KCNN and BiLSTM-Mhead Attention emotion analysis model of Micro-blogs Based on Multi-feature fusion

For important local features were ignored by the convolutional neural network in Micro-blogs emotion analysis, and contextual information could not be fully learned by the recurrent neural network, a mixed model of KCNNBiLSTM-MA based on multi-feature fusion was proposed. The K convolutional neural network was used to obtain the important Topk features of the partial features of sentences. Then the sentence sequence information and context connection were got from different directions through bidirectional long short-term memory network, and the Multi-head Attention is added to its foundation thus giving more attention to the important parts. Finally, the different features obtained are fused for classification. The experiment proves that the KCNNBiLSTM-MA model has better performance than other emotion analysis models.