面向地理课程自动解题的试题理解技术研究

发布时间：2018-10-23 20:55

【摘要】：作为当前人工智能和自然语言处理领域的研究热点之一,自动解题是利用计算机自动解答相关课程的试题。试题理解技术是让计算机自动去理解文本含义,通过对文本的词法分析、句法分析及语义理解,从而达到计算机自动解题效果。目前的自动解题研究主要针对数学类课程,对于其他类课程则研究较少。地理课程虽然作为文科类学科,但含有丰富的知识点,“上知天文,下知地理”也说明了地理课程的重要性,本文主要针对自动解题的试题理解技术进行研究,其研究内容如下:(1)本文针对地理课程试题进行分类,将SVM算法应用在地理试题分类领域,利用LIBSVM分类包中的Linear核函数对试题文本进行学习训练,将TFIDF提取关键词,生成特征向量,构建分类模型进行测试。在所收集的地理试题集上的实验结果表明,利用Linear核函数分类效果较好,其分类精度达到80%以上,实现了文本分类算法在地理课程领域中的应用。(2)本文针对地理课程自动解题,在构建地理实体本体的基础上,通过对试题题干及选项进行语义分析,进一步获取概念实体间的语义关系,并将这种语义关系转化为用于试题解答的解题规则。在所收集的地理试题集上的实验结果表明,本文获取的解题规则对文字类试题的解答有明显的提升作用,对图表类试题的解答也有一定的辅助作用。
[Abstract]:As one of the research hotspots in the field of artificial intelligence and natural language processing, automatic problem solving is the use of computer to solve the test questions of related courses automatically. The technique of question comprehension is to make the computer understand the meaning of text automatically. By analyzing the lexical, syntactic and semantic meaning of the text, the computer can solve the problem automatically. At present, the research on automatic problem solving is mainly aimed at mathematics courses, but less on other courses. Although geography course is a liberal arts subject, it contains rich knowledge points. "knowing astronomy on top, knowing geography below" also explains the importance of geography course. The research contents are as follows: (1) this paper classifies the geography course test questions, applies the SVM algorithm to the geography examination questions classification field, uses the Linear kernel function in the LIBSVM classification package to study and train the text of the test questions, and extracts the key words from the TFIDF. The feature vector is generated and the classification model is constructed for testing. The experimental results on the collected set of geography test questions show that the classification accuracy of Linear kernel function is better than 80%, and the text classification algorithm is applied in the field of geography course. (2) this paper solves the problem automatically for geography course. Based on the construction of geographical entity ontology, the semantic relations between conceptual entities are further obtained by semantic analysis of questions and options, and the semantic relations are transformed into problem solving rules for test questions. The experimental results on the collection of geographical test questions show that the rules obtained in this paper can obviously improve the solution of the text test questions, and also have some auxiliary effect on the solution of the chart type test questions.
【学位授予单位】：安徽工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1;TP18

【参考文献】