机器翻译质量自动评估系统中若干关键技术的研究

Research on Several Key Technologies in Machine Translation Quality Automatic Evaluation System

本文主要讨论一种机器翻译质量评估系统,该系统能从源文本片段、对应的译文片段、外部资源(语料库、语言模型、主题模型等)以及语言分析处理工具(翻译解码器、部分语音标记等)中提取多种翻译质量特征参数指标,主要包括翻译质量的充分性、复杂性、可信性和流畅性等几个维度的指标。该系统还提出了建立质量评估模型的机器学习的实施建议。

This paper mainly discusses a machine translation quality evaluation system,which can extract a variety of transla-tion quality feature parameter indicators from source text frag-ments,corresponding translation fragments,external resources(corpus,language model,topic model,etc.)and language analysis processing tools(translation decoder,part of speech markers,etc.),mainly including indicators of several dimensions such as sufficiency,complexity,credibility and fluency of translation quality.The system also proposes implementation recommenda-tions for machine learning to establish a quality assessment model.