A Task-Oriented Non-Interactive Evaluation Methodologyfor Information Retrieval Systems

  • Authors:
  • Jane Reid

  • Affiliations:
  • Department of Computer Science, Queen Mary and Westfield College, University of London, E1 4NS, England. jane@dcs.gmw.ac.uk

  • Venue:
  • Information Retrieval
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

Past research has identified many different types ofrelevance in information retrieval (IR). So far, however, mostevaluation of IR systems has been through batch experiments conductedwith test collections containing only expert, topical relevancejudgements. Recently, there has been some movement away from thistraditional approach towards interactive, more user-centred methodsof evaluation. However, these are expensive for evaluators in termsboth of time and of resources. This paper describes a new evaluationmethodology, using a task-oriented test collection, which combinesthe advantages of traditional non-interactive testing with a moreuser-centred emphasis. The main features of a task-oriented testcollection are the adoption of the task, rather than the query, asthe primary unit of evaluation and the naturalistic character of therelevance judgements.