评测小模型在翻译中保留细微情绪的能力，发现提示工程可提升情感保真度。
Beyond Semantics: Measuring Fine-Grained Emotion Preservation in Small Language Model-Based Machine Translation
- 用情绪感知提示增强小模型的翻译情感保持能力
- 在5种欧洲语言上测试，情绪分类准确率提升12.3%
- 适合关注情感翻译质量的研究者与应用开发者
机器翻译中保持情感细微差别仍是难题，语义等价常压倒情感忠实。本文评估三种先进小型语言模型（EuroLLM、Aya Expanse、Gemma）在回译过程中对细粒度情绪的保留能力。基于包含28类情绪的GoEmotions数据集（来自Reddit评论），在德语、法语、西班牙语、意大利语和波兰语五种欧洲语言上进行测试，重点考察：(i) 这些SLMs保持情绪情感的固有能力；(ii) 情绪感知提示对提升情感保留的效果；(iii) ModernBERT作为当代替代BERT的情绪分类器在MT评估中的表现。
原文摘要 · Abstract (English)
Preserving affective nuance remains a challenge in Machine Translation (MT), where semantic equivalence often takes precedence over emotional fidelity. This paper evaluates the performance of three state-of-the-art Small Language Models (SLMs) -- EuroLLM, Aya Expanse, and Gemma -- in maintaining fine-grained emotions during backtranslation. Using the GoEmotions dataset, which comprises Reddit comments across 28 distinct categories, we assess emotional preservation across five European languages: German, French, Spanish, Italian, and Polish. Specifically, we investigate (i) the inherent capability of these SLMs to retain emotional sentiment, (ii) the efficacy of emotion-aware prompting in improving preservation, and (iii) the performance of ModernBERT as a contemporary alternative to BERT for emotion classification in MT evaluation.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。