基于中文分词的主观题自动评分算法研究  

Automatic Scoring Algorithm for Subjective Questions Based on Chinese Word Segmentation

作　　者：宋雪亚[1] 王传安[1] 

机构地区：[1]安徽科技学院,安徽凤阳233100

出　　处：《河北北方学院学报（自然科学版）》2017年第9期7-11,共5页Journal of Hebei North University:Natural Science Edition

基　　金：安徽科技学院校级一般项目:"互联网+技术下的无线移动视频智能监控技术研究"(ZRC2016495)

摘　　要：目的为了对在线考试系统中主观题进行更合理的评分,提出一种基于中文分词的算法对主观题进行评分。方法对中文分词进行了详细介绍,并对已有的算法进行研究和改进,利用基于中文分词技术并结合文本相似度对主观题进行自动评分,从文本串长度相似度、文本串词形相似度和文本串词序相似度,再结合影响因子,形成最终的综合相似度。结果通过综合考虑考试科目的特征,合理的设置3个相对影响因子的值,对试卷通过4个实验进行测试,试卷题目分别为4个Office简答题,标准答案控制在100字内,每个实验回收电子试卷50份,与使用原算法的实验结果进行比对。实验测试表明,优化后的算法准确率有了很大提高。结论优化后的算法准确率有明显提高,在词形相似度较高的情况下评分效果与原算法差距不大,依然有改进的空间。Objective To improve the scoring validity of subjective test questions,a Chinese word segmentation algorithm was presented.Methods The algorithm,proposed after detailed introduction of Chinese word segmentation and literature review of the previous algorithms,helped to realize automatic scoring of subjective test questions.The similarities of text string length,text string formation and text string sequence,combined with the impact factor,led to the final comprehensive similarity.Results Fifty experiment papers,which included same four Office short answer questions with answers no more than one hundred words,were scored with both proposed algorithm and unmodified algorithm,with values of the three relative influencing factors properly set according to the characteristics of the test subject.Conclusion The accuracy of the optimized algorithm is obviously improved.In the condition of higher similarity,the scoring effect is not much different from that of the original algorithm,and there is still room for improvement.

关 键 词：中文分词 自动评分 文本相似度 

分 类 号：TP306.1[自动化与计算机技术—计算机系统结构]