Abstract
The integration of syntactic structures into Transformer machine translation has shown positive results, but to our knowledge, no work has attempted to do so with semantic structures. In this work we propose two novel parameter-free methods for injecting semantic information into Transformers, both rely on semantics-aware masking of (some of) the attention heads. One such method operates on the encoder, through a Scene-Aware Self-Attention (SASA) head. Another on the decoder, through a Scene-Aware Cross-Attention (SACrA) head. We show a consistent improvement over the vanilla Transformer and syntax-aware models for four language pairs. We further show an additional gain when using both semantic and syntactic structures in some language pairs.- Anthology ID:
- 2022.starsem-1.3
- Volume:
- Proceedings of the 11th Joint Conference on Lexical and Computational Semantics
- Month:
- July
- Year:
- 2022
- Address:
- Seattle, Washington
- Editors:
- Vivi Nastase, Ellie Pavlick, Mohammad Taher Pilehvar, Jose Camacho-Collados, Alessandro Raganato
- Venue:
- *SEM
- SIG:
- SIGLEX
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 28–43
- Language:
- URL:
- https://aclanthology.org/2022.starsem-1.3
- DOI:
- 10.18653/v1/2022.starsem-1.3
- Cite (ACL):
- Aviv Slobodkin, Leshem Choshen, and Omri Abend. 2022. Semantics-aware Attention Improves Neural Machine Translation. In Proceedings of the 11th Joint Conference on Lexical and Computational Semantics, pages 28–43, Seattle, Washington. Association for Computational Linguistics.
- Cite (Informal):
- Semantics-aware Attention Improves Neural Machine Translation (Slobodkin et al., *SEM 2022)
- PDF:
- https://preview.aclanthology.org/emnlp-22-attachments/2022.starsem-1.3.pdf
- Data
- Universal Dependencies