語の出現頻度情報に基づく適合度順検索の問題点

  • 相良 佳弘
    慶應義塾大学大学院 文学研究科図書館情報学専攻

書誌事項

タイトル別名
  • Some problems on ranking retrieval systems based on term frequencies
  • ゴ ノ シュツゲン ヒンド ジョウホウ ニ モトヅク テキゴウドジュン ケンサク ノ モンダイテン

この論文をさがす

抄録

Ranking retrieval systems based on term frequencies present users a sequence of documents ranked in descending order of similarity between query and document. These systems come into use in online database retrieval or WWW search engines. Most preceding researches pointed out many advantages of ranking retrieval systems. Whether ranking by systems satisfy users or not, however, has not been examined. In this research, an experiment comparing the ranking by system with the ranking made by user on his relevance or utility judgement was carried out. From this result, existing ranking retrieval systems based on term frequencies have some problems in ranking. The conditions to be fulfilled in order that the system ranking should be similar to user ranking were identified. Under following four conditions, system ranking is different from user ranking.<br/>  1) When various fields are covered by a database<br/>  2) When record lengths in a database vary<br/>  3) When many topics are treated in one record<br/>  4) When vague query or keywords are used Ranking retrieval systems using only term frequencies are not enough to make ranking similar to ranking by user. These problems of the ranking retrieval systems may be attributed to the fact that the main part of the process is based on keywords that used at conventional Boolean retrieval. Ranking retrieval systems shall be improved by the use of some methods that can reflect user's information needs in addition to term frequencies.

収録刊行物

被引用文献 (1)*注記

もっと見る

参考文献 (5)*注記

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ