arXiv cs.CL· Jenny Kunz·· 6 天前AI 评分22
翻译腔的方言如何塑造语言模型学习
Dialects of Translationese Shape Language Model Learning
AI 导读
研究用24种类型学和资源差异悬殊的源语言翻译出的英文文本训练小型英语语言模型，发现源语言会明显影响模型行为：通用困惑度更多由翻译语料的词汇多样性驱动，而语法表现若训练数据充足则与源语言和英语的类型学相似度强相关，翻译质量也是语言建模性能的强预测因子。
正文 · 原文
Abstract:Machine-translated data is widely used in multilingual NLP, particularly where native text is scarce. However, translated text differs systematically from native text. This phenomenon is known as translationese, and it reflects both traces of the source language and characteristic properties of translation itself. In this paper, we study how training on machine-translated data affects small English language models, focusing on how translationese from different source languages shapes linguistic acceptability judgments and language modeling for different domains. We train models on English text translated from 24 typologically and resource-diverse source languages, enabling a systematic analysis of how source language and corpus properties influence what models learn. Our results show that the source language has a clear impact on model behavior: general perplexity is more driven by the lexical diversity of the translated corpus, but grammatical performance is strongly correlated to typological similarity to English if trained on enough data. Even translation quality is a strong predictor of language modeling performance.
| Comments: | To appear at the Findings of EMNLP 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2602.16469 [cs.CL] |
| (or arXiv:2602.16469v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2602.16469 arXiv-issued DOI via DataCite |
Submission history
From: Jenny Kunz [view email]
[v1]
Wed, 18 Feb 2026 13:59:08 UTC (152 KB)
[v2]
Mon, 24 Aug 2026 12:28:51 UTC (172 KB)
[v3]
Fri, 9 Oct 2026 08:00:06 UTC (172 KB)
来源：arXiv cs.CL · arxiv.org