Have you tried Neural Topic Models? Comparative Analysis of Neural and Non-Neural Topic Models with Application to COVID-19 Twitter Data

TL;DR

This paper compares neural and non-neural topic models on COVID-19 Twitter data, showing neural models are superior in coherence and performance, and introduces a new regularization to improve neural models.

Contribution

It provides a comprehensive comparison of neural and traditional topic models and proposes a novel regularization to enhance neural model stability.

Findings

Neural topic models outperform classical models on evaluation metrics.

Neural models produce more coherent and interpretable topics.

The proposed regularization effectively addresses mode collapse.

Abstract

Topic models are widely used in studying social phenomena. We conduct a comparative study examining state-of-the-art neural versus non-neural topic models, performing a rigorous quantitative and qualitative assessment on a dataset of tweets about the COVID-19 pandemic. Our results show that not only do neural topic models outperform their classical counterparts on standard evaluation metrics, but they also produce more coherent topics, which are of great benefit when studying complex social problems. We also propose a novel regularization term for neural topic models, which is designed to address the well-documented problem of mode collapse, and demonstrate its effectiveness.