关键词: 学生评教, 教师专业发展, 情感分析, 极性分类, 观点抽取, 观点聚类

Abstract: Concern the problem that the low utilization rate of students?? online evaluation of teaching texts due to their unstructured characteristics，a complete set of text analysis methods for student evaluation of teaching based on sentiment analysis technology was proposed. Firstly，the texts were classified into positive and negative categories by sentiment polarity classification techniques. Secondly，the core views of each comment was got by view extraction technology based on part of speech. Thirdly，text vectorization and distance calculation were done by one hot representation combined with Jaccard distance and word similarity algorithm based on synonym forest. Fourthly，the texts expressing the same views were classified by clustering algorithm，and the number of comments contained in each cluster was calculated. Finally，the statistical analysis results were visualized in the form of word cloud，presenting the students?? feedback intuitively and succinctly. The text of teaching evaluation of Nanjing University of Aeronautics and Astronautics was taken as experimental data，the output results were compared with the original text. The accuracy rate of sentiment classification was 89%，which exceeded the application of most classification algorithms in students?? evaluation of teaching texts，and the effectiveness of the method was verified，which made up for the lack of the complete process from the original text of teaching evaluation to the final application and had practical significance for promoting teacher governance and teacher education.

Key words: students ?? evaluation of teaching, teachers ?? professional development, sentiment analysis, polarity classification, viewpoint extraction, viewpoint clustering