CS Colloquium
September 13, 2012
4:00 p.m., AKW 200

Refreshments will be available at 3:45

Speaker: David Mimno
Title: Finding thousands of topics in millions of books

Abtract: Statistical topic models have become popular in domains as distinct as biomedical research, political science, and literary scholarship. These methods represent text documents as combinations of themes, or topics, which are themselves probability distributions over a vocabulary. This low-dimensional topic representation is robust to variation in word choice and ambiguity in word sense, allowing users to analyze trends in large text collections. Existing methods for training topic models, however, have not kept pace with the size of today's document corpora. In this talk I will describe a new method that combines the best aspects of two inference methods, stochastic online inference and Markov chain Monte Carlo. I will demonstrate the scalability of this algorithm on a corpus of 1.2 million out-of-copyright books.

Bio: David Mimno is a postdoctoral researcher in the Computer Science department at Princeton University. He received his PhD from the University of Massachusetts, Amherst. Before graduate school, he served as Head Programmer at the Perseus Project, a digital library for cultural heritage materials, at Tufts University. He is supported by a CRA Computing Innovation fellowship.