Abstract
This paper presents a winning submission to the SemEval 2022 Task 1 on two sub-tasks: reverse dictionary and definition modelling. We leverage a recently proposed unified model with multi-task training. It utilizes data symmetrically and learns to tackle both tracks concurrently. Analysis shows that our system performs consistently on diverse languages, and works the best with sgns embeddings. Yet, char and electra carry intriguing properties. The two tracks’ best results are always in differing subsets grouped by linguistic annotations. In this task, the quality of definition generation lags behind, and BLEU scores might be misleading.- Anthology ID:
- 2022.semeval-1.8
- Volume:
- Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)
- Month:
- July
- Year:
- 2022
- Address:
- Seattle, United States
- Venue:
- SemEval
- SIG:
- SIGLEX
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 75–81
- Language:
- URL:
- https://aclanthology.org/2022.semeval-1.8
- DOI:
- 10.18653/v1/2022.semeval-1.8
- Cite (ACL):
- Pinzhen Chen and Zheng Zhao. 2022. Edinburgh at SemEval-2022 Task 1: Jointly Fishing for Word Embeddings and Definitions. In Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 75–81, Seattle, United States. Association for Computational Linguistics.
- Cite (Informal):
- Edinburgh at SemEval-2022 Task 1: Jointly Fishing for Word Embeddings and Definitions (Chen & Zhao, SemEval 2022)
- PDF:
- https://preview.aclanthology.org/nodalida-main-page/2022.semeval-1.8.pdf
- Code
- pinzhenchen/unifiedrevdicdefmod