Construction of the Corpus of Everyday Japanese Conversation: An Interim Report

Construction of the Corpus of Everyday Japanese Conversation: An Interim Report Hanae Koiso†, Yasuharu Den‡,†, Yuriko Iseki†, Wakako Kashino†, Yoshiko Kawabata†, Ken’ya Nishikawa†, Yayoi Tanaka†, Yasuyuki Usuda† † National Institute for Japanese Language and Linguistics 10-2 Midoricho, Tachikawa, Tokyo 190-8561, Japan {koiso, iseki, waka, kawabata, nishikawa, yayoi, usuda}@ninjal.ac.jp ‡ Graduate School of Humanities, Chiba University 1-33 Yayoicho, Inage-ku, Chiba 263-8522, Japan den@chiba-u.jp Abstract In 2016, we launched a new corpus project in which we are building a large-scale corpus of everyday Japanese conversation in a balanced manner, aiming at exploring characteristics of conversations in contemporary Japanese through multiple approaches. The corpus targets various kinds of naturally occurring conversations in daily situations, such as conversations during dinner with the family at home, meetings with colleagues at work, and conversations while driving. In this paper, we first introduce an overview of the corpus, including corpus size, conversation variations, recording methods, structure of the corpus, and annotations to be included in the corpus. Next, we report on the current stage of the development of the corpus and legal and ethical issues discussed so far. Then we present some results of the preliminary evaluation of the data being collected. We focus on whether or not the 94 hours of conversations collected so far vary in a balanced manner by reference to the survey results of everyday conversational behavior that we conducted previously to build an empirical foundation for the corpus design. We will publish the whole corpus in 2022, consisting of more than 200 hours of recordings. Keywords: Corpus of everyday Japanese conversation, corpus design, legal and ethical issues, corpus evaluation 1. Introduction Everyday conversation is the most basic form of human communication. In order to understand our diverse and situated interactional behavior, it is needed to collect and analyze various kinds of conversations in our daily life. Although several corpora of Japanese conversations have been developed, most of them are biased in terms of speaker attributes and situations, mainly targeting conversations in experimental settings, such as map task dialogs, and artificial situations, such as chats among university students recruited for recording purposes. There are few corpora of Japanese conversations that covers real situations in daily life.1 In 2016, we launched a new corpus project, in which we are building a large-scale corpus of everyday Japanese conversation, the Corpus of Everyday Japanese Conver- sation, CEJC. The main features of the CEJC are i) that we target conversations embedded in naturally occurring activities in daily life, without the exogenous intervention of researchers imposing topics or displacing the context of action (Mondada, 2012); ii) that we collect various kinds of everyday conversations in a balanced manner so as to capture the diversity of everyday conversations and to observe natural conversational behavior in our daily life; and iii) that we collect and publish not only audio but also video data in order to precisely understand the mechanism of our real-life social behavior. In this paper, we first introduce an overview of the corpus, 1For corpora of other …

Content truncated to 500 words