Best-match retrieval of sequential information

Abstract

This dissertation reports an investigation into a number of search and retrieval techniques for new generation of information retrieval systems. It discusses the use of automatic indexing techniques to obtain document and query representatives, this includes methods for the selection of terms from document texts and the use of conflation techniques to deal with variations in word forms. It also discusses the use of best match retrieval techniques as an alternative to Boolean retrieval mechanisms, which characterize current free-text retrieval (FTX) systems. This is foliowed by a description of term weighting techniques, which can be used to reflect the ability of search tenns to discriminate between relevant and non-relevant documents. Other techniques, which may, with further development, figure in future retrieval systems, are also briefly described.