対話者間の音声特徴類似度と対話の情報伝達効果の関係 Relationship between Speakers' Characteristics and the Information Transmission Quality in Dialog

この論文にアクセスする

この論文をさがす

著者

抄録

本稿では,対話者間の音声特徴類似度と対話の情報伝達効果の関係を調査する.そのため HCRC マップタスクのデータを利用する.韻律特徴については,GMM と bigram を用いて話者の韻律特徴をモデル化して類似度を測る.言語特徴について,bigram モデルとキーワードの相対頻度を用いて対話者間の言語類似度を測る.結果として,対話者間の韻律特徴類似度と対話の情報伝達の効果の間に,正の相関があった.対話者間の言語特徴類似度と対話の情報伝達の効果の間には相関が見られなかった.We investigate the correlation between speakers' characteristics similarity and their information transmission efficiency. Data from the HCRC map task corpus is used in our study. GMM and bigram model are used to model and compare speakers' prosodic similarity. Bigram model and keywords' relative frequency are used to measure the speakers' idiolectal similarity. Results show that speakers' prosodic similarity has a positive correlation with their information transmission efficiency, while their idiolectal similarity shows no correlation.

We investigate the correlation between speakers' characteristics similarity and their information transmission efficiency. Data from the HCRC map task corpus is used in our study. GMM and bigram model are used to model and compare speakers' prosodic similarity. Bigram model and keywords' relative frequency are used to measure the speakers' idiolectal similarity. Results show that speakers' prosodic similarity has a positive correlation with their information transmission efficiency, while their idiolectal similarity shows no correlation.

収録刊行物

  • 研究報告音声言語情報処理(SLP)

    研究報告音声言語情報処理(SLP) 2014-SLP-104(27), 1-6, 2014-12-08

    一般社団法人情報処理学会

各種コード

  • NII論文ID(NAID)
    110009850971
  • NII書誌ID(NCID)
    AN10442647
  • 本文言語コード
    JPN
  • 資料種別
    Technical Report
  • データ提供元
    NII-ELS  IPSJ 
ページトップへ