Exploiting user context and preferences for intelligent web search

Abstract

Seeking information relevant to a topic of interest has become a common task in our daily activities. However, searching the Web using current technologies still presents many limitations. One of the main limitations is that existing tools for searching the Web restrict user queries to a small number of terms. As a result, a single query may not reflect the user information needs at a sufficient level of detail. In addition, even if longer queries were allowed, the user may not find the right terms to supply appropriate queries, or may not be willing to put the effort required to explicitly describe his or her information needs. Another limitation of today’s search tools is that they are not capable of performing qualitative inference on the suggestions they offer. For certain domains, such as news or scientific articles, a good amount of structural information can be usefully exploited to extract meaningful content. This can help sort out the material returned by a search engine and to perform a qualitative analysis to warrant some of the search results. This paper shows how to enhance current search engines capabilities by (1) taking advantage of the user context, and (2) ranking search results based on preferential criteria provided by the user. We describe ongoing research on the use of context-specific terms to refine Web search and on the use of a defeasible argumentation framework to prioritize search results.