【摘要】 自动作文评分中的核心问题是对高信度的、机器可用的评分特征的选取。大学英语作文自动评分中的一个基本特征是词汇分布,而词汇分布的衡量需要一个描述清晰、刻画准确的大学英语作文分级词表,也就是一个效度较高的词表。实验证明目前已有的词表难以达到这个要求,但通过改进或者调整,可以取得对不同作文质量效度较高的词表。更多还原

【Abstract】 The key in automated essay scoring (AES) is the selection of machine-usable and high-reliable features for scoring. Lexical distribution (LD) is a basic feature in AES for non-English major college students in our country. The scaling of LD needs a word list with high validity, which can only be achieved from a detailed study of the research subject — college English writing. Experiment proves that the widely available word list does not possess the validity, but with an adaptation, a new word list with a high validity for college English writing can be achieved.更多还原