Japanese Daily Dialogue, or 日本語日常対話コーパス in Japanese, is a high-quality multi-turn dialogue dataset containing daily conversations on five topics: dailylife, school, travel, health, and entertainment. All dialogues are written in standard Japanese using basic vocabulary and word order. The dataset has been manually created and processed.
Here are the statistics for the dataset:
| Topic | # of dialogue | # of utterance |
|---|---|---|
| Topic 1 - Dailylife | 1,070 | 8,462 |
| Topic 2 - School | 1,058 | 8,197 |
| Topic 3 - Travel | 1,021 | 8,459 |
| Topic 4 - Health | 1,061 | 8,344 |
| Topic 5 - Entertainment | 1,051 | 8,318 |
| Total | 5,261 | 41,780 |
The dataset is structured with separate JSON files for each topic, which are stored in the data directory. Each JSON file follows the format:
{
"topic_id": 3,
"topic_name": "Travel",
"dialogue_id": 611,
"dialogue_length": 8,
"utterances": [
{
"turn_num": 1,
"speaker": "A",
"utterance": "おはようございます。高原の朝は冷えますね。"
},
{
"turn_num": 2,
"speaker": "B",
"utterance": "おはようございます。本当ですね。羽織るものが欲しいです。"
},
,,,
{
"turn_num": 8,
"speaker": "B",
"utterance": "ロッジのオーナーに聞いてみましょう。"
}
]
} For more details, please refer to the following paper:
@InProceedings{akama2023jdd,
title = {{日本語日常対話コーパスの構築}},
author = {赤間 怜奈 and 磯部 順子 and 鈴木 潤 and 乾 健太郎},
booktitle = {言語処理学会 第29回年次大会 発表論文集},
pages = {108--113},
year = {2023};
url = {https://www.anlp.jp/proceedings/annual_meeting/2023/pdf_dir/H1-1.pdf}
}
All rights to this dataset belong to our research group.
The dataset is only for research purposes. It is accessible to commercial companies for research and evaluation, but it may not be utilized as a service, such as a chatbot or any other similar application.
The dataset may not be distributed to others.
Japanese Dialy Dialogue dataset is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
