Research on Intelligent Classification Method of Seismic Information Text Based on BERT-BiLSTM Optimization Algorithm

Abstract

With the development of science and technology, it is possible to quickly obtain massive disaster information after the earthquake, but because the earthquake information not only has the characteristics of strong timeliness, but also is always in the process of dynamic change, it can quickly classify and analyze the earthquake information, which is of great significance for earthquake emergency decision-making. In this paper, an earthquake news text intelligent classification model based on the BERT-BiLSTM optimization algorithm is proposed. First, based on the BERT (Bidirectional Encoder Representation from Transformers) pre-trained model, the algorithm performs a sentence-level feature vector representation of the seismic news text, and enters the feature vector into the BiLSTM layer to extract the global features of the seismic news text, and then enters the SoftMax classifier for classification. Finally, the control experiment of earthquake news text data in Qinghai and Yunnan was passed. Experimental results show that the model is improved by 2 percentage points compared with the traditional Bert model method. Therefore, the intelligent classification model of earthquake information text proposed in this paper can effectively and accurately determine the category of earthquake news and help earthquake emergency rescue decision-making.