Research On Teaching Evaluation Based On Text Classification

Teaching evaluation is an activity that judges the value of teaching process and results according to teaching objectives and serves for teaching decision-making,it is a process of evaluate teachers' teaching work and students' learning achievement.Teaching evaluation includes two core links: teachers' evaluation and students' evaluation: teachers' evaluation refers to the evaluation of teachers' teaching work(teaching design,organization,implementation,etc.),and students' evaluation is the evaluation of students' learning achievement(learning score,course discussion,homework completion).At present,the teaching evaluation of universities mostly uses students' achievements and teachers' evaluation scores as the evaluation criteria,and the situation reflected in the evaluation text is not included in the evaluation results;in fact,students' discussion texts on the curriculum and comments on teachers can reflect the teaching effect.This thesis collects a large number of texts in teaching activities,constructs a corpus of teaching evaluation,and uses the method of text classification to predict students' learning performance and evaluate teachers' teaching effect,the main contents include:(1)Built the discussion area text corpus and the teachers' evaluation text corpus.Based on the text extraction of a course discussion area,a discussion area text corpus containing 8,143 texts is constructed according to the sentence-level categories of course class,teaching class,platform class and irrelevant class,it is used to analyze the influence of students' speech texts on their academic performance.Combined with the marking standard of the teachers evaluation text of a university,the evaluation text is marked according to the three aspects of attitude,curriculum and emotion,and a corpus of teachers evaluation text including 17,861 texts is constructed for the research of teaching evaluation.(2)A learning analysis model based on text classification in discussion area is constructed.Firstly,the text classification model based on rule and random forest fusion(RRF)is used to divide the text of the discussion area into four categories.Then,with the text categories and online learning data as the characteristics,the learning prediction model of support vector machine(SVM)is used to predict students' achievement.The experimental results show that the accuracy of the prediction model based on SVM is 95.26%.The learning analysis and prediction model has been applied in the mixed teaching based on MOOC to predict students' course scores and successfully reduce the failure rate of students' final grades.(3)An automatic text classification algorithm based on vocabulary matching(VM)and based on vocabulary matching and dependency parsing(VMDP)is proposed.Based on the corpus of the teachers' evaluation text,the domain dictionary of teaching and emotion is established,and the VM algorithm is used to classify the curriculum and emotion of the evaluation text.In order to solve the problem that the distance between the teacher or the curriculum subject and the evaluation words in the text is too long,on the basis of the VM algorithm,the dependency parsing relationship is integrated to improve the classification effect of the evaluation text in the curriculum.Experiments show that the classification accuracy of VM algorithm in emotion and curriculum is 84.16% and 83.67% respectively.After the integration of dependency parsing,the accuracy of curriculum classification is improved to 85.81%.At present,the text classification algorithm of teaching evaluation has been applied in the educational administration platform to complete the automatic scoring of teaching evaluation text.