Dialogue Collection for Recording the Process of Building Common Ground in a Collaborative Task

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5749–5758 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5749 Dialogue Collection for Recording the Process of Building Common Ground in a Collaborative Task Koh Mitsuda1, Ryuichiro Higashinaka1, Yuhei Oga2∗, Sen Yoshida1 1NTT Human Informatics Laboratories, NTT Corporation, Japan 2Graduate School of Systems and Information Engineering, University of Tsukuba, Japan {koh.mitsuda.td,ryuichiro.higashinaka.tp,sen.yoshida.tu}@hco.ntt.co.jp s2020716@s.tsukuba.ac.jp Abstract To develop a dialogue system that can build common ground with users, the process of building common ground through dialogue needs to be clarified. However, the studies on the process of building common ground have not been well conducted; much work has focused on finding the relationship between a dialogue in which users perform a collaborative task and its task performance represented by the final result of the task. In this study, to clarify the process of building common ground, we propose a data collection method for automatically recording the process of building common ground through a dialogue by using the intermediate result of a task. We collected 984 dialogues, and as a result of investigating the process of building common ground, we found that the process can be classified into several typical patterns and that conveying each worker’s understanding through affirmation of a counterpart’s utterances especially contributes to building common ground. In addition, toward dialogue systems that can build common ground, we conducted an automatic estimation of the degree of built common ground and found that its degree can be estimated quite accurately. Keywords: dialogue systems, data collection, common ground, collaborative task 1. Introduction Dialogue systems need to be able to build common ground with users accurately for them to successfully perform joint activities (Clark, 1996; Traum, 1994; Kopp and Kramer, 2021). To develop dialogue systems able to handle common ground, we believe the key is to reveal the process of building common ground. In previous studies, the process of building common ground has been investigated by analyzing a dialogue between people who accomplish a collaborative task (Benotti and Blackburn, 2021; Chandu et al., 2021). These studies mainly focused on finding the relation- ship between a dialogue and the final result of the task (Anderson et al., 1991; Foster et al., 2008; He et al., 2017); however, the studies on the process of building common ground itself have not been well conducted. A few exceptions include the study by Udagawa and Aizawa (2020), who associated reference expressions with objects in a collaborative task to analyze the pro- cess of building common ground. Bara et al. (2021) recorded intermediate common ground by having each worker answer questions about his/her understanding of his/her partner’s belief and behavior in real time dur- ing a collaborative task in a virtual space. These studies manually record common ground, which is costly. In this paper, we look into the process of building com- mon ground by automatically recording the interme- diate result of the task as a proxy for the common ground being built. …

Content truncated to 500 words