コロキアムB発表

日時: 9月13日(火)4限(15:10-16:40)


会場: L2

司会: SOUFI Mazen
福田 りょう D, 中間発表 知能コミュニケーション 中村 哲, 渡辺 太郎, 須藤 克仁
title: Speech Segmentation Optimization for End-to-End Speech Translation
abstract: Speech segmentation, which splits long speech into short segments, is essential for speech translation (ST). Popular VAD tools have generally relied on pause-based segmentation. Unfortunately, pauses in speech do not necessarily match sentence boundaries, and sentences can be connected by a very short pause that is difficult to detect by VAD. In this study, we propose a speech segmentation method using a binary classification model trained using a segmented bilingual speech corpus. We also propose a hybrid method that combines VAD and the above speech segmentation method. Experimental results reveal that the proposed method is more suitable for cascade and end-to-end ST systems than conventional segmentation methods. The hybrid approach further improves the translation performance.
language of the presentation: Japanese
 
KIYOMOTO BARBARA M, 2回目発表 知能コミュニケーション 中村 哲, 渡辺 太郎, 須藤 克仁
title: Word Emotion Classification Using Movie Scripts
abstract: A common issue that language learners face is identifying word emotion in their non-native language, translations that they reference may provide a literal translation without consideration for the emotional information a word carries. There is a limited amount of word emotion lexicons that are available, therefore in this task we attempt to create a system that automatically classifies word emotion using movie scripts as training data.
language of the presentation: English
 
加賀見 彰吾 M, 2回目発表 知能コミュニケーション 中村 哲, 渡辺 太郎, 宮尾 知幸, 須藤 克仁
title: Question generation system that asks for missing information in text
abstract: When people have questions about the contents of documents, it is often because there are parts of the documents that are not mentioned or not explained well enough. Questions about such parts are often based on the knowledge of the questioner, but few questioning systems hold prior knowledge. Therefore, in this study, we are trying a question generation system that can ask appropriate questions to those parts of the text that are not described or not explained well enough.
language of the presentation: Japanese
発表題目: 文章中の不足している情報を問う質問生成システム
発表概要: 文書の内容に対して人間が疑問を抱く場合は文章中に記載されていない箇所や説明不足な箇所があることが原因であることが多い. そのような箇所について質問する際には質問者の知識が基となっていることが多いが, 質問システムに予備知識を与えているものは少ない. そこで本研究では記載されていない箇所や説明不足な箇所に対して適切な質問が出来るような質問生成システムの開発に取り組んでいる.
 
奥田 由佳 M, 2回目発表 知能コミュニケーション 中村 哲, 渡辺 太郎, 須藤 克仁, 品川 政太朗
title:Conversational recommendation system that considers users multiple interests based on knowledge structure
abstract:In recent years, with the development of dialogue systems, research on systems that recommend items to users using natural conversation has attracted attention.However, existing research may cause the ‘’missing interest problem'' in which one of the interests cannot be considered when there are multiple interests of the user. Therefore, we propose and evaluate a system that considers that users have multiple interests.
language of the presentation: Japanese
発表題目: 知識の構造に基づきユーザの複数の興味を考慮する対話推薦システム
発表概要: 近年、対話システムの発展に伴い自然な会話を用いてユーザにアイテムを推薦するシステムが注目されている。しかし、既存の研究はユーザの興味が複数にまたがっていた場合どちらか一方の興味を考慮できない「興味の取りこぼし問題」を起こす可能性がある.そこで,ユーザが興味を複数にもつことを考慮したシステムを提案し評価する.