Research On Automatical Generation Of Student Reviews Based On Data Mining Method

Reviews on students given by teachers are an important means to reflect the various aspects of students, such as virtue, intelligence and labor. The review is not only a comprehensive assessment of the student’s performance in a specific time at school, but also is an important communication bridge between parents and teachers. With these reviews, students can find their weaknesses in time and obtain the direction indicator to improve themselves. This thesis aims to research intelligent students’ reviews generation and the research can be summarized as the following three aspects:Firstly, combined with the requirements and characteristics of current education system and purposes, the thesis uses association rules to extract evaluation indicators that meet the minimum support from a large number of reviews after word segmentation and labeling. By doing this, the thesis sets a multi-dimensional, multi-level and multi-content evaluation indicators system that conforms the current education system. According to the basic principles of the extraction, some indicators are classified and some are abandoned. The system mainly has guide function and incentive function. To use the system to fully evaluate a student is in line with the requirements and development of modern education, meanwhile, the system can provide reference for teachers to inspect the students’ performance in their daily school life.Secondly, based on the comprehensive evaluation indicators system with objectivity and reality and rule templates, the thesis extract emotional words that modifies indicators.According to the fuzzy set theory, we construct the membership function, use the membership degree to identify the polarity strength of emotional words and degree adverbs, and establish the fully degree adverb corpus and emotional words corpus.Thirdly, the thesis uses Analytic Hierarchy Process method to calculate the weight of every evaluation indicator and uses the weighted average method to obtain a composite score of students. Meanwhile, with the weight of every indicator, the thesis sets a template of student reviews, and then uses the score of every indicator to extract corresponding degree adverbs and emotional words from corpus and add them to the generated template according to the grammatical structure. Generated ideal reviews on students can not only improve the accuracy and objectivity, but also reduce the negative impact of human factors.The intelligent students’ reviews generation can help teachers to reduce the work burden and can play a positive role in promoting more scientific and standard student reviews.