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 2410.06338cs.CL
收藏

大语言模型是用户生成内容机器翻译的最先进质量评估器吗?

Are Large Language Models State-of-the-art Quality Estimators for Machine Translation of User-generated Content?

Centre for Translation Studies(翻译研究中心);Institute for People-Centred AI(以人为本人工智能研究院);University of Surrey(萨里大学)

Shenbin Qian, Constantin Orăsan, Diptesh Kanojia, Félix do Carmo

AI总结:

本文探究大语言模型能否作为无参考译文的用户生成内容情感类机器翻译的最优质量评估器,实验显示参数高效微调的LLM预测表现更优且可解释,但仍存在拒答、输出不稳定问题。

摘要:

本文研究在不使用参考译文的情况下,大语言模型(LLMs)是否可作为含情感表达的用户生成内容(UGC)机器翻译的最先进质量评估器。为此,我们采用一个已有的带人工标注错误的情感相关数据集,并基于多维度质量指标(Multi-dimensional Quality Metrics)计算质量评估分数。我们在上下文学习和参数高效微调(PEFT)场景下,对比了若干LLMs与我们的微调基线模型的准确率。研究发现,LLMs的PEFT在分数预测上表现优于微调模型,且能提供人类可解释的说明。然而,对LLM输出的人工分析显示,它们在评估UGC机器翻译时仍存在拒绝响应提示、输出不稳定等问题。

英文摘要:

This paper investigates whether large language models (LLMs) are state-of-the-art quality estimators for machine translation of user-generated content (UGC) that contains emotional expressions, without the use of reference translations. To achieve this, we employ an existing emotion-related dataset with human-annotated errors and calculate quality evaluation scores based on the Multi-dimensional Quality Metrics. We compare the accuracy of several LLMs with that of our fine-tuned baseline models, under in-context learning and parameter-efficient fine-tuning (PEFT) scenarios. We find that PEFT of LLMs leads to better performance in score prediction with human interpretable explanations than fine-tuned models. However, a manual analysis of LLM outputs reveals that they still have problems such as refusal to reply to a prompt and unstable output while evaluating machine translation of UGC.

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。

↑