Credit data acquisition method and system based on natural language interaction

Through the credit audit data acquisition method combined with natural language interaction and big data, the problems of low efficiency of manual credit audit and information overload in credit audit are solved, the decision-making efficiency and automated risk control capabilities of credit audit are improved, and efficient and intelligent Credit audit processing.

CN119027229APending Publication Date: 2024-11-26CITIC AIBANK CORPORATION LIMITED
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411127434.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

In the existing technology, the credit audit process relies on the low-coupling processing method of manual credit audit and big data risk control, resulting in low efficiency of manual credit audit, overload information, and inability to fully utilize big data support, and unable to effectively balance the accuracy of credit audit and efficiency.

Method used

Using the method of obtaining credit review data based on natural language interaction, through the extraction and logical calculation of structured and unstructured data, a collection of key information is generated, combined with the plug-in knowledge base, the automated processing capabilities of credit review data are improved, and the credit review personnel can be assisted in processing. Large amounts of data reduce analysis costs.

Benefits of technology

It improves the decision-making efficiency of manual information review, reduces the problem of information overload, enhances the professional skills reserve and information acquisition capabilities of credit review personnel, forms automated rules, and improves the capabilities and flexibility of automated risk control systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027229A_ABST
    Figure CN119027229A_ABST
Patent Text Reader

Abstract

The invention relates to a natural language interaction-based credit review data acquisition method and system, and the method comprises the steps: carrying out the fragmentation operation of first structured data and second structured data, carrying out the combination and recognition, generating a call data set, matching the call data set according to a keyword set, generating a credit review data set, and feeding back an interaction information request, character summarization and intention recognition are carried out through natural language interaction of credit auditing personnel, key information extraction and logic calculation are achieved, the analysis cost of manual auditing under large information is reduced, a complex process is simplified, no invasion is caused to a service scene, an efficient, intelligent, automatic and flexible mode is provided, and the method is suitable for popularization and application. The credit auditing system assists credit auditing staff in processing a large amount of data sets of credit auditing, solves the problem of information overload faced by professional credit auditing staff, and improves the decision-making efficiency of manual credit auditing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of financial technology, financial risk control and data processing technology, and in particular to a method and system for obtaining credit review data based on natural language interaction. Background Art

[0002] In the process of handling financial business, credit review is a necessary check procedure for the qualifications of the business subject. Credit review refers to the review and evaluation of the credit loan application of individuals, enterprises or other entities to determine whether to approve the loan or credit application and how much credit limit to grant. This process aims to evaluate the credit reliability and repayment ability of the applicant to reduce credit risk.

[0003] In traditional processing methods, the main processing methods for credit review are manual credit review and big data risk control. Among them, manual credit review needs to be handled by experienced credit reviewers, and decisions are made strictly in accordance with the institution's credit policies and standards. Under this processing method, the final credit review and approval results may vary depending on the credit reviewer's personal judgment and professional knowledge. Credit reviewers need to read and identify a large amount of data, facing problems such as information overload, low processing efficiency, and a large amount of repetitive work. They are overly dependent on the credit reviewer's personal experience, have a low degree of standardization, and experience is difficult to replicate. At the same time, the manual review process often takes a long time.

[0004] Big data risk control is a method of assessing and managing risks using large-scale data sets and advanced technologies. It is mainly used in finance, insurance, loans and other fields, and aims to help institutions better identify, quantify and manage different types of risks, such as credit risk, market risk, operational risk and compliance risk. Under existing technologies, big data risk control usually uses automated algorithms and models to process and analyze large-scale data sets, thereby reducing costs, improving efficiency and accuracy. Such intelligent risk control systems usually hope to automate risk decision-making processes, including credit decisions, portfolio management and transaction execution, to maximize the efficiency of decision-making and reduce the possibility of human errors. However, in actual applications, in order to achieve a high degree of standardization, big data risk control often has problems with fixed rules and poor flexibility. Therefore, it cannot adapt well to the use of complex credit projects and is easily affected by environmental changes (such as changes in the economic environment), resulting in a decline in the effectiveness of audit decisions.

[0005] In order to balance the accuracy and efficiency of credit review in the financial field, the existing technology often adopts a practical processing method that combines manual credit review with big data risk control. Taking the loan scenario as an example, it is preferred to use an automated system with big data risk control to process a large number of credit applications, and match special application cases based on conditions, such as high-risk loan applications or non-standard situations, and transfer special application cases to manual credit review to ensure the accuracy and reliability of credit decisions.

[0006] However, the processing method of the existing technology is essentially still a low-coupling processing method in which manual credit review and big data risk control operate independently. The manual credit review process cannot obtain higher data support through big data risk control. The actual review process still relies entirely on manual processing by credit reviewers based on their personal experience, and has not substantially improved the defects of the existing technology. Summary of the invention

[0007] In order to address the deficiencies of the prior art, the present invention proposes a credit review data acquisition method and system based on natural language interaction, which performs text summarization and intent recognition through natural language interaction of credit review personnel, realizes key information extraction and logical calculation, reduces the analysis cost of manual review under large information, simplifies complex processes without intrusion into business scenarios, and provides an efficient, intelligent, automated and flexible way to assist credit review personnel in processing large data sets for credit review, solves the information overload problem faced by professional credit review personnel, and improves the decision-making efficiency of manual credit review.

[0008] To achieve the above objectives, the technical solutions adopted by the present invention include:

[0009] A method for obtaining credit review data based on natural language interaction, characterized by comprising:

[0010] S1. Acquire source data, perform a structured data capture operation on the source data, generate first structured data, and determine whether the source data contains unstructured data;

[0011] S2. When it is determined that the source data contains unstructured data, extract the unstructured data and perform a structured conversion operation to generate second structured data;

[0012] S3, performing a slicing operation on the first structured data and the second structured data respectively to generate a first text information set and a second text information set, and matching the first text information set and the second text information set to generate a third text information set;

[0013] S4, extracting prompt word combinations according to the third text information set and configuring text information combination rules to generate a call data set;

[0014] S5. Obtain an interactive information request, perform a keyword extraction operation on the interactive information request, and generate a keyword set;

[0015] S6. Generate a credit review data set feedback interactive information request based on keyword set matching and calling data set.

[0016] Further, the matching the first text information set and the second text information set to generate a third text information set includes:

[0017] Taking the first text information set as a benchmark, determining whether the matching degree between each second text information in the second text information set and the first text information set is higher than a preset threshold;

[0018] When it is determined that the matching degree between the second text information and the first text information set is higher than a preset threshold, deleting the corresponding second text information in the second text information set to generate an adjusted second text information set;

[0019] The first text information set is combined with the adjusted second text information set to generate a third text information set.

[0020] Furthermore, the step S4 further includes:

[0021] Using the calling data set, configuring the first model load and performing the first model parameter conversion operation, calling the first preset model to generate a first feedback information set;

[0022] The first preset model includes a generative dialogue model.

[0023] Furthermore, the step S6 comprises:

[0024] Matching the first feedback information set according to the keyword set to obtain first credit review data;

[0025] Determining whether there is an unmatched keyword in the keyword set that does not match the first feedback information set;

[0026] When it is determined that there is an unmatched keyword that does not match the first feedback information set, use the unmatched keyword, configure the second model load and perform a second model parameter conversion operation, and call the first preset model to generate a second feedback information set;

[0027] The first credit review data and the second feedback information set are combined to generate a credit review data set.

[0028] Furthermore, the method further comprises:

[0029] Obtaining a plug-in knowledge base according to the interactive information request;

[0030] Match the plug-in knowledge base according to the keyword set to obtain the second credit review data;

[0031] A revision operation is performed on the set of credit review data using the second credit review data.

[0032] Furthermore, the method further comprises:

[0033] The credit review dataset is used as a training dataset and / or a verification dataset to perform improved training operations on the big data risk control model.

[0034] The present invention also relates to a credit review data acquisition system based on natural language interaction, which is characterized by comprising:

[0035] A source data processing module, configured to perform a structured data capture operation on the source data to generate first structured data, and to extract unstructured data and perform a structured conversion operation to generate second structured data;

[0036] A text information processing module, used to perform a segmentation operation on the first structured data and the second structured data respectively, generate a first text information set and a second text information set, and match the first text information set and the second text information set to generate a third text information set;

[0037] A data set management module, used to extract prompt word combinations according to the third text information set and configure text information combination rules to generate a call data set;

[0038] A request extraction module, used to perform a keyword extraction operation on the interactive information request to generate a keyword set;

[0039] The request feedback module is used to generate a credit review data set feedback interactive information request based on the keyword set matching call data set.

[0040] The present invention also relates to a computer-readable storage medium, characterized in that a computer program is stored on the storage medium, and the computer program implements the above method when executed by a processor.

[0041] The present invention also relates to an electronic device, characterized in that it comprises a processor and a memory;

[0042] The memory is used to store the first structured data, the second structured data and the calling data set;

[0043] The processor is used to execute the above method by calling the first structured data, the second structured data and the calling data set.

[0044] The present invention also relates to a computer program product, comprising a computer program and / or instructions, characterized in that the computer program and / or instructions implement the steps of the above method when executed by a processor.

[0045] The beneficial effects of the present invention are:

[0046] The credit review data acquisition method and system based on natural language interaction of the present invention are adopted, and the large model capability is utilized to perform text summarization and intent recognition through the natural language interaction of credit review personnel, so as to realize key information extraction and logical calculation, reduce the analysis cost of manual review under large information, simplify the complex process and have no invasion to the business scene, and provide an efficient, intelligent, automated and flexible way to assist credit review staff in handling the problem of large data sets of credit review, solve the problem of information overload faced by professional credit review personnel, and improve the decision-making efficiency of manual credit review; at the same time, by adding a plug-in knowledge base system and other methods, the relevant professional skill reserves and information acquisition capabilities of credit review personnel can be effectively improved, the strategy can be effectively identified with valuable common features, and effective decision rules can be recommended for automatic decision-making, and the automated risk control model can be fed back, so as to finally reduce the proportion of manual credit review intervention. Using the method and system of the present invention, the personal experience of professional credit review personnel can be continuously summarized, new automated rules can be formed, the automated risk control system can be supplemented, the capability of the automated risk control system can be improved, and flexible customized rules and automated loading and execution can be supported, which can be easily used as a bypass to assist business personnel in real-time processing of information. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The figure is a flow chart of the method for obtaining credit review data based on natural language interaction of the present invention.

[0048] Figure 2 This is a schematic diagram of the structure of the credit review data acquisition system based on natural language interaction of the present invention. DETAILED DESCRIPTION

[0049] In order to more clearly understand the content of the present invention, it will be described in detail with reference to the drawings and embodiments.

[0050] The first aspect of the present invention relates to a process of steps as follows Figure 1 The method for obtaining credit review data based on natural language interaction shown includes:

[0051] S1. Acquire source data, perform a structured data capture operation on the source data, generate first structured data, and determine whether the source data contains unstructured data.

[0052] In specific implementation, the manual credit review page is usually opened through a browser, and can interact with the original credit review system through a browser plug-in. For example, based on the interactive design of the browser plug-in, the content of the manual credit review page is obtained in real time, and the page is interacted with. After obtaining the page content, the key information is extracted based on the text summary capability of the large model and written back to the web page.

[0053] S2. When it is determined that the source data contains unstructured data, extract the unstructured data and perform a structured conversion operation to generate second structured data.

[0054] In specific implementation, some rules can be built into the browser plug-in. These rules are written in natural language by professional credit reviewers and rely on the intent recognition of the generative large model. After the web page content is loaded, they are automatically executed to generate relevant text summaries.

[0055] S3. Perform a segmentation operation on the first structured data and the second structured data respectively to generate a first text information set and a second text information set, and match the first text information set and the second text information set to generate a third text information set.

[0056] Preferably, taking the first text information set as a benchmark, determine whether the matching degree between each second text information in the second text information set and the first text information set is higher than a preset threshold; when it is determined that the matching degree between the second text information and the first text information set is higher than the preset threshold, delete the corresponding second text information in the second text information set to generate an adjusted second text information set; and combine the first text information set and the adjusted second text information set to generate a third text information set.

[0057] S4. Extract prompt word combinations according to the third text information set and configure text information combination rules to generate a call data set.

[0058] Preferably, a calling data set is used to configure a first model load and perform a first model parameter conversion operation, and a first preset model is called to generate a first feedback information set; the first preset model includes a generative dialogue model.

[0059] In specific implementation, the above-mentioned text information sets can be designed as bypasses as needed to directly generate notes in page interactions, assist professional credit reviewers in content extraction, and automatically interact with large models to generate content.

[0060] S5. Obtain an interactive information request, perform a keyword extraction operation on the interactive information request, and generate a keyword set.

[0061] Preferably, the plug-in knowledge base is obtained based on the interactive information request; the plug-in knowledge base is matched based on the keyword set to obtain the second credit review data.

[0062] S6. Generate a credit review data set feedback interactive information request based on keyword set matching and calling data set.

[0063] Preferably, the first feedback information set is matched based on the keyword set to obtain the first credit review data; it is determined whether there are unmatched keywords in the keyword set that do not match the first feedback information set; when it is determined that there are unmatched keywords that do not match the first feedback information set, the unmatched keywords are used to configure the second model load and perform a second model parameter conversion operation, and the first preset model is called to generate a second feedback information set; the first credit review data and the second feedback information set are combined to generate a credit review data set.

[0064] Preferably, the second credit review data is used to perform a revision operation on the credit review data set.

[0065] Preferably, the credit review dataset is used as a training dataset and / or a verification dataset to perform an improved training operation on the big data risk control model.

[0066] A typical process of using the above method to assist in obtaining credit review data may include:

[0067] The manual credit reviewer opens the credit review page;

[0068] After the page is loaded, the plug-in's resident process hits the preset keyword information;

[0069] The plugin resident process simply formats the text information of the page and then passes it to the backend;

[0070] After the backend processes the text, it slices and stores the relevant text information in the vector library;

[0071] The page performs manually customized tasks, such as "sorting out the credit limit of each credit card and the corresponding number of overdue payments";

[0072] This task is done by asking questions, recalling the vectors, combining them and sending them to the big model;

[0073] The big model performs intent recognition and logic calculations and generates results;

[0074] The generated structure is obtained by the plug-in and then embedded in the web page.

[0075] Another aspect of the present invention also relates to a credit review data acquisition system based on natural language interaction, the structure of which is as follows: Figure 2 As shown, including:

[0076] A source data processing module, configured to perform a structured data capture operation on the source data to generate first structured data, and to extract unstructured data and perform a structured conversion operation to generate second structured data;

[0077] A text information processing module, used to perform a segmentation operation on the first structured data and the second structured data respectively, generate a first text information set and a second text information set, and match the first text information set and the second text information set to generate a third text information set;

[0078] A data set management module, used to extract prompt word combinations according to the third text information set and configure text information combination rules to generate a call data set;

[0079] A request extraction module, used to perform a keyword extraction operation on the interactive information request to generate a keyword set;

[0080] The request feedback module is used to generate a credit review data set feedback interactive information request based on the keyword set matching call data set.

[0081] By using this system, the above-mentioned operation processing method can be executed and the corresponding technical effects can be achieved.

[0082] An embodiment of the present invention also provides a computer-readable storage medium capable of implementing all the steps in the method in the above embodiment, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, all the steps in the method in the above embodiment are implemented.

[0083] An embodiment of the present invention also provides an electronic device for executing the above method. As an implementation device of the method, the electronic device has at least a processor and a memory, and in particular, the memory stores data required for executing the method and related computer programs, such as first structured data, second structured data, and a calling data set, etc., and all steps of the implementation method are executed by the processor calling the data and program in the memory to obtain the corresponding technical effect.

[0084] Preferably, the electronic device may include a bus architecture, the bus may include any number of interconnected buses and bridges, and the bus will include various circuits linked together by one or more processors and memories. The bus may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are all well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the receiver and transmitter. The receiver and transmitter may be the same component, i.e., a transceiver, providing a unit for communicating with various other systems over a transmission medium. The processor is responsible for managing the bus and general processing, while the memory may be used to store data used by the processor when performing operations.

[0085] In addition, the electronic device may further include components such as a communication module, an input unit, an audio processor, a display, and a power supply. The processor (or controller, operation control) used may include a microprocessor or other processor device and / or a logic device, which receives input and controls the operation of various components of the electronic device; the memory may be one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices, which may store the above-mentioned related data information, and may also store a program for executing related information, and the processor may execute the program stored in the memory to realize information storage or processing, etc.; the input unit is used to provide input to the processor, for example, it may be a key or a touch input device; the power supply is used to provide power to the electronic device; the display is used to display display objects such as images and text, for example, it may be an LCD display. The communication module is a transmitter / receiver that sends and receives signals via an antenna. The communication module (transmitter / receiver) is coupled to the processor to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal. Based on different communication technologies, multiple communication modules may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless local area network module, etc. The communication module (transmitter / receiver) is also coupled to a speaker and a microphone via an audio processor to provide an audio output via the speaker and receive an audio input from the microphone, thereby realizing a common telecommunication function. The audio processor may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor is also coupled to a central processor, so that recording can be performed on the machine through the microphone, and the sound stored on the machine can be played through the speaker.

[0086] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or multiple boxes.

[0088] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0089] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in a box or multiple boxes. Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the attached claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0090] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for obtaining credit review data based on natural language interaction, characterized in that: include: S1. Acquire source data, perform a structured data capture operation on the source data, generate first structured data, and determine whether the source data contains unstructured data; S2. When it is determined that the source data contains unstructured data, extract the unstructured data and perform a structured conversion operation to generate second structured data; S3, performing a slicing operation on the first structured data and the second structured data respectively to generate a first text information set and a second text information set, and matching the first text information set and the second text information set to generate a third text information set; S4, extracting prompt word combinations according to the third text information set and configuring text information combination rules to generate a call data set; S5. Obtain an interactive information request, perform a keyword extraction operation on the interactive information request, and generate a keyword set; S6. Generate a credit review data set feedback interactive information request based on keyword set matching and calling data set.

2. The method according to claim 1, characterized in that The matching of the first text information set and the second text information set to generate a third text information set includes: Taking the first text information set as a benchmark, determining whether the matching degree between each second text information in the second text information set and the first text information set is higher than a preset threshold; When it is determined that the matching degree between the second text information and the first text information set is higher than a preset threshold, deleting the corresponding second text information in the second text information set to generate an adjusted second text information set; The first text information set is combined with the adjusted second text information set to generate a third text information set.

3. The method according to claim 1, characterized in that The step S4 further comprises: Using the calling data set, configuring the first model load and performing the first model parameter conversion operation, calling the first preset model to generate a first feedback information set; The first preset model includes a generative dialogue model.

4. The method according to claim 3, characterized in that The step S6 comprises: Matching the first feedback information set according to the keyword set to obtain first credit review data; Determining whether there is an unmatched keyword in the keyword set that does not match the first feedback information set; When it is determined that there is an unmatched keyword that does not match the first feedback information set, use the unmatched keyword, configure the second model load and perform a second model parameter conversion operation, and call the first preset model to generate a second feedback information set; The first credit review data and the second feedback information set are combined to generate a credit review data set.

5. The method according to claim 3, characterized in that The method further comprises: Obtaining a plug-in knowledge base according to the interactive information request; Match the plug-in knowledge base according to the keyword set to obtain the second credit review data; A revision operation is performed on the set of credit review data using the second credit review data.

6. The method according to claim 1, characterized in that The method further comprises: The credit review dataset is used as a training dataset and / or a verification dataset to perform improved training operations on the big data risk control model.

7. A credit review data acquisition system based on natural language interaction, characterized in that: include: A source data processing module, configured to perform a structured data capture operation on the source data to generate first structured data, and to extract unstructured data and perform a structured conversion operation to generate second structured data; A text information processing module, used to perform a segmentation operation on the first structured data and the second structured data respectively, generate a first text information set and a second text information set, and match the first text information set and the second text information set to generate a third text information set; A data set management module, used to extract prompt word combinations according to the third text information set and configure text information combination rules to generate a call data set; A request extraction module, used to perform a keyword extraction operation on the interactive information request to generate a keyword set; The request feedback module is used to generate a credit review data set feedback interactive information request based on the keyword set matching call data set.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store the first structured data, the second structured data and the calling data set; The processor is used to execute the method according to any one of claims 1 to 6 by calling the first structured data, the second structured data and the calling data set.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.