Scalable inference of topic evolution via models for latent geometric structures

TL;DR

This paper introduces a scalable, nonparametric Bayesian method for modeling the evolution of topics over time, capable of discovering new topics efficiently in large document collections.

Contribution

It presents a novel geometric and Bayesian framework that significantly accelerates topic evolution inference compared to existing methods.

Findings

Achieves several orders of magnitude speedup in inference time.

Successfully handles millions of documents within minutes.

Effectively models dynamic topic polytopes over time.

Abstract

We develop new models and algorithms for learning the temporal dynamics of the topic polytopes and related geometric objects that arise in topic model based inference. Our model is nonparametric Bayesian and the corresponding inference algorithm is able to discover new topics as the time progresses. By exploiting the connection between the modeling of topic polytope evolution, Beta-Bernoulli process and the Hungarian matching algorithm, our method is shown to be several orders of magnitude faster than existing topic modeling approaches, as demonstrated by experiments working with several million documents in under two dozens of minutes.