The Business Scene Dialogue corpus

©2020, The University of Tokyo

Updates

November 10, 2021: Further fix for the speaker information.
November 2, 2021: The data are updated by fixing incorrect speaker information and some misspellings in the conversation text.

Corpus Description

The Japanese-English business conversation corpus, namely Business Scene Dialogue (BSD) corpus, was constructed in 3 steps: 1) selecting business scenes, 2) writing monolingual conversation scenarios according to the selected scenes, and 3) translating the scenarios into the other language. Half of the monolingual scenarios were written in Japanese and the other half were written in English. The whole construction process was supervised by a person who satisfies the following conditions to guarantee the conversations to be natural:

• has the experience of being engaged in language learning programs, especially for business conversations
• is able to smoothly communicate with others in various business scenes both in Japanese and English
• has the experience of being involved in business

We provide balanced training, development and evaluation splits from BSD corpus. The documents in these sets are balanced in terms of scenes and original languages. In this repository we publicly share the full development and evaluation sets and a part of the training data set.

	Training	Development	Evaluation
Sentences	20,000	2,051	2,120
Scenarios	670	69	69

Corpus Statistics

Data Set	Scene	Scenarios	Sentences	Scenarios	Sentences
		JA-EN		EN-JA
Training	Face-to-face	122	3525	103	2986
	Phone call	68	1944	75	2175
	General chatting	61	1915	72	1883
	Meeting	56	1964	58	1787
	Training	12	562	19	463
	Presentation	6	607	18	189
	Total	325	10,000	345	10,000
Development	Face-to-face	11	319	12	314
	Phone call	6	176	7	185
	General chatting	7	223	8	248
	Meeting	7	240	7	219
	Training	1	40	1	23
	Presentation	1	31	1	33
	Total	34	997	35	1054
Evaluation	Face-to-face	12	381	11	345
	Phone call	6	163	7	212
	General chatting	7	211	8	212
	Meeting	7	228	7	229
	Training	1	38	1	30
	Presentation	1	31	1	40
	Total	34	1052	35	1068

Corpus Structure

The corpus is structured in json format consisting of documents, which consist of sentence pairs. Each sentence pair has a sentence number, speaker name in English and Japanese, text in English and Japanese, original language, scene of the scenario (tag), and title of the scenario (title).

[
    {
        "id": "190315_E001_17",
        "tag": "training",
        "title": "Training: How to do research",
        "original_language": "en",
        "conversation": [
            {
                "no": 1,
                "en_speaker": "Mr. Ben Sherman",
                "ja_speaker": "ベン シャーマンさん",
                "en_sentence": "I will be teaching you how to conduct research today.",
                "ja_sentence": "今日は調査の進め方についてトレーニングします。"
          },
            ...
	      ]
      },
	...
]

License

Our dataset is released under the Creative Commons Attribution-NonCommercial-ShareAlike (CC BY-NC-SA) license.

Reference

If you use this dataset, please cite the following paper: Matīss Rikters, Ryokan Ri, Tong Li, and Toshiaki Nakazawa (2019). "Designing the Business Conversation Corpus." In Proceedings of the 6th Workshop on Asian Translation, 2019.

@inproceedings{rikters-etal-2019-designing,
    title = "Designing the Business Conversation Corpus",
    author = "Rikters, Mat{\=\i}ss  and
      Ri, Ryokan  and
      Li, Tong  and
      Nakazawa, Toshiaki",
    booktitle = "Proceedings of the 6th Workshop on Asian Translation",
    month = nov,
    year = "2019",
    address = "Hong Kong, China",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D19-5204",
    doi = "10.18653/v1/D19-5204",
    pages = "54--61"
}

Acknowledgements

This work was supported by "Research and Development of Deep Learning Technology for Advanced Multilingual Speech Translation", the Commissioned Research of National Institute of Information and Communications Technology (NICT), JAPAN.