Automatic Web Searching and Categorizing Using Query Expansion and Focusing

Abstract

Abstract : We are in the process of build a prototype system that improves precision and recall rates for web search using query expansion and focusing techniques. We use linguistic analysis and co-occurrence information to analyze syntactic structures of the users' queries to improve search results. One standard method of improving internet search is through query expansion. The major query expansion techniques add terms using (i) lexical semantic relations and (ii) relevance feed back. The lexical semantic relations in WordNet have been used widely as a main lexical resource for approach (i). Past research results indicate that using WordNet did not significantly improve information retrieval effectiveness. Our query expansion system also uses WordNet in a query expansion stage. However, instead of just adding all related terms from WordNet (synonyms, hypernyms, hyponyms, etc.) directly into user's queries, our system selects only useful additional terms. This selection process uses syntactic analysis combined with collocation and co-occurrence information from a large corpus collected from our domain of interest (IT).