Query reformulation, search performance, and term suggestion devices in question-answering tasks

  • Authors:
  • Ying-Hsang Liu;Nicholas J. Belkin

  • Affiliations:
  • Rutgers University, New Brunswick, NJ;Rutgers University, New Brunswick, NJ

  • Venue:
  • Proceedings of the second international symposium on Information interaction in context
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Capturing context within query in query reformulation tasks has been identified as a promising technique for supporting users who are engaged with interactive information retrieval systems. User queries represent the evolution of information problems. A deeper understanding of the structure and process of query reformulation, in particular, could provide further information for system adaptations. The present study characterizes the query reformulation process in two types of term suggestion devices, relevance feedback (RF) and Local Context Analysis (LCA), in simulated question-answering tasks using Rutgers' TREC-8 Interactive Track dataset. Four types of query reformulation were identified on the basis of semantic contents and relations, as well as sequences in users' modifications of queries. We found a significant relationship between the types of query reformulations and the use of term suggestion devices. But we did not find significant correlations between types of query reformulations and search performance. Some issues regarding systematic biases in query reformulations and capturing context within queries in interactive IR system are discussed.