Abstract
Modern natural language understanding models depend on pretrained subword embeddings, but applications may need to reason about words that were never or rarely seen during pretraining. We show that examples that depend critically on a rarer word are more challenging for natural language inference models. Then we explore how a model could learn to use definitions, provided in natural text, to overcome this handicap. Our model’s understanding of a definition is usually weaker than a well-modeled word embedding, but it recovers most of the performance gap from using a completely untrained word.- Anthology ID:
- 2021.starsem-1.27
- Volume:
- Proceedings of *SEM 2021: The Tenth Joint Conference on Lexical and Computational Semantics
- Month:
- August
- Year:
- 2021
- Address:
- Online
- Editors:
- Lun-Wei Ku, Vivi Nastase, Ivan Vulić
- Venue:
- *SEM
- SIG:
- SIGLEX
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 288–293
- Language:
- URL:
- https://aclanthology.org/2021.starsem-1.27
- DOI:
- 10.18653/v1/2021.starsem-1.27
- Cite (ACL):
- Christopher Malon. 2021. Overcoming Poor Word Embeddings with Word Definitions. In Proceedings of *SEM 2021: The Tenth Joint Conference on Lexical and Computational Semantics, pages 288–293, Online. Association for Computational Linguistics.
- Cite (Informal):
- Overcoming Poor Word Embeddings with Word Definitions (Malon, *SEM 2021)
- PDF:
- https://preview.aclanthology.org/naacl24-info/2021.starsem-1.27.pdf
- Data
- SNLI