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 2302.06459cs.CL
收藏

在基于拼接的上下文感知神经机器翻译中编码句子位置

Encoding Sentence Position in Context-Aware Neural Machine Translation with Concatenation

Université Grenoble Alpes(格勒诺布尔阿尔卑斯大学);Naver Labs Europe(Naver欧洲实验室)

Lorenzo Lupo, Marco Dinarelli, Laurent Besacier

AI总结:

本文研究在基于拼接的上下文感知机器翻译中,通过多种方法将句子位置显式编码至 token 表示,发现特定编码方法在上下文折扣损失下可提升英译俄效果,但英译德未见相同益处。

摘要:

上下文感知翻译可通过使用标准 Transformer 架构处理连续句子的拼接来实现。本文研究了向模型提供拼接窗口中句子位置显式信息这一直观想法。我们比较了将句子位置编码到 token 表示中的各种方法,包括新方法。结果表明,若使用上下文折扣损失(Lupo et al., 2022)进行训练,Transformer 在英译俄任务中能从某些句子位置编码方法中获益。然而,在英译德任务中未观察到相同益处。需进一步实证研究以界定该方法有效的条件。

英文摘要:

Context-aware translation can be achieved by processing a concatenation of consecutive sentences with the standard Transformer architecture. This paper investigates the intuitive idea of providing the model with explicit information about the position of the sentences contained in the concatenation window. We compare various methods to encode sentence positions into token representations, including novel methods. Our results show that the Transformer benefits from certain sentence position encoding methods on English to Russian translation if trained with a context-discounted loss (Lupo et al., 2022). However, the same benefits are not observed in English to German. Further empirical efforts are necessary to define the conditions under which the proposed approach is beneficial.

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。

↑