Disfluent but effective? A quantitative study of disfluencies and conversational moves in team discourse

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3359–3369, Osaka, Japan, December 11-17 2016. Disfluent but effective? A quantitative study of disfluencies and conversational moves in team discourse Felix Gervits1 Kathleen Eberhard2 Matthias Scheutz1 1Tufts University, Human-Robot Interaction Laboratory, Medford, MA 2University of Notre Dame, Department of Psychology, Notre Dame, IN {felix.gervits, matthias.scheutz}@tufts.edu eberhard.1@nd.edu Abstract Situated dialogue systems that interact with humans as part of a team (e.g., robot teammates) need to be able to use information from communication channels to gauge the coordination level and effectiveness of the team. Currently, the feasibility of this end goal is limited by several gaps in both the empirical and computational literature. The purpose of this paper is to address those gaps in the following ways: (1) investigate which properties of task-oriented discourse corre- spond with effective performance in human teams, and (2) discuss how and to what extent these properties can be utilized in spoken dialogue systems. To this end, we analyzed natural language data from a unique corpus of spontaneous, task-oriented dialogue (CReST corpus), which was annotated for disfluencies and conversational moves. We found that effective teams made more self-repair disfluencies and used specific communication strategies to facilitate grounding and coordination. Our results indicate that truly robust and natural dialogue systems will need to interpret highly disfluent utterances and also utilize specific collaborative mechanisms to facili- tate grounding. These data shed light on effective communication in performance scenarios and directly inform the development of robust dialogue systems for situated artificial agents. 1 Introduction Effective coordination is fundamental to teamwork in many fields, particularly for teams working under stress. Though a variety of team structures exist, actions teams are among the most demanding, due to the need for the teammates to engage in goal-oriented, interdependent tasks, and to dynamically adapt their decision-making, communication, and planning strategies (Serfaty et al., 1993; Sundstrom et al., 2000). Recently, action teams have begun to incorporate artificial agents in mixed-initiative roles (see Sycara and Sukthankar (2006) for a thorough review). Such teams are largely employed in performance scenarios (e.g., search and rescue missions, military squads, surgical teams), and require strong team communication to achieve complex objectives in a time-sensitive context. Among the many elements needed for coordination in human-agent teams, perhaps the most important for team success is establish- ing common ground. Common ground is a mutual understanding between teammates, involving shared knowledge of the task environment, goals, perspectives, and other factors. For common ground to be established, it is important that teammates not only share information, but also ensure that the informa- tion was understood the way it was intended. This process, known as grounding, results in a mutual recognition by both parties of the shared information being a part of their common ground and forms the basis for coordination in both dialogue and action (Clark, 1996). 1.1 Task-oriented remote dialogue Though there is no objective way to measure common ground, evidence for it may be found in the team …

Content truncated to 500 words