Query Expansion with Wikipedia

To address the task of automated query expansion with compound queries we make use of the highly structured Wikipedia corpus. We consider every article linked to by a specific disambiguation page as a unique sense. Using language modeling analysis over the articles referred to in order to determine which sense is most likely intended by considering the co-occurence of the different query terms in articles associated with candidate senses. Expansion terms are selected from the Wikipedia disambiguation sentence and added to the user defined query. An evaluation of this method is performed with the CLEF english collection and the TREC defined relevance assessments metrics. We observe a notable improvement in mean average precision in both of our experimental conditions, which differ only in the weighting of original query terms and expansion terms.