来华留学预科汉语考试作文评分研究--基于概化理论和多面Rasch模型  

An Analysis of Composition Scoring in Preparatory Chinese Examination for International Students:Based on Generalizability Theory and Many-facet Rasch Model

作　　者：孔傅钰 Kong Fuyu(Beijing Language and Culture University,Beijing,100083)

机构地区：[1]北京语言大学国际学生教育政策与评价研究院,北京100083

出　　处：《考试研究》2022年第4期41-51,共11页Examinations Research

基　　金：国家社科重点项目“基于HSK大数据挖掘的汉语习得研究”(17AYY011);北京语言大学中外研究生创新基金项目“基于概化理论和多面Rasch模型的来华留学汉语预科考试作文评分研究”(21YCX189)的工作成果

摘　　要：为了探究来华留学生预科汉语教育结业汉语综合统一考试的作文评分信度,本研究采用概化理论和多面Rasch模型分析5名评分员对120篇实考作文样本的评分情况。概化理论的研究表明:考生能力是得分总变异的最大来源,一位评分员进行评分时,其结果即可达到可接受的概化系数;两位评分员进行评分时,信度系数提高的幅度最大,因此应保持目前的双评状态。多面Rasch模型的分析显示:评分量表基本能区分考生能力,评分员的严厉性差异显著,存在对高水平考生偏严而对低水平考生偏宽松的趋势,个别评分员自身一致性较差。In order to explore the reliability of the composition scoring of the Preparatory Chinese Examination for International Students,this study uses the Generalizability Theory and the Many-facet Rasch Model to analyze the scores given by five raters of 120 actual test compositions.The research of Generalizability Theory shows that the Chinese ability of test takers is the biggest source of the total variation of scores.When one rater scores,the results can have the acceptable generalizability coefficient.When two raters score,the reliability coefficient increases the most.Therefore,the current state of double rating should be maintained.Many-facet Rasch Model analysis shows that the rating scale can basically distinguish the ability of test takers and a significant difference in the severity of raters exits.There is a tendency to be stricter for high-level test takers and looser for low-level ones,and the consistency of individual raters is poor.

关 键 词：作文评分 汉语预科考试 概化理论 多面RASCH模型 

分 类 号：G424.74[文化科学—课程与教学论]