Generating status hierarchies from meeting transcripts using the author-topic model

Abstract

Topic models may be applied to solve problems of interest to many subfields of social-science. This paper expands the social-science uses of topic modeling to the analysis of group decision-making. In particular, we study committees of experts in the U.S. Food and Drug Administration. The output of the analysis is a set of directed social networks that reveal meaningful status hierarchies within these committees.

Publication
In Proceedings of the Workshop: Applications for Topic Models: Text and Beyond, Neural Information Processing Systems (NIPS) 2009