Abstract
Positional Encodings (PEs) are used to inject word-order information into transformer-based language models. While they can significantly enhance the quality of sentence representations, their specific contribution to language models is not fully understood, especially given recent findings that various positional encodings are insensitive to word order. In this work, we conduct a systematic study of positional encodings in Bidirectional Masked Language Models (BERT-style) , which complements existing work in three aspects: (1) We uncover the core function of PEs by identifying two common properties, Locality and Symmetry; (2) We show that the two properties are closely correlated with the performances of downstream tasks; (3) We quantify the weakness of current PEs by introducing two new probing tasks, on which current PEs perform poorly. We believe that these results are the basis for developing better PEs for transformer-based language models.- Anthology ID:
- 2023.findings-emnlp.955
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2023
- Month:
- December
- Year:
- 2023
- Address:
- Singapore
- Editors:
- Houda Bouamor, Juan Pino, Kalika Bali
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 14313–14331
- Language:
- URL:
- https://aclanthology.org/2023.findings-emnlp.955
- DOI:
- 10.18653/v1/2023.findings-emnlp.955
- Cite (ACL):
- Lihu Chen, Gael Varoquaux, and Fabian Suchanek. 2023. The Locality and Symmetry of Positional Encodings. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 14313–14331, Singapore. Association for Computational Linguistics.
- Cite (Informal):
- The Locality and Symmetry of Positional Encodings (Chen et al., Findings 2023)
- PDF:
- https://preview.aclanthology.org/emnlp-22-attachments/2023.findings-emnlp.955.pdf