Automatic multidocument summarization of research abstracts: Design and user evaluation

  • Authors:

  • Shiyan Ou;Christopher S. G. Khoo;Dion H. Goh

  • Affiliations:

  • Division of Information Studies, School of Communication & Information, Nanyang Technological University, Singapore, 637718;Division of Information Studies, School of Communication & Information, Nanyang Technological University, Singapore, 637718;Division of Information Studies, School of Communication & Information, Nanyang Technological University, Singapore, 637718

  • Venue:
  • Journal of the American Society for Information Science and Technology
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

The purpose of this study was to develop a method for automaticconstruction of multidocument summaries of sets of researchabstracts that may be retrieved by a digital library or searchengine in response to a user query. Sociology dissertationabstracts were selected as the sample domain in this study. Avariable-based framework was proposed for integrating andorganizing research concepts andrelationships as well as researchmethods and contextual relations extractedfrom different dissertation abstracts. Based on the framework, anew summarization method was developed, which parses the discoursestructure of abstracts, extracts research concepts andrelationships, integrates the information across differentabstracts, and organizes and presents them in a Web-basedinterface. The focus of this article is on the user evaluation thatwas performed to assess the overall quality and usefulness of thesummaries. Two types of variable-based summaries generated usingthe summarization methodwith or without the use of a taxonomywerecompared against a sentence-based summary that lists only theresearch-objective sentences extracted from each abstract andanother sentence-based summary generated using the MEAD system thatextracts important sentences. The evaluation results indicate thatthe majority of sociological researchers (70%) and general users(64%) preferred the variable-based summaries generated with the useof the taxonomy. © 2007 Wiley Periodicals, Inc.