Rafael Alvarado on How to Read a Topic Model

October 31, 2017

Rafael Alvarado, University of Virginia

Rafael Alvarado will discuss a system for viewing and accessing the results of topic modeling this Friday, November 3, at 1:30pm in Alderman 317. Prof. Alvarado is Research Faculty at Data Science Institute and Associate Director of SHANTI, both at the University of Virginia. His presentation is sponsored by IATH.

Abstract: Since the invention of LDA in 2002 (Blei, Ng, and Jordan), topic models have become a mainstay of digital humanities and text analytic research. However, the use of topic models remains narrowly conceived, and the method continues to raise epistemological concerns. In this presentation, Prof. Alvarado will demonstrate an interactive topic model browser designed to address one of these concerns, the problem of topic reference. Do topics refer to anything in the world, such as collective representations, or do they merely provide a convenient means of dimensionality reduction for users of large text databases? Drawing from various corpora, including online newspapers of various nationalist movements, he will present some conclusions on how to make sense of topics and how we might integrate these models into scholarly arguments.