An evaluation of query expansion by the addition of clustered terms for a document retrieval system

Although seemingly meaningful clusters can be obtained, the results indicate that the effort involved in finding clusters and adding the clustered terms to queries is far too great to warrant their use in an operational system.

Abstract

An evaluation of graph theoretical clusters of index terms which can be extracted from an automatically indexed document collection, and the effects of employing such clusters in automatic document retrieval is described. The graph theoretical clusters which were developed from six data bases under two different cluster definitions were analyzed for average size and related data. The clusters were also used to expand the queries in each of six data bases to determine the effect of the expansions on the document retrieval results. Although a large variety of clusters and associated query expansions were obtained, no significant improvements in the document retrieval performance were achieved. In some cases, however, significant degradations in the retrieval performance occurred. Although seemingly meaningful clusters can be obtained, the results indicate that the effort involved in finding clusters and adding the clustered terms to queries is far too great to warrant their use in an operational system. The data bases employed were relatively small, and the authors caution against generalizing these results to larger data bases or other situations.