Teaching evaluation sentiment analysis method fusing cognitive migration

Abstract

The invention relates to a teaching evaluation sentiment analysis method fusing cognitive migration, and belongs to the field of education informatization. The method comprises the following steps: cleaning and preprocessing collected data; performing word embedding on an unstructured teaching evaluation text by using a BERT language model, inputting the unstructured teaching evaluation text into a bidirectional GRU network for learning, and outputting context hidden memory information of the text; the method comprises the following steps: modeling historical practice records of students by using a neurocognitive diagnosis model, and obtaining personalized cognitive competence vectors of the students as prior knowledge characterization; the extracted text hidden memory information and the prior knowledge information are fused into a cognitive migration matrix, and the prior knowledge is migrated into the text memory through an attention mechanism, so that the real emotional characteristics contained in the text are enhanced. According to the method, sentiment classification can be accurately carried out on the teaching evaluation text carrying the conflicting sentiment features, and the problem of sentiment misclassification caused by existence of the conflicting sentiment features in teaching evaluation is solved.

Description

technical field [0001] The invention belongs to the field of educational informatization, and relates to a teaching evaluation emotion analysis method integrating cognitive transfer. Background technique [0002] As a branch of natural language processing technology, text sentiment analysis technology can automatically extract and recognize emotions and opinions from texts, so as to realize the analysis, processing, induction and reasoning of subjective texts with emotional color. Especially in recent years, the introduction of deep learning to empower text sentiment analysis technology has made great progress in solving unstructured complex semantic data. This technology has been widely used in product reviews, movie reviews and other fields. However, different from general application scenarios, the emotions contained in the teaching evaluation in the teaching evaluation scene are more euphemistic and subtle, which increases the difficulty of emotional recognition; in addi...

AI Technical Summary

Problems solved by technology

[0005] The existing teaching evaluation sentiment analysis models only consider the emotional features contained in the text. When the emotional features in the text conflict, simply extracting features from the text semantic space to identify the true emotion expressed in the teaching evaluation, the recognition is accurate. The rate will decrease, and the model will have misclassification problems

Images