中国学生大规模汉译英测试机助评分模型的研究与构建

A model for the computer-assisted scoring of Chinese EFL learners' Chinese-English translation

以往的翻译自动评价系统局限于机器译文评判,缺乏人工译文评价系统。本研究旨在研制信度可靠、运行稳定的汉英人工译文自动评分系统,实现大规模汉译英测试的评分自动化。本研究构建了三种比例训练集的大规模测试评分模型,模型预测分值与人工评分的相关系数均高于0.85,当训练集达到100篇时,模型预测分值与人工评分基本一致,不存在显著性差异。研究结果表明,本研究提取的变量预测能力较强,机助测试评分模型表现良好,能够比较准确地预测中国二语学习者的汉译英成绩。

Although there are multiple algorithms applied in the evaluation of machine translation,no system is available for human translation.This study aims to construct a reliable and stable statistical model for the computer-assisted scoring of Chinese EFL learners' Chinese-English (C-E) translation.It is hoped that the model proposed in this study,once implemented in the form of a computer program,can be used to score C-E translation papers in large-scale examinations.This study constructed 3 tentative selection models with different sizes of training sets.The result of the Paired-Samples T-Test showed that the computer scores produced by the 100 training set and the 150 training set bore no statistically significant difference from the human scores.Statistical results indicate that the variables extracted in this research are very effective with high predictive power,and the selection models can produce reliable scores for Chinese EFL learners' C-E translation.