Coordinated Topic Modeling

TL;DR

This paper introduces coordinated topic modeling, a novel approach that uses reference axes to produce more interpretable and comparable topic representations across corpora, leveraging embedding-based methods and self-training.

Contribution

It proposes ECTM, an embedding-based coordinated topic model that incorporates reference axes and self-training to improve interpretability and corpus comparison.

Findings

ECTM outperforms baseline models in multiple domain experiments.

The model effectively captures corpus-specific semantics while maintaining global topic coherence.

Self-training enhances the model's ability to align topics with reference axes.

Abstract

We propose a new problem called coordinated topic modeling that imitates human behavior while describing a text corpus. It considers a set of well-defined topics like the axes of a semantic space with a reference representation. It then uses the axes to model a corpus for easily understandable representation. This new task helps represent a corpus more interpretably by reusing existing knowledge and benefits the corpora comparison task. We design ECTM, an embedding-based coordinated topic model that effectively uses the reference representation to capture the target corpus-specific aspects while maintaining each topic's global semantics. In ECTM, we introduce the topic- and document-level supervision with a self-training mechanism to solve the problem. Finally, extensive experiments on multiple domains show the superiority of our model over other baselines.