Recognizing Authority in Dialogue with an Integer Linear Programming Constrained Model

Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics, pages 1018–1026, Portland, Oregon, June 19-24, 2011. c⃝2011 Association for Computational Linguistics Recognizing Authority in Dialogue with an Integer Linear Programming Constrained Model Elijah Mayfield Language Technologies Institute Carnegie Mellon University Pittsburgh, PA 15213 elijah@cmu.edu Carolyn Penstein Ros´e Language Technologies Institute Carnegie Mellon University Pittsburgh, PA 15213 cprose@cs.cmu.edu Abstract We present a novel computational formula- tion of speaker authority in discourse. This notion, which focuses on how speakers posi- tion themselves relative to each other in dis- course, is first developed into a reliable cod- ing scheme (0.71 agreement between human annotators). We also provide a computational model for automatically annotating text using this coding scheme, using supervised learning enhanced by constraints implemented with In- teger Linear Programming. We show that this constrained model’s analyses of speaker au- thority correlates very strongly with expert hu- man judgments (r2 coefficient of 0.947). 1 Introduction In this work, we seek to formalize the ways speak- ers position themselves in discourse. We do this in a way that maintains a notion of discourse structure, and which can be aggregated to evaluate a speaker’s overall stance in a dialogue. We define the body of work in positioning to include any attempt to formal- ize the processes by which speakers attempt to influ- ence or give evidence of their relations to each other. Constructs such as Initiative and Control (Whittaker and Stenton, 1988), which attempt to operationalize the authority over a discourse’s structure, fall under the umbrella of positioning. As we construe posi- tioning, it also includes work on detecting certainty and confusion in speech (Liscombe et al., 2005), which models a speaker’s understanding of the in- formation in their statements. Work in dialogue act tagging is also relevant, as it seeks to describe the ac- tions and moves with which speakers display these types of positioning (Stolcke et al., 2000). To complement these bodies of work, we choose to focus on the question of how speakers position themselves as authoritative in a discourse. This means that we must describe the way speakers intro- duce new topics or discussions into the discourse; the way they position themselves relative to that topic; and how these functions interact with each other. While all of the tasks mentioned above focus on specific problems in the larger rhetorical question of speaker positioning, none explicitly address this framing of authority. Each does have valuable ties to the work that we would like to do, and in section 2, we describe prior work in each of those areas, and elaborate on how each relates to our questions. We measure this as an authoritativeness ratio. Of the contentful dialogue moves made by a speaker, in what fraction of those moves is the speaker po- sitioned as the primary authority on that topic? To measure this quantitatively, we introduce the Nego- tiation framework, a construct from the field of sys- temic functional linguistics (SFL), which addresses specifically the concepts that we …

Content truncated to 500 words