Research on Internet Text Sentiment Classification Based on BERT and CNN-BiGRU

In order to classify Internet text according to different sentiments, a new text sentiment classification model BERT-CBA is proposed, which combines BERT pretrained model, Convolutional Neural Network (CNN), Bidirectional Gated Recurrent Unit (BiGRU) and Attention Mechanism. First, the text data is input into BERT to generate word vector representation of the fusion context text context, the output of BERT is simultaneously input into CNN and BiGRU to perform parallel operations, and the operation results are fused and spliced into the attention mechanism layer.Finally, the Softmax excitation function is used to obtain emotional probability distribution of the text. The experimental results on the test set show that the model proposed in this paper has obtained the expected results on various evaluation indicators, and has certain feasibility and effectiveness for text sentiment classification.