Beyond Sentence-Level End-to-End Speech Translation: Context Helps
| Autor: | Barry Haddow, Biao Zhang, Ivan Titov, Rico Sennrich |
|---|---|
| Přispěvatelé: | University of Zurich |
| Rok vydání: | 2021 |
| Předmět: |
Machine translation
Computer science Speech recognition Concatenation 410 Linguistics Feature selection Context (language use) 000 Computer science knowledge & systems computer.software_genre Robustness (computer science) 10105 Institute of Computational Linguistics Speech translation computer Decoding methods Sentence |
| Zdroj: | ACL/IJCNLP (1) Zhang, B, Titov, I, Haddow, B & Sennrich, R 2021, Beyond Sentence-Level End-to-End Speech Translation: Context Helps . in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) . Online, pp. 2566-2578, The Joint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, Bangkok, Thailand, 1/08/21 . https://doi.org/10.18653/v1/2021.acl-long.200 |
| Popis: | Document-level contextual information has shown benefits to text-based machine translation, but whether and how context helps end-to-end (E2E) speech translation (ST) is still under-studied. We fill this gap through extensive experiments using a simple concatenation-based context-aware ST model, paired with adaptive feature selection on speech encodings for computational efficiency. We investigate several decoding approaches, and introduce in-model ensemble decoding which jointly performs document- and sentence-level translation using the same model. Our results on the MuST-C benchmark with Transformer demonstrate the effectiveness of context to E2E ST. Compared to sentence-level ST, context-aware ST obtains better translation quality (+0.18-2.61 BLEU), improves pronoun and homophone translation, shows better robustness to (artificial) audio segmentation errors, and reduces latency and flicker to deliver higher quality for simultaneous translation. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|