Search system

[0001] The present invention relates to a system for searching information stored electronically. In particular, though not exclusively, the present invention relates to a system for searching information found on the world wide web.

[0002] A number of Internet search engines are presently available which allow a user to retrieve information, such as web pages, images, news group postings and the like, based upon one or more user specified keywords. Perhaps the most widely known of these search engines is the Google search engine. The Google search engine includes a fixed designation interface wherein an inputted search term results in the production of a list of documents or “hits” which contain the search term. The Google ranking system typically orders the hits depending upon the number of times the search term occurs in a given document. It will be appreciated however that such a system leads to documents in different categories being listed next to one another. The Google search engine does permit a degree of tailoring of the search hits by using predefined fixed search criteria. These criteria can be selected either before or after a search term has been inputted. The predefined search criteria provided by the Google search engine include “Web”, “Images”, “Groups” and “Sews”, however these fixed criteria are often not sufficient to enable a user to tailor a search in a meaningful manner.

[0003] According to the present invention there is provided a method of searching a plurality of semi-structured or unstructured documents comprising the steps of:

• [0004] providing a search interface having a search term input window, a search initiation button and a plurality search result display window, the search interface further including one or more search term clarification buttons;
• [0005] providing a plurality of documents having one or more search identifier terms associated therewith inputting a search term into the search term input window and initiating a search;
• [0006] searching for documents related to the search term, and displaying documents identified by the search in the search result display windows, wherein each search result display window displays search results related to a particular identifier term, and wherein further the search identifier term assigned to each search result display window is dependent upon the results of prior searches.

[0007] It will be understood that the terms “semi-structured” and “unstructured” relate to documents such as web pages and news articles which are not comprised of data only from known data types that are specified in a plurality of predefined fields (such as those data from relational databases).

[0008] The search term clarification buttons may be utilised to assign specific additional search terms to a search term inputted by a user into the search window. The additional search terms assigned to the search clarification buttons may correspond to search identifier terms associated with documents which may be searched. Preferably the search terms assigned to the search clarification buttons correspond to the most popular search terms utilised in prior searches. A user may utilise one or more of the search term clarification buttons either before or after inputting a search term into the search window.

[0009] The identifier terms assigned to a searchable document may comprise words or phrases found in the body of the document. Alternatively the identifier terms may comprise words or phrases attached to but not visible in the document.

[0010] According to a further aspect of the present invention there is provided a search interface having a search term input window, a search initiation button and a search result display window, the search interface further including one or more search term clarification buttons, wherein the operation of the or each search term clarification button is dependent upon prior searches undertaken by the search interface. The operation of the or each search term clarification button may be dependent upon the occurrence of search terms used in prior searches. Alternatively the operation of the or each search term clarification button may be dependent upon the occurrence of terms appearing in the results of prior searches. Alternatively the operation of the or each search term clarification button may be dependent upon the occurrence of identifier terms assigned to the results of prior searches.

[0011] The search interface may be provided with a plurality of search result display windows which, in use, display the results of a search.

[0017] Referring firstly to FIG. 1 there is shown an interface, generally designated 10, for an Internet based search system. The interface 10 is presented to a user on a display means such as a screen, and the user may interact with the interface 10 by known interface means such as a keyboard and mouse. The interface 10 includes a search window 12 into which a user may type one or more search terms and a search button 14 which a user may “press” to initiate a search based on one or more terms typed into the window 12. The interface 10 further includes four topic buttons 16, 18, 20, 22 which are marked “Doc Type”, “Topic”, “Products” and “Market” respectively. The topic buttons 16, 18, 20, 22 may be employed to refine the results of a search and will be described in greater detail below.

[0018] The interface also includes a number of result display windows 24 of which only the uppermost is shown in FIG. 1. Each window 24 is provided with a tab 26 bearing the name of a class which relates to the results displayed in that particular window. In the embodiment shown there are provided six windows 24 having the class names “General”, “Report”, “Product”, “News”, “Slides” and “Spec”. The windows 24 are configured such that when a particular tab 24 is selected, for example by clicking on the tab 26 with a cursor, the window 24 is presented uppermost.

[0019] FIG. 1 shows the situation where a user inputs a single search term into the search window 12, for example the term “Accord” 28 and presses the search button 14. The search has identified a number of hits which are perceived to be relevant and are presented in lists on the result display windows 24. The results presented in each display window 24 are related to the class described on the tab 26 of the respective window 24. The various classes, and hence the order in which the search result display windows 24 are presented, are ranked based on an aggregated similarity score. In the embodiment shown documents falling to “General” class are perceived to be of more relevance to the user than documents falling into the “Product” class.

[0020] The order in which the classes, and hence the display windows 24, is presented is changeable dependent upon the prior actions of users of the search system. For example, should users searching the term “Accord” exhibit a tendency to immediately go to a display window 24 having a particular class heading, for example “Product” then order of the classes can be changed such that subsequent users are presented with the “Product” class display window 24 uppermost. Additionally where users subsequently choose to view “Accord” documents associated with a class which is not: presently included in the class headings, such as “Vehicle”, then this class may be included in subsequent searches if it is deemed more popular than an existing class heading.

[0021] As described above, the interface 10 is provided with four topic buttons 16, 18, 20, 22 which are marked “Doc Type”, “Topic”, “Products” and “Market” respectively. These topic buttons may be employed either before or after a search term has been inputted into the search window 12 to narrow the results of the search. FIG. 2 shows the result where a search term “Accord” has been inputted as described with reference to FIG. 1 and thereafter the topic button marked “Topic” has been pressed. As before, the user is provided with a plurality of search result display windows 24 however the classes displayed on the tabs 26 are different. In the embodiment shown in FIG. 2 the class headings are, in order of perceived relevance to the user, “Organization”, “Entertainment”, “Gov”, “Edu”, “Slides” and “Spec”. As before, the order in which the class headings are presented may be changed depending upon the documents which are subsequently viewed by users of the system.

[0022] In a similar manner to the class headings the nature and order of the topic buttons may also be varied as a result of the habits of users of the system. For example, should there be a flurry of travel related searches where the term “Travel” is utilised as a search term, then a topic button marked “Travel” may be provided for subsequent searchers to utilise. In the embodiment shown the interface 10 is provided with four topic buttons 16,18,20,22, however it will be appreciated that additional topic buttons may be provided.

[0023] FIG. 3 illustrates the situation where a search term is used which is considered to be so vague or broad in subject matter that the results generated thereby are deemed to fall into a multitude of classes. In such an instance the results are divided into a number of super classes 30 which in turn have a plurality of sub classes. Each super class and its associated subclasses are arranged in a result display window 24 as described above. The super class and subclasses are presented in the form of a two dimensional array. In the embodiment shown in FIG. 3 the use of a broad search term has generated four super classes 30 headed “Language”, “Topics”, “Media” and “Doc Type” with each having a range of related subclasses. For example the “Language” super class has subclasses headed “Japanese”, “Chinese”, “French”, “Indian” and “German”, while the “Media” super class has subclasses headed “Image”, “Audio”, “Video”, “Animation” and “Drawing”.

[0024] The super classes are presented to the user in order of perceived importance as estimated by the system based on the aggregate document similarity scores within the super or sub classes, as well as combining these scores with scores based on the actions and preferences of previous users of the system. As before the subclasses are ranked in order of this perceived importance estimate and the order of the super classes and associated subclasses may be changed based upon previous user actions.

[0025] FIGS. 4a to 4b show a simplified example of how the system of the present invention analyses and subsequently presents documents identified by a particular search term. Looking firstly at FIG. 4a a search for the term “Honda” may result in the identification of five documents identified as d1 to d5. Each document contains a number of identifier terms or labels 32 which may either be contained within the text of the document or attached thereto For the sake of simplicity each document is shown to have three identifier terms 32. For example, document d1 has the identifier terms “car”, “product” and “sales”, while document d4 has the identifier terms “car”, “magazine” and “review”. The identifier terms 32 may be either contained in the text of the document or attached to the document, for example in the form of a metatag, in a manner which is not visible to a reader of the document. It will be appreciated that a document may include a combination of each type of term 32. In the instance of a non-visible term, this may be assigned to a particular document by a reviewer of the document or assigned automatically by an automated means such as, for example, a text classifier.

[0026] Having identified documents d1 to d5 as being particularly relevant to the requested search, the system then orders them for presentation to the user. The ordering of the results is based on an analysis of the identifier terms attached to or contained within the documents. In the present example the term “car” is present in all five documents, the term “magazine” in three documents (d3, d4 and d5), the term “product” in two documents (d1, d2) and the term “review” in two documents (d2, d4) and the terms “sales”, “special issue” and “comparison” in one document each. As the term “car” appears most often then documents having this term are perceived to be of the most relevance. These documents are thus presented in a result display window 24 with the term “car” appearing on a tab 26 to denote the class of the documents displayed. Documents having the term “product” are perceived to be the second most relevant class of documents and documents having the term “magazine” the third.

[0027] The relevance of the various terms and the documents associated therewith may be judged on a combination of the number of times a given term occurs and the ranking of the term within a given document. In the present example it will be appreciated that the term “magazine” occurs three times and the term “product” two times, however the relevance of the “product” documents is perceived to be greater than that of the “magazine” documents. It will be noted that “product” is the top term in one document and the second term in the other document, while the term “magazine” is consistently the second term.

[0028] As has been described above, a user may refine their search, either before or after inputting a keyword, by using the topic buttons 16,18,20,22 on the interface 10. A user may also input a topic or class term into the search window 12 along with a search term. For example, a user wishing to identify documents relating to travel and Hawaii may input “\travel Hawaii” into the search window before pressing the search button. This would instruct the system to identify all documents having the identifier label “travel” associated therewith before then searching these documents for references to Hawaii. Additionally the system may be configured such that inputting “\travel” instructs the system to identify documents having identifier labels which are deemed to be related to travel such as, for example, “hotel”, “flights” and “car hire”