Continuous Time Dynamic Topic Models

TL;DR

This paper introduces the continuous time dynamic topic model (cDTM), which models evolving topics over time using Brownian motion, offering efficient inference and finer temporal resolution compared to discrete models.

Contribution

The paper presents the cDTM, a novel continuous-time dynamic topic model that improves inference efficiency and temporal granularity over existing discrete-time models.

Findings

cDTM outperforms discrete models in predictive perplexity

cDTM effectively predicts timestamps of documents

Demonstrated on two news corpora

Abstract

In this paper, we develop the continuous time dynamic topic model (cDTM). The cDTM is a dynamic topic model that uses Brownian motion to model the latent topics through a sequential collection of documents, where a "topic" is a pattern of word use that we expect to evolve over the course of the collection. We derive an efficient variational approximate inference algorithm that takes advantage of the sparsity of observations in text, a property that lets us easily handle many time points. In contrast to the cDTM, the original discrete-time dynamic topic model (dDTM) requires that time be discretized. Moreover, the complexity of variational inference for the dDTM grows quickly as time granularity increases, a drawback which limits fine-grained discretization. We demonstrate the cDTM on two news corpora, reporting both predictive perplexity and the novel task of time stamp prediction.