Knowledge Graph Construction Research For Financial Entities

Finance is one of the fields with the most abundant and accurate historical data.It is characterized by data and information intensive,huge data volume,and diverse data types.In the context of the rapid development of artificial intelligence,practitioners in the financial industry also hope to use artificial intelligence technology to build efficient and intelligent information processing systems in the financial field to help financial practitioners obtain information more quickly and efficiently,so that they can grasp industry trends in advance,track industry development trends,capture opportunities in vast amounts of data and improve their competitiveness.An efficient intelligent information processing system needs a knowledge base or knowledge map in the vertical domain as the support of the knowledge base,just as an expert needs the knowledge of the professional field as the support.Knowledge map has become one of the key technologies for the implementation of artificial intelligence.Therefore,this article takes investment events as the starting point and digs the potential relationship between investment and financing by digging investment events to build a knowledge map in the financial field,supplementing the gaps in the knowledge map in the financial field,and helping financial practitioners better analyze mining investment The investment relationship in the event.The main work of this article are given below:1.Obtaining the data set.This paper uses web crawler technology to crawl semistructured investment event text data from Zero2 IPO Research website,and provides basic text data for the construction of knowledge map in the financial field.2.Acquisition of investment event semantic vector.This paper first uses the Language Technology Platform(LTP)of Harbin Institute of Technology's Social Computing and Information Retrieval Research Center to segment the text data,and then based on the BIOES sequence labeling rules,customizes the label coding to further improve the part of speech.They use the word2 vec model to train the data in this article,and more fully learn the semantic vector representation of words in the field of investment events.3.Named Entity Recognition.A word-Bi LSTM-Attention-CRF recurrent neural network model structure is proposed.In this paper,based on the traditional method of entity extraction using the Bi LSTM-CRF recurrent neural network model,the word combination vector is used as the input vector,and an attention mechanism is added to the network structure.The attention mechanism gives corresponding weights of different feature vectors,which can effectively improve the results of named entity recognition in the financial field.They use the trained model to entity extract the semistructured text that has been labeled with part-of-speech,and finally obtain structured corpus text data.4.Construction and application of knowledge graph.After the first few steps,the py2 neo module in the python program is used to create entity attributes and define the entity relationship for the structured text data.Finally,the Neo4 j graph database is used for entity attributes,entity relationships and entity data for storage and display.And provide query function.