潜在语义分析技术在自动评卷系统中的应用

提出了一种基于潜在语义分析（LSA）的相似文本匹配算法，并将其应用于自动评卷系统中．首先，在充分考虑词项之间相关性的基础上，在低维空间中表示学生答案文本与标准答案文本，然后利用奇异值分解方法模型对其进行了改进；其次，利用LSA技术，以学生答案文本与标准答案文本之间的余弦相似度作为相似性准则，根据相似度值确定该题的得分．实验结果表明，该算法充分考虑了文本语义信息，评分效果较好，是实现基于语义评卷系统的有益探索．

Application of Latent Semantic Analysis in Auto-Grading System

Based on the method of latent semantic analysis （LSA）, a text matching algorithm was proposed for applying to automatic grading system. Firstly, by fully considering the correlation between terms, texts of ex- aminee~s answers and standard answers were represented in lower-dimensional space and the model was im- proved using the way of singular value decomposition. Secondly, using LSA, the cosine similarity between the texts of examinee~s answers and standard answers was taken as similarity criterion to determine the score of answer of each examination question. Experimental results show that by considering the semantic information of text, the proposed algorithm has satisfactory scoring results, and the work presented in this paper is a bene- ficial exploration for implementing semantic based automatic grading system.