Considering Student's Cognitive Abilities-Entity Level and Entirety Level Sentiment Analysis of Teaching Evaluation

Student Teaching Feedback Sentiment Analysis (STFSA) plays a crucial role in evaluating teaching effectiveness, and its analysis results are influenced by both teaching process organization and students’ course cognition. However, current sentiment analysis of students’ feedback faces the challenges of integrating fine-grained entity-level sentiment and multi-polar sentiment in overall sentiment analysis. Therefore, we propose a novel sentiment analysis strategy for Entity-level and Entirety-level Teaching evaluation Sentiment analysis that considers students’ Cognitive abilities (EETSC). This strategy utilizes two-level networks: one for the entity level network based on dual attention mechanism to conduct fine-grained sentiment analysis of students’ feedback, and the other for the overall sentiment analysis network that integrates students’ personalized cognitive abilities to adjust the sentiment characteristics of different entities and obtain more accurate and reasonable emotions at the overall level. Experimental results show that, compared to the state-of-the-art baseline, EETSC achieves an accuracy of 86.34% and an F1 score of 77.13% in the recognition of six types of teaching entity sentiments, representing improvements of 2.19% and 2.18%, respectively. For entirety-level sentiment recognition, EETSC achieves an accuracy of 87.36% and an F1 score of 78.02%, with improvements of 0.76% and 1.21%. Further experimental analysis indicates that EETSC can alleviate the problem of sentiment polarity conflicts in teaching evaluations and provides a solution for integrating students’ cognitive states into teaching sentiment analysis in the field of educational natural language processing.