Multiformer: A Head-Configurable Transformer-Based Model for Direct Speech Translation

Gerard Sant, Gerard I. Gállego, Belen Alastruey, Marta Ruiz Costa-jussà


Abstract
Transformer-based models have been achieving state-of-the-art results in several fields of Natural Language Processing. However, its direct application to speech tasks is not trivial. The nature of this sequences carries problems such as long sequence lengths and redundancy between adjacent tokens. Therefore, we believe that regular self-attention mechanism might not be well suited for it. Different approaches have been proposed to overcome these problems, such as the use of efficient attention mechanisms. However, the use of these methods usually comes with a cost, which is a performance reduction caused by information loss. In this study, we present the Multiformer, a Transformer-based model which allows the use of different attention mechanisms on each head. By doing this, the model is able to bias the self-attention towards the extraction of more diverse token interactions, and the information loss is reduced. Finally, we perform an analysis of the head contributions, and we observe that those architectures where all heads relevance is uniformly distributed obtain better results. Our results show that mixing attention patterns along the different heads and layers outperforms our baseline by up to 0.7 BLEU.
Anthology ID:
2022.naacl-srw.34
Volume:
Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: Student Research Workshop
Month:
July
Year:
2022
Address:
Hybrid: Seattle, Washington + Online
Editors:
Daphne Ippolito, Liunian Harold Li, Maria Leonor Pacheco, Danqi Chen, Nianwen Xue
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
277–284
Language:
URL:
https://aclanthology.org/2022.naacl-srw.34
DOI:
10.18653/v1/2022.naacl-srw.34
Bibkey:
Cite (ACL):
Gerard Sant, Gerard I. Gállego, Belen Alastruey, and Marta Ruiz Costa-jussà. 2022. Multiformer: A Head-Configurable Transformer-Based Model for Direct Speech Translation. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: Student Research Workshop, pages 277–284, Hybrid: Seattle, Washington + Online. Association for Computational Linguistics.
Cite (Informal):
Multiformer: A Head-Configurable Transformer-Based Model for Direct Speech Translation (Sant et al., NAACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.naacl-srw.34.pdf
Video:
 https://aclanthology.org/2022.naacl-srw.34.mp4