Evaluation of an automatic weighting in a boolean retrieval system with thesaurus : I. Weighting of index terms

DOI

Bibliographic Information

Other Title
  • シソーラス型文献検索システムにおける自動的重みづけ法の評価 : I.索引語への重みづけ
  • シソーラス型文献検索システムにおける自動的重みづけ法の評価-1-索引語への重みづけ
  • シソーラスガタ ブンケン ケンサク システム ニ オケル ジドウテキ オモミ ヅケ ホウ ノ ヒョウカ 1 サクインゴ エ ノ オモミ ヅケ

Search this article

Abstract

Retrieval performance by the weighting scheme developped is demonstrated with a set of 10 search topics and 33,614 documents for half a year in the INIS data base. Conventional man prepared medium or tighter queries are clearly superior to weighted loosest queries but slightly inferior to weighted looser ones in average performance. The weight generation sample extracted within the limits of 100 documents from a data file for cver two months can be roughly applied to weighting in another residual file and a more realistic data file of 241,708 documents for three and a half years. It is tried to estimate approximately the degrees of improvement in precision and loss in recall at each ranked score position in other weight applied outputs, from rank-precision curves for some weight generation samples. This relevance ranking scheme leads to considerable benefits in practical query formulation. Compared with the present weighting of both manually indexed and machine up-posted terms, weighting of manually indexed terms only may possibly bring about a much more promising application of the scheme.

Journal

  • Dokumenteshon kenkyu

    Dokumenteshon kenkyu 30 (6), 247-253, 1980

    Information Science and Technology Association, Japan

Details

  • CRID
    1390282679152571264
  • NII Article ID
    110002729560
  • NII Book ID
    AN00172819
  • DOI
    10.18919/dokumen.30.6_247
  • ISSN
    2189826X
    00125180
  • NDL BIB ID
    2183932
  • Text Lang
    ja
  • Data Source
    • JaLC
    • NDL
    • CiNii Articles
  • Abstract License Flag
    Disallowed

Report a problem

Back to top