Using attribute inheritance to identify crawl paths

[0001] A web crawler is a software application that can crawl the both the Internet and enterprise intranets. A unit of crawling is called a cruise, and includes a branching tree of paths where each node is a Web page. For each cruise, a web crawler can take a number of inputs, including certain initial or seed URLs, a maximum depth of nodes to be crawled, and/or a set of one or more regular expressions to which the crawled URLs must adhere.

[0002] FIG. 1 shows an example input 10 for a web crawler. During a cruise, the web crawler visits each of the seed URLs, parses the HTML content obtained from each URL for links contained in each URL. The user can override the general parser to locate or ignore specific types of links, such as advertising links. Next, the web crawler visits the contained URLs, iterating the process to make sure that no link is visited twice. The input 10 causes the web crawler to crawl the specified seed websites, i.e. to a specified depth of 100, and for all HTML, PPT, DOC, and JPG files, but cuts short a cruise branch at the CNN link. FIG. 2 shows a result 12 of a cruise according to input 10.

[0003] The results of a cruise by the web crawler can be used for various applications or scenarios. A text mining system, for example as implemented in the Text Retrieval and Extraction (TREX) component of the SAP NetWeaver technology platform, can index the HTML contents obtained from the links visited by the web crawler to enable a full text search over these contents and/or documents. The text mining system can also add attributes to the indexed pages and documents to enable a search over these attributes. For example, the attributes may be metadata provided with a document, such as author, title, and so on.

[0004] For these and other applications of web crawler results, it may be useful to know which seed URL led to a given content or document. However, for many Web pages, it is difficult to know from the page's URL which seed URL led to the page. For example, a URL may contain an IP addresses that has no association with the seed URL, or contain no path information to show the path by which the URL was accessed. Further, the number of URLs and documents crawled during a cruise can number in the millions, making a quick determination of the seed URL very difficult.

[0005] This document discloses systems and methods for executing a web cruise operation and for identifying crawl paths of the web cruise. Each crawl path represents successive uniform resource locator (URL) nodes in a parent/child relationship. In one aspect, a method of identifying crawl paths of a web cruise operation includes the step of identifying one or more seed URLs for the web cruise operation, where each seed URL defines an origination of at least one crawl path.

[0006] The method further includes the steps of identifying a set of attributes of each parent URL in each crawl path to be inherited by one or more child URLs found in the web cruise operation, and associating each child URL with the set of attributes identified for all parent URLs in the crawl path. In another aspect, a method further includes the step of receiving cruise parameters for the web cruise operation, where the cruise parameters define uniform resource locator (URL) access actions to generate the one or more crawl paths.

[0007] In yet another aspect, a system is provided for executing a web cruise operation. The system includes a cruise service providing at least one web crawler tool, the cruise service being configured to receive cruise parameters defining uniform resource locator (URL) access actions to generate one or more crawl paths. Each crawl path represents successive URL nodes in a parent/child relationship. The web crawler is configured to identify a set of attributes of each parent URL in each crawl path to be inherited by one or more child URLs found in the web cruise operation.

[0008] The system further includes an index server having a text mining engine for generating a text-based index of text-based data objects associated with each child URL found in the web cruise operation, the index server also having an attribute engine for generating an attribute index of a set of attributes inherited by each child URL found in the web cruise operation.

[0009] The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features and advantages will be apparent from the description and drawings, and from the claims.

[0015] Like reference symbols in the various drawings indicate like elements.

[0016] When a web crawler user specifies one or more seed URLs for a cruise, as shown by means of an example in FIG. 1, the user can also specify a set of attributes with values that are to be inherited by each URL crawled from particular seed URLs. As an example, and with reference to FIG. 2, a user can specify that any URL on a path starting from S1 (www.sap.com) may have an attribute “server” with value “SAP,” and any URL on a path starting from S2 (www.ibm.com) may have attribute “server” with value “IBM.”

[0017] FIG. 3 shows a computer communication network 100 having a system 102 for executing a cruise of the internet 108 or other URL-based networks such as a corporate intranet, for an application 104 using attribute inheritance to identify crawl paths for the cruise. The system 102 includes a preprocessor 114 and a cruise service 118. The system 102 also includes an index server 120 having a search engine 122 configured to search web pages based on key terms, a text mining engine 124 that works with the search engine 122 to perform intelligent text-based searching and retrieval on text-based objects associated with web pages, and an attribute engine 126 configured to extract attributes from searched web pages. The exact configuration of these three engines is inessential to the invention and their functions may be combined into one engine in other embodiments.

[0018] The application 104 initiates a cruise by sending seed URLs to the cruise service 118, which instantiates a web cruise to execute a specified cruise. The web crawler sends the seed URLs to the preprocessor 114, which generates “HTTP Gets” for each seed URL to retrieve the contents from the URLs over the internet 108, where these contents include text documents in various formats and the document attributes.

[0019] Crawl paths for a cruise are identified as follows. The cruise is specified by a regular expression, an example of which is shown in FIG. 1. The first seed URLs are sent to the preprocessor 114, which gives a callback to the cruise service 118 upon initiation of the cruise. The cruise service then sends the regular expression specifying the cruise parameters for that particular cruise to the preprocessor. The preprocessor 114 compiles the regular expression and uses it to steer the cruise along the desired crawl paths. The preprocessor 114 uses the regular expression to analyze the retrieved web pages and to extract any URLs that match the cruise parameters, and retrieves the web pages corresponding to those further URLs.

[0020] The web crawler in the cruise service 118 stores both the list of URLs that define the crawl path and the attributes and their values for each cruised page. As each node is cruised, the preprocessor 114 sends the retrieved text and attributes to the web crawler. The web crawler sends the text and attributes to the index server 120. In the index server 120, the text mining engine 124 indexes the text and the attribute engine 126 indexes the attributes. The indexes created are stored in the database 110.

[0021] During a cruise, the web crawler in the cruise service 118 remembers these attributes and their values for each and every URL crawled during the cruise, and attaches the remembered attribute values to the page or document obtained from each URL. The attributes and their values are inherited by every child node in each path from the parent nodes. In case a page is reached from two seed URLs, such that a crawled node inherits multi-valued attributes, the attributes from the seed URL with the shorter path, i.e. the lesser depth, are taken for the reached URL.

[0022] When the text mining engine 124 indexes the crawled pages, it also indexes the attribute values. The system 102 may also receive from a user the attributes to be inherited during runtime, after creation of the cruise. These new attributes are assigned to all subsequently crawled pages. For any attribute specified anew during runtime, the new specification supersedes any existing specification. Accordingly, for any page crawled during a cruise, the system 102 can discover which seed URL led to it by performing an attribute search.

[0023] FIG. 4 illustrates a method 200 for executing a cruise using attribute inheritance to identify crawl paths in accordance with an exemplary embodiment. At 202, cruise parameters are specified. The cruise parameters include seed URLs, cruise depth, file types to be cruised, as well as exclusions to be omitted from the cruise. The cruise parameters are used by a web crawler to execute the specified cruise. At 204, the attributes and values to be inherited by all child nodes, or URLs, of parent nodes are specified and received.

[0024] At 206, the web crawler crawls the internet or other URL-based network to build up a cruise tree. The cruise tree includes all links originating from a seed URL and visited as specified by the cruise parameters. At 208, each node in the cruise tree inherits attributes and values from its seed, or in other words the URL from which it was found.

[0025] At 210, an analysis is made whether the seed URL for a given found URL is unique. If not, at 212 the found URL inherits attributes and associated values from the seed with the shortest path, or with the lowest number of nodes along a branch of the cruise tree. If the seed is unique, the full text of each crawled page and documents therein is indexed by the text mining engine of the index server, while the attributes and associated values of each crawled page are indexed by the attribute engine of the index server.

[0026] At 216, the web crawler receives instructions to discover a seed URL for a given page. The web crawler uses the indexed attributes and text to traverse the cruise tree to quickly return the seed with the same attribute values as the given page. Thus, for any page in the cruise tree generated by a cruise, a user perform an attribute search to discover which seed URL led to the page.

[0027] Although a few embodiments have been described in detail above, other modifications are possible. The logic flow depicted in FIG. 4 may be executed in an order other than that shown and still be within the scope of the embodiments described herein. Other embodiments may be within the scope of the following claims.