【摘要】 通过集成学习方法,探索影响英语学习者作文质量的语言特征,提高现有作文自动评分系统的准确率。基于剑桥FCE考试数据集,使用支持向量回归和随机森林算法筛选特征,构建并评测自动评分模型。与现有方法对比实验的结果表明,基于集成学习的评分模型准确率有所提升。该方法可以有效评估英语学习者的作文质量,有助于开发面向大规模机考和网络自主学习平台的作文自动评阅系统。更多还原

【Abstract】 This paper aims to find the linguistic features that affect the quality of English learners’ writing and to improve the accuracy of current automatic essay scoring systems through the ensemble learning. Based on the Cambridge FCE test dataset, support vector regression and random forest were used to select features, and the automatic scoring model was constructed and evaluated. Compared with the existing methods, the accuracy of the scoring model based on ensemble learning is improved. It can effectively evaluate the quality of English learner’s essay and help to develop the automatic essay scoring systems for large-scale computer-based tests and online self-learning platforms.更多还原