Research On The Construction Of University Teacher Competency Indicators Based On BERT Mode

The competency index of college teachers is a standard to measure college teachers to become excellent students.It includes both explicit and implicit factors.The explicit factor refers to surface features such as disciplinary expertise,while the implicit factor refers to deep features such as values.At present,the task of constructing the competency index of college teachers mainly stays in the stage of manual processing text repeatedly.With the increasing number of relevant texts,this method will consume a lot of manpower and time costs,and usually the classification results are subjective and have high professional requirements.This paper considers using natural language processing technology to analyze the relevant texts of college teachers,so as to acquire the competency index knowledge of college teachers.It is hoped that the research of this paper can provide reference value for the research of relevant text fields and the improvement of the competence of college teachers.This paper extracts the text on the topic of college teachers from CNKI,preprocesses it by paragraphing-clause,data cleaning and Chinese word segmentation,and numerically represents the text,and uses the model for classification.The main purpose of this work is to classify the content containing the competency factors of college teachers from a large number of texts,and finally construct competency indicators based on the classification results.Specifically,support vector machine,naive Bayes,logistic regression,and BERT model were respectively used to compare the classification effect,and BERT model achieved a good classification result(F1=0.8286).Then the BERT model is improved,firstly based on feature output;Second,the improvement based on the integration of external knowledge;Experiments show that the improved classification model is more accurate when the feature of BERT model is the last four layers splice(F1=0.8337),and the BERT model combines LSTM and GRU(F1=0.8450).Finally,key words were extracted from the text labeled "1",and Kmeans algorithm was used for clustering.Five competency indicators of college teachers were obtained,which were political literacy,knowledge literacy,teaching ability,professional attitude and personality traits.