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.
@inproceedings{DBLP:conf/starsem/SlobodkinCA22,
author = {Aviv Slobodkin and
Leshem Choshen and
Omri Abend},
editor = {Vivi Nastase and
Ellie Pavlick and
Mohammad Taher Pilehvar and
Jos{\'{e}} Camacho{-}Collados and
Alessandro Raganato},
title = {Semantics-aware Attention Improves Neural Machine Translation},
booktitle = {Proceedings of the 11th Joint Conference on Lexical and Computational
Semantics, *SEM@NAACL-HLT 2022, Seattle, WA, USA, July 14-15, 2022},
pages = {28--43},
publisher = {Association for Computational Linguistics},
year = {2022},
url = {https://doi.org/10.18653/v1/2022.starsem-1.3},
doi = {10.18653/V1/2022.STARSEM-1.3},
timestamp = {Mon, 26 Sep 2022 16:58:59 +0200},
biburl = {https://dblp.org/rec/conf/starsem/SlobodkinCA22.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
See the full reference list in the paper.