基于链语法的英语作文自动评分研究

Research on Automated English Essay Scoring Based on Link Grammar

分析研究当前国内外英语作文自动评分系统的研究成果,提出了一种基于链语法(Link Grammar)的能为学生提供实时训练的英语作文自动评分系统。系统首先通过依据Link Grammar对句子分析后的成本向量(cost vector)计算文章的语法权值,其次则利用文章核心词、非核心词熵值和有效句比例来对文章主题分档,最后则是结合语法权值和主题分档结果计算文章分数。测试结果表明,10~15分段作文达到了88.9%的准确率,说明该系统在10~15分段作文具有较好的适应性,同时系统的分析时间也能很好的满足实时性的要求。

An automated essay scoring(AES) system based on the Link Grammar is proposed after analyzing the relative results on AES both at home and abroad, which can provide real-time writing training for English learners. Firstly, the grammar's weight of this article is calculated on the basis of cost vector which is the sentence's parsing result of Link Grammar. Then the topic of this article is partitioned using the entropy of keywords, the entropy of other words except keywords and the proportion of effective sentences. Finally, the points of the article is calculated by the grammar's weight and the topic partition of the article. Tests' results show that the system has achieved 88.9% accuracy, which indicates the system has a good adaptability for these articles whose point is range from 10 points to 15 points. Besides, the parsing time of the system can meet the demand of real-time performance.