Semi-supervised Based On Multiple Classifiers Ensemble Model For Semantic Classification Of Teaching Evaluation

In recent years,with the continuous reform and development of colleges and universities,the quality of teaching has become the core work of the development of colleges and universities.The quality and effectiveness of teaching directly affect the quality of school personnel training.Therefore,it is a key problem for every school to manage well the evaluation of teachers' teaching quality.Students' evaluation of teaching is one of teaching quality evaluation methods.There are a large number of students' teaching text information in the existing educational administration system and it is difficult to quantify the qualitative analysis of the text information and manual analysis of the workload is too large,so these texts have not been used,not play their real value.This paper is mainly about the sentiment classification of students' teaching evaluation text.By judging the emotional tendencies of the evaluation text,we can be more objective to get students' satisfaction with teachers,so as to better promote the quality of teaching in Colleges and universities.The main research results of this paper are as follows:1.We construct an emotional dictionary in the field of teaching evaluation,the EduNet.The text uses semantic double propagation to expand emotional dictionaries and subject dictionaries.By analyzing the dependency relations between words,we define the extended rules and get the final emotional dictionary and subject dictionary,which makes up the blank of emotional dictionary in the field of teaching evaluation2.We construct a short text representation model based on word vector similarity.In view of the few words and short content particularity of student teaching evaluation text.In this paper,we use Word2 Vec model to train the word vector and extend the feature of short text by finding word vector similarity.3.We construct the model of the teaching evaluation model based on semi-supervised and ensemble learning.Firstly,we use the ensemble learning method to comprehensive the incompleteness of a single classifier.Then,the semi-supervised learning methods used to compensate for the problem of too small initial training set.Finally,the two learning methods are integrated together,and a semi-supervised based on ensemble learning framework is constructed.