arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2412.02101cs.CL
收藏

提升多语言神经机器翻译中 Decoder-only 架构的语言迁移能力

Improving Language Transfer Capability of Decoder-only Architecture in Multilingual Neural Machine Translation

Nara Institute of Science and Technology(奈良科学技术大学);National Institute of Information and Communications Technology(日本信息通信研究机构)

Zhi Qu, Yiran Wang, Chenchen Ding, Hideki Tanaka, Masao Utiyama, Taro Watanabe

AI总结:

本文针对 decoder-only 架构在多语言神经机器翻译中语言迁移不足的问题,提出两阶段解码与翻译指令对比学习,在零样本翻译上显著超越 encoder-decoder 基线。

摘要:

现有的多语言神经机器翻译(MNMT)方法主要聚焦于改进采用 encoder-decoder 架构的模型,以翻译多种语言。然而,decoder-only 架构在 MNMT 中探索较少,原因是仅使用平行数据训练时其表现不佳。在本研究中,我们将 decoder-only 架构的问题归因于其缺乏语言迁移能力。具体而言,decoder-only 架构在利用目标语言特征编码源语言 token 方面存在不足。我们提出将解码过程划分为两个阶段,使目标 token 在第一阶段被显式排除,从而隐式增强跨语言迁移能力。此外,我们对翻译指令施加对比学习,从而提升零样本翻译性能。我们在 TED-19 和 OPUS-100 数据集上进行实验,同时考虑从头训练和微调两种场景。实验结果表明,与 encoder-decoder 架构相比,我们的方法不仅在有监督翻译中表现具有竞争力,还在零样本翻译中取得最高 3.39 BLEU、6.99 chrF++、3.22 BERTScore 和 4.81 COMET 的提升。

英文摘要:

Existing multilingual neural machine translation (MNMT) approaches mainly focus on improving models with the encoder-decoder architecture to translate multiple languages. However, decoder-only architecture has been explored less in MNMT due to its underperformance when trained on parallel data solely. In this work, we attribute the issue of the decoder-only architecture to its lack of language transfer capability. Specifically, the decoder-only architecture is insufficient in encoding source tokens with the target language features. We propose dividing the decoding process into two stages so that target tokens are explicitly excluded in the first stage to implicitly boost the transfer capability across languages. Additionally, we impose contrastive learning on translation instructions, resulting in improved performance in zero-shot translation. We conduct experiments on TED-19 and OPUS-100 datasets, considering both training from scratch and fine-tuning scenarios. Experimental results show that, compared to the encoder-decoder architecture, our methods not only perform competitively in supervised translations but also achieve improvements of up to 3.39 BLEU, 6.99 chrF++, 3.22 BERTScore, and 4.81 COMET in zero-shot translations.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑