Research on the mechanism of teaching culture in Civics course based on deep learning model

Abstract

Abstract This paper uses convolutional neural networks to construct a deep learning-based teaching mechanism for question-and-answer on civic culture. The one-hot representation and the distributed representation of words in the distributed expression form of natural language are analyzed. Word2Vec, a word embedding training tool, is used to generate word embedding representations, and CNN is applied to store and classify word semantics so as to generate semantic matrices. The attention mechanism is combined to simplify the model processing steps and reduce the complexity of operations. The deep learning hierarchical semantic extraction model is tested by loss function and relative similarity to judge the accuracy and performance of the model’s operation. The results showed that the mean composite score of the experimental group was 5.2 points higher than that of the control group, and the t-test results P=0.003<0.05 indicated that there was a significant difference between the overall assessment scores of the two groups, and the teaching of Civic Culture Quiz was beneficial to the improvement of students’ performance.