AUTOMATIC QUERY EXPANSION BASED ON FUZZY THESAURUS FOR INFORMATION RETRIEVAL

摘要

Query expansion (QE) has been proved to be one of effective methods for improving the performance of the information retrieval (IR) system. Query expansion in the traditional keyword-based search methods is appending new associated words or terms to the separated words that are used to represent the user's query. However introducing such separated words into the search process would return a number of irrelevant documents as results to the user. This motivates us to combine the modifier with its corresponding headword. The combined term is then employed for the following search process. So expanded terms in our method are not attached to the separated words/terms but the combined terms. As a consequence, more relevant documents can be found out, and the precision and recall of the IR system will be increased by use of the proposed query expansion method. Quite often, the expansion process is conducted with the use of a thesaurus that contains a group of synonyms for each term. Considering that not all the synonyms are with the equal importance for a word/term, they are usually only related to each other to some degree. To appropriately specify the degree of relationships between two synonymous words/terms stored in the thesaurus, constructing a fuzzy thesaurus is a good choice we think, in which the grade of relevance for each pair of related words/terms can be mapped into the interval. By using the developed fuzzy synonym thesaurus, the combined term can be expanded. Through the proposed approaches, we can retrieve relevant documents in a relatively narrow search space and meanwhile widen the coverage of the original query.