対話文生成のための Web を用いた話題語の抽出  [in Japanese] Extracting Topic Words from the Web for Dialogue Sentence Generation  [in Japanese]

Access this Article

Search this Article

Author(s)

    • Rafal Rzepka RZEPKA Rafal
    • 北海道大学大学院情報科学研究科 Graduate School of Information Science and Technology Hokkaido University
    • 荒木 健治 ARAKI Kenji
    • 北海道大学大学院情報科学研究科 Graduate School of Information Science and Technology Hokkaido University

Abstract

本研究では,Internet Relay Chat の対話文の 1 文中で話題となっている話題語の抽出を行う.インターネット上の対話文は,blog や Web ページに比べ,口語に近い表現が多数使用される.このことから,Web 上の対話においてその入力文が意味するところを把握するには,ユーザの意図にそった入力文の解釈が必要となる.しかし,名詞だけを対象にユーザの入力文から,その文中で話題になっていることを把握ことは困難である.そこで本稿では,形容詞,動詞も考慮した Web からの話題語の抽出手法を提案する.実験の結果,名詞のみを対象にした場合よりも形容詞を考慮した場合の方が多様な表現の話題語を抽出できることが明らかとなった.In this paper we extract topic words from Internet Relay Chat utterances. In such dialogues there are many more spoken language expressions than in blogs or usual Web pages and we presume that the always changing topic is difficult to determine only by nouns which are usually used for topic recognition. In this paper we propose a method for determining a conversation topic considering also association adjectives and verbs retrieved from the Web. Our first experiments show that extracting association words using nouns and adjectives leads to determining topic labels of higher diversity.

In this paper we extract topic words from Internet Relay Chat utterances. In such dialogues there are many more spoken language expressions than in blogs or usual Web pages and we presume that the always changing topic is difficult to determine only by nouns which are usually used for topic recognition. In this paper we propose a method for determining a conversation topic considering also association adjectives and verbs retrieved from the Web. Our first experiments show that extracting association words using nouns and adjectives leads to determining topic labels of higher diversity.

Journal

  • IPSJ SIG Notes

    IPSJ SIG Notes 2009(2(2009-FI-93)), 121-126, 2009-01-15

    Information Processing Society of Japan (IPSJ)

References:  9

Codes

  • NII Article ID (NAID)
    110007123982
  • NII NACSIS-CAT ID (NCID)
    AN10114171
  • Text Lang
    JPN
  • Article Type
    Technical Report
  • ISSN
    09196072
  • NDL Article ID
    9791771
  • 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.
    Z14-1121
  • Data Source
    CJP  NDL  NII-ELS  IR  IPSJ 
Page Top