). In Tap-and-talk multi-modal dialogs contain a usable GUI without speech output. System prompts are generally not provided and the interaction is managed by the user's click events on the GUI. In Hands-free multi-modal, dialogs use a GUI display and speech input and output. The dialog may be authored for Tap-and-talk, but may still use the RunSpeech algorithm, or other speech controls features, to enable system driven voice prompting, while confirmation is provided visually or aurally, depending on the active mode of input at the time of confirmation. Switching between multi-modal/hands-free and voice-only is done by detecting the type of client the controls are talking to. Generally, Hands-free is switched on on-demand.

[0249] Although the present invention has been described with reference to preferred embodiments, workers skilled in the art will recognize that changes may be made in form and detail without departing from the spirit and scope of the invention.

[0250] The QA control adds speech functionality to the primary control to which it is attached. Its object model is an abstraction of the content model of the exemplary tags in Appendix A.

1.1 QA Control

[0251]

	<Speech:QA
	id=“...”
	controlsToSpeechEnable=“...”
	speechIndex=“...”
	ClientTest=“...”
	runat=“server” >
	<Question ...>
	<Statement ...>
	...
	<Answer ...>
	<Confirm ...>
	...
	<Command ...>
	...
	</Speech:QA>

1.1.1 Core Properties
String ControlsToSpeechEnable

[0252] ControlsToSpeechEnable specifies the list of IDs of the primary controls to speech enable. IDs are comma delimited.

1.1.2 Activation Mechanisms

Int SpeechIndex

[0253] SpeechIndex specifies the ordering information of the QA control—this is used by RunSpeech. Note: If more than one QA control has the same SpeechIndex, RunSpeech will execute them in source order. In situations where some QA controls have SpeechIndex specified and some QA controls do not, RunSpeech will order the QA controls first by SpeechIndex, then by source order.

String ClientTest

[0254] ClientTest specifies a client-side script function which returns a boolean value to determine when the QA control is considered available for selection by the RunSpeech algorithm. The system strategy can therefore be changed by using this as a condition to activate or de-activate QA controls more sensitively than SpeechIndex. If not specified, the QA control is considered available for activation.

1.1.3 Questions, Statements, Answers, Confirms and Commands

Question[ ] Questions

[0255] QA control contains an array of question objects or controls, defined by the dialog author. Each question control will typically relate to a function of the system, eg asking for a value, etc. Each question control may specify an activation function using the ClientTest attribute, so an active QA control may ask different kinds of questions about its primary control under different circumstances. For example, the activation condition for main question Q_Main may be that the corresponding primary control has no value, and the activation condition for a Q_GiveHelp may be that the user has just requested help. Each Question may specify answer controls from within the QA control which are activated when the question control is outputted.

Statement[ ] Statement

[0256] QA control contains an array of statement objects or controls. Statements are used to provide information to the listener, such as welcome prompts.

Answer[ ] Answers

[0257] QA control contains an array of answer objects or controls. An answer control is activated directly by a question control within the QA control, or by a StartEvent from the Primary control. Where multiple answers are used, they will typically reflect answers to the system functions, e.g. A_Main might provide a value in response to Q_Main, and A_Confirm might providing a yes/no+correction to Confirm.

Confirm[ ] Confirm

[0258] QA control may contain a confirm object or control. This object is a mechanism provided to the dialog authors which simplify the authoring of common confirmation subdialogs.

Command[ ] Command

[0259] A Command array holds a set of command controls. Command controls can be thought of as answer controls without question controls, whose behavior on recognition can be scoped down the control tree.

1.2 Question Control

[0260] The question control is used for the speech output relating to a given primary control. It contains a set of prompts for presenting information or asking a question, and a list of ids of the answer controls, which may provide an answer to that question. If multiple answer controls are specified, these grammars are loaded in parallel when the question is activated. An exception will be thrown if no answer control is specified in the question control.

[0261]

	<Question
	id=“...”
	ClientTest=“...”
	Answers=“...”
	Count=“...”
	initialTimeout=“...”
	babbleTimeout=“...”
	maxTimeout=“...”
	Modal=“...”
	PromptFunction=“...”
	OnClientNoReco=“...” >
	<prompt ... />
	...
	</Question>

String ClientTest

[0262] ClientTest specifies the client-side script function returning a boolean value which determines under which circumstances a question control is considered active within its QA control (the QA control itself must be active for the question to be evaluated). For a given QA control, the first question control with a true condition is selected for output. For example, the function may be used to determine whether to output a question which asks for a value (“Which city do you want?”) or which attempts to confirm it (“Did you say London?”). If not specified, the question condition is considered true.

Prompt[ ] Prompts

[0263] The prompt array specifies a list of prompt objects, discussed below. Prompts are also able to specify conditions of selection (via client functions), and during RunSpeech execution only the first prompt with a true condition is selected for playback.

String Answers

[0264] Answers is an array of references by ID to controls that are possible answers to the question. The behavior is to activate the grammar from each valid answer control in response to the prompt asked by the question control.

Integer initialTimeOut

[0265] The time in milliseconds between start of recognition and the detection of speech. This value is passed to the recognition platform, and if exceeded, an on Silence event will be thrown from the recognition platform. If not specified, the speech platform will use a default value.

Integer babbleTimeout

[0266] The period of time in milliseconds in which the recognition server or other recognizer must return a result after detection of speech. For recos in “tap-and-talk” scenarios this applies to the period between speech detection and the recognition result becoming available. For recos in dictation scenarios, this timeout applies to the period between speech detection and each recognition return—i.e. the period is restarted after each return of results or other event. If exceeded, the onClientNoReco event is thrown but different status codes are possible. If there has been any kind of recognition platform error that is detectable and the babbleTimeout period has elapsed, then an onClientNoReco is thrown but with a status code −3. Otherwise if the recognizer is still processing audio—e.g. in the case of an exceptionally long utterance or if the user has kept the pen down for an excessive amount of time—the onClientNoReco event is thrown, with status code −15. If babbleTimeout is not specified, the speech platform will default to an internal value.

Integer maxTimeout

[0267] The period of time in milliseconds between recognition start and results returned to the client device browser. If exceeded, the on MaxTimeout event is thrown by the browser—this caters for network or recognizer failure in distributed environments. For recos in dictation scenarios, as with babbleTimeout, the period is restarted after the return of each recognition or other event. Note that the maxTimeout attribute should be greater than or equal to the sum of initialTimeout and babbleTimeout. If not specified, the value will be a browser default.

Bool Modal

[0268] When modal is set to true, no answers except the immediate set of answers to the question are activated (i.e. no scoped Answers are considered). The defaults is false. For Example, this attribute allows the application developer to force the user of the client device to answer a particular question.

String PromptFunction(prompt)

[0269] PromptFunction specifies a client-side function that will be called once the question has been selected but before the prompt is played. This gives a chance to the application developer to perform last minute modifications to the prompt that may be required. PromptFunction takes the ID of the target prompt as a required parameter.

String OnClientNoReco

[0270] OnClientNoReco specifies the name of the client-side function to call when the NoReco (mumble) event is received.

1.2.1 Prompt Object

[0271] The prompt object contains information on how to play prompts. All the properties defined are read/write properties.

[0272]

	<prompt
	id=“...”
	count=“...”
	ClientTest=“...”
	source=“...”
	bargeIn=“...”
	onClientBargein=“...”
	onClientComplete=“...”
	onClientBookmark=“...” >
	. . . text/markup of the prompt . . .
	</prompt>

Int Count

[0273] Count specifies an integer which is used for prompt selection. When the value of the count specified on a prompt matches the value of the count of its question control, the prompt is selected for playback. Legal values are 0-100.

[0274]

	<Question id=Q_Ask”>
	<prompt count=“1”> Hello </prompt>
	<prompt count=“2”> Hello again </prompt>
	</Question>

[0275] In the example, when Q_Ask.count is equal to 1, the first prompt is played, and if it is equal to 2 (i.e. the question has already been output before), the second prompt is then played.

String ClientTest

[0276] ClientTest specifies the client-side script function returning a boolean value which determines under which circumstances a prompt within an active question control will be selected for output. For a given question control, the first prompt with a true condition is selected. For example, the function may be used to implement prompt tapering, eg (“Which city would you like to depart from?” for a function returning true if the user if a first-timer, or “Which city?” for an old hand). If not specified, the prompt's condition is considered true.

String InlinePrompt

[0277] The prompt property contains the text of the prompt to play. This is defined as the content of the prompt element. It may contain further markup, as in TTS rendering information, or <value> elements. As with all parts of the page, it may also be specified as script code within <script> tags, for dynamic rendering of prompt output.

String Source

[0278] Source specifies the URL from which to retrieve the text of the prompt to play. If an inline prompt is specified, this property is ignored.

Bool BargeIn

[0279] BargeIn is used to specify whether or not barge-in (wherein the user of the client device begins speaking when a prompt is being played) is allowed on the prompt. The defaults is true.

String onClientBargein

[0280] onClientBargein specifies the client-side script function which is invoked by the bargein event.

String onClientComplete

[0281] onClientComplete specifies the client-side script function which is invoked when the playing of the prompt has competed.

String OnClientBookmark

[0282] OnClientBookmark accesses the name of the client-side function to call when a bookmark is encountered.

1.2.2 Prompt Selection

[0283] On execution by RunSpeech, a QA control selects its prompt in the following way:

• [0284] ClientTest and the count attribute of each prompt are evaluated in order. The first prompt with both ClientTest and count true is played. A missing count is considered true. A missing ClientTest is considered true.
  1.3 Statement Control

[0285] Statement controls are used for information-giving system output when the activation of grammars is not required. This is common in voice-only dialogs. Statements are played only once per page if the playOnce attribute is true.

[0286]

	<Statement
	id=“...”
	playOnce=“...”
	ClientTest=“...”
	PromptFunction=“...” >
	<prompt ... />
	...
	</Statement >

Bool playOnce

[0287] The playOnce attribute specifies whether or not a statement control may be activated more than once per page. playOnce is a Boolean attribute with a default (if not specified) of TRUE, i.e., the statement control is executed only once. For example, the playOnce attribute may be used on statement controls whose purpose is to output email messages to the end user. Setting playOnce=“False” will provide dialog authors with the capability to enable a “repeat” functionality on a page that reads email messages.

String ClientTest

[0288] ClientTest specifies the client-side script function returning a boolean value which determines under which circumstances a statement control will be selected for output. RunSpeech will activate the first Statement with ClientTest equal to true. If not specified, the ClientTest condition is considered true.

String PromptFunction

[0289] PromptFunction specifies a client-side function that will be called once the statement control has been selected but before the prompt is played. This gives a chance to the authors to do last minute modifications to the prompt that may be required.

Prompt[ ] Prompt

[0290] The prompt array specifies a list of prompt objects. Prompts are also able to specify conditions of selection (via client functions), and during RunSpeech execution only the first prompt with a true condition is selected for playback.

[0291]

	<Speech:QA
	id=“QA_Welcome”
	ControlsToSpeechEnable=“Label1”
	runat=“server” >
	<Statement id=“WelcomePrompt” >
	<prompt bargeIn=“False”> Welcome </prompt>
	</Statement>
	</Speech:QA>

1.4 Confirm Control

[0292] Confirm controls are special types of question controls. They may hold all the properties and objects of other questions controls, but they are activated differently. The RunSpeech algorithm will check the confidence score found in the confirmThreshold of the answer control of the ControlsToSpeechEnable. If it is too low, the confirm control is activated. If the confidence score of the answer control is below the confirmThreshold, then the binding is done but the onClientReco method is not called. The dialog author may specify more than one confirm control per QA control. RunSpeech will determine which confirm control to activate based on the function specified by ClientTest.

[0293]

	<Answer ConfirmThreshold=... />
	<Confirm>
	...all attributes and objects of Question...
	</Confirm>

1.5 Answer Control

[0294] The answer control is used to specify speech input resources and features. It contains a set of grammars related to the primary control. Note that an answer may be used independently of a question, in multimodal applications without prompts, for example, or in telephony applications where user initiative may be enabled by extra-answers. Answer controls are activated directly by question controls, by a triggering event, or by virtue of explicit scope. An exception will be thrown if no grammar object is specified in the answer control.

[0295]

	<Answer
	id=“...”
	scope=“...”
	StartEvent=“...”
	StopEvent=“...”
	ClientTest=“...”
	onClientReco=“...”
	onClientDTMF=“...”
	autobind=“...”
	server=“...”
	ConfirmThreshold=“...”
	RejectThreshold=“...” >
	<grammar ... />
	<grammar ... />
	...
	<dtmf ... />
	<dtmf ... />
	...
	<bind ... />
	<bind ... />
	...
	</Answer>

String Scope

[0296] Scope holds the id of any named element on the page. Scope is used in answer control for scoping the availability of user initiative (mixed task initiative: i.e. service jump digressions) grammars. If scope is specified in an answer control, then it will be activated whenever a QA control corresponding to a primary control within the subtree of the contextual control is activated.

String StartEvent

[0297] StartEvent specifies the name of the event from the primary control that will activate the answer control (start the Reco object). This will be typically used in multi-modal applications, eg on MouseDown, for tap-and-talk.

String StopEvent

[0298] StopEvent specifies the name of the event from the primary control that will de-activate the answer control (stop the Reco object). This will be typically used in multi-modal applications, eg on MouseUp, for tap-and-talk.

String ClientTest

[0299] ClientTest specifies the client-side script function returning a boolean value which determines under which circumstances an answer control otherwise selected by scope or by a question control will be considered active. For example, the test could be used during confirmation for a ‘correction’ answer control to disable itself when activated by a question control, but mixed initiative is not desired (leaving only accept/deny answers controls active). Or a scoped answer control which permits a service jump can determine more flexible means of activation by specifying a test which is true or false depending on another part of the dialog. If not specified, the answer control's condition is considered true.

Grammar[ ] Grammars

[0300] Grammars accesses a list of grammar objects.

DTMF[ ] DTMFs

[0301] DTMFs holds an array of DTMF objects.

Bind[ ] Binds

[0302] Binds holds a list of the bind objects necessary to map the answer control grammar results (dtmf or spoken) into control values. All binds specified for an answer will be executed when the relevant output is recognized. If no bind is specified, the SML output returned by recognition will be bound to the control specified in the ControlsToSpeechEnable of the QA control.

String OnClientReco

[0303] OnClientReco specifies the name of the client-side function to call when spoken recognition results become available.

String OnClientDTMF

[0304] OnClientDTMF holds the name of the client-side function to call when DTMF recognition results become available.

Boolean Autobind

[0305] The value of autobind determines whether or not the system default bindings are implemented for a recognition return from the answer control. If unspecified, the default is true. Setting autobind to false is an instruction to the system not to perform the automatic binding.

String Server

[0306] The server attribute is an optional attribute specifying the URI of the speech server to perform the recognition. This attribute over-rides the URI of the global speech server attribute.

Integer ConfirmThreshold

[0307] Holds a value representing the confidence level below which a confirm control question will be automatically triggered immediately after an answer is recognized within the QA control. Legal values are 0-100.

[0308] Note that where bind statements and onClientReco scripts are both specified, the semantics of the resulting Tags are that binds are implemented before the script specified in onClientReco.

Integer RejectThreshold

[0309] RejectThreshold specifies the minimum confidence score to consider returning a recognized utterance. If overall confidence is below this level, a NoReco event will be thrown. Legal values are 0-100.

1.5.1 Grammar

[0310] The grammar object contains information on the selection and content of grammars, and the means for processing recognition results. All the properties defined are read/write properties.

[0311]

	<Grammar
	ClientTest=“...”
	Source=“...”
	>
	...grammar rules...
	</Grammar>

String ClientTest

[0312] The ClientTest property references a client-side boolean function which determines under which conditions a grammar is active. If multiple grammars are specified within an answer control (e.g. to implement a system/mixed initiative strategy, or to reduce the perplexity of possible answers when the dialog is going badly), only the first grammar with a true ClientTest function will be selected for activation during RunSpeech execution. If this property is unspecified, true is assumed.

String Source

[0313] Source accesses the URI of the grammar to load, if specified.

String InlineGrammar

[0314] InlineGrammar accesses the text of the grammar if specified inline. If that property is not empty, the Source attribute is ignored.

1.5.2 Bind

[0315] The object model for bind follows closely its counterpart client side tags. Binds may be specified both for spoken grammar and for DTMF recognition returns in a single answer control.

[0316]

	<bind
	Value=“...”
	TargetElement=“...”
	TargetAttribute=“...”
	Test=“...”
	/>

String Value

[0317] Value specifies the text that will be bound into the target element. It is specified as an XPath on the SML output from recognition.

String TargetElement

[0318] TargetElement specifies the id of the primary control to which the bind statement applies. If not specified, this is assumed to be the ControlsToSpeechEnable of the relevant QA control.

String TargetAttribute

[0319] TargetAttribute specifies the attribute on the TargetElement control in which bind the value. If not specified, this is assumed to be the Text property of the target element.

String Test

[0320] The Test attribute specifies a condition which must evaluate to true on the binding mechanism. This is specified as an XML Pattern on the SML output from recognition.

1.5.2.1 Automatic Binding

[0321] The default behavior on the recognition return to a speech-enabled primary control is to bind certain properties into that primary control. This is useful for the dialog controls to examine the recognition results from the primary controls across turns (and even pages). Answer controls will perform the following actions upon receiving recognition results:

• • [0322] 1. bind the SML output tree into the SML attribute of the primary control
  • [0323] 2. bind the text of the utterance into the SpokenText attribute of the primary control
  • [0324] 3. bind the confidence score returned by the recognizer into the Confidence attribute of the primary control.

[0325] Unless autobind=“False” attribute is specified on an answer control, the answer control will perform the following actions on the primary control:

• • [0326] 1. bind the SML output tree into the SML attribute;
  • [0327] 2. bind the text of the utterance into the SpokenText attribute;
  • [0328] 3. bind the confidence score returned by the recognizer into the Confidence attribute;

[0329] Any values already held in the attribute will be overwritten. Automatic binding occurs before any author-specified bind commands, and hence before any onClientReco script (which may also bind to these properties).

1.5.3 DTMF

[0330] DTMF may be used by answer controls in telephony applications. The DTMF object essentially applies a different modality of grammar (a keypad input grammar rather than a speech input grammar) to the same answer. The DTMF content model closely matches that of the client side output Tags DTMF element. Binding mechanisms for DTMF returns are specified using the targetAttribute attribute of DTMF object.

[0331]

	<DTMF
	firstTimeOut=“...”
	interDigitTimeOut=“...”
	numDigits=“...”
	flush=“...”
	escape=“...”
	targetAttribute=“...”
	ClientTest=“...”>
	<dtmfGrammar ...>
	</DTMF>

Integer firstTimeOut

[0332] The number of milliseconds to wait between activation and the first key press before raising a timeout event.

Integer interDigitTimeOut

[0333] The number of milliseconds to wait between key presses before raising a timeout event.

Int numDigits

[0334] The maximum number of key inputs permitted during DTMF recognition.

Bool Flush

[0335] A flag which states whether or not to flush the telephony server's DTMF buffer before recognition begins. Setting flush to false permits DTMF key input to be stored between recognition/page calls, which permits the user to ‘type-ahead’.

String Escape

[0336] Holds the string value of the key which will be used to end DTMF recognition (eg ‘#’).

String targetAttribute

[0337] TargetAttribute specifies the property on the primary control in which to bind the value. If not specified, this is assumed to be the Text property of the primary control.

String ClientTest

[0338] The ClientTest property references a client-side boolean function which determines under which conditions a DTMF grammar is active. If multiple grammars are specified within a DTMF object, only the first grammar with a true ClientTest function will be selected for activation during RunSpeech execution. If this property is unspecified, true is assumed.

1.5.4 DTMFGrammar

[0339] DTMFGrammar maps a key to an output value associated with the key. The following sample shows how to map the “1” and “2” keys to text output values.

[0340]

	<dtmfgrammar>
	<key value=“1”>Seattle</key>
	<key value=“2”>Boston</key>
	</dtmfgrammar>

1.6 Command Control

[0341] The command control is a special variation of answer control which can be defined in any QA control. Command controls are forms of user input which are not answers to the question at hand (eg, Help, Repeat, Cancel), and which do not need to bind recognition results into primary controls. If the QA control specifies an activation scope, the command grammar is active for every QA control within that scope. Hence a command does not need to be activated directly by a question control or an event, and its grammars are activated in parallel independently of answer controls building process. Command controls of the same type at QA controls lower in scope can override superior commands with context-sensitive behavior (and even different/extended grammars if necessary).

[0342]

	<Command
	id=“...”
	scope=“...”
	type=“...”
	RejectThreshold=“...”
	onClientReco=“...” >
	<Grammar ...>
	<dtmf ... >
	...
	</Command>

String Scope

[0343] Scope holds the id of a primary control. Scope is used in command controls for scoping the availability of the command grammars. If scope is specified for a command control, the command's grammars will be activated whenever a QA control corresponding to a primary control within the subtree of the contextual control is activated.

String Type

[0344] Type specifies the type of command (eg ‘help’, ‘cancel’ etc.) in order to allow the overriding of identically typed commands at lower levels of the scope tree. Any string value is possible in this attribute, so it is up to the author to ensure that types are used correctly.

Integer RejectThreshold

[0345] RejectThreshold specifies the minimum confidence level of recognition that is necessary to trigger the command in recognition (this is likely to be used when higher than usual confidence is required, eg before executing the result of a ‘Cancel’ command). Legal values are 0-100.

String onClientReco

[0346] on Command specifies the client-side script function to execute on recognition of the command control's grammar.

Grammar Grammar

[0347] The grammar object which will listen for the command.

DTMF DTMF

[0348] The dtmf object which will activate the command.

[0349] Using the control described above, various forms of initiatives can be developed, some examples are provided below:

2.1 Mixed initiative Dialogs

[0350] Mixed initiative dialogs provide the capability of accepting input for multiple controls with the asking of a single question. For example, the answer to the question “what are your travel plans” may provide values for an origin city textbox control, a destination city textbox control and a calendar control (“Fly from Puyallup to Yakima on September 30^th”).

[0351] A robust way to encode mixed initiative dialogs is to handwrite the mixed initiative grammar and relevant binding statements, and apply these to a single control.

[0352] The following example shows a single page used for a simple mixed initiative voice interaction about travel. The first QA control specifies the mixed initiative grammar and binding, and a relevant prompt asking for two items. The second and third QA controls are not mixed initiative, and so bind directly to their respective primary control by default (so no bind statements are required). The RunSpeech algorithm will select the QA controls based on an attribute “SpeechIndex” and whether or not their primary controls hold valid values.

[0353]

<%@ Page language=“c#” AutoEventWireup=“false”
inherits=“SDN.Page” %>
<%@ Register tagPrefix=“SDN” Namespace=“SDN” Assembly=“SDN”
%>
<html>
<body>
<Form id=“WebForm1” method=post runat=“server”>
<ASP:Label id=“Label1” runat=“server”>Departure
city</ASP:Label>
<ASP:TextBox id=“TextBox1” runat=“server” />
<br>
<ASP:Label id=“Label2” runat=“server”>Arrival
city</ASP:Label>
<ASP:TextBox id=“TextBox2” textchanged=“TextChanged”
runat=“server” />
<!-speech information -->
<Speech:QA id=“QAmixed” controlsToSpeechEnable=“TextBox1”
speechIndex=“1” runat=“server”>
	<Question id=“Q1” Answers=“A1”>
	<prompt>“Please say the cities you want to fly
from and to”</prompt>
	</Question>
	<Answer id=“A1” >
	<grammar src=“...”/>
	<bind targetElement=“TextBox1”
value=“/sml/path1”/>
	<bind targetElement=“TextBox2”
value=“/sml/path2”/>
	</Answer>
</Speech:QA>
<Speech:QA id=“QA1” controlsToSpeechEnable=“TextBox1”
speechIndex=“2” runat=“server”>
	<Question id=“Q1” Answers=“A1”>
	<prompt>“What's the departure city?”</prompt>
	</Question>
	<Answer id=“A1”>
	<grammar src=“...”/>
	</Answer>
</Speech:QA>
<Speech:QA id=“QA2” controlsToSpeechEnable=“TextBox2”
speechIndex=“3” runat=“server”>
	<Question id=“Q1” Answer=“A1”>
	<prompt>“What's the arrival city”</prompt>
	</Question>
	<Answer id=“A1” >
	<grammar src=“...”/>
	</Answer>
</Speech:QA>
</Form>
</body>
</html>

2.2 Complex Mixed Initiative

[0354] Application developers can specify several answer to the same question control with different levels of initiatives. Conditions are specified that will select one of the answers when the question is asked, depending on the initiative settings that they require. An example is provided below:

[0355]

	<Speech:QA
	id=“QA_Panel2”
	ControlsToSpeechEnable=“Panel2”
	runat=“server” >
	<Question	answers=“systemInitiative,
	mixedInitiative” .../>
	<Answer	id=“systemInitiative”
	ClientTest=“systemInitiativeCond”
	onClientReco=“SimpleUpdate” >
	<grammar src=“systemInitiative.gram” />
	</Answer>
	<Answer	id=“mixedInitiative”
	ClientTest=“mixedInitiativeCond”
	onClientReco=“MixedUpdate” >
	<grammar src=“mixedInitiative.gram” />
	</Answer>
	</Speech:QA>

[0356] Application developers can also specify several question controls in a QA control. Some question controls can allow a mixed initiative style of answer, whilst others are more directed. By authoring conditions on these question controls, application developer can select between the questions depending on the dialogue situation.

[0357] In the following example the mixed initiative question asks the value of the two textboxes at the same time (e.g., ‘what are your travel plans?’) and calls the mixed initiative answer (e.g., ‘from London to Seattle’). If this fails, then the value of each textbox is asked separately (e.g., ‘where do you leave from’ and ‘where are you going to’) but, depending on the conditions, the mixed-initiative grammar may still be activated, thus allowing users to provide both values.

[0358]

	<Speech:QA
	id=“QA_Panel2”
	ControlsToSpeechEnable=“TextBox1, TextBox2”
	runat=“server” >
	<Question
	ClientTest=“AllEmpty( )”
	answers=“AnsAll”
	.../>
	<Question
	ClientTest=“TextBox1IsEmpty( )”
	answers=“AnsAll, AnsTextBox1” .../>
	<Question
	ClientTest=“TextBox2IsEmpty( )”
	answers=“AnsAll, AnsTextBox2” .../>
	<Answer
	id=“AnsTextBox1”
	onClientReco=“SimpleUpdate”>
	<grammar src=“AnsTextBox1.gram” />
	</Answer>
	<Answer
	id=“AnsTextBox2”
	onClientReco=“SimpleUpdate” >
	<grammar src=“ AnsTextBox2.gram” />
	</Answer>
	<Answer
	id=“AnsAll”
	ClientTest=“IsMixedInitAllowed( )”
	onClientReco=“MixedUpdate”
	>
	<grammar src=“AnsAll.gram” />
	</Answer>
	</Speech:QA>

2.3 User Initiative

[0359] Similar to the command control, a standard QA control can specify a scope for the activation of its grammars. Like a command control, this QA control will activate the grammar from a relevant answer control whenever another QA control is activated within the scope of this context. Note that its question control will only be asked if the QA control itself is activated.

[0360]

	<Speech:QA
	id=“QA_Panel2”
	ControlsToSpeechEnable=“Panel2”
	runat=“server” >
	<Question ... />
	<Answer	id=“AnswerPanel2”
	scope=“Panel2”
	onClientReco=“UpdatePanel2( )” >
	<grammar src=“Panel2.gram” />
	</Answer>
	</Speech:QA>

[0361] This is useful for dialogs which allow ‘service jumping’—user responses about some part of the dialog which is not directly related to the question control at hand.

2.4 Short Time-Out Confirms

[0362] Application developers can write a confirmation as usual but set a short time-out. In the timeout handler, code is provided to that accept the current value as exact.

[0363]

	<Speech:QA
	id=“QA_Panel2”
	ControlsToSpeechEnable=“Panel2”
	runat=“server” >
	<Confirm timeOut=“20”
	onClientTimeOut=“AcceptConfirmation”... />
	<Answer	id=“CorrectPanel2”
	onClientReco=“UpdatePanel2( )” >
	<grammar src=“Panel2.gram” />
	</Answer>
	</Speech:QA>

2.5 Dynamic Prompt Building and Editing

[0364] The promptFunction script is called after a question control is selected but before a prompt is chosen and played. This lets application developers build or modify the prompt at the last minute. In the example below, this is used to change the prompt depending on the level of experience of the users.

[0365]

	<script language=javascript>
	function GetPrompt( ) {
	if(experiencedUser == true)
	Prompt1.Text = “What service do you
	want?”;
	else
	Prompt1.Text = “Please choose between
	e-mail,
	calendar and news”;
	return;
	}
	</script>
	<Speech:QA
	id=“QA_Panel2”
	ControlsToSpeechEnable=”Panel2”
	runat=“server” >
	<Question PromptFunction=“GetPrompt”... >
	<Prompt id=“Prompt1” />
	</Question>
	<Answer ... />
	</Speech:QA>

2.6 Using Semantic Relationships

[0366] Recognition and use of semantic relationships can be done by studying the result of the recognizer inside the on Reco event handler.

[0367]

	<script language=“javascript”>
	function Reco( ) {
	/*

[0368] Application developers can access the SML returned by the recogniser or recognition server. If a semantic relationship (like sport-news) is identified, the confidence of the individual elements can be increased or take any other appropriate action.

[0369]

	*/
	}
	</script>
	<Speech:QA
	id=“QA_Panel2”
	ControlsToSpeechEnable=“Panel2”
	runat=“server” >
	<Question ... />
	<Answer onClientReco=“Reco” >
	<grammar src=“Panel2.gram” />
	</Answer>
	</Speech:QA>

[0370] A mechanism is needed to provide voice-only clients with the information necessary to properly render speech-enabled pages. Such a mechanism must provide the execution of dialog logic and maintain state of user prompting and grammar activation as specified by the application developer.

[0371] Such a mechanism is not needed for multimodal clients. In the multimodal case, the page containing speech-enabled controls is visible to the user of the client device. The user of the client device may provide speech input into any visible speech-enabled control in any desired order using the a multimodal paradigm.

[0372] The mechanism used by voice-only clients to render speech-enabled pages is the RunSpeech script or algorithm. The RunSpeech script relies upon the SpeechIndex attribute of the QA control and the SpeechGroup control discussed below.

3.1 SpeechControl

[0373] During run time, the system parses a control script or webpage having the server controls and creates a tree structure of server controls. Normally the root of the tree is the Page control. If the control script uses custom or user control, the children tree of this custom or user control is expanded. Every node in the tree has an ID and it is easy to have name conflict in the tree when it expands. To deal with possible name conflict, the system includes a concept of NamingContainer. Any node in the tree can implement NamingContainer and its children lives within that name space.

[0374] The QA controls can appear anywhere in the server control tree. In order to easily deal with SpeechIndex and manage client side rendering, a SpeechGroup control is provided. The Speechgroup control is hidden from application developer.

[0375] One SpeechGroup control is created and logically attached to every NamingContainer node that contain QA controls in its children tree. QA and SpeechGroup controls are considered members of its direct NamingContainer's SpeechGroup. The top level SpeechGroup control is attached to the Page object. This membership logically constructs a tree—a logical speech tree—of QA controls and SpeechGroup controls.

[0376] For simple speech-enabled pages or script (i.e., pages that do not contain other NamingContainers), only the root SpeechGroup control is generated and placed in the page's server control tree before the page is sent to the voice-only client. The SpeechGroup control maintains information regarding the number and rendering order of QA controls on the page.

[0377] For pages containing a combination of QA control(s) and NamingContainer(s), multiple SpeechGroup controls are generated: one SpeechGroup control for the page (as described above) and a SpeechGroup control for each NamingContainer. For a page containing NamingContainers, the page-level SpeechGroup control maintains QA control information as described above as well as number and rendering order of composite controls. The SpeechGroup control associated with each NamingContainer maintains the number and rendering order of QAs within each composite.

[0378] The main job of the SpeechGroup control is to maintain the list of QA controls and SpeechGroups on each page and/or the list of QA controls comprising a composite control. When the client side markup script (e.g. HTML) is generated, each SpeechGroup writes out a QACollection object on the client side. A QACollection has a list of QA controls and QACollections. This corresponds to the logical server side speech tree. The RunSpeech script will query the page-level QACollection object for the next QA control to invoke during voice-only dialog processing.

[0379] The page level SpeechGroup control located on each page is also responsible for:

• • [0380] Determining that the requesting client is a voice-only client; and
  • [0381] Generating common script and supporting structures for all QA controls on each page.

[0382] When the first SpeechGroup control renders, it queries the System.Web.UI.Page.Request.Browser property for the browser string. This property is then passed to the RenderSpeechHTML and RenderSpeechScript methods for each QA control on the page. The QA control will then render for the appropriate client (multimodal or voice-only).

3.2 Creation of SpeechGroup Controls

[0383] During server-side page loading, the on Load event is sent to each control on the page. The page-level SpeechGroup control is created by the first QA control receiving the on Load event. The creation of SpeechGroup controls is done in the following manner: (assume a page containing composite controls).

[0384] Every QA control will receive on Load event from run time code. onLoad for a QA:

• • [0385] Get the QA's NamingContainer N1
  • [0386] Search for SpeechGroup in the N1's children
    • [0387] If already exists, register QA control with this SpeechGroup. onLoad returns.
    • [0388] If not found:
      • [0389] Create a new SpeechGroup G1, inserts it into the N1's children
      • [0390] If N1 is not Page, find N1's NamingContainer N2
      • [0391] Search for SpeechGroup in N2's children, if exists, say G2, add G1 to G2. If not, create a new one G2, inserts in to N2's children
      • [0392] Recursion until the NamingContainer is the Page (top level)

[0393] During server-side page rendering, the Render event is sent to the speech-enabled page. When the page-level SpeechGroup control receives the Render event, it generates client side script to include RunSpeech.js and inserts it into the page that is eventually sent to the client device. It also calls all its direct children to render speech related HTML and scripts. If a child is SpeechGroup, the child in turn calls its children again. In this manner, the server rendering happens along the server side logical speech tree.

[0394] When a SpeechGroup renders, it lets its children (which can be either QA or SpeechGroup) render speech HTML and scripts in the order of their SpeechIndex. But a SpeechGroup is hidden and doesn't naturally have a SpeechIndex. In fact, a SpeechGroup will have the same SpeechIndex as its NamingContainer, the one it attaches to. The NamingContainer is usually a UserControl or other visible control, and an author can set SpeechIndex to it.

3.3 RunSpeech

[0395] The purpose of RunSpeech is to permit dialog flow via logic which is specified in script or logic on the client. In one embodiment, RunSpeech is specified in an external script file, and loaded by a single line generated by the server-side rendering of the SpeechGroup control, e.g.:

[0396]

	<script language=“javascript”
	src=“/scripts/RunSpeech.js” />

[0397] The RunSpeech.js script file should expose a means for validating on the client that the script has loaded correctly and has the right version id, etc. The actual validation script will be automatically generated by the page class as inline functions that are executed after the attempt to load the file.

[0398] Linking to an external script is functionally equivalent to specifying it inline, yet it is both more efficient, since browsers are able to cache the file, and cleaner, since the page is not cluttered with generic functions.

3.4 Events

3.4.1 Event Wiring

[0399] Tap-and-talk multimodality can be enabled by coordinating the activation of grammars with the on MouseDown event. The wiring script to do this will be generated by the Page based on the relationship between controls (as specified in the ControlsToSpeechEnable property of the QA control in).

[0400] For example, given an asp:TextBox and its companion QA control adding a grammar, the <input> and <reco> elements are output by each control's Render method. The wiring mechanism to add the grammar activation command is performed by client-side script generated by the Page, which changes the attribute of the primary control to add the activation command before any existing handler for the activation event:

[0401]

	<!-- Control output -->
	<input id=“TextBox1” type=“text” .../>
	<reco id=“Reco1” ... />
	<grammar src=“...” />
	</reco>
	<!-- Page output -->
	<script>
	TextBox1.onMouseDown =
	“Reco1.Start( );”+TextBox1.onMouseDown;
	</script>

[0402] By default, hook up is via onmousedown and onmouseup events, but both StartEvent and StopEvent can be set by web page author.

[0403] The textbox output remains independent of this modification and the event is processed as normal if other handlers were present.

3.4.2 Page Class Properties

[0404] The Page also contains the following properties which are available to the script at runtime:

• [0405] SML—a name/value pair for the ID of the control and it's associated SML returned by recognition.
• [0406] SpokenText—a name/value pair for the ID of the control and it's associated recognized utterance
• [0407] Confidence—a name/value pair for the ID of the control and it's associated confidence returned by the recognizer.

[0408] The RunSpeech algorithm is used to drive dialog flow on the client device. This may involve system prompting and dialog management (typically for voice-only dialogs), and/or processing of speech input (voice-only and multimodal dialogs). It is specified as a script file referenced by URI from every relevant speech-enabled page (equivalent to inline embedded script).

[0409] Rendering of the page for voice only browsers is done in the following manner:

• [0410] The RunSpeech module or function works as follows
• [0411] (RunSpeech is called in response to document.onreadystate becoming “complete”):
  • [0412] (1) Find the first active QA control in speech index order (determining whether a QA control is active is explained below).
  • [0413] (2) If there is no active QA control, submit the page.
  • [0414] (3) Otherwise, run the QA control.

[0415] A QA control is considered active if and only if:

• • [0416] (1) The QA control's ClientTest either is not present or returns true, AND
  • [0417] (2) The QA control contains an active question control or statement control (tested in source order), AND
  • [0418] (3) Either:
    • [0419] a. The QA control contains only statement controls, OR
    • [0420] b. At least one of the controls referenced by the QA control's ControlsToSpeechEnable has an empty or default value.

[0421] A question control is considered active if and only if:

• • [0422] (1) The question control's ClientTest either is not present or returns true, AND
  • [0423] (2) The question control contains an active prompt object.

[0424] A prompt object is considered active if and only if:

• • [0425] (1) The prompt object's ClientTest either is not present or returns true, AND
  • [0426] (2) The prompt object's Count is either not present, or is less than or equal to the Count of the parent question control.

[0427] A QA control is run as follows:

• • [0428] (1) Determine which question control or statement control is active and increment its Count.
  • [0429] (2) If a statement control is active, play the prompt and exit.
  • [0430] (3) If a question control is active, play the prompt and start the Recos for each active answer control and command control.

[0431] An answer control is considered active if and only if:

• • [0432] (1) The answer control's ClientTest either is not present or returns true, AND
  • [0433] (2) Either:
    • [0434] a. The answer control was referenced in the active question contol's Answers string, OR
    • [0435] b. The answer control is in Scope

[0436] A command control is considered active if and only if:

• • [0437] (1) It is in Scope, AND
  • [0438] (2) There is not another command control of the same Type lower in the scope tree.

[0439] RunSpeech relies on events to continue driving the dialog—as described so far it would stop after running a single QA control. Event handlers are included for Prompt.OnComplete, Reco.OnReco, Reco.OnSilence, Reco.OnMaxTimeout, and Reco.OnNoReco. Each of these will be described in turn.

[0440] RunSpeechOnComplete works as follows:

• • [0441] (1) If the active Prompt object has an OnClientComplete function specified, it is called.
  • [0442] (2) If the active Prompt object was contained within a statement control, or a question control which had no active answer controls, RunSpeech is called.

[0443] RunSpeechOnReco works as follows:

• • [0444] (1) Some default binding happens—the SML tree is bound to the SML attribute and the text is bound to the SpokenText attribute of each control in ControlsToSpeechEnable.
  • [0445] (2) If the confidence value of the recognition result is below the ConfidenceThreshold of the active answer control, the Confirmation logic is run.
  • [0446] (3) Otherwise, if the active answer control has on OnClientReco function specified, it is called, and then RunSpeech is called.

[0447] RunSpeechOnReco is responsible for creating and setting the SML, SpokenText and Confidence properties of the ControlsToSpeechEnable. The SML, SpokenText and Confidence properties are then available to scripts at runtime.

[0448] RunSpeechOnSilence, RunSpeechOnMaxTimeout, and RunSpeechOnNoReco all work the same way:

• • [0449] (1) The appropriate OnClientXXX function is called, if specified.
  • [0450] (2) RunSpeech is called.

[0451] Finally, the Confirmation logic works as follows:

• • [0452] (1) If the parent QA control of the active answer control contains any confirm controls, the first active confirm control is found (the activation of a confirm control is determined in exactly the same way as the activation of a question control).
  • [0453] (2) If no active confirm control is found, RunSpeech is called.
  • [0454] (3) Else, the QA control is run, with the selected confirm control as the active question control.

[0455] For multi-modal browsers, only the grammar loading and event dispatching steps are carried out.

[0456] In this embodiment, there is no concept of primary control to speech-enable as it existed in Appendix A. The speech layer provides input to the visual layer as well as explicit support for dialog flow management. The semantic layer implements the logic needed for confirmation and validation. In a multimodal interaction, the semantic layer does not need to be used as confirmation and validation are visual and implemented using standard ASP.NET constructs. If desired though, the sematic layer can be updated with value changes made through visual or GUI interfaces in order that confirmation and validation can be still implemented.

[0457] FIG. 12 illustrates the speech controls inheritance diagram.

[0458] The following provides examples of various forms of application scenarios.

2.1 Multimodal App, Tap-and-Talk

[0459]

	<speech:QA id=“qa1” runat=“server”>
	<Answers>
	<speech:Answer SemanticItem=“siText” ID=“answer1”
	XpathTrigger=“/sml/value”	runat=“server”>
	</speech:Answer>
	</Answers>
	<Reco StartEvent=“textbox1.onmousedown”
	StopEvent=“textbox1.onmouseup” ID=“reco1”
	Mode=“Single”>
	<Grammars>
	<speech:Grammar
	Src=“http://mysite/mygrammar.grxml”
	ID=“Grammar1”	runat =“server”>
	</speech:Grammar>
	</Grammars>
	</Reco>
	</speech:QA>

2.2 Multimodal App, Click-and-Wait-for-Recognition

[0460]

	<speech:QA id=“qa1” runat=“server”>
	<Reco id=“reco1” StartEvent=“textbox1.onmousedown”
	mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“htp://mysite/mygrammar.grxml”
	rnat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/value” SemanticItem=“siText”
	runat=“server”>
	</speech:answer>
	</Answers>
	</speech:QA>

2.3 Multimodal App, Do-Field

[0461]

	<speech:QA id=“qa1” runat=“server”>
	<Reco id=“reco1”
	StartEvent=“dofieldButton.onmousedown”
	StopEvent=“dofieldButton.onmouseup”
	mode=“multiple”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/mylargegrammar.xml” runat=“server”>
	</speech:grammar>
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/value1” SemanticItem=“siOne”
	runat=“server”>
	</speech:answer>
	<speech:answer id=“answer2”
	XpathTrigger=“/sml/value2” SemanticItem=“siTwo”
	runat=“server”>
	</speech:answer>
	speech:answer id=“answer3”
	XpathTrigger=“/sml/value3” SemanticItem=“siThree”
	runat=“server”>
	</speech:answer>
	<speech:answer id=“answer4”
	XpathTrigger=“/sml/value4” SemanticItem=“siFour”
	runat=“server”>
	</speech:answer>
	<speech:answer id=“answer5”
	XpathTrigger=“/sml/value5” SemanticItem=“siFive”
	runat=“server”>
	</speech:answer>
	</Answers>
	</speech:QA>

2.4 Voice Only App, Statement

[0462]

	<speech:QA id=“welcome” PlayOnce=“true” runat=“server”>
	<Prompt InLineprompt=“Hello there!”></Prompt>
	</speech:QA>

2.5 Voice Only App, Simple Question

[0463]

	<speech:QA id=“qa1” runat=“server”>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/citygrammar.grxml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Prompt InLinePrompt=“Which city do you want to fly
	to?”></Prompt>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/city” SemanticItem=“siCity”
	runat=“server”>
	</speech:answer>
	</Answers>
	</speech:QA>

2.6 Voice Only App, Question With Mixed-Initiative (Optional Answers)

[0464]

	<speech:QA id=“qa1” runat=“server”>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/cityANDstate.xml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Prompt InLinePrompt=“Which city do you want to fly
	to?”></Prompt>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/city” SemanticItem = “siCity”
	runat=“server”>
	</speech:answer>
	</Answers>
	<ExtraAnswers>
	<speech:answer id=“answer2”
	XpathTrigger=“/sml/state” SemanticItem = “siState”
	runat=“server”>
	</speech:answer>
	</ExtraAnswers>
	</speech:QA>

2.7 Voice Only App, Explicit Confirmation

[0465]

	<speech:QA id=“qa1” runat=“server”>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/citygrammar.xml” runat=“server”>
	</speech:grammar>
	</Grammars>
	</Reco>
	<Prompt InLinePrompt=“Which city do you want to fly
	to?”></Prompt>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/city” SemanticItem=“siCity”
	confirmThreshold=“0.75”	runat=“server”>
	</speech:answer>
	</Answers>
	</speech:QA>
	<speech:QA id=“qa2” runat=“server”
	xpathAcceptConfirms=“/sml/accept”
	xpathDenyConfirms=“/sml/deny”>
	<Prompt InLinePrompt=“Did you say
	<SALT:value>textbox1.value</SALT:value>”></Prompt>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/yes_no_city.xml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Confirms>
	<speech:answer id=“answer2”
	XpathTrigger=“/sml/city” SemanticItem=“siCity”
	confirmThreshold=“0.75”	runat=“server”>
	</speech:answer>
	</Confirms>
	</speech:QA>

2.8 Voice Only App, Short Time-Out Confirmation

[0466]

	<speech:QA id=“qa1” runat=“server”
	xpathAcceptConfirms=“/sml/accept”
	xpathDenyConfirms=“/sml/deny”
	firstInitialTimeout=“500”>
	<Prompt InLinePrompt=“Did you say
	<SALT:value>textbox1.value</SALT:value>”></Prompt>
	<Reco id=“reco1” InitialTimeout=“350”
	mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/yes_no_city.grxml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Confirms>
	<speech:answer XpathTrigger=“/sml/city”
	SemanticItem=“siCity” confirmThreshold=“0.75”
	runat=“server”>
	</speech:answer>
	</Confirms>
	</speech:QA>

2.9 Voice Only App, Commands

[0467]

	<speech:QA id=“qa1” runat=“server”>
	<Prompt id=“prompt1” InLinePrompt=“Where do you want
	to fly to?”></Prompt>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/city.grxml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/city” SemanticItem=“siCity”
	runat=“server”></speech:answer>
	</Answers>
	</speech:QA>
	<speech:Command id=“command1” type=“cancel” scope=“qa1”
	OnClientCommand=“myCommand”
	runat=“server”></speech:Command>
	<script>
	function myCommand( )
	{	CallControl.Hangup( ); }
	</script>

2.10 Voice Only App, Prompt Selection

[0468]

	<speech:QA id=“qa1” runat=“server”>
	<Prompt id=“prompt1” InLinePrompt=“Where do you want
	to fly to?”></Prompt>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/city.grxml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/city” SemanticItem=“siCity”
	runat=“server”></speech:answer>
	</Answers>
	</speech:QA>
	<speech:Command id=“command1” type=“cancel” scope=“qa1”
	OnClientCommand=“myCommand”
	runat=“server”></speech:Command>
	<script>
	function myCommand( )
	{	CallControl.Hangup( ); }
	</script>
	<speech:qa id=“qa1” runat=“server”>
	<Prompt id=“prompt1”
	PromptSelectFunction=“promptSelection” />
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar src=“http://mysite/city.xml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/city” SemanticItem=“siCity”
	runat=“server”></speech:answer>
	</Answers>
	</speech:qa>
	<script>
	function promptSelection (lastCommandOrException, count,
	answerArray)
	{
	if (lastCommandOrException == “Silence”)
	{
	return “Sorry, I couldn't hear you. Please speak
	louder. Where do you want to fly to?”;
	}
	else if (count>3)
	{
	return “Communication problems are preventing me
	from hearing the arrival city. Please try again later.”;
	}
	return “Where do you want to fly to?”; //Default
	prompt
	}
	}
	</script>

2.11 Voice Only App, Implicit Confirmation

[0469]

	<speech:qa id=“qa1” runat=“server”
	xpathDenyConfirms=“/sml/deny”
	xpathAcceptConfirms=“/sml/accept”>
	<Prompt id=“prompt1”
	PromptSelectFunction=“promptSelection”></Prompt>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/yes_no_city.xml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/date” SemItem=“siDate”
	runat=“server”></speech:answer>
	</Answers>
	<Confirms>
	<speech:answer id=“confirm1”
	XpathTrigger=“/sml/city” SemItem=“siCity”
	runat=“server”></speech:answer>
	</Confirms>
	</speech:qa>
	<script>
	function promptSelection (lastCommandOrException, count,
	SemanticItemList)
	{
	var myPrompt = “ ”;
	if (SemanticItemList[“siCity”].value != null)
	{
	myPrompt = “Flying from ” +
	SemanticItemList[“siCity”].value + “. ”;
	myPrompt += “On what date?”;
	}
	else
	{ myPrompt = “On what date?”;
	}
	return myPrompt;
	}
	</script>

2.12 Voice Only App, QA With Reco and dtmf

[0470]

	<speech:qa id=“qa1” runat=“server”>
	<Prompt id=“prompt1” InLinePrompt=“Press or say one if
	you accept the charges, two	if you
	don't.”></Prompt>
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar
	src=“http://mysite/acceptCharges.xml”
	runat=“server”></speech:grammar>
	</Grammars>
	</Reco>
	<Dtmf smlContext=“sml/accept”></Dtmf>
	<Answers>
	<speech:answer id=“answer1”
	XpathTrigger=“/sml/accept”
	SemanticItem=“siAccept”
	runat=“server”></speech:answer>
	</Answers>
	</speech:qa>

2.13 Voice-Only App, Record-Only QA

[0471]

	<speech:qa id=“qa1” runat=“server”>
	<Answers>
	<speech:answer id=“a1”
	XpathTrigger=“/SML/@recordlocation”
	SemanticItem = “foo”
	runat=“server”></speech:answer>
	</Answers>
	<Reco id=“recordonly”>
	<record beep=“true”></record>
	</Reco>
	</speech:qa></FORM>

3.1 QA Activation (Voice-Only)

[0472] QA are tested for activeness in SpeechIndex order (see run-time behavior).

[0473] A QA is active when clientActivationFunction returns true AND

[0474] If the Answers array is non empty, the SemanticItems pointed to by the set of Answers are empty OR

[0475] If the answers array is empty, at least one item in the Confirm array does need confirmation.

[0476] A QA can have only Answers (normal question: Where do you want to go?), only Confirms (explicit confirmation: Did you say Boston? or short time-out confirmation: Boston.), both (implicit confirmation: When do you want to fly to Boston?) or none (statement: Welcome to my application!).

[0477] A QA can have extra answers even if it has no answers (e.g., mixed initiative).

3.2 Answer, Confirm.

[0478] Upon recognition, commands are processed first, followed by Answers, ExtraAnswers and Confirms.

[0479] A target element (e.g. textbox1.value) can be in one of these states: empty, invalid, needsConfirmation, confirmed. A target is empty before any recognition result is associated with this item, or if the item has been cleared. A target is in needsConfirmation state when a recognition result has been associated with it, but the confidence level is below the confirmationThreshold for this item. And a target is confirmed when either a recognition result has been associated with it with a confidence level high enough or a confirmation loop set it to this state explicitly.

[0480] Answers are therefore responsible for setting the value in the target element and the confidence level (this is done in a semantic layer). Confirms are responsible for confirming the item, clearing it or setting it to a new value (with a new confidence level).

3.3 Command Execution (and Scope)

[0481] Commands specify a scope and are active for all QA's within that scope. The default processing of a command is to set the current QA's lastCommandException to the command's type. If the command specifies a Grammar, this grammar is activated in parallel with any grammars in the current Reco object. QAs can be modal (allowCommands=false), in which case, no commands will be processed for that particular QA.

3.4 Validators

[0482] A CompareValidator will be active when the value of the SemanticItemToValidate it refers to has not been validated by this validator. If SemanticItemToCompare is specified (rather than ValueToCompare), then the CompareValidator will only be active if the value of the SemanticItemToCompare is non-empty (i.e. if it has been assigned a value by a previous QA).

[0483] A CustomValidator will be active when the value of the SemanticItemToValidate it refers to has not been validated by this validator.

4.1 Client Detection

[0484] The speech controls do pay attention to the variety of client that they are rendering for. If the client doesn't support SALT, the controls won't render any speech-related tags or script. Client detection is done by checking the browser capabilities and detecting whether it's a voice-only client (browser is Quadrant), or multimodal (IE, PocketIE, etc, with SALT support).

[0485] Hands-free is not a mode in the client, but rather an application-specific modality, and therefore the only support required is SALT (as in multimodal). Hands-free operation is therefore switched-on by application logic.

4.2 Multimodal

[0486] Support for multimodal applications is built in the speech controls. In multimodal operations commands, dtmf, confirm, prompts, etc do not make sense from an interaction point of view, so they won't be rendered. Tap-and-talk (or any other type of interaction, like click-and-wait-for-recognition) is enabled by hooking up the calls to start and stop recognition with GUI events using the Reco object attributes startElement/startEvent and stopElement/stopEvent, plus the Reco object mode attribute.

[0487] During render time, the speech controls are passed information specifying whether the client is a voice-only client or multimodal client. If the client is multimodal, the rendering process hooks the call to start recognition to the GUI event specified by the StartEvent attribute of the Reco object. The rendering process also hooks the call to stop recognition to the GUI event specified by the StopEvent attribute of the Reco object.

[0488] The multimodal client needs a mechanism which will invoke author-specified functions to handle speech-related events (e.g., timeouts) or recognition processing. This mechanism is the Multimodal.js script. Multimodal.js is specified in an external script file and loaded by a single line generated by server-side rendering, e.g.,

[0489]

	<script language=’“javascript”
	src=“/scripts/Multimodal.js” />

[0490] This method mirrors the ASP.NET way of generating ‘system’ client-side script loaded via URI. Linking to an external script is functionally equivalent to specifying it inline, yet is more efficient since clients are able to cache the file, and cleaner, since the page is not cluttered with generic functions.

4.3 Voice-Only

4.3.1 Runtime Script (RunSpeech)

[0491] Unlike in a multimodal interaction, where the user initiates all speech input by clicking/selecting visual elements in the GUI, a mechanism is needed to provide voice-only clients with the information necessary to properly render speech-enabled ASP.NET pages. Such a mechanism must guarantee the execution of dialog logic and maintain state of user prompting and grammar activation as specified by the author.

[0492] The mechanism used by the Speech Controls is a client-side script (RunSpeech.js) that relies upon the SpeechIndex attribute of the QA control, plus the flow control mechanisms built in the framework (ClientActivationFunction, default activation rules, etc.). RunSpeech is loaded via URI similar to the loading mechanism of Multimodal.js as described above.

4.3.2 SpeechIndex

[0493] SpeechIndex is an absolute ordering index within a naming container.

[0494] If more than one speech control has the same SpeechIndex, they are activated in source order. In situations where some controls have SpeechIndex specified and some controls do not, those with SpeechIndex will be activated first, then the rest in source order.

[0495] NOTE: Speech index is automatically set to 0 for new controls. Dialog designers should leave room in their numbering scheme to insert new QA's later. Begin with a midrange integer and increment by 100, for example. For example number QA's 1000, 1100, 1200 instead of 1, 2, 3. this leaves room for a large number of QA's at any point the dialog and plenty of room to add QA's at the beginning.

4.3.3 ClientActivationFunction

[0496] clientActivationFunction specifies a client-side script function which returns a boolean value to determine when this control is considered available for selection by the run-time control selection algorithm. If not specified, it defaults to true (control is active).

[0497] The system strategy can therefore be changed by using this as a condition to activate or de-activate QAs more sensitively than SpeechIndex. If not specified, the QA is considered available for activation.

4.3.4 Count

[0498] Count is a property of the QA control that indicates how many times that control has been activated consecutively. This Count property will be reset if the previously active QA is different that the current QA (same applies for Validators), otherwise, it is incremented by one. The Count property is exposed to application developers through the PromptSectionFunction of the Prompt object.

Controls Reference

General Authoring Notes

1. Script References are not Validated at Render Time.

[0499] The Speech Controls and objects described in this section contain attributes whose values are references to script functions written by the dialog author. These functions are executed on client devices in response to speech-related events (e.g. expiration of timeout) or as run time processing (e.g. modification of prompt text prior to playback). Render time validation is not performed on script references, i.e., no checks for existence of script functions is done during rendering of controls. If an attribute contains a reference to a client-side script function and the function does not exist, client-side exceptions will be thrown.

[0500] In voice-only mode, script functions generating exceptions during runtime will cause a redirection to the error page defined in the Web.config file. If no error page is defined, RunSpeech will continue to execute without reporting the exception.

2. All Speech Controls Should be Contained Within ASP.NET <form> Tag or Equivalant.

[0501] The Speech Control described in this section must all be placed in ASP.NET web pages inside the <form> tag. Behavior of controls placed outside the <form> tag is undefined.

3. Client-Side Script References Must Refer to Function and Not Include Parenthes.

[0502] Using the PromptSelectFunction as an example. the following is correct syntax:

• [0503] <Prompt id=“P1” PromptSelectFunction=“mySelectFunction”/>//using “mySelectFunction( )” is incorrect syntax
  4. IE Requires Exact Cases when Running Jscript.

[0504] Therefore, the case for event values specified in the StartEvent and StopEvent attributes of the Prompt object must be exactly as those events are defined. This happens to be all lowercase letters for most standard IE events. For example, the onmouseup and onmousedown events must be specified in all lowercase letters.

5. All Speech Controls Expose the Common Attribute id.

6. Behavior of Visible and Enabled Properties of Speech Controls.

[0505] Setting the visible or enabled properties of Speech Controls to “False” will cause them not to render.

7. Minimum Client Requirements

[0506] In one embodiment, clients must be running IE6.0 or greater and JScript 5.5 or greater for speech controls and associated script functions to work properly.

8. Rendering <smex> to Telserver

[0507] The speech controls automatically handle rendering <smex> tags to the telephony server on every page as is required by the server. In one embodiment, smex tags are rendered whether the client is the tel server or the desktop client.

[0508] Speech Controls provide mechanisms that allow dialog authors to specify values to control properties on an application or page basis.

5.1 Application-Level Settings

5.1.1 Application Global Variables

[0509] Dialog authors may use their application's Web.config file to set values of global variables for speech-enabled web applications. The values of the global variables persist throughout the entire lifetime of the web application. ‘Errorpage’ is the only global variable that may be specified and is set for the application during render time.

[0510]

	<appSettings>
	<add key=“errorpage” value=“...” />
	</appSettings>

[0511] The <appSettings> tag must be placed one level inside the <configuration> tag within the Web.config file.

[0512] The errorpage key specifies a URI to a default error page. Redirection to this error page will occur during run time when the speech platform or the DTMF engine returns an error. A default error page is included with the SDK; the user can also create a custom error page.

[0513] Note: Developers who create their own error page must call window.close at the bottom of the error page in the voice only case in order to release the call.

5.1.2 Application-Level Setting of Common Control Properties

[0514] Dialog authors may use their application's Web.config file to set values of common control properties and have those values persist during the lifetime of the web application. For example, an author may wish use the Web.config file to set the maxTimeout value for Reco objects in their application. The properties are set in the Web.config file using the following syntax:

[0515]

	<configuration>
	<SpeechStyleSheet>
	<Style id=“style1” >
	<QA allowCommands=“false”>
	...
	<Prompt bargein=“false” ... />
	<Reco maxTimeout=“5000”... />
	<Dtmf preFlush=“true” ... />
	<Answers confirmThreshold=“0.80” ... />
	<ExtraAnswers confirmThreshold=“0.80”
	.../>
	<Confirms confirmThreshold=“0.80”... />
	</QA>
	<Command .../>
	<CustomValidator .../>
	<CompareValidator .../>
	<SemanticItem .../>
	</Style>
	</SpeechStyleSheet>
	</configuration>

[0516] The Reco corresponding Reco object would reference the “style1” Style:

• [0517] <Reco id=“reco1” . . . StyleReference=“style1” . . . />

[0518] If the Style id is “globalStyle,” the property values set in the Style apply application-wide to pertinent controls. So, in the above example, if id=“ ” (or the property is omitted from the Style tag), a maxTimeout of 5000 milliseconds will be used for all Reco objects in the application (unless overridden).

[0519] For a complete list of properties which are settable through the SpeechStyleSheet, see below.

[0520] The StyleSheet control allows dialog authors to set values to common control properties at a page-level scope. The StyleSheet control is a collection of Style objects. The Style object exposes properties of each control that are settable on a page-level basis. The StyleSheet control is rendered for both multimodal and voice-only modes. An exception will be thrown if the StyleSheet control contains an object which is not a Style object.

[0521]

	class Stylesheet : SpeechControl
	{
	string	id{get; set;};
	StyleCollection	Styles{get;};
	}

6.1 StyleSheet Properties
Styles

[0522] Optional. Used in both multimodal and voice-only modes. The Styles property is a collection of Style objects used to set property values for Speech Controls and their objects. The property values last during the lifetime of the current page.

[0523] The Style object is used to set property values for Speech Controls and their objects. The property values last during the lifetime of the current page.

[0524]

	class Style : Control
	{
		string	id{get; set;};
		string	StyleReference{get; set;};
		QAStyle	QA{get; set;};
		CommandStyle	Command{get; set;};
		CustomValidatorStyle	CustomValidator{get; set;};
		CompareValidatorStyle	CompareValidator{get; set;};
		SemanticItemStyle	SemanticItem{get; set;};
	}

7.1 Style Properties
id

[0525] Required. The programmatic name of the Style object.

StyleReference

[0526] Optional. Used in both multimodal and voice-only modes. Specifies the name of a Style object. At render time, the StyleSheet control will search for the named Style object and also set property values specified in the named Style. An exception is thrown for an invalid StyleReference.

[0527] For every property of a speech control with a StyleReference, the value is determined as follows:

• • [0528] 1. the value is set directly in the speech control
  • [0529] 2. the style object directly referenced
  • [0530] 3. any style referenced by a style
  • [0531] 4. the global style object
  • [0532] 5. the speech control default value.

[0533] The following example sets shows two QA properties are set using StyleReference:

[0534]

	<speech:Stylesheet id=“SS”>
	<speech:Style id=“base_style” >
	<QA OnClientActive=“myOnClientActive”/>
	</speech:Style>
	<speech:Style id=“derived_style”
	StyleReference=“base_style”>
	<QA PlayOnce=“true”/>
	</speech:Style>
	</speech:Stylesheet>

QA

[0535] Optional. The QA property of the Style object is used to set property values for all QA controls on a page that reference this Style. The following example shows how to set the AllowCommands and PlayOnce properties for the QA controls that reference this Style:

[0536]

	<speech:Stylesheet id=“SS1”>
	<speech:Style id=“WelcomePageQA_Style” >
	<QA AllowCommands=“false” PlayOnce=“true”/>
	</speech:Style>
	</speech:StyleSheet>
	<QA id=“...” StyleReference=“WelcomePageQA_Style” .../>

[0537] The next example shows how to set the bargein property of all Prompt objects on a given page using Params:

[0538]

<speech:Stylesheet id=“SS2”>
<Style Name=“Style1”>
	<QA>
	<Answers ConfirmThreshold=“0.8” Reject=“0.4”/>
	<Prompt>
	<Params>
	<Param name=“BargeinType” value=“grammar”/>
	<Param name=“foo” value=“bar” />
	<Params>
	</Prompt>
	</QA>
</Style>
</speech:StyleSheet>

Command

[0539] Optional. The Command property of the Style object is used to set property values for all Command controls on a page that reference this Style.

CustomValidator

[0540] Optional. The CustomValidator property of the Style object is used to set property values for all CustomValidator controls on a page that reference this Style.

CompareValidator

[0541] Optional. The CompareValidator property of the Style object is used to set property values for all CompareValidator controls on a page that reference this Style.

SemanticItem

[0542] Optional. The SemanticItem property of the Style object is used to set property values for all SemanticItem controls on a page the reference this Style.

[0543] The following properties may be set using the Style object.

QA Properties

• [0544] AllowCommands
• [0545] PlayOnce
• [0546] XpathAcceptConfirms
• [0547] XpathDenyConfirms
• [0548] AcceptRejectThreshold
• [0549] DenyRejectThreshold
• [0550] FirstInitialTimeout
• [0551] ConfirmByOmission
• [0552] ConfirmIfEqual
• [0553] OnClientActive
• [0554] OnClientListening
• [0555] OnClientComplete.
  Prompt Properties

[0556] These apply to Prompts in QA, CompareValidator, CustomValidator and Command controls.

• [0557] Bargein
• [0558] OnClientBookmark
• [0559] OnClientError
• [0560] Prefetch
• [0561] Type
• [0562] Lang
• [0563] Params
  Reco Properties
• [0564] StartEvent
• [0565] StopEvent
• [0566] Mode
• [0567] InitialTimeout
• [0568] BabbleTimeout
• [0569] MaxTimeout
• [0570] EndSilence
• [0571] Reject
• [0572] OnClientSpeechDetected
• [0573] OnClientSilence
• [0574] OnClientNoReco
• [0575] OnClientError
• [0576] Lang
• [0577] Params
  Grammar Properties

[0578] These apply to both Reco and Dtmf grammars.

• [0579] Type
• [0580] Lang
  Dtmf Properties
• [0581] InitialTimeout
• [0582] InterDigitTimeout
• [0583] OnClientSilence
• [0584] OnClientKeyPress
• [0585] OnClientError
• [0586] Params
  Answer Properties

[0587] These apply to the Answers, ExtraAnswers and Confirms collections.

• [0588] ConfirmThreshold
• [0589] Reject
  Command Properties
• [0590] Scope
• [0591] AcceptCommandThreshold
  CompareValidator Properties
• [0592] ValidationEvent
• [0593] Operator
• [0594] Type
• [0595] InvalidateBoth
  CustomValidator Properties
• [0596] ValidationEvent
  SemanticItem Properties
• [0597] BindOnChange

[0598] The QA control is responsible for querying the user with a prompt, starting a corresponding recognition object and processing recognition results.

[0599] The QA control is rendered for both multimodal and voice-only modes.

[0600]

	class QA : IndexedStyleReferenceSpeechControl
	{
		string	id{get; set;};
		int	SpeechIndex{get; set;};
		string	ClientActivationFunction{get; set;};
		string	OnClientActive{get; set;};
		string	OnClientComplete{get; set;};
		string	OnClientListening{get; set;};
		bool	AllowCommands{get; set;};
		bool	PlayOnce{get; set;};
		string	XpathAcceptConfirms{get; set;};
		string	XpathDenyConfirms{get; set;};
		float	AcceptRejectThreshold{get; set;};
		float	DenyRejectThreshold{get; set;};
		float	FirstInitialTimeout{get; set;};
		string	StyleReference{get; set;};
		bool	ConfirmByOmission{get; set;};
		bool	ConfirmIfEqual{get; set;};
		AnswerCollection	Answers{get;};
		AnswerCollection	ExtraAnswers{get;};
		AnswerCollection	Confirms{get;};
		Prompt	Prompt{get;};
		Reco	Reco{get;};
		Dtmf	Dtmf{get;};
	}.

8.1 QA Properties

[0601] All properties of the QA control are available to the application developer at design time.

SpeechIndex

[0602] Optional. Default is Zero, which is equivalent to no SpeechIndex. Only used in voice-only mode. Specifies the activation order of speech controls on a page and the activation order of composite controls. All controls with SpeechIndex>0 will be run and then controls with SpeechIndex=0 will be run in source order. If more than one control has the same SpeechIndex, they are activated in source order. In situations where some controls specify SpeechIndex and some controls do not, those with SpeechIndex specified will be activated first, then the rest in source order. SpeechIndex values start at 1. An exception will be thrown for non-valid values of SpeechIndex.

ClientActivationFunction

[0603] Optional. Only used in voice-only mode. Specifies a client-side script function which returns a Boolean value to determine when a QA control is considered available for selection by the run-time control selection algorithm. If not specified, it defaults to true (control is active). The signature for ClientActivationFunction is as follows:

• • [0604] bool ClientActivationFunction (object lastActiveObj, string lastCommandOrException, int count)
    where:
• [0605] lastActiveobj is the last active control, e.g. QA, CustomValidator or CompareValidator. For the first activated QA on a page, lastActiveObj will be null.
• [0606] lastCommandOrException is a Command type (e.g., “Help”) or a Reco event (e.g., “Silence” or “NoReco”) of the last active control. For the first activated QA on a page or if the last active control is a validator, lastCommandOrException will be an empty string.
• [0607] count number of times the last active QA has been activated consecutively, 1 if this is the first active QA on the page. Count starts at 1 and has no limit. However, for the first activated QA on a page, count will be set to zero.
  OnClientActive

[0608] Optional. Used in both multimodal and voice-only modes. Specifies a client-side script that will be called after RunSpeech determines this QA is active (voice-only mode) or after the startEvent is fired (in multimodal) and before processing the QA (e.g., playing a prompt or starting recognition). The onClientActive function does not return values. The signature for onClientActive is as follows:

• [0609] function onClientActive(string eventsource, string lastCommandOrException, int Count, object SemanticItemList)
  where:
• [0610] eventsource is the id of the object (specified by Reco.StartEvent) whose event started the Reco associated with the QA (for multimodal). eventsource will be null in voice-only mode.
• [0611] lastCommandOrException is a Command type (e.g., “Help”) or a Reco event (e.g., “Silence” or “NoReco”) for voice-only mode. lastCommandOrException is the empty string for multimodal.
• [0612] Count is the number of times the QA has been activated consecutively. Count starts at 1 and has no limit for voice-only mode. Count is zero for multimodal.
• [0613] SemanticItemList For voice-only mode, SemanticItemList is an associative array that maps semantic item id to semantic item objects. For multimodal, SemanticItemList will be null.
  OnClientComplete

[0614] Optional. Used in both multimodal and voice-only modes. Specifies a client-side script that will be called after execution of a QA (successfully or not) and before passing dialog control back to the RunSpeech algorithm (in voice-only) or the end user (in multimodal). The OnClientComplete function is called before postbacks to the server for QAs whose AutoPostBack attribute of the Answer object is set to true. The onClientComplete function does not return values. The signature for onClientComplete is as follows:

• [0615] function onClientComplete (string eventsource, string lastCommandOrException, int Count, object SemanticItemList)
  where:
• [0616] eventsource is the id of the object (specified by Reco.StopEvent) whose event stopped the Reco associated with the QA (for multimodal). eventsource will be null in voice-only mode.
• [0617] lastCommandOrException is a Command type (e.g., “Help”) or a Reco event (e.g., “Silence” or “NoReco”) for voice-only mode. lastCommandOrException is the empty string for multimodal.
• [0618] Count is the number of times the QA has been activated consecutively. Count starts at 1 and has no limit for voice-only mode. Count is zero for multimodal.
• [0619] SemanticItemList For voice-only mode, SemanticItemList is an associative array that maps semantic item id to semantic item objects. For multimodal, SemanticItemList will be null.
  OnClientListening

[0620] Optional. Used in both multimodal and voice-only modes. Specifies a client-side script (function) that will be called/executed after successful start of the reco object. The main use is so the GUI can change to show the user that they can start speaking. The function does not return any values. The signature for OnClientListening is as follows:

• [0621] function onClientListening(string eventsource, string lastCommandOrException, int Count, object SemanticItemList)
  where:
• [0622] eventsource is the id of the object (specified by Reco.StartEvent) whose event started the Reco associated with the QA (for multimodal). eventsource will be null in voice-only mode.
• [0623] lastCommandOrException is a Command type (e.g., “Help”) or a Reco event (e.g., “Silence” or “NoReco”) for voice-only mode. lastCommandOrException is the empty string for multimodal.
• [0624] Count is the number of times the QA has been activated consecutively. Count starts at 1 and has no limit for voice-only mode. Count is zero for multimodal.
• [0625] SemanticItemList For voice-only mode, SemanticItemList is an associative array that maps semantic item id to semantic item objects. For multimodal, SemanticItemList will be null.

[0626] Note: In multimodal mode OnClientListening is only available if author chooses to use StartEvent. If author decides to start reco programmatically, then onClientListening is not called for the author because the author can detect when reco.start returns successfully.

[0627] Note: OnClientListening is ignored when specified in QA's that do not contain reco objects.

AllowCommands

[0628] Optional. Only used in voice-only mode. Indicates whether or not Commands may be activated for a QA control. When AllowCommands is set to false, no commands may be activated. Defaults to true.

PlayOnce

[0629] Optional. Only used in voice-only mode. Specifies whether or not a QA may be activated more than once per page. If not specified, PlayOnce is set to false. PlayOnce=“true” may be used to author statements like welcoming prompts. When a QA is reduced to a statement (no reco), setting PlayOnce=“false” will provide dialog authors with the capability to enable a “repeat” functionality on a page that reads email messages.

XpathAcceptConfirms

[0630] Optional. Only used in voice-only mode. Specifies the path in the sml document (recognition result) that indicates the confirm items were accepted. Required if Confirms are specified. If XpathAcceptConfirms is specified without a Confirm being specified it is ignored. XpathAcceptConfirms must be a valid xml path. An invalid xml path will cause a redirection to the default error page during run time.

XpathDenyConfirms

[0631] Optional. Used only in voice-only mode. Specifies the path in the sml document that indicates the confirm items were denied. Required if Confirms are specified. If a Confirm is specified and XpathDenyConfirms is not set an exception is thrown. If XpathDenyConfirms is specified without a Confirm being specified it is ignored. XpathDenyConfirms must be a valid xml path. An invalid xml path will cause a redirection to the default error page during run time.

AcceptRejectThreshold

[0632] Optional. Used only in voice-only mode. If confidence for an accept confirm is not above this threshold no action will be taken. Legal values are 0-1 and are platform specific. An exception will be thrown for out of range AcceptRejectThreshold values. Default is zero.

DenyRejectThreshold

[0633] Optional. Used only in voice-only mode. If confidence for a deny confirm is not above this threshold no action will be taken. Legal values are 0-1 and are platform specific. An exception will be thrown for out of range DenyRejectThreshold values. Default is zero.

FirstInitialTimeout

[0634] Optional. Only used in voice-only mode. Specifies the initial timeout in msec for the QA when count==1. The status of the TargetElements specified in the Confirms answer list will be set to “Confirmed” if no speech is detected within firstInitialTimeout milliseconds. If not specified the default value of firstInitialTimeout is 0, which means that silence does not imply confirmation of the Answer. An exception will be thrown if firstInitialTimeout is specified for a QA that does not contain Confirms. An exception will be thrown for negative values of FirstInitialTimeout.

StyleReference

[0635] Optional. Used in both multimodal and voice-only modes. Specifies the name of a Style object. At render time, the QA control will search for the named Style control and will use any property values specified on the Style as default values for its own properties. Explicitly set property values on the control will override those set on the Style.

ConfirmByOmission

[0636] Optional. Used only in voice-only mode. Default is true. This flag controls confirmation of more than one item. If the flag is set to true, then any semantic items whose xpath is not present in the reco result, will be set to Confirmed. ConfirmByOmission enables the following scenario:

• [0637] (ConfirmByOmission=true)
• [0638] Q: Flying from?
• [0639] A: Boston.
• [0640] Q: Flying to?
• [0641] A: Seattle.
• [0642] Q: From Boston to Seattle?
• [0643] A: From NY.
• [0644] (Seattle is confirmed as destination city).
  ConfirmIfEqual

[0645] Optional. Used only in voice-only mode. Default is true. This flag controls the processing of corrections during confirmation. If ConfirmIfEqual is true and a recognized correction is the same value already in the semantic item, the item is maked confirmed. If ConfirmIfEqual is false and a recognized correction is the same value already in the semantic item, the item is maked as needing confirmation.

Answers

[0646] Optional. An array of answer objects. This list of objects is used both to determine activation, and to carry out semantic processing logic. An exception will be thrown if an Answers collection contains non-answer objects.

ExtraAnswers

[0647] Optional. An array of answer objects. These items are not used for activation, but they are taken into account when processing recognition results. If an ExtraAnswer is recognized, it will overwrite the semantic item it points to, even if it was previously confirmed.

Confirms

[0648] Optional. An array of answer objects. These items are used for activation if the answers array is empty and they affect the confirmation logic.

Prompt

[0649] Optional for multimodal. Required for voice-only. An exception is thrown if a Prompt is not specified in voice-only mode.

Reco

[0650] Optional for multimodal and voice-only. Typically, only one reco can be specified in a QA.

Dtmf

[0651] Optional. Only used in voice-only mode. Typically, only one Dtmf can be specified in a QA.

[0652] The Command control provides a way for obtaining user input that is not an answer to the question at hand (eg, Help, Repeat, Cancel), and which does not map to textual input into primary controls. A Command specifies an activation scope, which means that its grammar is active (in parallel with the current recognition grammar) for every QA within that scope. Commands have a type attribute which is used to implement a chain of events: Commands of the same type at QAs lower in scope can override superior commands with context-sensitive behavior (and even different/extended grammars if necessary) and to notify the QA what command was uttered (via the reason parameter).

[0653] Commands are not rendered for multimodal mode.

[0654]

class Command : SpeechControl
{
	string	id{get; set;};
	string	Scope{get; set;};
	string	Type{get; set;};
	string	XpathTrigger{get; set;};
	float	AcceptCommandThreshold{get;
set;};
	string	OnClientCommand{get; set;};
	bool	AutoPostBack{get; set;};
	TriggeredEventHandler	OnTriggered;
	string	StyleReference{get; set;};
	Prompt	Prompt{get;};
	Grammar	Grammar{get;};
	Grammar	DtmfGrammar{get;};
}

9.1 Command Properties

[0655] All properties of the Command control are available to the application developer at design time.

Scope

[0656] Required. Only used in voice-only mode. Specifies the id of a QA or other ASP.NET control (e.g., form, panel, or table). Scope is used in Commands to specify when the Command's grammars will be active. Exceptions are thrown if Scope is invalid or not specified.

Type

[0657] Required. Only used in voice-only mode. Specifies the type of command (eg ‘help’, ‘cancel’ etc.) in order to allow the overriding of identically typed commands at lower levels of the scope tree. Any string value is possible in this attribute, so it is up to the author to ensure that types are used correctly. An exception is thrown if Type is not specified.

[0658] Note: An exception will be thrown if more than 1 Command of same Type has the same Scope. For example, 2 Type=“Help” Commands for the same QA (Scope=“QA1”).

AcceptCommandThreshold

[0659] Optional. Only used in voice-only mode. Specifies the minimum confidence level of recognition that is necessary to trigger the command (this is likely to be used when higher than usual confidence is required, e.g. before executing the result of a ‘Cancel’ command). Legal values are 0-1. Default value is 0. Exceptions will be thrown for out of range AcceptCommandThreshold values.

[0660] If a command is matched (its xpathTrigger is present in the recoResult) no further commands will be processed, and no Answers, ExtraAnswers, Confirms, etc. will be processed. Then, if the confidence of the node specified by XpathTrigger is greater than or equal to the acceptThreshold, the active QAs LastCommandOrException is set to the Command's type, and the Command's on Command function is called. Otherwise (if the confidence of the node is less than the acceptThreshold) the active QAs LastCommandOrException is set to “NoReco” and the active QAs Reco's OnClientNoReco function is called.

XpathTrigger

[0661] Required. Only used in voice-only mode. SML document path that triggers this command. An exception will be thrown if XpathTrigger is not specified. XpathTrigger must be a valid xml path. An invalid xml path will cause a redirection to the default error page during run time.

OnClientCommand

[0662] Optional. Only used voice-only mode. Specifies the client-side script function to execute on recognition of the Command's grammar. The function does not return any values. The signature for OnClientCommand is as follows:

• • [0663] function OnClientCommand(XMLNode smlNode)
    where: smlNode is the matched SML node.

[0664] Note: If AutoPostBack is set to true, the OnClientCommand function is executed before posting back to the server. If the author wishes to persist any page state across postback, the OnClientCommand function is a good place to invoke the ClientViewState object of RunSpeech.

AutoPostBack

[0665] Optional. Only used in voice-only mode. Specifies whether or not the Command control posts back to the server each time a Command grammar is recognized. Default is false. If set to true, the server-side Triggered event is fired.

[0666] The internal state of the voice-only page is maintained automatically during postback. Authors may use the ClientViewState object of RunSpeech to declare and set additional values they wish to persist across postbacks.

OnTriggered

[0667] Optional. Only used in voice-only mode. Specifies a server-side script function to be executed when the Triggered event is fired (see autopostback attribute above). This handler must have the form (in C#—the signature would look slightly different in other languages):

• [0668] void myFunction (object sender, CommandTriggeredEventArgs e);

[0669] The handler can be assigned to in two different ways—

• [0670] declaratively:
• [0671] <speech:Command . . . OnTriggered=“myFunction” . . . />
• [0672] or programmatically:
• [0673] Command.Triggered+=new TriggeredEventHandler(myFunction);
• [0674] TriggeredEventHandler is what is called a “delegate”—it basically specifies the signature of functions which can handle its associated event type. It looks like this:
  • [0675] public delegate void TriggeredEventHandler(object sender, TriggeredEventArgs e);
    where:
• [0676] TriggeredEventArgs is a class derived from System.EventArgs which contains one public property, string Value.

[0677] An exception will be thrown if AutoPostBack is set to true and no handler is specified for the Triggered event.

[0678] An exception will be thrown if AutoPostBack is set to false and a handler is specified for the Triggered event.

StyleReference

[0679] Optional. Only used in voice-only mode. Specifies the name of a Style object. At render time, the QA control will search for the named Style control and will use any property values specified on the Style as default values for its own properties. Explicitly set property values on the control will override those set on the Style.

Prompt

[0680] Optional. May be used to specify prompt to be played for global commands.

Grammar

[0681] Optional. The grammar object which will listen for the command.

[0682] Note: The grammar object is optional because the QA scoped by this command may contain the rule that generates this command's Xpath. The author has the flexibility of specifying the rule in the QA control or the Command control.

DtmfGrammar

[0683] Optional. The DtmfGrammar object which will activate the command. Available at run time.

[0684] Note: The DtmfGrammar object is optional because the QA scoped by this command may contain the rule that generates this command's Xpath. The author has the flexibility of specifying the rule in the QA control or the Command control. DtmfGrammars for all Commands along the QA's scope chain will be combined into the Grammars collection for the QA's Dtmf object.

[0685] Speech Controls does not provide a set of common commands—e.g., help, cancel, repeat.

[0686] This control compares two values, applying the operator, and if the comparison is false, invalidates the item specified by SemanticItemToValidate. Optionally, both items (ToCompare and ToValidate) are invalidated. The CompareValidator is triggered on the client by change or confirm events; however, validation prompts are played in SpeechIndex order.

[0687] The CompareValidator control is rendered for voice-only mode. For multimodal, ASP.NET validator controls may be used.

[0688]

class CompareValidator : IndexedStyleReferenceSpeechControl
{
	string	id{get; set;};
	int	SpeechIndex{get; set;};
	ValidationType	Type{get; set;};
	string	ValidationEvent{get; set;};
	string	SemanticItemToCompare{get;
set;};
	string	ValueToCompare{get; set;};
	string	SemanticItemToValidate{get;
set;};
	ValidationCompareOperator	Operator(get; set;};
	bool	InvalidateBoth{get; set;};
	string	StyleReference{get; set;};
	Prompt	Prompt{get;};
}

10.1 CompareValidator Properties

[0689] All properties of the CompareValidator control are only used in voice-only mode and are available to the application developer at design time.

SpeechIndex

[0690] Optional. Specifies the activation order of CompareValidator controls on a page. If more than one control has the same SpeechIndex, they are activated in source order. In situations where some controls specify SpeechIndex and some controls do not, those with SpeechIndex specified will be activated first, then the rest in source order. SpeechIndex values start at 1. An exception will be thrown for non-valid values of SpeechIndex.

Type

[0691] Required. Sets the datatype of the comparison. Legal values are “String”, “Integer”, “Double”, “Date”, and “Currency”. Default value is “String”.

ValidationEvent

[0692] Default is “onconfirmed”. ValidationEvent may be set to one of two values, either “onchange” or “onconfirmed”.

[0693] If ValidationEvent is set to “onchanged”, the CompareValidator will be run each time the value of the Text property of the associated SemanticItem changes. The CompareValidator control will be run before the SemanticItem's OnChanged handler is called. The SemanticItem's OnChanged handler will only be called if the CompareValidator does indeed validate the changed data. If the CompareValidator invalidates the data, the State of the SemanticItem is set to Empty and the OnChanged handler is not called.

[0694] If ValidationEvent is set to “onconfirmed”, the CompareValidator will be run each time the State of the associated SemanticItem changes to Confirmed. The CompareValidator control will be run before the SemanticItem's OnConfirmed handler is called. The SemanticItem's OnConfirmed handler will only be called if the CompareValidator does indeed validate the changed data. If the CompareValidator invalidates the data, the State of the SemanticItem is set to Empty and the OnConfirmed handler is not called.

[0695] After processing all SemanticItems involved a recognition turn, RunSpeech starts again. At that point, the previously failed validators will be active and RunSpeech will select the first QA/Validator that is active in SpeechIndex order. It is the author's responsibility to place the validator controls directly before the QA control that collects the answer for the SemanticItem in order to get the correct behavior.

SemanticItemToCompare

[0696] Optional. Either SemanticItemToCompare or ValueToCompare must be specified. Specifies the Id of the SemanticItem which will be used as the basis for the comparison. Available at design time and run time. An exception will be thrown if either SemanticItemToCompare or ValueToCompare is not specified.

ValueToCompare

[0697] Optional. Either SemanticItemToCompare or ValueToCompare must be specified. Specifies the value to be used as the basis for the comparison. The author may wish to specify the value here instead of taking the value from the semantic item. If both ValueToCompare and SemanticItemToCompare are set, SemanticItemToCompare takes precedence. An exception will be thrown if either SemanticItemToCompare or ValueToCompare is not specified. An exception will be thrown if ValueToCompare can not be converted to a valid Type.

SemanticItemToValidate

[0698] Required. Specifies the Id of the SemanticItem that is being validated against either ValueToCompare or SemanticItemToCompare. An exception will be thrown for unspecified SemanticItemToValidate.

Operator

[0699] Optional. One of “Equal”, “NotEqual”, “GreaterThan”, “GreaterThanEqual”, “LesserThan”, “LesserThanEqual”, “DataTypeCheck”. Default value is “Equal”. The values are compared in the following order: Value to Validate [operator] ValueToCompare.

InvalidateBoth

[0700] Optional. If true, both SemanticItemToCompare and SemanticItemToValidate are marked Empty. Default is false (i.e., invalidate only the SemanticItemtToInvalidate). If SemanticItemToValidate has not been set (i.e. ValueToCompare has been specified), InvalidateBoth is ignored.

[0701] The following example illustrates the usage of the InvalidateBoth attribute. The scenario is an itinerary application. The user has already been prompted and answered the question for departing city. At this point in the dialog an ASP.NET textbox control has been filled with the recognition results (assume txtDepartureCity.Value=“Austin”).

[0702] The next QA prompts the user for the arrival city, the SemanticItem object binds to txtArrivalCity.Value. In response to the prompt, the user says “Boston”. However, the recognition engine returns “Austin” (e.g. arrival city is same as departing city).

[0703] The CompareValidator control may be used to direct the dialog flow in this case to re-prompt the user for both departing and arriving cities:

[0704]

	<CompareValidator	id=“compareCities” SpeechIndex=“5”
	Type=“String”
	SemanticItemToCompare=“si_DepartureCity”
	SemanticItemToValidate=“si_ArrvivalCity”
	Operator=“NotEqual”
	InvalidateBoth=“True”
	runat=“server”
	</CompareValidator>

StyleReference

[0705] Optional. Specifies the name of a Style object. At render time, the QA control will search for the named Style control and will use any property values specified on the Style as default values for its own properties. Explicitly set property values on the CompareValidator control will override those set on the Style.

Prompt

[0706] Optional. Prompt to indicate the error.

[0707] The CustomValidator control is used to validate recognition results when complex validation algorithms are required. The control allows dialog authors to specify their own validation routines. The CustomValidator is triggered on the client by change or confirm events; however, validation prompts are played in SpeechIndex order.

[0708] The CustomValidator control is only rendered for voice-only mode. For multimodal, ASP.NET validator controls may be used.

[0709]

	class CustomValidator : IndexedStyleReferenceSpeechControl
	{
		string	id{get; set;};
		int	SpeechIndex{get; set;};
		string	ValidationEvent{get; set;};
		string	SemanticItemToValidate{get; set;};
		string	ClientValidationFunction{get; set;};
		string	StyleReference{get; set;};
		Prompt	Prompt{get;};
	}

11.1 CustomValidator Properties

[0710] All properties of the CustomValidator control are only used in voice-only mode and are available to the application developer at design time.

SpeechIndex

[0711] Optional. Only used in voice-only mode. Specifies the activation order of speech controls on a page and the activation order of composite controls. If more than one control has the same SpeechIndex, they are activated in source order. In situations where some controls specify SpeechIndex and some controls do not, those with SpeechIndex specified will be activated first, then the rest in source order. SpeechIndex values start at 1. An exception will be thrown for non-valid values of SpeechIndex.

ValidationEvent

[0712] Default is “onconfirmed”. ValidationEvent may be set to one of two values, either “onchange” or “onconfirmed”.

[0713] If ValidationEvent is set to “onchanged”, the CustomValidator will be run each time the value of the Text property of the associated SemanticItem changes. The CustomValidator control will be run before the SemanticItem's OnChanged handler is called. The SemanticItem's OnChanged handler will only be called if the CustomValidator does indeed validate the changed data. If the CustomValidator invalidates the data, the State of the SemanticItem is set to Empty and the OnChanged handler is not called.

[0714] If ValidationEvent is set to “onconfirmed”, the CustomValidator will be run each time the State of the associated SemanticItem changes to Confirmed. The CustomValidator control will be run before the SemanticItem's OnConfirmed handler is called. The SemanticItem's OnConfirmed handler will only be called if the CustomValidator does indeed validate the changed data. If the CustomValidator invalidates the data, the State of the SemanticItem is set to Empty and the OnConfirmed handler is not called.

[0715] After processing all SemanticItems involved a recognition turn, RunSpeech starts again. At that point, the previously failed validators will be active and RunSpeech will select the first QA/Validator that is active in SpeechIndex order. It is the author's responsibility to place the validator controls directly before the QA control that collects the answer for the SemanticItem in order to get the correct behavior.

SemanticItemToValidate

[0716] Required. Specifies the id of the SemanticItem that is being validated. An exception will be thrown for unspecified SemanticItem ToValidate.

ClientValidationFunction

[0717] Required. Specifies a function that checks the value of the SemanticItemToValidate.AttributeToValidate and returns true or false indicating whether the value is valid or invalid. The signature for ClientValidationFunction is as follows:

• • [0718] bool ClientValidationFunction (string value)
    where:
• [0719] value is the contents of ElementToValidate.AttributeToValidate.

[0720] An exception will be thrown if ClientValidationFunction is not specified.

StyleReference

[0721] Optional. Specifies the name of a Style object. At render time, the QA control will search for the named Style control and will use any property values specified on the Style as default values for its own properties. Explicitly set property values on the control will override those set on the Style.

Prompt

[0722] Optional. Prompt to indicate the error.

[0723] The Answer object contains information on how to process recognition results and bind the results to controls on an ASP.NET page.

How Answer Object is Used.

Voice-Only Mode.

[0724] The RunSpeech script uses the Answer object to perform answer processing on the client. Answer processing begins when the OnReco event fired by the speech platform is received by the client. The resultant SML document returned by the speech platform is searched for the node specified by the required XpathTrigger attribute. If the XpathTrigger node is found in the SML document and contains a non-null value, the value is filled into the semantic item specified in the SemanticItem property of the answer. For non-existent XpathTrigger in the SML document or null value of XpathTrigger, RunSpeech looks for the next QA to activate.

[0725] After the non-null value of the XpathTrigger node is found, RunSpeech invokes the ClientNormalization function (if specified). The ClientNormalizationFunction returns a text string that reflects the author-defined transformation of the value of the XpathTrigger node. For example, the author may wish to transform the date “Nov. 17, 2001” returned by the speech platform to “11/17/2001”. Semantic items are used for both simple and complex data binding.

[0726] The SML document returned by the speech platform may contain a platform-specific confidence rating for each XpathTrigger node. During answer processing, RunSpeech compares this confidence rating to the value specified in the ConfirmThreshold attribute of the Answer object. Results of the comparison are then used to set the internal confirmed state of the semantic item. This state information is subsequently used to determine whether or not an answer requires confirmation from the user.

[0727] RunSpeech internally marks an answer as needing confirmation if the confidence returned with the XpathTrigger is less than or equal to the value of the ConfirmThreshold attribute. Otherwise RunSpeech internally marks the semantic item associated with the answer as confirmed. This internal state information is used during confirmation processing.

Multimodal.

[0728] The Answer object is used in multimodal scenarios by the Multimodal.js script just as it is used by RunSpeech in voice-only (described above) with one exception. In multimodal, platform-specific confidence ratings are not compared to the ConfirmThreshold attribute of the Answer object, therefore internal state information of each answer is not maintained. Confirmation of results is done visually. If an incorrect result is bound to a visual control, the user senses the problem visually and may then initiate another speech input action to correct the error.

[0729] Rendered for both multimodal and voice-only modes

[0730]

	class Answer : Control
	{
	string	id{get; set;};
	float	Reject{get; set;};
	float	ConfirmThreshold{get; set;};
	string	XpathTrigger{get; set;};
	string	SemanticItem{get; set;};
	string	ClientNormalizationFunction{get; set;};
	string	StyleReference{get; set;};
	}

12.1 Answer Properties

[0731] All properties of the Answer object are available to the application developer at design time.

Reject

[0732] Optional. Used in both multimodal and voice-only modes. Specifies the rejection threshold for the Answer. Answers having confidence values below Reject will cause a noReco event to be thrown. If not specified, the value 0 will be used. Legal values are 0-1 and are platform specific. An exception will be thrown for out of range Reject values.

[0733] Rejected Answers are treated as if they were not present in the reco result to begin with. If, after this processing, no relevant information remains (no Answers, ExtraAnswers, Confirms, Commands, or xpathAcceptConfirms/xpathDenyConfirms), an onnoreco event is fired (which mimics exactly the tags version).

ConfirmThreshold

[0734] Optional. Used in voice-only mode. Specifies the minimum confidence level of recognition that is necessary to mark this item as confirmed. If the confidence of the matched item is less than or equal to this threshold, the item is marked as needing confirmation. Legal values are 0-1. Default value is 0. An exception will be thrown for out of range ConfirmThreshold values.

XpathTrigger

[0735] Required for Answers and ExtraAnswers. Optional for Confirms. Used in both multimodal and voice-only modes. Specifies what part of the SML document this answer refers to. It is specified as an XPath on the SML output from recognition. An exception will be thrown if XpathTrigger is not specified for Answers or ExtraAnswers. XpathTrigger must be a valid xml path. An invalid xml path will cause a redirection to the default error page during run time.

[0736] For Confirms, if XpathTrigger is not set or set to the empty string, the confirm won't allow for correction. Yes/no confirmations are enabled when XpathTrigger is used in this way.

SemanticItem

[0737] Optional. Used in both multimodal and voice-only modes.

ClientNormalizationFunction

[0738] Optional. Used in both multimodal and voice-only modes. Specifies a client-side function that will take the matched sml node as a parameter and returns a string that reflects author-specified normalization (transformation) of the recognized item. The signature for ClientNormalizationFunction is as follows:

[0739]

	string ClientNormalizationFunction (XMLNode SMLnode,
	object SemanticItem)

where:

• [0740] SMLnode is the node specified in the Xpath.
• [0741] SemanticItem is the client-side SemanticItem object specified in the Answer object.
  StyleReference

[0742] Optional. Used in both multimodal and voice-only modes. Specifies the name of a Style object. At render time, the Answer object will search for the named Style control and will use any property values specified on the Style as default values for its own properties. Explicitly set property values by the Answer object will override those set on the referenced Style.

[0743] SemanticMap is a container of SemanticItem objects.

[0744]

	class SemanticMap : SpeechControl
	{
	SemanticItemCollection SemItems{get;};
	SemanticItem GetSemanticItem (string name);
	}

13.1 SemanticMap Properties
SemItems

[0745] A collection of SemanticItem objects.

13.2 SemanticMap Methods

GetSemanticItem

[0746] This is a function that takes the id of a SemanticItem and returns the SemanticItem object. The signature of GetSemanticItem is:

• • [0747] function GetSemanticItem(string id)

[0748] The SemanticItem object describes where and when an Answer's recognition results are written to visual controls on a page. The object also keeps track of the current state of Answers, i.e., whether an Answer has changed or been confirmed.

[0749]

	class SemanticItem : Control
	{
		string	id{get; set;};
		string	TargetElement{get; set;};
		string	TargetAttribute{get; set;};
		bool	BindOnChanged{get; set;};
		string	BindAt{get; set;};
		bool	AutoPostBack{get; set;};
		string	OnClientChanged{get; set;};
		string	OnClientConfirmed{get; set;};
		SemanticEventHandler	Changed;
		SemanticEventHandler	Confirmed;
		string	Text{get;};
		SemanticState	State{get;};
		StringDictionary	Attributes{get; set;};
		string	StyleReference{get;};
	}

14.1 SemanticItem Properties
id

[0750] Required. The programmatic id of this semantic item.

TargetElement

[0751] Optional. Used in both multimodal and voice-only modes. Specifies the id of the visual control to which the recognition results should be written. If specified, default binding will occur when the value is changed or confirmed depending on the value of BindOnChanged. An exception is thrown if TargetElement is the id of multiple controls.

TargetAttribute

[0752] Optional. Used in both multimodal and voice-only modes. Specifies the property name of the TargetElement to which this answer should be written. The default value is null. An exception will be thrown if TargetElement is specified and TargetAttribute is not specified.

BindOnChanged

[0753] Optional. Used voice-only mode, ignored in multimodal. Default is false. In VoiceOnly mode, BindOnChanged controls when to bind recognition results to visual elements.

[0754] A value of true causes binding everytime the value of the SemanticItem changes.

[0755] A value of false causes binding only when the SemanticItem has been confirmed.

BindAt

[0756] Optional. Used in both multimodal and voice-only modes. Can be omitted or set to “server”. Default is null (omitted). If BindAt is set to “server”, it indicates that the TargetElement/TargetAttribute pair refers to a server-side control or property. An exception will be thrown when BindAt is set to an invalid value.

[0757] If BindAt is “server”, an exception will be thrown if:

• • [0758] SemanticItem.TargetElement is not a server-side control, or
  • [0759] SemanticItem.TargetAttribute is not a member of the control specified by SemanticItem.TargetElement, or
  • [0760] SemanticItem.TargetAttribute is a member of SemanticItem.TargetElement, but is not of type string, or
  • [0761] SemanticItem.TargetAttribute is a string, but is read-only.
    AutoPostBack

[0762] Optional. Used in both multimodal and voice-only modes. Specifies whether or not the control posts back to the server when the binding event is fired. The binding event can be on Changed or on Confirmed and is controlled by the value of BindOnChange. Default is false.

[0763] The state of the voice-only page is maintained automatically during postback. Authors may use the ClientViewState object of RunSpeech to declare and set any additional values they wish to persist across postbacks.

OnClientChanged

[0764] Optional. Used in both multimodal and voice-only modes. Specifies a client-side function to be called when the value of the Text property of this SemanticItem changes. The function does not return any values. The signature for OnClientChanged is as follows:

• • [0765] function OnClientChanged(object SemanticItem)
    where SemanticItem is the client-side SemanticItem object.

[0766] Note: If AutoPostBack is set to true, the OnClientChanged function is executed before posting back to the server. If the author wishes to persist any page state across postback, the OnClientChanged function is a good place to access the ClientViewState object of RunSpeech.

OnClientConfirmed

[0767] Optional. Used in both multimodal and voice-only modes. Specifies a client-side function to be called when this SemanticItem's [value is confirmed. The function does not return any values. The signature for OnClientConfirmed is as follows:

• • [0768] function OnClientConfirmed(object SemanticItem)
    where SemanticItem is the client-side SemanticItem object.

[0769] Note: If AutoPostBack is set to true, the OnClientConfirmed function is executed before posting back to the server. If the author wishes to persist any page state across postback, the OnClientConfirmed function is a good place to access the ClientViewState object of RunSpeech

Changed

[0770] Optional. Used in both multimodal and voice-only modes. Specifies a server-side script function to be executed when the Changed event is fired.

[0771] The signature of a SemanticEventHandler is: (in C#—the signature would look slightly different in other languages)

[0772]

	public delegate void SemanticEventHandler ( object
	sender, SemanticEventArgs e ) ;

where:

• [0773] SemanticEventArgs is a class derived from System.EventArgs.

[0774]

	public class SemanticEventArgs : EventArgs
	{
	public string	Text {get;};
	public StringDictionary Attributes {get;}
	}

[0775] Text

[0776] Returns the value that this SemanticItem has been set to.

[0777] State

[0778] Returns the state of this SemanticItem.

Confirmed

[0779] Optional. Used in both multimodal and voice-only modes. Specifies a server-side script function to be executed when the Confirmed event is fired. In multimodal mode, the Confirmed event will be fired immediately after the Changed event.

[0780] The signature of a SemanticEventHandler is: (in C#—the signature would look slightly different in other languages)

• • [0781] public delegate void SemanticEventHandler (object sender, SemanticEventArgs e);
    where:
• [0782] SemanticEventArgs is a class derived from System.EventArgs.

[0783]

	public class SemanticEventArgs : EventArgs
	{
	public string Text {get;}
	public StringDictionary Attributes {get;}
	}

[0784] Text

[0785] Read only. Returns the value that this SemanticItem has been set to.

[0786] State

[0787] Read only. Returns the state of this SemanticItem.

Text

[0788] The text value that this SemanticItem has been set to. Default is null.

State

[0789] The confirmation state of this SemanticItem. Values of State will be one of SemanticState.Empty, SemanticState.NeedsConfirmation or SemanticState.Confirmed.

Attributes

[0790] Optional. Used in both multimodal and voice-only modes. This is a collection of name/value pairs. Attributes is used to pass user defined information to the client-side semantic item and back to the server (they are kept synchronized). Attributes may only be set programmatically. For example:

• [0791] SemanticItem.Attributes [“myvarname”]=“myvarvalue”

[0792] Attributes are not cleared when the SemanticItem is reset by the system. If developers wish to reset the attributes, they must do so manually.

StyleReference

[0793] Optional. Used in both multimodal and voice-only modes. Specifies the name of a Style object. At render time, the QA SemanticItem object will search for the named Style control and will use any property values specified on the Style as default values for its own properties. Explicitly set property values by the SemanticItem object will override those set on the referenced Style.

14.2 SemanticItem Client-Side Object

[0794] //Notation doesn't imply programming language

[0795]

	class SemanticItem
	{
	SemanticItem (sco, id, targetElement, targetAttribute,
	bindOnChanged, bindAtServer, autoPostback,
	onClientChanged, onClientConfirmed,
	hiddenFieldID,
	value, state);
	SetText (string text, boolean isConfirmed);
	Confirm( );
	Clear( );
	Empty( );
	AddValidator (validator);
	IsEmpty( );
	NeedsConfirmation( );
	IsConfirmed( );
	Encode( );
	Object value; //Read only
	string state; //Read only
	object attributes;
	}

SetText (String Text, Boolean is Confirmed)

[0796] The SetText method of the client side semantic item object is used to alter the value property. The parameters are

• • [0797] string text the string which will become the value of the Semantic Item
  • [0798] Boolean is Confirmed determines whether the Semantic Item state property is “confirmed” (if true) or “needs confirmation” if false
    Confirm( )

[0799] This method sets the state property of the Semantic Item property to “confirmed.”

Clear( )

[0800] This method sets the value property of the Semantic Item to NULL and sets the state property to “empty.”

Empty( )

AddValidator (Validator)

IsEmpty( )

[0801] This method checks to see if the state property of the Semantic Item and returns true if it is “empty” and false if it is “needs confirmation” or “confirmed.”

NeedsConfirmation( )

[0802] This method checks to see if the state property of the Semantic Item and returns true if it is “needs confirmation” and false if it is “empty” or “confirmed.”

IsConfirmed( )

[0803] This method checks to see the state property of the Semantic Item and returns true if it is “confirmed” and false if it is “needs confirmation” or “empty.”

Encode( )

Object Value

[0804] ReadOnly.

String State

[0805] Read Only.

Object Attributes

14.3 Run-Time Behavior

[0806] As a general rule, the order of execution for every transition Empty->NeedsConfirmation or NeedsConfirmation->Confirmed:

• • [0807] Client-side binding (if needed)
  • [0808] Client-side event
  • [0809] If (Autopostback), trigger submit.

[0810] On the server, the order of execution is:

• • [0811] Server-side binding (if needed)
  • [0812] Server-side event.

[0813] If the semantic item is programmatically changed in the server, no events (server or client side) will be thrown. If (BindOnChanged=false) and (Autopostback=true) and we have both Changed and Confirmed handlers, both events will be triggered, in order.

[0814] Changed events will be thrown in the server (if needed and handlers are set) even if the server-side value is the same as the previous one (didn't change apparently).

[0815] If AutoPostBack is set to true, the controls will cause two postbacks, synchronized with on Changed, and on Confirmed.

[0816] The prompt object contains information on how to play prompts. All the properties defined are read/write properties.

[0817] Rendered for voice-only. Not rendered for multimodal.

How Prompt Object is Used

Voice-Only

[0818] The Prompt object is a required element of the QA control. RunSpeech uses the Prompt object to select the appropriate text for the prompt and then play the prompt on the client.

[0819] After RunSpeech determines which QA to activate it either increments or initializes the count attribute of the QA. The count attribute is incremented if the QA being activated was the same QA that was active during the last loop through RunSpeech. The count attribute is initialized to count=1 if this is the first time the QA has been activated. The QA's count attribute may be used by the script specified in the PromptSelectFunction attribute of the Prompt object.

[0820] RunSpeech then sets out to determine which text will be synthesized and played back to the user. The dialog author has the option of providing a script function for prompt text that is complex to build, or simply specifying the prompt text as content of the Prompt object. If RunSpeech detects the existence of an author-specified PromptSelectFunction, it passes the text returned from the PromptSelectFunction to the speech platform for synthesis and playback to the user. Otherwise RunSpeech will pass the text in the content of the Prompt object to the speech platform.

[0821] If a serious or fatal error occurs during the synthesis process, the speech platform will fire the on Error event. RunSpeech receives this event, sets lastCommandOrException to “PromptError” and calls the script function specified by the OnClientError attribute. The dialog author may then choose to take appropriate action based upon the type of error that occurred.

[0822] After the prompt playback has finished, the speech platform fires the oncomplete event which is caught by RunSpeech. RunSpeech then looks for the Reco object associated with the current QA. If a Reco object is found (i.e., the QA is not just a prompting mechanism), RunSpeech requests the speech platform to start the recognition process.

[0823] Finally, RunSpeech examines the value of the PlayOnce attribute of the QA containing the Prompt object. If PlayOnce is true, RunSpeech disables the Prompt object for subsequent activations of this same QA.

[0824] If speech is detected during the playing of the prompt, the playback of the prompt will be stopped automatically by the platform. RunSpeech catches the onbargein event and halts execution. Since a prompt.OnComplete event may not follow a bargein, RunSpeech resumes when a listen event is received.

[0825] If a bookmark is encountered, Runspeech activates the function specified by the OnClientBookmark property.

Multimodal.

[0826] The Prompt object is not used in multimodal scenarios.

PromptSelectFunction

[0827] The following three examples illustrate using the PromptSelectionFunction to select or modify prompt text using the parameters available to PromptSelectFunction.

[0828] The first example shows how to use the count parameter to select a prompt based upon the number of times the QA has been activated. The scenario is:

[0829] A user calls a menu based service, enters password. Server-side processing determines the user's first and last name and inserts the name information into hidden textboxes (txtFirstName.value and txtLastName.value) on the welcome page. The welcome page contains a QA which prompts the user to enter the desired service. The QA's Prompt object is built to handle 1) the prompt to play for a first time pass and 2) the prompt to play if the user fails to select a service at the first prompting (i.e., the same QA is activated after a timeout expires).

[0830]

	<speech:QA id=“welcomeQA” runat=“server”>
	<Prompt id=“welcomePrompt”
	PromptSelectFunction=“SelectWelcomePrompt”
	/>
	<Reco id=“welcomeReco” mode=“automatic”>
	<Grammars>
	<speech:grammar id=“welcomeGrammar”
	src=“http://mysite/services.xml”
	runat=“server” />
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“servicesAnswer” SemanticItem
	=	“siService” runat=“server” />
	</Answers>
	</speech:QA>
	<script>
	function SelectWelcomePrompt(lastCommandOrException, count,
	assocArray)
	{
	switch(count)
	{
	case 1: return “Welcome to Acme Services
	<SALT:value>txtFirstName.value</SALT:value>. Please select
	the Email, Calendar or Stock service.”;
	case 2: return “Welcome Please select the Email,
	Calendar or Stock service.”;
	case 3: return “Welcome Please select the Email,
	Calendar or Stock service.”;
	default: return “I'm sorry
	<SALT:value>txtFirstName.value</SALT:value>, we're having
	communication problems. Good Bye.”;
	}
	}
	</script>

[0831] The next example shows how to use the lastCommandOrException parameter to modify a prompt based upon a event previous event in the dialog. The scenario is:

[0832] A user is asked to provide the name of a departing airport. The QA contains a Prompt object that is built to handle the initial prompt, a prompt if the user asks for help, and a prompt if the user fails to respond (i.e. a timeout occurs).

[0833]

	<speech:qa id=“qa1” runat=“server”>
	<Prompt id=“prompt1”
	PromptSelectFunction=“SelectDepartingAirport” />
	<Reco id=“reco1” mode=“automatic”>
	<Grammars>
	<speech:grammar id=“gram1”
	src=“http://mysite/NYAirport.xml”
	runat=“server” />
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“ans1” SemanticItem=“siAns1”
	runat=“server” />
	</Answers>
	</speech:qa>
	<speech:command id=“command1” runat=“server”
	XpathTrigger=“/sml/help”	scope=“qa1” type=“HELP”>
	<Grammar src=“http://mysite/help.xml” runat=“server”
	/>
	</speech:command>
	<script>
	function SelectDepartingAirport (lastCommandOrException,
	count, assocArray)
	{
	if (count==1) return “From which airport would you
	like to depart?”;
	switch(lastCommandOrException)
	{
	case “SILENCE”: return “I'm sorry I didn't catch that.
	From which airport would you like to depart?”;
	case “HELP”:   return “You may choose from Kennedy,
	La Guardia, or that little airport on Long Island. From
	which airport would you like to depart?”;
	default return “What we have here is a failure to
	communicate. Good bye.”;
	}
	}
	</script>

[0834] The last example shows how to use the assocArray parameter to modify a prompt during a confirmation pass. The scenario is:

[0835] The user is asked to provide itinerary details: departing and arrival cities and travel date. The QA is constructed to implicitly confirm the departing and arrival city information and explicitly confirm the travel date. The Prompt object is built to provide appropriate prompting of items requiring confirmation.

[0836]

<speech:qa id=“qa1” runat=“server”>
	<Prompt id=“prompt1” InLinePrompt=“What is your
desired	itinerary?”></Prompt>
	<Reco id=“reco1” mode=“Automatic”>
	<Grammars>
	<speech:grammar id=“grm1”
	src=“http://mysite/city_date.xml”
	runat=“server” />
	</Grammars>
	</Reco>
	<Answers>
	<speech:answer id=“A1” XpathTrigger=“/sml/departCity”
	SemanticItem=“siTb1” ConfirmThreshold=“0.90”
runat=“server” />
	<speech:answer id=“A2” XpathTrigger=“/sml/arrivalCity”
	SemanticItem=“siTb2” ConfirmThreshold=“0.90”
runat=“server” />
	<speech:answer id=“A3” XpathTrigger=“/sml/departDate”
	SemanticItem=“siTb3” ConfirmThreshold=“1.00”
runat=“server” />
	</Answers>
</speech:qa>
<speech:qa id=“qa2” runat=“server”
XpathDenyConfirms=“/sml/deny”
XpathAcceptConfirms=“/sml/accept”>
	<Prompt id=“prompt2”
PromptSelectFunction=“myPromptFunction” />
	<Reco id=“reco2” mode=“automatic”>
	<Grammars>
	<speech:grammar id=“grm2”
	src=“http://mysite/cityANDdateANDyes_no.xml”
	runat=“server” />
	</Grammars>
	</Reco>
	<Confirms>
	<speech:answer id=“conf1”
XpathTrigger=“/sml/departCity”	SemanticItem=“siTb1”
ConfirmThreshold=“0.90” runat=“server” />
	<speech:answer id=“conf2”
XpathTrigger=“/sml/arrivalCity”	SemanticItem=“siTb2”
ConfirmThreshold=“0.90” runat=“server” />
	<speech:answer id=“conf3”
XpathTrigger=“/sml/departDate”
	SemanticItem=“siTb2” ConfirmThreshold=“1.00”
runat=“server” />
	</Confirms>
</speech:qa>
<script>
function myPromptFunction(lastCommandOrException, count,
assocArray)
{
var promptext = “Did you say ”;
	if (assocArray[“siTb1”] !=null && assocArray[“siTb1”]
!=“”)
	{
	promptText += “from” + assocArray[“siTb1”];
	return promptText;
	}
	if (assocArray[“siTb2”] !=null && assocArray[“siTb2”]
!=“”)
	{
	promptText += “to” + assocArray[“siTb2”];
	return promptText;
	}
	if (assocArray[“siTb1”] !=null && assocArray[“siTb3”]
!=“”)
	{
	promptText += “on” + assocArray[“siTb3”];
	return promptText;
	}
}
</script>
class Prompt : Control
{
	string	id{get; set;};
	string	type{get; set;};
	bool	prefetch{get; set};
	string	lang{get; set;};
	bool	bargein{get; set;};
	string	src{get; set;};
	string	PromptSelectFunction{get; set;};
	string	OnClientBookmark{get; set;};
	string	OnClientError{get; set;};
	string	InlinePrompt{get; set;};
	string	StyleReference{get; set;};
	ParamCollection	Params{get; set:};
}

15.1 Prompt Properties