Research And Implementation Of Search System Based On Joint Extraction Of Financial Entity And Trigger

In recent years,with the increase of public income,people's demand for investment and financial management has become increasingly strong.The characteristics of online financial information are mostly large-scale and idle,and the occurrence of financial events is gradual and uncertain.In addition,relevant information is mixed with a lot of irrelevant information,which leads to the difficulty and complexity of non-practitioners in this field in mastering professional knowledge and finding investment risks.On the other hand,many people don't have the awareness of risk prevention and control,investment failures,and even deceived incidents happen from time to time.Therefore,how to extract financial events from numerous network information has a high practical application value for the prevention and control of public financial risks.After fully investigating the research status at home and abroad,aiming at the problem of financial event extraction,on the basis of single event identification model and Bert-Bi LSTM-CRF event extraction model,this paper designs and implements a joint identification and extraction model based on financial entities and trigger words.In this model,the combination framework of deep learning is used to build an event recognition model that integrates the features of entities and trigger words and a joint extraction model with multiple convolutions and multiple attentions.The knowledge extraction of financial announcement text is completed by extracting trigger words,event elements and relationships.Through comparative experimental analysis,the results are evaluated,and the model adopted in this paper has a better effect on the extraction of financial events.The model of event recognition and joint extraction in this paper is superior to the extraction method of comparative experiment in accuracy and recall.Based on the above-mentioned financial event extraction model as the core technology,this paper designs and implements a financial event-oriented query and search system based on financial knowledge map.According to users' needs,the system can extract the information that users are interested in,and help users know the occurrence and development of financial events in a more intuitive way,so as to provide support for users to prevent financial risks.For the construction of financial database,the code written by scrapy framework is used to crawl the financial public information of Flush website.Then train and apply the event recognition model,classify the fuzzy financial events,and then import the events into the event extraction model according to the categories to complete the financial event extraction.The secondary graph database is used to store the graph and build the knowledge graph in the financial field.The overall functions of the system include account entry,model training,query and search,and data management.The system adopts B/S architecture,builds and develops the back-end of the system by python and java,and then develops the front-end by CSS and Java Script,which completes the realization of the system and adds practicality and convenience to the existing browsers in the market in the form of extended applications.Finally,through the related function test and performance test of the system,the basic functions of the system can be guaranteed,and the system can run stably,and the search rate and accuracy of users can be improved to some extent.