• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Book chapter

Measuring Topic Quality in Latent Dirichlet Allocation

P. 149-157.
Topic modeling is an important direction of study for modern text mining; unsupervised mining of collections of topics is intended to produce understanding and capture the essence of issues a dataset is devoted to. However, existing techniques of topic evaluation in topic models such as latent Dirichlet allocation (LDA) are still lacking in their ability to represent human interpretability and worth for qualitative studies. In this work, we propose a novel topic quality metric that more closely corresponds to human judgement than existing ones. We support this claim with the results of an experimental study where test subjects rate LDA topics on how interpretable they are.

In book

St. Petersburg: The Euler International Mathematical Institute, 2014.