Design and Evaluation of the Corpus of Everyday Japanese Conversation

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5587–5594 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5587 Design and Evaluation of the Corpus of Everyday Japanese Conversation Hanae Koiso†, Haruka Amatani†, Yasuharu Den‡, Yuriko Iseki†, Yuichi Ishimoto†, Wakako Kashino†, Yoshiko Kawabata†, Ken’ya Nishikawa†, Yayoi Tanaka†, Yasuyuki Usuda†, Yuka Watanabe† † National Institute for Japanese Language and Linguistics 10-2 Midoricho, Tachikawa, Tokyo 190-8561, Japan {koiso, h-amatani, iseki, yishi, waka, kawabata, nishikawa, yayoi, usuda, yuwatanabe}@ninjal.ac.jp ‡ Graduate School of Humanities, Chiba University 1-33 Yayoicho, Inage-ku, Chiba 263-8522, Japan den@chiba-u.jp Abstract We have constructed the Corpus of Everyday Japanese Conversation (CEJC) and published it in March 2022. The CEJC is designed to contain various kinds of everyday conversations in a balanced manner to capture their diversity. The CEJC features not only audio but also video data to facilitate precise understanding of the mechanism of real-life social behavior. The publication of a large-scale corpus of everyday conversations that includes video data is a new approach. The CEJC contains 200 hours of speech, 577 conversations, about 2.4 million words, and a total of 1675 conversants. In this paper, we present an overview of the corpus, including the recording method and devices, structure of the corpus, formats of video and audio files, transcription, and annotations. We then report some results of the evaluation of the CEJC in terms of conversant and conversation attributes. We show that the CEJC includes a good balance of adult conversants in terms of gender and age, as well as a variety of conversations in terms of conversation forms, places, activities, and numbers of conversants. Keywords: Corpus of everyday Japanese conversation, corpus design, corpus evaluation 1. Introduction The characteristics of languages and behaviors in con- versations occurring in daily life are an important re- search theme. Although many corpora of Japanese con- versations have been constructed, most have targeted conversations in artificially created settings in terms of topics and recording situations or were biased in terms of conversational situation. In addition, there is no cor- pus of Japanese conversations that provides video data, and some do not even provide audio data (see Table 1). Given this situation, we have engaged since 2016 in the construction of the Corpus of Everyday Japanese Con- versation, CEJC. The main features of the CEJC are i) a focus on conversations embedded in naturally oc- curring activities in daily life, ii) the variety of every- day conversations, collected in a balanced manner to capture the diversity of everyday conversations and ob- serve natural conversational behavior in our daily life; and iii) the publication of audio and video data in or- der to precisely understand the mechanism of real-life social behavior. We published a part of the CEJC, 50 hours of conversations, in 2018 on a trial basis, and will publish the whole corpus, comprising 200 hours, in March 2022. In this paper, we present an overview of the corpus, in- cluding the recording method and devices, …

Content truncated to 500 words