mT5: A Massively Multilingual Pre-trained Text-to-Text Transformer
| dc.contributor.author | Xue, Linting | |
| dc.contributor.author | Constant, Noah | |
| dc.contributor.author | Roberts, Adam | |
| dc.contributor.author | Kale, Mihir | |
| dc.contributor.author | Al‑Rfou, Rami | |
| dc.contributor.author | Siddhant, Aditya | |
| dc.contributor.author | Barua, Aditya | |
| dc.contributor.author | Raffel, Colin | |
| dc.date.accessioned | 2025-06-02T13:31:54Z | |
| dc.date.available | 2025-06-02T13:31:18Z | |
| dc.date.available | 2025-06-02T13:31:54Z | |
| dc.date.issued | 2020-10-22 | |
| dc.description | 영어 위주의 T5를 101개 언어로 확장한 mT5 모델을 제안하며, 다국어 벤치마크에서 우수한 결과를 보여줍니다. 특히 제로샷 번역 시 특정 언어로 전이되는 문제를 완화한 전략도 포함됩니다 ©2020 Google Research | |
| dc.description.abstract | The recent "Text-to-Text Transfer Transformer" (T5) leveraged a unified text-to-text format and scale to attain state-of-the-art results on a wide variety of English-language NLP tasks. In this paper, we introduce mT5, a multilingual variant of T5 that was pre-trained on a new Common Crawl-based dataset covering 101 languages. We detail the design and modified training of mT5 and demonstrate its state-of-the-art performance on many multilingual benchmarks. We also describe a simple technique to prevent "accidental translation" in the zero-shot setting, where a generative model chooses to (partially) translate its prediction into the wrong language. All of the code and model checkpoints used in this work are publicly available. | |
| dc.description.sponsorship | Google Research | |
| dc.identifier.uri | https://arxiv.org/abs/2010.11934 | |
| dc.identifier.uri | http://data.inu.ac.kr/handle/123456789/1960.2 | |
| dc.language.iso | en_US | |
| dc.publisher | arXiv | |
| dc.subject | mT5 | |
| dc.subject | Multilingual NLP | |
| dc.subject | Text-to-Text | |
| dc.subject | Transfer Learning | |
| dc.subject | Zero-shot | |
| dc.title | mT5: A Massively Multilingual Pre-trained Text-to-Text Transformer | |
| dc.type | Article |