Query expansion and noise treatment for information retrieval

Abstract

Most of the search engines available over the Web are based on mathematical approaches | classical techniques in the Information Retrieval area. Thereby, they are suitable for the retrieval of documents containing some or all the terms of a query, though not to retrieve the documents containing the meaning those terms were intended to express. This paper presents some advantages obtained from query expansion with WordNet and noise treatment with knowledge on top of Paraconsistent Logic. Both methods are semantically driven, allowing the retrieval of documents which do not contain any term of the original query. Noise treatment results from the combination of a smooth term comparison with knowledge about term authentication based on behaviors of features in the collection. Although query expansion recurs for every query, noise treatment is part of the indexing mechanism, causing no overhead in queries. The domain is retrieval of ontologies represented in Resource Description Framework.