基于GloVe-CNN算法的英语在线考试主观题自动评分模型  

Automatic scoring model of English online examination subjective questions based on GloVe-CNN algorithm

作　　者：黎秋艳[1] 刘佳祎 王鹏 王杰 LI Qiuyan;LIU Jiayi;WANG Peng;WANG Jie(College of Information Technology,Guilin Universiey of Electronic Technology,Guilin 541004,China;Network and Information Center,Guilin University of Technology,Guilin 541006,China;Network Information Center,Guangxi Normal University,Guilin 541006,China)

机构地区：[1]桂林电子科技大学信息科技学院,广西桂林541004 [2]桂林理工大学网络与信息中心,广西桂林541006 [3]广西师范大学网络信息中心,广西桂林541006

出　　处：《桂林理工大学学报》2023年第1期155-160,共6页Journal of Guilin University of Technology

基　　金：国家自然科学基金地区项目(62166004);广西哲学社会科学规划研究课题(21FGL040)。

摘　　要：针对英语在线考试系统的主观题评分本身具有的复杂性,引入单词表达式的词向量模型global vector(GloVe)与卷积神经网络(CNN)构建文本表达式特征,提出了基于GloVe-CNN算法的主观题自动评阅模型,以实现英语在线考试系统主观题自动评分。通过平方加权Kappa评价指标,与人工阅卷分数差对比发现,该模型在评分的一致性、准确性等指标上要优于卷积神经网络、KNN模型等传统的模型,整体性能较好,可进一步减轻教师的教学负担。The scoring of subjective questions in the English online exam system is complicated.Therefore,the word vector model of word expression global vector(GloVe)and convolutional neural network(CNN)are introduced to construct text expression features.A subjective question automatic scoring model based on the GloVe-CNN algorithm is proposed to achieve automatic scoring of subjective questions in the English online exam system.The model is evaluated by square weighted Kappa evaluation index and the difference between manual scoring.The experimental results show that the model outperforms traditional models such as convolutional neural networks and KNN models in terms of consistency and accuracy in grading.The overall performance is good,which can further reduce the teaching burden on teachers.

关 键 词：英语在线考试系统 自动评分 词向量 卷积神经网络 

分 类 号：T391.1[一般工业技术]