Construction of Online Course Resource Library Based on MEA-BP Neural Networks

Abstract

In this paper, the MEA algorithm is applied to the initial optimization of the weight threshold of BP neural network, which improves the stability of the classification of course resources through a two-stage optimization search process. Secondly, the effective recognition of knowledge point entities of teaching resources is realized by combining the LSTM-CRF entity recognition model, the BERT pre-trained language model and the text data enhancement technique based on the maximum bi-directional matching algorithm. In addition, the article proposes a candidate knowledge point entity generation method based on Levenshtein Distance and an entity characterization method based on BERT-TextCNN, which helps to alleviate the noise problem of the text data of curriculum resources and enhance the entity linking effect. The results of the study show that this newly constructed online course resource repository can significantly improve the allocation level of educational resources and reduce the variability of resource allocation among colleges and universities. The coefficient of educational resource allocation variation for SN217 in the university district was reduced from 0.764 to 0.518.In addition, the mean values of this repository in terms of learning path, perceived usefulness, course resources, and intention to use ranged from 1.45 to 1.66.The study in this paper provides valuable methods and insights for optimizing online educational resources.

Keywords

• mea-bp
• lstm-crf
• bert-textcnn
• levenshtein distance
• course repository
• 93c62