摘要： 针对传统主观题自动评分准确度低的问题，提出一种基于文本相似度计算的主观题评分方法。利用扩展的《同义词词林》计算词语之间的相似度，根据标准答案中的词语和学生答卷中的词语以及词语之间的相似度构造二部图，通过二部图的最大匹配算法获得标准答案和学生答案的相似度。实验结果表明，该方法可以给主观题评分提供一个较好的参考。
关键词:
自动评分,
文本相似度,
二部图,
匈牙利算法,
最大匹配
Abstract: To solve the problem of low precision rate in the traditional automated assessment of subjective tests, this paper improves a method to estimate subject tests. Words of semantic similarity computing are based on “TongYiCi CiLin”. Bipartite graph is constructed according to calculating similarity of words in standard answer and student’s answer. And the biggest matching algorithm is used to compute similarity between texts. Experimental results show that precision and recall rate are greatly improved.
Key words:
automated assessment,
text similarity,
bipartite graph,
Hungarian algorithm,
maximum matching
中图分类号:
程传鹏?, 齐晖. 文本相似度计算在主观题评分中的应用[J]. 计算机工程, 2012, 38(5): 288-290.
CHENG Chuan-Feng-?, JI Hui. Application of Text Similarity Computing in Subjective Question Assessment[J]. Computer Engineering, 2012, 38(5): 288-290.