Interpreting Character Embeddings With Perceptual Representations: The Case of Shape, Sound, and Color

Sidsel Boldsen, Manex Agirrezabal, Nora Hollenstein


Abstract
Character-level information is included in many NLP models, but evaluating the information encoded in character representations is an open issue. We leverage perceptual representations in the form of shape, sound, and color embeddings and perform a representational similarity analysis to evaluate their correlation with textual representations in five languages. This cross-lingual analysis shows that textual character representations correlate strongly with sound representations for languages using an alphabetic script, while shape correlates with featural scripts. We further develop a set of probing classifiers to intrinsically evaluate what phonological information is encoded in character embeddings. Our results suggest that information on features such as voicing are embedded in both LSTM and transformer-based representations.
Anthology ID:
2022.acl-long.470
Volume:
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
May
Year:
2022
Address:
Dublin, Ireland
Editors:
Smaranda Muresan, Preslav Nakov, Aline Villavicencio
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
6819–6836
Language:
URL:
https://aclanthology.org/2022.acl-long.470
DOI:
10.18653/v1/2022.acl-long.470
Bibkey:
Cite (ACL):
Sidsel Boldsen, Manex Agirrezabal, and Nora Hollenstein. 2022. Interpreting Character Embeddings With Perceptual Representations: The Case of Shape, Sound, and Color. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6819–6836, Dublin, Ireland. Association for Computational Linguistics.
Cite (Informal):
Interpreting Character Embeddings With Perceptual Representations: The Case of Shape, Sound, and Color (Boldsen et al., ACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.acl-long.470.pdf
Software:
 2022.acl-long.470.software.zip
Video:
 https://aclanthology.org/2022.acl-long.470.mp4
Code
 syssel/interpreting-character-embeddings