Classification of Microblog Users’ Sentiments Based on BERT-BiLSTM-CBAM

Abstract

This study proposes a method for more accurate classification of microblog users' sentiments based on the BERT-BiLSTM-CBAM hybrid model. First, the text information is pre-trained by the bidirectional encoder representation from transformers (BERT) model to get feature vectors. Then, the feature vectors are spliced and recombined using bidirectional long-short-term memory network (BiLSTM) and CBAM mechanism to obtain new feature vectors. Finally, these new feature vectors are input to the full connection layer and then processed by the softmax function to obtain the sentiment category of the text. The experiment conducted on the sample dataset demonstrates that the model proposed in this study yielded accurate and dependable result in classifying microblog texts in the sample dataset. The model based on BERT-BiLSTM-CBAM algorithm is more efficient than the traditional depth model in processing microblog contents