摘要
自动作文评分的标准除内容外,更重要的是语言的运用。中国大学生英语作文在语言运用上尤具自身的特点。为了探索该类文本的有效的自动评分方法,本文采用中国英语学习者语料库中大学英语四级考试作文子库的部分作文为样本,针对语言使用提取特征,采用多元回归、KNN和SVM方法分别训练评分模型并进行评分实验,结果表明多元回归方法能最准确地模拟人工评分,但KNN方法更适合日常教学使用。
The criteria of automated essay scoring include content and,more importantly,the use of language.Chinese college students＇ English writing has its unique attributes in language use.In order to find out an efficient automated scoring method for this type of text,in this study some essays from sub-corpus st3 of CLEC were used as the sample,attributes of language use were extracted,techniques of multi-regression,KNN,and SVM were adopted to train a scoring model respectively,and experiments of scoring were conducted.The result shows that the technique of multi-regression can reach the highest scoring precision while the technique of KNN is more suitable for daily teaching.
出处
《广东外语外贸大学学报》
2010年第3期87-90,共4页
Journal of Guangdong University of Foreign Studies
基金
广东外语外贸大学2009年度校级青年项目(面向大学英语教学的自动作文评分和反馈研究)
关键词
大学英语作文
自动作文评分
评分方法
college English writing
automated essay scoring
scoring technique
