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 2409.11274cs.CLcs.AI
收藏

面向语音翻译中语言扩展的任务算术方法

Task Arithmetic for Language Expansion in Speech Translation

University of Washington(华盛顿大学);Sony Group Corporation(索尼集团);University of Electro-Communications(电通大学);Carnegie Mellon University(卡内基梅隆大学)

Yao-Fei Cheng, Hayato Futami, Yosuke Kashiwagi, Emiru Tsunoo, Wen Shen Teo, Siddhant Arora, Shinji Watanabe

AI总结:

本研究针对语音翻译扩展语言对成本高的问题,提出融合语言控制模型的增强型任务算术方法,无需重训即可构建一对多ST系统,在多数据集上取得显著性能提升,还可扩展至无配对数据的语言对。

摘要:

大语言模型(LLMs)的最新进展引发了人们对语音-文本多模态基础模型的研究兴趣,这类模型在指令微调语音翻译(ST)任务上取得了优异性能。然而,由于需要在新旧组合数据集上重新训练,扩展语言对的成本很高。为解决这一问题,我们旨在利用任务算术方法,基于现有的一对一ST系统构建一对多ST系统,且无需重新训练。直接在ST中应用任务算术会导致语言混淆;因此,我们提出一种增强型任务算术方法,引入语言控制模型以确保生成正确的目标语言。我们在MuST-C和CoVoST-2数据集上的实验显示,BLEU分数分别提升了最高4.66和4.92,COMET分数分别提升8.87和11.83。此外,我们证明该框架可通过任务类比,基于现有机器翻译(MT)和ST模型合成ST模型,从而扩展到缺乏配对ST训练数据或预训练ST模型的语言对。

英文摘要:

Recent progress in large language models (LLMs) has gained interest in speech-text multimodal foundation models, achieving strong performance on instruction-tuned speech translation (ST). However, expanding language pairs is costly due to re-training on combined new and previous datasets. To address this, we aim to build a one-to-many ST system from existing one-to-one ST systems using task arithmetic without re-training. Direct application of task arithmetic in ST leads to language confusion; therefore, we introduce an augmented task arithmetic method incorporating a language control model to ensure correct target language generation. Our experiments on MuST-C and CoVoST-2 show BLEU score improvements of up to 4.66 and 4.92, with COMET gains of 8.87 and 11.83. In addition, we demonstrate our framework can extend to language pairs lacking paired ST training data or pre-trained ST models by synthesizing ST models based on existing machine translation (MT) and ST models via task analogies.

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。

↑