Research On Internet Financial Entity Recognition Based On BERT

With the continuous development of big data processing technology,Internet financial information has entered the era of big data.In the face of massive Internet financial texts,using natural language processing technology to identify the entities of Internet financial texts has become an inevitable trend of technological development.Named entity recognition refers to the recognition and classification of named entities in text which is an important foundation of natural language processing technology.The existing named entity recognition model can't identify financial entities well,and the Internet financial entities can't identify,so the financial supervision can't be started.This paper proposes an improved model which combines the advantages of sequence tagging model(STM)and BERT-Bi GRU-CRF.This model improves the accuracy of Internet financial entity recognition in text.The Le NER-Br data set is used to verify the model and other models,each detection index is better than other models.The F1 value of the model is 0.5% higher than STM and 2% higher than BERT-Bi GRU-CRF.In order to show the recognition effect of the model,the paper constructs a visualization system of Internet financial entity recognition.The system functions include Internet financial text entry,segmentation tagging,entity extraction,word frequency statistics,and Noun relation display,etc.The new model and visualization system can provide technical support for financial supervision,financial data annotation,etc.