Simple word matching between the user query and document is common, as are mis-matches of meaning that occur as a consequence, and errors in recall. These defects in the "bag of words" model are well known, and raising the semantic level of representation will improve retrieval. This can be done by expanding words and user queries using traditional reference sources such as gazetteers and synonym lists or ontologies. Categories and Subject Descriptors H.3.1 [Information Retrieval]: Content analysis and indexing
Judith Gelernter, Michael E. Lesk