Network Public Opinion Sentiment Analysis based on Bert Model

To solve the high complexity of traditional methods, this paper proposes a Bert-based network public opinion sentiment analysis method to improve the analysis efficiency of sentiment tendency. This method maps the input text sequence to the three spaces of Query, Key and Value to obtain the query vector, key vector and value vector. For each query vector, use the Softmax on the inner product of the query vector and the key vector to obtain the encoded vector. Then input the encoded vector into the trained classifier to obtain the recognition result. This method can overcome the shortcomings of ignoring context by RNN, and simplify the algorithm complexity to
$O(n)$
from
$O(n^{2})$
of RNN and Text-CNN. The experiment verifies the performance of the proposed method on the social network data set. Its public opinion classification accuracy rate is 98.72% and f1_score is 98.5%, which prove that this method can achieve a good performance on the network public opinion sentiment analysis.