An Intelligent Recommendation Model for ELT Resources Combining Improved LFM Model and Parallel CNN

To tackle the issues of the cold start and data scarcity in existing recommendation systems, this study presents an intelligent recommendation model for resources in English teaching, which merges an enhanced hidden semantic model with parallel convolutional neural network. In the conventional latent semantic model, the application of a Transformer-based bidirectional encoder representation together with the BM25 algorithm leads to a notable reduction in space complexity and training time. Meanwhile, the integration of a convolutional neural network efficiently tackles the cold start issue, thus enhancing both the accuracy and stability of the model, especially in data-scarce environments. The results of the experiment indicate that the suggested model requires only 26 iterations to achieve optimal performance, while the other models remain unstable. To summarize, the proposed model can effectively tackle the cold-start and data scarcity issues that are common in traditional recommendation models, and it also exhibits exceptional accuracy in practical situations.