Presentation 2022-11-30
Dialogue disfluency detection using context
Hiroto Nakashima, Kazutaka Shimada,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recent automatic speech recognition (ASR) techniques have been improved by a large amount of training data and machine learning, such as deep learning technology. Problems in the outputs from ASR are not only recognition errors but also outputs caused by disfluency from speakers. It is difficult to remove them automatically, and removing them by hand is costly. In this paper, we propose a disfluency detection model with BERT. The model utilizes context information of target utterances. We introduce two types of context information. The first one is real utterances that appear around the target utterance. We compare several sequence lengths of the previous and following utterances. The second one is a generated utterance by GPT-2. Our model adds the utterance generated from the target utterance as the following context. In the experiment, the long sequence improves the disfluency detection accuracy, and real context outperforms generated context.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) disfluency detection / disfluency / dialogue generation / context complement
Paper # NLC2022-13,SP2022-33
Date of Issue 2022-11-22 (NLC, SP)

Conference Information
Committee NLC / IPSJ-NL / SP / IPSJ-SLP
Conference Date 2022/11/29(3days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Mitsuo Yoshida(Univ. of Tsukuba) / 須藤 克仁(奈良先端科学技術大学院大学) / Tomoki Toda(Nagoya Univ.) / 戸田 智基(名古屋大学)
Vice Chair Hiroki Sakaji(Univ. of Tokyo) / Takeshi Kobayakawa(NHK)
Secretary Hiroki Sakaji(NTT) / Takeshi Kobayakawa(Hiroshima Univ. of Economics) / (株式会社デンソーアイティーラボラトリ) / (北海学園大学) / (東京農工大学)
Assistant Kanjin Takahashi(Sansan) / Yasuhiro Ogawa(Nagoya Univ.) / / Ryo Aihara(Mitsubishi Electric) / Daisuke Saito(Univ. of Tokyo)

Paper Information
Registration To Technical Committee on Natural Language Understanding and Models of Communication / Special Interest Group on Natural Language / Technical Committee on Speech / Special Interest Group on Spoken Language Processing
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Dialogue disfluency detection using context
Sub Title (in English)
Keyword(1) disfluency detection
Keyword(2) disfluency
Keyword(3) dialogue generation
Keyword(4) context complement
1st Author's Name Hiroto Nakashima
1st Author's Affiliation Kyusyu Institute of Technology(KIT)
2nd Author's Name Kazutaka Shimada
2nd Author's Affiliation Kyusyu Institute of Technology(KIT)
Date 2022-11-30
Paper # NLC2022-13,SP2022-33
Volume (vol) vol.122
Number (no) NLC-287,SP-288
Page pp.pp.21-26(NLC), pp.21-26(SP),
#Pages 6
Date of Issue 2022-11-22 (NLC, SP)