【摘要】 文章首先回顾了信度和效度的概念以及检测信度和效度的方法,以此为依据,将收集到的电脑评分和专家人工评分的数据进行了相关性分析、信度检验、重复性方差分析、独立样本t检验以及定性分析等各项分析,多方位地多元评分系统的信度和效度进行了验证。结果表明,本系统具有良好的内部一致性,信度较好,但是初评分比例较高时,信度较低;与专家评分的结果对比研究表明,自动评分系统结果对说明文和应用文体两种文体写作能力解释力较差。更多还原

【Abstract】 The paper sets out to analyze the reliability and validity of automated essay scoring with E- scorer.It begins with a review on previous empirical researches on the reliability and validity of automated essay scoring systems. The evidence based approach is used in this study to examine the reliability of automated scoring and compare the automated and human scores. The relevance test,reliability statistics,repeated measures and independent samples T- test show that the automated scoring system is highly reliable when the proportion of preliminary scores is below 50%. However,the statistical and quality analyses show that the scoring results from automated scoring system are not able to provide good interpretation for practical and informative writing ability.更多还原