Water accidents severity classification based on Prompt-BERT
Jiachen Yu, Jiayi Ouyang, Xiongguan Bao · 2023
Abstarct—Accident report severity classification is a hot topic in the field of water transportation. Accident reports have the characteristics of strong professionalism and large amount of text data, which leads to the poor classification effect of traditional text classification models in this field. To address the above problems, Prompt-Bert is proposed in this paper.First, BERT is used to vectorize words, and then the probability of accident severity is predicted using the accident type as the prompt word, and the maximum probability of accident severity is selected as the output, and compared with three baseline models, Word2Vec-CNN, Word2Vec-LSTM, and BERT The accuracy is improved by 2.50%, 5.00%, and 11.00%, respectively, when compared on the same data set. This experiment demonstrates the feasibility of the model proposed in this paper.