主观试题自动批改技术作为自然语言理解中重要课题之一，对于学生自侧作业、大规模在线考试方面具有非常重要的意义。目前针对客观题的自动批改技术己经相当成熟，但是，对于论述、简答、写作等没有统一答案的主观试题，涉及到人工智能、模式识别和自然语言理解诸多方面，又由于汉语的特点，在实现上有其自勇的困难，相对来说比较复余一直制约中文主观试题自动批改技术的发展、因此。研究如何实现中文主观题的自动批改具有很重要的现实意义。 本文根据汉语理解的特点，模拟教师入工批改的过程，通过对自动批改中涉及到的中文知识表示、分词方法、旬法分析、语义相似度等方面的研究，利用统计与规则相结合分词方法、改进线图分析算法、非线性函数和“成对比较法”相结合的关键词权值汁算方法和加权语义相似度计算方法分析和实现主观试题的自动批改。在语义相似度计算方面，利用知网提供的语义信息，从义原到概念、从概念到词语的相似度计算方怯，根据词语相似度的计算结果，利用词语加权语义相似度计算句于的相似度，最后将相似度的值转化为最终的分数。根据以上自动批改的思想，设计出一个主观题自动批改试验系统行测试，验证本文分析算法可行有效，达到一定批改效果。

Assessment t∞lmology ofsubjective tests based on Chinese is one ofthe co他problems in Natural Language Understanding．massive online examil埘OllS has very important significance,the objective view that the automatic marking technology has been very mature． But,when refers to the subject questio璐，such as short al'lSWer questions、essay questions and writing etc，is no single a／iswer to the questions，because the subject questions’ atRo-asscs8 techniqms are restricted by the development of the techniques in Artificial Intelligence area,the Natural Language Understanding area andthe Pattern Recognition area, and SO 011 its own involving Chinese language featur鼯，the realization has its own difficulties and characteristics of relatively achieve is relatively complex,This,to a certain extent, constrains automated assessment technology of subjective tests based on Chinese．Therefore, the study of how the automated assessment technology of subjet,re tests based on Clline∞ that has great practical significance． Based 011 the understanding of the characteristics of Chinese．1his paper simulates in manual process，computer automatecl assessment technology should take On Chinese automatic word segmentation,knowledge denotation,syntax analysis，semantic similarity,to realize that Subjective Qu妇Auto-assess．Analysis ofthe process ofusing role-based and corpus based statistical methods，improvements On Chinese syntax parsing have given by adopting methods of both bottom l|p and top down,the non-linear function and the"double compafing method"to calculate the term weiglmng。in terms of semantic smgarity,谢th the abundant semantic information supplied by HowNet,It calculates the sentence similarity with weighted semantic simiL,Ⅱity based on Chinese words semantic similarity,From primary to the original concepts and terms from concept to the similarity calculation method．Then we can get the final scores depend on the value ofsimilarity． According to the above idea,n把paper designs a Subjective Question Auto-assess system and applies it into small—scale Chinese text,from the test results,it proves that the algorithm mentioned above is efficiency and feasibility．The auto-assess result was pretty good．