Application of the Capsule Network With Attention-Mechanism-Based Dynamic Routing in the Analysis of Utopian Online Literature

This study applies the dynamic routing algorithm based on the attention mechanism to the capsule network model, aiming to address the problem of multi - label entity relation extraction in utopian online literature. The research constructs an improved capsule network model and elaborates on relevant theories, technical foundations, and the model construction process. Through experiments, by comparing with models such as the convolutional neural network with average pooling (Avg - pooling+CNN), the recurrent neural network (RNN), and the attention - based recurrent neural network (Att+RNN), the results show that the improved Att+capsule model performs better in terms of indicators such as precision, recall, and F1 - score, and is also more stable during the training process. The case analysis further verifies the effectiveness of the model in entity relation extraction, theme mining, and sentiment analysis. Future research can be carried out in directions such as expanding the dataset, optimizing the model structure, and expanding the application fields.