Modeling Topic Control in Conversations using Speaker-centric Nonparametric Topic Models

Identifying influential speakers in multi-party conversations has been the focus of research in communication, sociology, and psychology for decades. It has been long acknowledged qualitatively that controlling the topic of a conversation is a sign of influence. To capture who introduces new topics in conversations, we introduce SITS—Speaker Identity for Topic Segmentation—a nonparametric hierarchical Bayesian model that is capable of discovering (1) the topics used in a set of conversations, (2) how these topics are shared across conversations, (3) when these topics change during conversations, and (4) a speaker-specific measure of “topic control”. We validate the model via evaluations using multiple datasets, including work meetings, online discussions, and political debates. Experimental results confirm the effectiveness of SITS in both intrinsic and extrinsic evaluations. On an intrinsic evaluation, SITS outperforms existing methods on topic segmentation. In an extrinsic evaluation, we annotated influencers in conversations and showed that methods for V.-A. Nguyen Computer Science, University of Maryland E-mail: [email protected] J. Boyd-Graber iSchool and UMIACS, University of Maryland E-mail: [email protected] P. Resnik Linguistics and UMIACS, University of Maryland E-mail: [email protected] D. Cai Communications, Temple University E-mail: [email protected] J. Midberry Communications, Temple University E-mail: [email protected] Y. Wang Communications, Temple University E-mail: [email protected] 2 Nguyen, Boyd-Graber, Resnik, et al. detecting influencers based on SITS outperform traditional, popular methods using structural patterns of conversations.