Exploiting the Similarity of Non-Matching Terms at RetrievalTime

  • Authors:
  • Fabio Crestani

  • Affiliations:
  • Department of Computing Science, University of Glasgow, Scotland, UK. fabio@dcs.gla.ac.uk

  • Venue:
  • Information Retrieval
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

In classic Information Retrieval systems a relevantdocument will not be retrieved in response to a query if the documentand query representations do not share at least one term. Thisproblem, known as “term mismatch”, has been recognised for a longtime by the Information Retrieval community and a number of possiblesolutions have been proposed. Here I present a preliminaryinvestigation into a new class of retrieval models that attempt tosolve the term mismatch problem by exploiting complete or partialknowledge of term similarity in the term space. The use of termsimilarity enables to enhance classic retrieval models by taking intoaccount non-matching terms. The theoretical advantages and drawbacksof these models are presented and compared with other models tacklingthe same problem. A preliminary experimental investigation into theperformance gain achieved by exploiting term similarity with theproposed models is presented and discussed.