Word Embeddings as Metric Recovery in Semantic Spaces
<jats:p> Continuous word representations have been remarkably useful across NLP tasks but remain poorly understood. We ground word embeddings in semantic spaces studied in the cognitive-psychometric literature, taking these spaces as the primary objects to recover. To this end, we relate log c...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
MIT Press - Journals
2021
|
| Online Access: | https://hdl.handle.net/1721.1/135063 |
| _version_ | 1811093595613036544 |
|---|---|
| author | Hashimoto, Tatsunori B Alvarez-Melis, David Jaakkola, Tommi S |
| author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
| author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Hashimoto, Tatsunori B Alvarez-Melis, David Jaakkola, Tommi S |
| author_sort | Hashimoto, Tatsunori B |
| collection | MIT |
| description | <jats:p> Continuous word representations have been remarkably useful across NLP tasks but remain poorly understood. We ground word embeddings in semantic spaces studied in the cognitive-psychometric literature, taking these spaces as the primary objects to recover. To this end, we relate log co-occurrences of words in large corpora to semantic similarity assessments and show that co-occurrences are indeed consistent with an Euclidean semantic space hypothesis. Framing word embedding as metric recovery of a semantic space unifies existing word embedding algorithms, ties them to manifold learning, and demonstrates that existing algorithms are consistent metric recovery methods given co-occurrence counts from random walks. Furthermore, we propose a simple, principled, direct metric recovery algorithm that performs on par with the state-of-the-art word embedding and manifold learning methods. Finally, we complement recent focus on analogies by constructing two new inductive reasoning datasets—series completion and classification—and demonstrate that word embeddings can be used to solve them as well. </jats:p> |
| first_indexed | 2024-09-23T15:47:37Z |
| format | Article |
| id | mit-1721.1/135063 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T15:47:37Z |
| publishDate | 2021 |
| publisher | MIT Press - Journals |
| record_format | dspace |
| spelling | mit-1721.1/1350632023-01-10T19:54:46Z Word Embeddings as Metric Recovery in Semantic Spaces Hashimoto, Tatsunori B Alvarez-Melis, David Jaakkola, Tommi S Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory <jats:p> Continuous word representations have been remarkably useful across NLP tasks but remain poorly understood. We ground word embeddings in semantic spaces studied in the cognitive-psychometric literature, taking these spaces as the primary objects to recover. To this end, we relate log co-occurrences of words in large corpora to semantic similarity assessments and show that co-occurrences are indeed consistent with an Euclidean semantic space hypothesis. Framing word embedding as metric recovery of a semantic space unifies existing word embedding algorithms, ties them to manifold learning, and demonstrates that existing algorithms are consistent metric recovery methods given co-occurrence counts from random walks. Furthermore, we propose a simple, principled, direct metric recovery algorithm that performs on par with the state-of-the-art word embedding and manifold learning methods. Finally, we complement recent focus on analogies by constructing two new inductive reasoning datasets—series completion and classification—and demonstrate that word embeddings can be used to solve them as well. </jats:p> 2021-10-27T20:10:34Z 2021-10-27T20:10:34Z 2016 2019-05-31T16:27:18Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/135063 en 10.1162/TACL_A_00098 Transactions of the Association for Computational Linguistics Creative Commons Attribution 4.0 International license https://creativecommons.org/licenses/by/4.0/ application/pdf MIT Press - Journals MIT Press |
| spellingShingle | Hashimoto, Tatsunori B Alvarez-Melis, David Jaakkola, Tommi S Word Embeddings as Metric Recovery in Semantic Spaces |
| title | Word Embeddings as Metric Recovery in Semantic Spaces |
| title_full | Word Embeddings as Metric Recovery in Semantic Spaces |
| title_fullStr | Word Embeddings as Metric Recovery in Semantic Spaces |
| title_full_unstemmed | Word Embeddings as Metric Recovery in Semantic Spaces |
| title_short | Word Embeddings as Metric Recovery in Semantic Spaces |
| title_sort | word embeddings as metric recovery in semantic spaces |
| url | https://hdl.handle.net/1721.1/135063 |
| work_keys_str_mv | AT hashimototatsunorib wordembeddingsasmetricrecoveryinsemanticspaces AT alvarezmelisdavid wordembeddingsasmetricrecoveryinsemanticspaces AT jaakkolatommis wordembeddingsasmetricrecoveryinsemanticspaces |