Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
audio
audioduration (s)
676
762
label
class label
10 classes
0japanese-themed-dialogue01
0japanese-themed-dialogue01
0japanese-themed-dialogue01
1japanese-themed-dialogue02
1japanese-themed-dialogue02
1japanese-themed-dialogue02
2japanese-themed-dialogue03
2japanese-themed-dialogue03
2japanese-themed-dialogue03
3japanese-themed-dialogue04
3japanese-themed-dialogue04
3japanese-themed-dialogue04
4japanese-themed-dialogue05
4japanese-themed-dialogue05
4japanese-themed-dialogue05
5japanese-themed-dialogue06
5japanese-themed-dialogue06
5japanese-themed-dialogue06
6japanese-themed-dialogue07
6japanese-themed-dialogue07
6japanese-themed-dialogue07
7japanese-themed-dialogue08
7japanese-themed-dialogue08
7japanese-themed-dialogue08
8japanese-themed-dialogue09
8japanese-themed-dialogue09
8japanese-themed-dialogue09
9japanese-themed-dialogue10
9japanese-themed-dialogue10
9japanese-themed-dialogue10

Japanese Themed Casual Conversation Corpus

Overview

The Japanese Themed Casual Conversation Corpus is a collection of Japanese two-speaker casual conversations based on predefined themes.

The dataset contains 10 dialogues covering a broad range of everyday and knowledge-related topics.

Each dialogue includes two-speaker mixed audio, speaker-separated audio, and a manually created, time-aligned ELAN (.eaf) annotation file.

The conversations were recorded for speech and dialogue research. Participants were given a broad theme and engaged in natural conversation rather than reading a fixed script.

This repository serves as a public sample demonstrating IngCrowd's expertise in Japanese dialogue-data collection, multi-track audio production, manual transcription, and ELAN-based time-aligned annotation.


Publicly Available Data

The dataset currently contains:

  • 10 Japanese themed casual conversations
  • 10 two-speaker mixed audio recordings
  • 20 speaker-separated audio recordings
  • 10 manually created ELAN (.eaf) annotation files
  • Approximately 121 minutes of dialogue

Audio Format

  • WAV
  • 16 kHz sampling rate
  • 16-bit
  • Mono

Video recordings are not included in this repository.

Corresponding video recordings may be provided for research purposes upon request, subject to availability and applicable conditions.


Key Features

  • Natural Japanese casual conversations based on predefined themes
  • Two-speaker mixed audio and speaker-separated audio
  • Manually created ELAN annotations
  • Human-generated verbatim transcriptions
  • Speaker-level time-aligned segmentation
  • Preservation of spontaneous speech phenomena such as fillers, hesitations, repetitions, laughter, and unintelligible speech
  • Broad topic coverage ranging from everyday life to technology and science
  • Suitable for speech processing, dialogue analysis, and Japanese language research

Data Collection

The dialogues were collected through online conversations conducted via Zoom.

Each dialogue consists of a natural conversation between two participants. The participants were assigned a broad theme and asked to converse freely about related topics. The conversations were not based on a fixed script.


Conversation Themes

ID Conversation Type Theme Participant 1 Participant 2 Duration Availability
01 Themed casual conversation Entertainment and hobbies Female, 40s Male, 20s 11:15 Available
02 Themed casual conversation Daily life and living Female, 40s Male, 20s 11:47 Available
03 Themed casual conversation Health, beauty, and fashion Female, 40s Male, 20s 12:42 Available
04 Themed casual conversation Manners, weddings, funerals, and ceremonial occasions Female, 40s Male, 20s 12:39 Available
05 Themed casual conversation Life, romance, and interpersonal concerns Female, 40s Female, 40s 12:16 Available
06 Themed casual conversation Parenting and school Female, 40s Female, 40s 12:36 Available
07 Themed casual conversation Internet and communications Female, 40s Female, 30s 11:35 Available
08 Themed casual conversation Smart devices, PCs, and home appliances Female, 40s Female, 30s 11:49 Available
09 Themed casual conversation Computer technology Female, 40s Female, 30s 11:46 Available
10 Themed casual conversation Education, academic subjects, and science Female, 40s Female, 30s 12:27 Available

The total duration across all 10 dialogues is 120:52.

Participant information is presented only at a broad demographic level. No participant names or other personally identifiable information are included.

Participant 1 corresponds to _individual_audio1.wav, and Participant 2 corresponds to _individual_audio2.wav.


Data Structure

Each dialogue directory contains the following files:

japanese-themed-dialogueXX/
├── japanese-themed-dialogueXX_audio.wav
├── japanese-themed-dialogueXX_individual_audio1.wav
├── japanese-themed-dialogueXX_individual_audio2.wav
└── japanese-themed-dialogueXX.eaf
  • _audio.wav — two-speaker mixed audio
  • _individual_audio1.wav and _individual_audio2.wav — speaker-separated audio
  • .eaf — manually created ELAN annotation with time-aligned transcription

The repository includes directories japanese-themed-dialogue01 through japanese-themed-dialogue10.


Additional Resources

The dataset includes audio recordings and ELAN annotation files. Corresponding video recordings are not included in this repository.

Researchers who wish to obtain synchronized video recordings for academic or research purposes are encouraged to contact us at aidatasets@ingcrowd.jp.


Transcription and Annotation

All transcriptions were manually created by human annotators using ELAN.

Each utterance was manually segmented and aligned with the corresponding speech.

Rather than editing spoken language into polished written language, the annotations preserve fillers, hesitations, repetitions, laughter, and unintelligible speech as faithfully as possible.

No automatic speech recognition (ASR) system was used during transcription or annotation.


Potential Uses

This dataset is suitable for research and development in:

  • Automatic Speech Recognition (ASR)
  • Speaker diarization
  • Spoken dialogue systems
  • Dialogue and discourse analysis
  • Turn-taking analysis
  • Topic and conversation analysis
  • Japanese speech corpus research
  • ELAN-based annotation research
  • Speech segmentation and alignment
  • Conversational AI

Commercial Support

IngCrowd provides custom Japanese speech-data collection and annotation services, including:

  • Conversation theme and scenario design
  • Participant recruitment
  • Remote or on-site recording
  • Mixed and speaker-separated audio production
  • Manual verbatim transcription
  • Human-created ELAN annotation
  • Time-aligned speech segmentation
  • Audio, text, and video dataset production

For collaboration, custom dataset production, additional data, or corresponding video recordings, please contact us at aidatasets@ingcrowd.jp.


License

This dataset is released under the CC BY-NC 4.0 license.


日本語

概要

Japanese Themed Casual Conversation Corpus は、あらかじめ設定されたテーマに基づいて2名の参加者が会話した、日本語のテーマ型雑談コーパスです。

日常生活からテクノロジー、学問・サイエンスまで、幅広いテーマを扱う10件の対話を収録しています。

各対話には、2話者ミックス音声、話者別音声、および人手で作成した時間同期ELAN(.eaf)アノテーションを収録しています。

参加者はあらかじめ設定された大きなテーマに沿って会話しており、台本の読み上げではなく、自然な雑談として収録しています。

本リポジトリは、株式会社イングクラウドが提供する日本語対話データ収集、マルチトラック音声制作、人手による書き起こし、ELANを用いた時間同期アノテーションのサンプルとして公開しています。


公開データ

現在、本データセットには以下のデータを収録しています。

  • 日本語テーマ型雑談:10件
  • 2話者ミックス音声:10本
  • 話者別音声:20本
  • 人手で作成したELAN(.eaf)アノテーション:10本
  • 総収録時間:約121分

音声仕様

  • WAV形式
  • サンプリング周波数:16 kHz
  • 量子化ビット数:16 bit
  • モノラル

対応する映像データは本リポジトリには含まれていません。

研究目的で利用を希望される場合は、お問い合わせいただければ、提供可能な範囲および条件に応じて個別にご案内いたします。


特長

  • あらかじめ設定されたテーマに基づく自然な日本語雑談
  • 2話者ミックス音声および話者別音声
  • 人手で作成したELANアノテーション
  • 人手による発話忠実な書き起こし
  • 発話単位で時間同期されたアノテーション
  • フィラー、言いよどみ、繰り返し、笑い、聞き取り困難な発話など、自然発話の特徴を保持
  • 日常生活からテクノロジー、学問・サイエンスまで幅広いテーマを収録
  • 音声処理、対話分析、日本語研究などに利用可能

データ収集

対話はZoomを用いたオンライン会話として収録しました。

各対話は2名の参加者による自然な雑談で構成されています。参加者には大きなテーマを提示し、そのテーマに関連する内容について自由に会話してもらいました。固定された台本は使用していません。


収録テーマ

ID 対話形式 テーマ 参加者1 参加者2 収録時間 公開状況
01 テーマ型雑談 エンターテインメントと趣味 40代女性 20代男性 11:15 公開中
02 テーマ型雑談 暮らし、生活 40代女性 20代男性 11:47 公開中
03 テーマ型雑談 健康、美容とファッション 40代女性 20代男性 12:42 公開中
04 テーマ型雑談 マナー、冠婚葬祭 40代女性 20代男性 12:39 公開中
05 テーマ型雑談 生き方と恋愛、人間関係の悩み 40代女性 40代女性 12:16 公開中
06 テーマ型雑談 子育てと学校 40代女性 40代女性 12:36 公開中
07 テーマ型雑談 インターネット、通信 40代女性 30代女性 11:35 公開中
08 テーマ型雑談 スマートデバイス、PC、家電 40代女性 30代女性 11:49 公開中
09 テーマ型雑談 コンピュータテクノロジー 40代女性 30代女性 11:46 公開中
10 テーマ型雑談 教養と学問、サイエンス 40代女性 30代女性 12:27 公開中

全10件の総収録時間は120分52秒です。

参加者情報は年代・性別のみを掲載しており、参加者名など個人を特定できる情報は含めていません。

参加者1は _individual_audio1.wav、参加者2は _individual_audio2.wav に対応しています。


データ構成

各対話フォルダには以下のファイルが含まれています。

japanese-themed-dialogueXX/
├── japanese-themed-dialogueXX_audio.wav
├── japanese-themed-dialogueXX_individual_audio1.wav
├── japanese-themed-dialogueXX_individual_audio2.wav
└── japanese-themed-dialogueXX.eaf
  • _audio.wav:2話者ミックス音声
  • _individual_audio1.wav_individual_audio2.wav:話者別音声
  • .eaf:人手で作成したELAN形式の時間同期アノテーション

japanese-themed-dialogue01 から japanese-themed-dialogue10 までを収録しています。


追加データ

本データセットには音声ファイルおよびELANアノテーションファイルを収録しています。対応する映像データは、本リポジトリには含まれていません。

研究目的で対応する映像データの利用を希望される場合は、aidatasets@ingcrowd.jp までお問い合わせください。


書き起こし・アノテーション

書き起こしはすべて人手によりELANを用いて作成しています。

各発話は人手で区切り、音声との時間対応を付与しています。

読みやすい文章へ整形するのではなく、フィラー、言いよどみ、繰り返し、笑い、聞き取りが困難な発話を含め、実際の発話内容をできる限り忠実に転記しています。

書き起こしおよびアノテーションには、自動音声認識(ASR)は使用していません。


想定用途

本データセットは以下の研究・開発用途を想定しています。

  • 日本語音声認識(ASR)
  • 話者分離
  • 音声対話システム
  • 対話・談話分析
  • ターンテイキング分析
  • トピック・会話分析
  • 日本語音声コーパス研究
  • ELANアノテーション研究
  • 音声区間分割・時間同期
  • 対話AI

カスタムデータセット制作

株式会社イングクラウドでは、プロジェクトの要件に応じた日本語音声データセットの制作を行っています。

対応内容の例:

  • 会話テーマ・シナリオ設計
  • 参加者募集・収録進行
  • 対面・リモート収録
  • 2話者ミックス音声・話者別音声作成
  • 人手による発話忠実な書き起こし
  • 人手によるELANアノテーション
  • 発話区間の時間同期
  • 音声・テキスト・映像データセット制作

カスタムデータセット制作、追加データ、対応する映像データについては、aidatasets@ingcrowd.jp までお問い合わせください。


ライセンス

本データセットは CC BY-NC 4.0 ライセンスで公開しています。

Custom Dataset Services / データセット制作のご相談

IngCrowd provides custom Japanese participant recruitment, speech and dialogue recording, transcription, time-aligned ELAN annotation, data collection, and human evaluation for AI development and academic research.

For custom dataset creation or research support, please visit our website or contact us by email.

株式会社イングクラウドでは、AI開発・学術研究向けに、研究参加者や話者の募集、日本語の音声・対話収録、文字起こし、ELANアノテーション、データ収集、人手評価などを提供しています。

独自のデータセット制作や研究支援については、ウェブサイトまたはメールよりお問い合わせください。

Downloads last month
83