Towards a Conversation-Analytic Taxonomy of Speech Overlap

Towards A Conversation-Analytic Taxonomy of Speech Overlap Felix Gervits and Matthias Scheutz Human-Robot Interaction Laboratory, Tufts University 200 Boston Ave. Medford, MA 02155 {felix.gervits, matthias.scheutz}@tufts.edu Abstract We present a taxonomy for classifying speech overlap in natural language dialogue. The scheme classifies overlap on the basis of several features, including onset point, local dialogue history, and management behavior. We describe the various dimensions of this scheme and show how it was applied to a corpus of remote, collaborative dialogue. Moving forward, this will serve as the basis for a computational model of speech overlap, and for use in artificial agents that interact with humans in social settings. Keywords: overlap, turn-taking, conversation analysis 1. Introduction 1.1. Background Speech overlap is a common phenomenon found in human dialogue across the world (Schegloff, 2000). Overlap is not the same as interruption, as the former is considered to be a product of turn-taking organization while the latter a violation of conversational norms (Drew, 2009; Drum- mond, 1989). Much of the past work on speech overlap comes from the field of Conversation Analysis (CA). CA emphasizes talk-in-interaction and aims to study the ways in which social interaction is managed by the participants through dialogue. At the heart of CA is the model proposed by Sacks et al. (1974) (hereafter referred to as SSJ) which elegantly describes the turn-taking organization at the core of human social interaction. The SSJ model makes some important predictions about the structure of turn-taking and how it relates to speech overlap. The most important of these predictions is that while speakers exchange turns in the course of a typical dialogue, they tend to follow the “one-speaker-at-a-time” rule. Another prediction is that speaker changes occur with minimal gap or overlap, and when overlap does occur, it is usually resolved very quickly. 1.2. Speech Overlap Much of our understanding of the structure of overlap comes from the work of Gail Jefferson (Jefferson, 1982; Jefferson, 1986; Jefferson, 2004). Jefferson showed that overlap is an orderly process characterized by precise onset times with regards to turn-taking structure. She identified several types of overlap based on these onset points (de- scribed in Section 3.2.1 below), and found that they gener- ally coincide with the point in a turn at which a speaker change can occur. This suggests that overlap is a con- sequence of peoples’ adherence to the one-speaker rule and the goal to minimize gaps in between turns. On this account, overlap in collaborative dialogue is not seen as rude, but rather supportive. It indicates that conversational partners are receptive to one another and attempt to make smooth and efficient turn transitions - all predictions of the SSJ model. In addition to characterizing overlap onset, there has also been some work on understanding how people manage and recover from overlap. Jefferson (2004) addresses some as- pects of overlap management by describing ways in which people can drop out or hold the turn during overlap, as well as how people deal with segments of speech in overlap …

Content truncated to 500 words