Prediction Focused Topic Models via Feature Selection
Abstract
Supervised topic models are often sought to balance prediction quality and interpretability. However, when models are (inevitably) misspecified, standard approaches rarely deliver on both. We introduce a novel approach, the prediction-focused topic model, that uses the supervisory signal to retain only vocabulary terms that improve, or at least do not hinder, prediction performance. By removing terms with irrelevant signal, the topic model is able to learn task-relevant, coherent topics. We demonstrate on several data sets that compared to existing approaches, prediction-focused topic models learn much more coherent topics while maintaining competitive predictions.
Cite
Text
Ren et al. "Prediction Focused Topic Models via Feature Selection." Artificial Intelligence and Statistics, 2020.Markdown
[Ren et al. "Prediction Focused Topic Models via Feature Selection." Artificial Intelligence and Statistics, 2020.](https://mlanthology.org/aistats/2020/ren2020aistats-prediction/)BibTeX
@inproceedings{ren2020aistats-prediction,
title = {{Prediction Focused Topic Models via Feature Selection}},
author = {Ren, Jason and Kunes, Russell and Doshi-Velez, Finale},
booktitle = {Artificial Intelligence and Statistics},
year = {2020},
pages = {4420-4429},
volume = {108},
url = {https://mlanthology.org/aistats/2020/ren2020aistats-prediction/}
}