Prediction Constrained Topic Models

Open source software for training and evaluating prediction-constrained topic models, and comparing these to baseline classification methods (such as logistic regression or ensembles of decision trees).

Available on Github:

https://github.com/dtak/prediction-constrained-topic-models/

Underlying methods have been published in the AISTATS 2018 paper:

"Prediction-constrained semi-supervised topic models" M. C. Hughes, L. Weiner, G. Hope, T. H. McCoy, R. H. Perlis, E. B. Sudderth, and F. Doshi-Velez. Artificial Intelligence & Statistics (AISTATS), 2018.

Paper PDF: https://www.michaelchughes.com/papers/HughesEtAl_AISTATS_2018.pdf

Supplement PDF: https://www.michaelchughes.com/papers/HughesEtAl_AISTATS_2018_supplement.pdf