CompareLDA: A Topic Model for Document Comparison

Comparative Latent Dirichlet Allocation (CompareLDA) learns predictive topic distributions that comply with the pairwise comparison observations.

Installation

This is a GoLang project and can be assembled using standard GoLang infrastructure:

• cmd/eval/chibeval.go is a Chib-style estimator of the predictive log-likelihood;
• cmd/fit/rldafit.go is a CompareLDA trainer;
• cmd/inf/rldainf.go is a CompareLDA predictor.