Discovering Discrete Latent Topics with Neural Variational Inference

TL;DR

This paper introduces neural variational inference methods for topic modeling, enabling flexible, scalable discovery of an unbounded number of topics with improved efficiency over traditional approaches.

Contribution

It proposes neural variational inference for topic models, including a recurrent network with stick-breaking construction for discovering unlimited topics.

Findings

Effective on multiple datasets

Faster inference compared to traditional methods

Capable of discovering many topics

Abstract

Topic models have been widely explored as probabilistic generative models of documents. Traditional inference methods have sought closed-form derivations for updating the models, however as the expressiveness of these models grows, so does the difficulty of performing fast and accurate inference over their parameters. This paper presents alternative neural approaches to topic modelling by providing parameterisable distributions over topics which permit training by backpropagation in the framework of neural variational inference. In addition, with the help of a stick-breaking construction, we propose a recurrent network that is able to discover a notionally unbounded number of topics, analogous to Bayesian non-parametric topic models. Experimental results on the MXM Song Lyrics, 20NewsGroups and Reuters News datasets demonstrate the effectiveness and efficiency of these neural topic models.