DC Field | Value | Language |
---|---|---|
dc.contributor.author | Park, Sungjoon | ko |
dc.contributor.author | Bak, JinYeong | ko |
dc.contributor.author | Oh, Alice Haeyun | ko |
dc.date.accessioned | 2023-08-15T07:01:15Z | - |
dc.date.available | 2023-08-15T07:01:15Z | - |
dc.date.created | 2023-07-07 | - |
dc.date.issued | 2017-09 | - |
dc.identifier.citation | 2017 Conference on Empirical Methods in Natural Language Processing, EMNLP 2017, pp.401 - 411 | - |
dc.identifier.uri | http://hdl.handle.net/10203/311551 | - |
dc.description.abstract | Vector representation of words improves performance in various NLP tasks, but the high-dimensional word vectors are very difficult to interpret. We apply several rotation algorithms to the vector representation of words to improve the interpretability. Unlike previous approaches that induce sparsity, the rotated vectors are interpretable while preserving the expressive performance of the original vectors. Furthermore, any pre-built word vector representation can be rotated for improved interpretability. We apply rotation to skip-grams and glove and compare the expressive power and interpretability with the original vectors and the sparse overcomplete vectors. The results show that the rotated vectors outperform the original and the sparse overcomplete vectors for interpretability and expressiveness tasks. | - |
dc.language | English | - |
dc.publisher | Association for Computational Linguistics (ACL) | - |
dc.title | Rotated word vector representations and their interpretability | - |
dc.type | Conference | - |
dc.identifier.scopusid | 2-s2.0-85049130647 | - |
dc.type.rims | CONF | - |
dc.citation.beginningpage | 401 | - |
dc.citation.endingpage | 411 | - |
dc.citation.publicationname | 2017 Conference on Empirical Methods in Natural Language Processing, EMNLP 2017 | - |
dc.identifier.conferencecountry | DK | - |
dc.identifier.conferencelocation | Copenhagen | - |
dc.identifier.doi | 10.18653/v1/d17-1041 | - |
dc.contributor.localauthor | Oh, Alice Haeyun | - |
dc.contributor.nonIdAuthor | Park, Sungjoon | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.