中国外语

本研究基于翻译比赛语料，利用自然语言处理工具抽取了译文中的词汇丰富度、句法复杂度和语篇衔接等154个汉语语言计量特征，考察了这些特征与英译汉翻译质量之间的关系及其对译文分数的预测力。结果表明：第一，词汇层面的复杂词汇占比、词汇多样性、名词密度及简单词汇占比指标，句法层面的汉语搭配和依存句法特征多样性指标以及篇章层面的中心句子、主语语义类占比、主语密度及代词相关指标与英译汉翻译质量有显著相关关系；第二，逐步线性回归模型的16个指标能够解释译文分数的52.40%，随机森林回归模型的40个指标能够解释译文分数的47.88%，且两个模型在指标选择上有高度一致性。本研究揭示了汉语语言计量特征对译文质量的影响，也为英译汉翻译质量评估提供了有效的研究手段，同时为翻译实践提供了参考和支持。

Based on a corpus of English-to-Chinese translation competition data,this study employs natural language processing tools to extract 154 quantitative linguistic features from the translations,covering lexical richness,syntactic complexity,and discourse cohesion.It investigates the relationship between these features and translation quality,as well as their predictive power for translation scores.The results show that several features are significantly correlated with translation quality.At the lexical level,the proportion of complex words,lexical diversity,noun density,and the proportion of simple words exhibit strong associations with translation scores.At the syntactic level,the diversity of Chinese collocations and dependency structures emerges as key indicators.At the discourse level,features such as the proportion of central sentences,the distribution of semantic categories of subjects,subject density,and pronoun-related metrics also demonstrate significant correlations.Furthermore,a stepwise linear regression model incorporating 16 features accounts for 52.40% of the variance in translation scores,while a random forest regression model using 40 features explains47.88% of the variance.Notably,both models show a high degree of consistency in feature selection.This study highlights the influence of Chinese linguistic features on the quality of English-to-Chinese translations,proposes an effective approach for translation quality assessment based on quantitative linguistic analysis,and offers both theoretical insights and practical support for translation evaluation and practice.