検索キーワード間の修飾?被修飾関係の詳細な分析に基づくWWW検索性能の向上  [in Japanese] Improvement in Performance of WWW Search Engines Based on Detailed Analysis of Dependency Relation between Input Keywords  [in Japanese]

Access this Article

Search this Article

Author(s)

Abstract

ウェブ検索エンジンに,ユーザが検索キーワードとして2 つの語を入力した場合に,その2 語の修飾?被修飾関係を意味と文法カテゴリの両面から詳細に分析し,特定の関係が文書中に出現しているか否かを判定することにより,ウェブ検索エンジンの性能を向上させる手法を提案する.どのような関係を使うことが有効になるかを判定する基礎として,どのようなキーワードが実際に用いられるのかの検索の実態の調査を行い,その結果を用いて有効な関係を選ぶなどを手法の実現に反映させた.提案手法をフィルタリングツールとして構築し,評価実験を行った結果,単なる修飾?被修飾関係を用いる検索手法に比べ,精度,再現率ともに向上した.また,広く使われている検索エンジンを使った上位20 位における適合ページ数の実験でも,適合ページ数が平均1.5~2 ページほど増えることが示され,この手法の有効性が確かめられた.In this paper, we propose a method to improve performance of WWW search engines. We focus on the case that two keywords are input to web search engine. We think that the keywords input by users have a semantic relation. We try to improve retrieval accuracy by checking whether two keywords have a dependency relation in the candidate pages or not. We construct a filtering tool which accept output of an ordinary search engine and select plausible ones by checking dependency relation. We show an experimental evaluation of our method. As a result, we compared our method with ones using just direct dependency relation and showed to raise the precision and recall. Furthermore, the best 20 pages by our system contained about 1.5窶骭2 more relevant pages than general search engines. Therefore, we could confirm the validity of our method.

In this paper, we propose a method to improve performance of WWW search engines. We focus on the case that two keywords are input to web search engine. We think that the keywords input by users have a semantic relation. We try to improve retrieval accuracy by checking whether two keywords have a dependency relation in the candidate pages or not. We construct a filtering tool which accept output of an ordinary search engine and select plausible ones by checking dependency relation. We show an experimental evaluation of our method. As a result, we compared our method with ones using just direct dependency relation and showed to raise the precision and recall. Furthermore, the best 20 pages by our system contained about 1.5-2 more relevant pages than general search engines. Therefore, we could confirm the validity of our method.

Journal

  • IPSJ journal

    IPSJ journal 48(10), 3386-3404, 2007-10-15

    Information Processing Society of Japan (IPSJ)

References:  19

Cited by:  1

Codes

  • NII Article ID (NAID)
    110006402787
  • NII NACSIS-CAT ID (NCID)
    AN00116647
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • ISSN
    1882-7764
  • NDL Article ID
    8973076
  • NDL Source Classification
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL Call No.
    Z14-741
  • Data Source
    CJP  CJPref  NDL  NII-ELS  IR  IPSJ 
Page Top