Are learned topics more useful than subject headings

Authors:
Youn Noh;Katrina Hagedorn;David Newman
Affiliations:
Yale University, New Haven, CT, USA;University of Michigan, Ann Arbor, MI, USA;University of California, Irvine, Irvine, CA, USA
Venue:
Proceedings of the 11th annual international ACM/IEEE joint conference on Digital libraries
Year:
2011

Citing 4
Cited 0

Latent dirichlet allocation

The Journal of Machine Learning Research
Subject metadata enrichment using statistical topic models

Proceedings of the 7th ACM/IEEE-CS joint conference on Digital libraries
Organizing the OCA: learning faceted subjects from a library of digital books

Proceedings of the 7th ACM/IEEE-CS joint conference on Digital libraries
Evaluating topic models for digital libraries

Proceedings of the 10th annual joint conference on Digital libraries

Quantified Score

Hi-index	0.00

Visualization

Abstract

Topic models, through their ability to automatically learn and assign topics to documents in a collection, have the potential to greatly improve how content is organized and searched in digital libraries. However, much remains to be done to assess the value of topic models in digital library applications. In this work, we present results from a user study, in which participants evaluated the similarity of books clustered using matched topics and Library of Congress Subject Headings (LCSH). Topics outperformed LCSH in 11 cases; LCSH outperformed topics in 4. These results suggest that topics are a viable alternative to LCSH.