Is Automated Topic Model Evaluation Broken?: The Incoherence of Coherence

TL;DR

This paper questions the validity of automated topic coherence metrics by comparing them with human judgments and analyzing their consistency across classical and neural models.

Contribution

It highlights the validation gap in automated coherence measures for neural models and systematically evaluates models to reveal discrepancies with human assessments.

Findings

Automated coherence often disagrees with human judgments.

Neural models outperform classical models on automated metrics but not necessarily on human evaluations.

There is a significant standardization gap in topic model benchmarking.

Abstract

Topic model evaluation, like evaluation of other unsupervised methods, can be contentious. However, the field has coalesced around automated estimates of topic coherence, which rely on the frequency of word co-occurrences in a reference corpus. Contemporary neural topic models surpass classical ones according to these metrics. At the same time, topic model evaluation suffers from a validation gap: automated coherence, developed for classical models, has not been validated using human experimentation for neural models. In addition, a meta-analysis of topic modeling literature reveals a substantial standardization gap in automated topic modeling benchmarks. To address the validation gap, we compare automated coherence with the two most widely accepted human judgment tasks: topic rating and word intrusion. To address the standardization gap, we systematically evaluate a dominant classical model and two state-of-the-art neural models on two commonly used datasets. Automated evaluations declare a winning model when corresponding human evaluations do not, calling into question the validity of fully automatic evaluations independent of human judgments.