Interpretable Word Embeddings via Informative Priors
Computation and Language
2019-09-05 v1 Machine Learning
Machine Learning
Abstract
Word embeddings have demonstrated strong performance on NLP tasks. However, lack of interpretability and the unsupervised nature of word embeddings have limited their use within computational social science and digital humanities. We propose the use of informative priors to create interpretable and domain-informed dimensions for probabilistic word embeddings. Experimental results show that sensible priors can capture latent semantic concepts better than or on-par with the current state of the art, while retaining the simplicity and generalizability of using priors.
Cite
@article{arxiv.1909.01459,
title = {Interpretable Word Embeddings via Informative Priors},
author = {Miriam Hurtado Bodell and Martin Arvidsson and Måns Magnusson},
journal= {arXiv preprint arXiv:1909.01459},
year = {2019}
}
Comments
10 pages, 2 figures, EMNLP 2019