Financial Tranxaction Information Extraction System Based On Rules And Statistical Models

For the financial industry,how to obtain the customer’s purchase intention of financial products largely affects the achievement rate of financial transactions.In this state,it is a subject of great research value to obtain valuable information efficiently and accurately from various types of transaction Information.At present,many financial institutions get useful information from transaction information by means of manual treatment or combined with the principle of single rule,so that the workload and labor cost are relatively high,and the efficiency and accuracy are not high.In this case,through continuous exploration and research,it is found that through the combination of rules and statistical models of information extraction technology,can be very good to deal with this problem,so that enterprises can quickly obtain valuable information from a large data group in a short period of time,better service for users.This thesis discusses the rules and statistics-based information extraction technology in detail,and combines the actual business situation of a financial technology company to describe in detail the design and implementation of rules and statistics on this system.The purpose of this system research is to extract financial transaction trading information from the customer’s chat records in the financial transaction QQ group.The main content of the system design includes the design of three classifiers(chat record classifier,transaction record classifier,transaction class classifier),and the design of regular and non-regular chat record processing modules.1.The chat record classifier filters the non-transaction information chat records according to the rules;the transaction record classifier determines whether the chat records are regular according to the chat record characteristics;the transaction type classifier corresponds to different attribute lists and different attributes according to different financial types.The corresponding attribute values have their own characteristics to identify the financial transaction type category.2.The regular chat record processing module design.Rule-based information extraction is implemented by designing transaction type classifiers and attribute recognition methods.3.The non-regular chat record processing module design.This part is divided into two parts.First,the annotation model is trained,and then the training annotation entity is used to mark the transaction information of the chat record.The system uses ICTCLAS word segmentation tool,identification entity naming,feature selection and conditional random field labeling to realize information extraction based on statistical model.Finally,the attribute and attribute value list are extracted by the annotation model.The results show that the accuracy of the information extraction system based on rules and statistical models for identifying the corresponding values of different attributes corresponding to financial types is 80%.It can be seen from this that the research of information extraction technology can significantly reduce the cost of using human resources and time.