Interpretable Word Embeddings via Informative Priors

Miriam Hurtado Bodell, Martin Arvidsson, Måns Magnusson

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

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.
Original languageEnglish
Title of host publicationThe 2019 Conference on Empirical Methods in Natural Language Processing And the 9th International Joint Conference on Natural Language Processing
Subtitle of host publicationProceedings of System Demonstrations
Pages6324-6330
Publication statusPublished - 2019
MoE publication typeA4 Article in a conference publication
EventConference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing - Hong Kong, China
Duration: 3 Nov 20197 Nov 2019

Conference

ConferenceConference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing
Abbreviated titleMNLP/IJCNLP
Country/TerritoryChina
CityHong Kong
Period03/11/201907/11/2019

Fingerprint

Dive into the research topics of 'Interpretable Word Embeddings via Informative Priors'. Together they form a unique fingerprint.

Cite this