Data processing: database and file management or data structures – Database design – Data structure types
Reexamination Certificate
1999-05-05
2001-12-04
Breene, John (Department: 2177)
Data processing: database and file management or data structures
Database design
Data structure types
Reexamination Certificate
active
06327590
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to information retrieval systems and more particularly, to a system and method of collaboratively ranking results returned from a search engine using user and group profiles.
BACKGROUND OF THE INVENTION
The World Wide Web (the “web” or “WWW”) is an architectural framework for accessing documents (or web pages) stored on a worldwide network of distributed servers called the Internet. Documents stored on the Internet are defined as web pages. The architectural framework of the web integrates web pages stored on the Internet using links. Web pages consist of elements that may include text, graphics, images, video and audio. A web page, which points to the location of another web page, is said to be linked to that other web page. Links that are set forth in a web page usually take the form of a text fragment or an image. A user follows a link by selecting it.
With the advent of networking technology and the World Wide Web, the ability to access information from external sources has greatly increased. Various search engines enable a user to submit a query, which returns a collection of items or documents. A well-crafted query may return a manageable set of documents, typically from 30 to 50 documents. A less narrow query may return over 1000 documents. An overly narrow query may return no documents (in which case no ranking is required). Various techniques are available for assisting the user in refining or narrowing his/her search query. However, once the search result has been properly narrowed significant problem in information retrieval is how to rank the results returned by the search engine or the combination of search engines.
For individual search engines, there are many different techniques for ranking results, ranging from counting the frequency of the appearance of the various search terms in the search query to calculating vector similarities between a search term vector and each returned document vector. In a networked environment such as the World Wide Web, meta-searchers access different and often heterogeneous search engines and face the additional difficulty of combining the ranking information returned by the individual engines. Meta-searcher is a Web information retrieval system aimed at searching answers to a user's query in the heterogeneous information providers distribute over the Web. When a meta-searcher receives responses (usually in the form of HTML files) from the information providers, a special component of a meta-searcher called a wrapper, process the responses to answer the original query. Since many search engines, including meta-searchers, hide the mechanism used for document ranking, the problem of merging search results is compounded. A problem common to both individual search engines and meta-search engines is that these approaches ignore, or knowing nothing about, the user conducting the search, or the user's context for conducting the search.
Relevance feedback is one approach that elicits information about the user and his/her search context. Relevance feedback techniques re-rank the search results by using user feedback to recalculate the relative importance of key words in the query. While powerful from a technical point of view, relevance feedback approaches suffer from user interface issues. The relevance information required is often difficult to elicit successfully from users during the search process. U.S. Pat. No. 4,996,642 to Hey, System and Method for Recommending Items, describes a system for providing recommendations to users based on others items previously sampled by the user and the availability of the items.
Knowledge Pump, a Xerox system, provides community-based recommendations by initially allowing users to identify their interests and “experts” in the areas of those interests. Knowledge Pump is then able to “push” relevant information to the users based on those preferences. This is accomplished by monitoring network traffic to create profiles of the users, including their interests and “communities of practice,” thus refining the community specifications. However, monitored or automatically created profiles for establishing context may not accurately reflect the user's context at all times.
There is a need for a system and method of ranking search results which does not require user solicited relevance information. There is also a need for a system of ranking search results which takes into account a predetermined user context profile. There is also a need for a system and method of ranking search results which ranks results based on a user selected context. There is also a need for a system and method of ranking search results which takes into account a group or community to which the user belongs. There is a further need for a system and method of creating a user and community profile for ranking search results.
SUMMARY OF THE INVENTION
A system for ranking search results obtained from an information retrieval system, according to the invention, includes a search pre-processor, a search engine and a search post-processor. The search preprocessor, responsive to a search query, determines the context of the search query by comparing the terms in the search query with a predetermined user context profile. The user's context profile may include, for example, the user's identity, the community or set of communities applicable to the search, and the point of view the user wishes to adopt (e.g., that of a domain expert) for the search. Preferably, the context profile is a user profile or a community profile, which includes a hierarchical set of terms that have been rated by the user or community. Also, a recommender system may be used to generate the user or community context profile.
The search engine, responsive to the search query, generates a search result comprising at least one item obtained from the information retrieval system. (If no items are returned, such as when the search is overly narrow, no ranking is required.) Generally, a great number of items will be generated, which the search engine will provide in its own predetermined form of hierarchical valuation. The search post-processor, responsive to a non-empty search result, ranks each item returned in the search result in accordance with the context of the search query. The ranked results may then be provided or displayed in any normal fashion, such as on a computer display or printed out. If more than one search engine is used, each search engine returns its own list of search results. The post-search processor then ranks all items returned, regardless of search engine, in accordance with the context of the search query.
A method of ranking search results obtained from an information retrieval system, according to the invention, includes providing a predetermined user context profile, generating a search query, and applying the context profile to the search query to generate a context of the search query. A search is then performed based on the search query, which includes at least one item obtained from the information retrieval system. The search results are then ranked in accordance with the context of the search query.
The system and method according to the invention couples a predetermined user context profile (e.g., user profiling, community profiling or recommender profiling) with the search process. By coupling context profiling with the search process, search results are no longer an isolated event, but are ranked within the context of a particular user or community or recommender system point of view. Depending on the user's context for the search, a different predetermined context profile may be selected, thus customizing the ranking of each particular search.
The user and community profiles are built by analyzing document collections put together by the users and the communities to which the users belong. If any of the retrieved search results are considered relevant to the user or the community, they can be used to tune or modify the particular user or community
Chidlovskii Boris
Glance Natalie S.
Grasso Antonietta
Breene John
Rayyan Susan S.
Walder Jeannette M.
Xerox Corporation
LandOfFree
System and method for collaborative ranking of search... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for collaborative ranking of search..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for collaborative ranking of search... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2584246