【作 者】：
吴林静， 刘清堂， 毛 刚， 黄焕， 黄景修
【关 键 词 】：
慕课；课程评论；语义特征；分类；语义挖掘
【栏 目】：
教育大数据
【中文摘要】：
文章从大数据的视角，应用语义分析的方法对慕课评论进行了分析和挖掘。针对慕课评论数量大且信息混杂的特点，文章提出了一种面向大数据的慕课评论语义分析模型。在该模型中，慕课评论被分为三种主要的类别：内容相关类、情感相关类和其他类。针对不同类别的差异，文章提出了基于词类的语义特征用于对评论进行表征和分类。以爱课程慕课上的四门课程评论作为实例进行分析发现：（1）以词类作为语义特征进行评论分类，单课程内部分类精度可达到84.36%，跨课程分类精度可达到79.72%以上；（2）针对内容相关类评论，通过词云分析可发现学习者的关注热点；（3）针对情感相关类评论，通过情感分析可评价学习者对课程的情感倾向；（4）针对其他类评论，通过关键词过滤和句式分析，可挖掘出学习者求助信息，完善课程支持服务。
【英文摘要】：
This paper uses a semantic analysis method to analyze and mine the MOOC comments from the perspective of big data. Aiming at the huge number and chaos of the MOOC comments, this paper proposes a semantic analysis model of MOOC comments. In this model, MOOC comments are classified into three types: comments related to course content, comments related to emotions and other comments. According to the differences among those types, this paper proposes a semantic feature description method based on parts of speech to characterize and classify comments. In order to evaluate the effectiveness of the model, this paper takes comments of four courses from the icourse MOOC as examples. The research results indicate that (1) Taking parts of speech as the semantic feature, the classification accuracy in a single course could reach 84.36%, and above 79.72% in cross-courses.(2) As for the comments related to contents, word cloud analysis could mine the hot spots which students are interested in.(3) For comments related to emotions, sentiment analysis could evaluate the emotional tendency of students.(4) For other comments, keywords filtering and sentence patterns analysis are used to mine the help information to improve the curriculum support services.