采用多面Rasch模型提高英语运用测试的评分质量

Multifaceted Rasch Model in the Rating of English Performance Assessment

摇目前电大系统英语考试的口试和作文部分多采用语言运用测试的方式。语言运用测试由于引入评分者而使评分的主观性变大。如何控制评分者差异对考生分数的影响成为保证语言运用测试评分质量的重要环节。本文在比较了行为测试中评分质量控制方面常用的三种理论的基础上，着重介绍了多面Rasch模型在提高评分质量方面的贡献，并探讨了在电大系统如何采用该模型对英语运用测试中的评分者进行培训，以控制评分质量和提高考试信度。

Radio & TV University (RTVU) system in China adopted language performance assessment to test students' oral and written proficiency. Due to the use of human raters in scoring the test, how to control the impact of rater variations on scoring has become a critical issue. This paper introduces three commonly used approaches in measuring rater performance in performance assessment with a focus on multifaceted rasch model. In addition to introducing the advantages that this model brings to rater training, the paper discusses the possibility and benefits of using the model in the practice of RTVU rater training. Using this model might have to enhance the rating quality in performance assessment of English language proficiency.