Evaluation methods for topic models

Hanna M. Wallach, Iain Murray, Ruslan Salakhutdinov, David Mimno

Research output: Chapter in Book/Report/Conference proceedingConference contribution


A natural evaluation metric for statistical topic models is the probability of held-out documents given a trained model. While exact computation of this probability is intractable, several estimators for this probability have been used in the topic modeling literature, including the harmonic mean method and empirical likelihood method. In this paper, we demonstrate experimentally that commonly-used methods are unlikely to accurately estimate the probability of held-out documents, and propose two alternative methods that are both accurate and efficient.
Original languageEnglish
Title of host publicationProceedings of the 26th Annual International Conference on Machine Learning (ICML '09)
Place of PublicationNew York, NY, USA
Number of pages8
ISBN (Print)978-1-60558-516-1
Publication statusPublished - 2009


Dive into the research topics of 'Evaluation methods for topic models'. Together they form a unique fingerprint.

Cite this