Power Text Mining and Classification Based on Deep Learning Hybrid Models

The proposed deep learning hybrid model for power text mining and classification addresses the limitation of traditional methods that fail to reflect the significance of each word in the text during feature extraction. The model combines Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, along with an attention mechanism, to effectively extract features from power text.Firstly, CNN is used to extract local information from power text, followed by integrating the semantics of the entire text. This step helps in capturing important patterns and relationships within the text. Secondly, LSTM is employed to extract contextual features from power text, which are then enhanced by adding an attention mechanism that calculates the attention score of output information. This mechanism enables the model to focus on crucial words and phrases in the text.Finally, the output of LSTM Attention is fused with the output of CNN to create a more comprehensive representation of power text features. This approach ensures that both local and global information is considered when classifying power text. Experimental results on public datasets demonstrate that the proposed model outperforms LSTM, CNN, and their improved models in terms of accuracy and effectiveness in power text classification. The model's ability to focus on important words and phrases leads to better performance in identifying relevant information from power text.