自然対話における聞き手の反応と話し手のうなずき・言語情報・韻律情報との関係に関する予備的検討  [in Japanese] Analysis of Speaker's Nod, Word, Prosody and Listener's Reaction in Japanese Spontaneous Speech  [in Japanese]

Access this Article

Search this Article

Author(s)

Abstract

自然対話でのマンマシンインタフェースの実現を考えたとき、人間のように自然な対話を実現するためには、音声言語情報ばかりでなく、ジェスチャーなどのマルチモーダルな情報を利用することが有効であると考えられる.本研究では日本語の自然対話をマルチモーダルな環境で収録し、人間同士の自然対話におけるジェスチャーの分析を行なった。日本語の自然対話ではとくにうなずきが多用される。収録した対話を分析した結果、うなずきにはあいづち同様の機能だけでなく、話し手の発話末でうなずきが生じるケースが非常に多いことが示され、また、発話末のうなずきに対し、聞き手がうなずきやあいづちなどの反応を返す現象が数多くみられた。一方、過去の研究から発話末の言語情報、韻律情報などが聞き手の反応と関係があることが示唆されていることを踏まえ、発話末の話し手のうなずき・言語情報・韻律的特微に対し、聞き手の反応がどのような傾向を示すか分析した。結果として、発話末のうなずきは言語情報や韻律情報以上に聞き手の反応をうながす機能があることが示唆された。People use gestures like nods for smooth communication in spontaneous speech dialogue. Usual dialogue continues exchanging interlocutor's information with each other using speech and gestures and therefore it is supposed that there is correlation between speakers' gestures and listeners' reaction. In this paper, we focused the relationship between the speaker's nods, syntactic features and prosodic features in the final part of each utterance and the listener's reaction like nods or backchannels. 4 dialogues by four pairs of good friends were recorded, where they can look at each other via two prompters. The prompter can record the interlocutor's gesture on videotape and project the partner's image through a half mirror. As a result, it was suggested that speakers' nod is caused frequently in the final part of utterances and that listeners show reaction like nods or backchannels frequently when the speaker nodded, said some typical words, or spoke with the typical intonation sin the final part of utterance.

People use gestures like nods for smooth communication in spontaneous speech dialogue. Usual dialogue continues exchanging interlocutor's information with each other using speech and gestures and therefore it is supposed that there is correlation between speakers' gestures and listeners' reaction. In this paper, we focused the relationship between the speaker's nods, syntactic features and prosodic features in the final part of each utterance and the listener's reaction like nods or backchannels. 4 dialogues by four pairs of good friends were recorded, where they can look at each other via two prompters. The prompter can record the interlocutor's gesture on videotape and project the partner's image through a half mirror. As a result, it was suggested that speakers' nod is caused frequently in the final part of utterances and that listeners show reaction like nods or backchannels frequently when the speaker nodded, said some typical words, or spoke with the typical intonation sin the final part of utterance.

Journal

  • IPSJ SIG Notes

    IPSJ SIG Notes 2004(74(2004-SLP-052)), 93-98, 2004-07-17

    Information Processing Society of Japan (IPSJ)

Cited by:  5

Codes

  • NII Article ID (NAID)
    110002913863
  • NII NACSIS-CAT ID (NCID)
    AN10442647
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • Data Source
    CJPref  NII-ELS  IPSJ 
Page Top