Parallel computing for term selection in routing/filtering

  • Authors:

  • Andy MacFarlane;Stephen E. Robertson;Julie A. McCann

  • Affiliations:

  • Centre for Interactive Systems Research, City University, London;Centre for Interactive Systems Research, City University, London and Microsoft Research Ltd, Cambridge;Department of Computing, Imperial College, London

  • Venue:
  • ECIR'03 Proceedings of the 25th European conference on IR research
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

It has been postulated that a method of selecting terms in either routing or filtering using relevance feedback would be to evaluate every possible combination of terms in a training set and determine which combination yields the best retrieval results. Whilst this is not a realistic proposition because of the enormous size of the search space, some heuristics have been developed on the Okapi system to tackle the problem which are computationally intensive. This paper describes parallel computing techniques that have been applied to these heuristics to reduce the time it takes to select to select terms.