話者の心の状態遷移モデルに基づく対話音声認識  [in Japanese] Speech Recognition in Spoken Dialogue Based on State Transition Model of Speaker's Mind  [in Japanese]

Access this Article

Search this Article

Author(s)

Abstract

音声認識において話題や話者の内部状態に適応した言語モデルを用いる場合、生起し得る単語列の候補数が減少するため、認識率は高くなることが予想される。このような観点から、筆者らは、一般的なコーパスから作成した平均的な言語モデルを用いた場合を基準とし、話題を"学術情報検索"に限定した場合のユーザの発話の言語モデル、および、ユーザの内部状態に適応した言語モデルの採用による認識率の改善の効果を実験により検証した。その結果、平均的な言語モデルよりも、話題に適応し、さらに状態に適応した言語モデルの採用が最も効果的であることを確認した。The performance of a speech recognition/understanding system is expected to be higher if the language model can be adapted to the current topic of the utterance. It is expected to be still higher if the model can be adapted to the current sated of the mind of the speaker. From this point of view, the present study examines the merits of these adapted language models over a language model obtained from a general corpus, by restricting the topic to "academic information retrieval" and by adopting a representation of the speaker's mind in terms of a probabilistic finite-state automaton. The experimental results confirmed the advantages of these models in quantitative terms.

The performance of a speech recognition / understanding system is expected to be higher if the language model can be adapted to the current topic of the utterance. It is expected to be still higher if the model can be adapted to the current state of the mind of the speaker. From this piont of view, the present study examines the merits of these adapted language models over a language model obtained from a general corpus, by restricting the topic to "academic information retrieval" and by adopting a representation of the speaker's mind in terms of a probabilistic finite-state automaton. The experimental results confirmed the advantages of these models in quantitative terms.

Journal

  • IPSJ SIG Notes

    IPSJ SIG Notes 2001(11(2000-SLP-035)), 79-84, 2001-02-02

    Information Processing Society of Japan (IPSJ)

References:  9

Cited by:  2

Codes

  • NII Article ID (NAID)
    110002917227
  • NII NACSIS-CAT ID (NCID)
    AN10442647
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • ISSN
    09196072
  • NDL Article ID
    5699446
  • 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  CJPref  NDL  NII-ELS  IPSJ 
Page Top