Policy Text Classification Algorithm Based on Bert

With the development of the Internet, the policy text classification model of deep learning is used to improve the effect of policy text classification, and to play and use the huge value contained in the policy text. In order to more accurately determine the policy field described by the text, a BERT-based policy text classification algorithm is proposed. First, the algorithm uses the BERT (Bidirectional Encoder Representations from Transformers) pre-trained language model to vectorize the sentence-level feature of the policy field text, and then the obtained feature vector is input into the classifier for classification, and finally the homology policy field is used. The text data set is verified. The experimental results show that the classification of the trained model on the test set recorded the highest F1 value of 93.25%. It is nearly 6% higher than the classification task of the BERT model for the MRPC task. Therefore, the proposed policy domain text classification algorithm can more accurately and efficiently judge the domain of the policy text, which is helpful for further analysis of the text data in the policy domain and extract more valuable information.