Bank contract compliance evaluation method and system based on multiple intelligent agents

Abstract

The invention provides a bank contract compliance assessment method and system based on multiple intelligent agents. The method comprises the following steps: constructing a bank contract compliance assessment knowledge base; training an open source pre-training model by adopting the constructed bank contract compliance evaluation knowledge base to obtain a compliance large model; intelligent agents corresponding to evaluation task flows are configured for evaluation items of bank contract compliance, and multiple intelligent agents are obtained; setting a scoring rule of each intelligent agent and configuring a scoring weight; and according to the multiple intelligent agents, based on the constructed compliance evaluation large model, in combination with the set scoring rule of each intelligent agent and the configured scoring weight, obtaining a compliance score of the to-be-evaluated bank contract. According to the method, the special compliance large model is obtained through training, multiple intelligent agents are designed, the bank contract compliance evaluation result is obtained, digital and intelligent system construction is achieved, and the work pressure of bank contract compliance evaluation personnel is relieved fundamentally.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a method and system for evaluating the compliance of bank contracts based on multiple intelligent agents. Background Art

[0002] The current intelligent contract evaluation technology mainly uses traditional natural language for analysis and a scoring method that compares with historical data. These methods have limitations in comprehensively evaluating contract compliance; in addition, after the policy changes, the adjustment of the evaluation plan often lacks timeliness, resulting in insufficient practicality. Therefore, it is impossible to effectively achieve the construction of digital and intelligent systems, nor can it fundamentally relieve the work pressure of personnel. Summary of the Invention

[0003] This application provides a method for evaluating the compliance of bank contracts based on multiple intelligent agents, which can solve the technical problems that the current intelligent contract evaluation technology mainly uses traditional natural language for analysis and compares with historical data, and cannot effectively achieve the construction of digital and intelligent systems, nor can it fundamentally relieve the work pressure of bank contract evaluators.

[0004] In a first aspect, this application provides a method for evaluating the compliance of bank contracts based on multiple intelligent agents, including the following steps:

[0005] Construct a knowledge base for evaluating the compliance of bank contracts;

[0006] Use the constructed knowledge base for evaluating the compliance of bank contracts to train an open-source pre-trained model to obtain a compliance large model;

[0007] Configure intelligent agents with corresponding evaluation task flows for each evaluation item of bank contract compliance;

[0008] Set the scoring rules for each intelligent agent and configure the scoring weights;

[0009] Based on multiple intelligent agents, and based on the constructed compliance evaluation large model, combine the set scoring rules of each intelligent agent and configure the scoring weights to obtain the compliance score of the bank contract to be evaluated.

[0010] Combined with the first aspect, in an implementation, the construction of the knowledge base for evaluating the compliance of bank contracts specifically includes the following steps:

[0011] Obtain the knowledge base required for evaluating the compliance of bank contracts:

[0012] Segment the knowledge text in the knowledge base required for evaluating the compliance of bank contracts to obtain a knowledge base after segmenting the knowledge text.

[0013] In combination with the first aspect, in one implementation, the knowledge base required for obtaining the bank contract compliance assessment includes an internal business process knowledge base, an internal system knowledge base, a financial regulation knowledge base, an internal contract document knowledge base, an internal violation document knowledge base, and a regulatory policy knowledge base.

[0014] In combination with the first aspect, in one implementation, after segmenting and processing the knowledge text in the knowledge base required for the bank contract compliance assessment to obtain a vector database, the following steps are further included:

[0015] Update the bank contract compliance assessment knowledge base according to the changing financial regulatory policies.

[0016] In combination with the first aspect, in one implementation, training an open-source pre-trained model using the constructed bank contract compliance assessment knowledge base to obtain a compliance large model specifically includes the following steps:

[0017] Select a natural language pre-trained large model or a model supporting fine-tuning as the open-source pre-trained model;

[0018] Process the knowledge text in the constructed bank contract compliance assessment knowledge base into a knowledge text in the form of a dataset;

[0019] Based on the open-source pre-trained model, use the knowledge text in the form of a dataset to train and obtain compliance large models with different parameter scales.

[0020] In combination with the first aspect, in one implementation, the intelligent agents in the step of configuring intelligent agents with corresponding evaluation task flows for each evaluation item of the bank contract compliance include:

[0021] Subject qualification verification intelligent agent;

[0022] Legal content review intelligent agent;

[0023] Program compliance monitoring intelligent agent;

[0024] Intelligent agent for regulatory policy evaluation items;

[0025] Environment interaction and coordination intelligent agent.

[0026] In combination with the first aspect, in one implementation, setting the scoring rules for each intelligent agent and configuring the scoring weights specifically includes the following steps;

[0027] Score separately with different evaluation items of the bank contract as the scoring items;

[0028] The full score of the basic score for a single scoring item is defined as a;

[0029] a i Marked as the i-th scoring item, 0 < a i<a;

[0030] b i The weight marked for the i-th scoring item, b 1 +b 2 +b 3 +···+b i = 1.

[0031] Combined with the first aspect, in one implementation, based on the multi-intelligent agent, based on the constructed compliance evaluation large model, combined with the scoring rules set for each intelligent agent and configured scoring weights, the compliance score of the bank contract to be evaluated is obtained, which specifically includes the following steps:

[0032] Extract the embedding layer of the compliance large model with different parameter scales for average pooling operation to obtain the embedding model of the compliance large model with different parameter scales;

[0033] Use the embedding models of the compliance large models with different parameter scales to vectorize the content in the bank contract compliance evaluation knowledge base, and obtain the vector representation of the knowledge text in the knowledge base required for bank contract compliance evaluation;

[0034] According to the task flow in the multi-intelligent agent, based on the vector representation of the knowledge text in the knowledge base required for bank contract compliance evaluation, obtain the scores of each evaluation item of the bank contract to be evaluated;

[0035] Perform numerical transformation on the scores of each evaluation item of the bank contract to be evaluated according to the comprehensive score calculation formula, and calculate and obtain the compliance score of the bank contract to be evaluated.

[0036] In the second aspect, the present application provides a bank contract compliance evaluation system based on multi-intelligent agents, including:

[0037] An evaluation knowledge base acquisition module, used to construct a bank contract compliance evaluation knowledge base;

[0038] A compliance large model acquisition module, communicatively connected to the evaluation knowledge base acquisition module, used to train an open-source pre-trained model with the constructed bank contract compliance evaluation knowledge base to obtain a compliance large model;

[0039] A multi-intelligent agent configuration module, used to configure intelligent agents with corresponding evaluation task flows for each evaluation item of bank contract compliance;

[0040] A scoring rule and weight setting module, used to set the scoring rules of each intelligent agent and configure scoring weights;

[0041] A compliance scoring module, communicatively connected to the compliance large model acquisition module, the multi-intelligent agent configuration module, and the scoring rule and weight setting module, is configured to obtain the compliance score of a bank contract to be evaluated based on multi-intelligent agents, a constructed compliance evaluation large model, in combination with the set scoring rules for each intelligent agent and configured scoring weights.

[0042] Combined with the second aspect, in an implementation, the compliance large model acquisition module includes:

[0043] A pre-trained model selection unit, configured to select a natural language pre-trained large model or a model supporting fine-tuning as an open-source pre-trained model;

[0044] A knowledge text form processing unit, configured to process the knowledge text in the constructed bank contract compliance evaluation knowledge base into a knowledge text in dataset form;

[0045] A compliance large model acquisition unit, communicatively connected to the pre-trained model selection unit and the knowledge text form processing unit, is configured to train and obtain compliance large models with different parameter scales based on the open-source pre-trained model and using the knowledge text in dataset form.

[0046] The beneficial effects brought by the technical solutions provided in the embodiments of the present application at least include:

[0047] By training to obtain a dedicated compliance large model to replace the traditional natural language processing model, the generalization ability of the compliance large model is stronger and it can perform data analysis and evaluation work for different compliance evaluation items of bank contracts; using the compliance large model, it is designed as multi-intelligent agents to conduct multi-dimensional compliance evaluations on bank contracts to obtain the scores of each evaluation item, and finally calculate the final contract compliance evaluation result by combining the weight values of each evaluation item in the bank contract compliance evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic flowchart of the method for evaluating the compliance of bank contracts based on multi-intelligent agents in the present application;

[0049] Figure 2 It is a functional module block diagram of the system for evaluating the compliance of bank contracts based on multi-intelligent agents in the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0051] In the description of the specification, claims and the above drawings of this application, the terms "comprising", "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices. Descriptions such as "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequential order, nor do they limit that "first", "second" and "third" are of different types.

[0052] In the description of the embodiments of this application, terms such as "exemplary", "for example" or "for instance" are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "exemplary", "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary", "for example" or "for instance" is intended to present relevant concepts in a specific manner.

[0053] In the description of the embodiments of this application, unless otherwise specified, " / " means "or". For example, A / B may mean A or B; "and / or" in the text is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "a plurality of" means two or more than two.

[0054] In some processes described in the embodiments of this application, a plurality of operations or steps appear in a specific order. However, it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of this application or may be executed in parallel. The serial numbers of the operations are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in order or in parallel, and these operations or steps may be combined.

[0055] To make the objectives, technical solutions and advantages of this application clearer, the following will further describe the embodiments of this application in detail with reference to the drawings.

[0056] In a first aspect, as Figure 1 shown, the method for evaluating bank contract compliance based on multiple intelligent agents provided by this application includes the following steps:

[0057] Step S1: Construct a knowledge base for evaluating bank contract compliance;

[0058] Step S2: Train an open-source pre-trained model using the constructed bank contract compliance assessment knowledge base to obtain a compliance large model;

[0059] Step S3: Configure intelligent agents for each evaluation item of bank contract compliance, each with a corresponding evaluation task flow;

[0060] Step S4: Set the scoring rules for each intelligent agent and configure the scoring weights;

[0061] Step S5: Based on multiple intelligent agents, using the constructed compliance assessment large model, combined with the set scoring rules and configured scoring weights of each intelligent agent, obtain the compliance score of the bank contract to be evaluated.

[0062] In this application, a dedicated compliance large model is obtained through training to replace the traditional natural language processing model. The generalization ability of the compliance large model is stronger and can perform data analysis and evaluation work for different compliance evaluation items of bank contracts; using the compliance large model, it is designed as multiple intelligent agents to conduct multi-dimensional compliance evaluations on bank contracts to obtain the scores of each evaluation item, and finally calculate the final contract compliance evaluation result by combining the weight values of each evaluation item of the bank contract compliance assessment.

[0063] In one embodiment, the step S1: Construct a bank contract compliance assessment knowledge base specifically includes the following steps:

[0064] Step S11: Obtain the knowledge base required for bank contract compliance assessment:

[0065] Step S12: Segment the knowledge text in the knowledge base required for bank contract compliance assessment to obtain the knowledge base after segmented processing of the knowledge text.

[0066] Based on Step S11 and Step S12, obtain the knowledge compliance constraint basis of bank contracts.

[0067] In one embodiment, the step S11: Obtain the following knowledge bases required for bank contract compliance assessment:

[0068] A: Internal business process knowledge base: Obtain various business process data from the bank's internal office system, which includes process corresponding business and node data. The node data mainly includes: node start and end dates, node corresponding department levels, node handlers, node approval opinions, etc.;

[0069] B: Internal system knowledge base: Include system documents such as organizational structure and responsibility division system, internal control and risk management system, loan approval and risk assessment system, customer information management and protection system, business criterion system, compliance risk management guidelines, employee behavior standards, procurement management methods, etc.

[0070] C: Financial Regulations Knowledge Base: including the compliance management measures for financial institutions, the guidelines for compliance risk management of commercial banks, laws, regulations and regulatory requirements, etc.;

[0071] D: Internal Contract Document Knowledge Base: including contract templates, various contract requirement documents, and historical contract data, etc.;

[0072] E: Internal Violation Document Knowledge Base, including processes, contract documents, penalty documents, and rectification documents where problems occurred in previous internal control compliance;

[0073] F: Regulatory Policy Knowledge Base, which is updated regularly to adapt to changes in financial regulatory policies;

[0074] G: Other Reference Document Knowledge Base.

[0075] In one embodiment, the step S12: segment the knowledge text in the knowledge base required for the compliance assessment of bank contracts, and obtain the database after the segmented processing of the knowledge text, which is specifically implemented as follows:

[0076] During the process of creating the knowledge base for the compliance assessment of bank contracts, the knowledge text in each knowledge base is segmented. The segmentation can be carried out using schemes such as line breaks, special symbol marking, word count, etc., and the following segmentation rules shall be observed:

[0077] Two adjacent segmented texts need to have a certain semantic similarity;

[0078] The segmented text should be marked with the theme, and the knowledge text with different theme contents cannot be stored in the same segment.

[0079] In one embodiment, after the step S12: segment the knowledge text in the knowledge base required for the compliance assessment of bank contracts and obtain the vector database, the following steps are further included

[0080] Update the knowledge base for the compliance assessment of bank contracts according to the changing financial regulatory policies, mainly in two aspects, one is the internal and external regulatory policies; the update mechanism is divided into two types, internal and external: internally, it mainly includes changes in compliance regulations, contract templates, process changes, etc., which are pushed to the knowledge base in real time; externally, it mainly includes using web crawler technology to regularly monitor relevant policy release websites for regular comparison. If a new policy is released, it will be updated to the knowledge base in a timely manner.

[0081] Based on step S12, the policy effectiveness of the knowledge base for the compliance assessment of bank contracts is ensured.

[0082] In one embodiment, the step S2: train the open-source pre-trained model using the constructed knowledge base for the compliance assessment of bank contracts to obtain the compliance large model, which specifically includes the following steps:

[0083] Step S21: Select a suitable large natural language pre-trained model or a model supporting fine-tuning as the open-source pre-trained model, such as a model based on the Transformer architecture;

[0084] Step S22: Process the knowledge texts in the constructed bank contract compliance assessment knowledge base into dataset-form knowledge texts;

[0085] Step S23: Based on the open-source pre-trained model, use the dataset-form knowledge texts to train compliance large models with different parameter scales to adapt to different task requirements. The training method selects full-scale fine-tuning, Lora, STF, etc. training schemes according to the amount of training data.

[0086] Based on steps S21 - S213, train a compliance large model dedicated to banks to replace the traditional natural language processing model. The compliance large model has stronger generalization ability and can perform analysis work in different modules. When the financial supervision policies are updated and changed, only the knowledge base needs to be updated, and there is no need to retrain the compliance large model.

[0087] In one embodiment, in step S3: An intelligent agent corresponding to the evaluation task flow is configured for each evaluation item of bank contract compliance, which specifically includes the following steps:

[0088] Step S3A: The intelligent agent for subject qualification verification. The subject mainly refers to the two parties of the contract. The tasks of the intelligent agent for subject qualification verification mainly include data collection, qualification verification, qualification risk scoring, and qualification scoring of the two parties of the contract. Among them, the data collection is specifically implemented as obtaining the qualification information of the two parties of the contract from the enterprise registration database, credit assessment agencies, etc.; the qualification verification is to check the validity of qualifications such as enterprise business licenses, personal identification documents, and special industry licenses; the qualification risk scoring is to give a subject qualification risk score based on historical records and the current status; the qualification scoring of the two parties of the contract is to give the qualification scores of the two parties of the contract, and give deduction items and suggestions;

[0089] Step S3B: The intelligent agent for legal content review. The legal content review mainly refers to the review of whether the contract terms comply with laws and regulations. The tasks of this intelligent agent include regulation matching, unfair clause review, professional term explanation, and scoring. Among them, the regulation matching is to compare the contract terms with relevant legal articles; unfair clause review: identify clauses that may violate the consumer protection law or unequal clauses; the professional term explanation is to provide a concise explanation of the contract terms for non-legal professionals; the scoring is to give a regulation compliance score, and give deduction items and suggestions;

[0090] Step S3C: Program compliance monitoring intelligent agent. Program compliance monitoring mainly reviews the compliance of contract processes. The tasks of this intelligent agent include process verification and scoring. Among them, the process verification is to confirm whether the contract signing follows the necessary legal procedures and internal approval processes; the scoring is to give a process compliance score, list the deduction items and suggestions for score improvement;

[0091] Step S3D: Intelligent agent for regulatory policy evaluation items. This intelligent agent is used to evaluate the adaptability of contracts to financial policies. The tasks of this intelligent agent include regulatory update detection and policy suggestions. Among them, the regulatory update monitoring is to conduct a detailed comparative analysis of contract terms with the latest regulations and guiding principles of financial regulatory agencies; the policy suggestion is to propose suggestions for adjusting contract terms based on regulatory changes; the scoring is the regulatory policy compliance score, list the deduction items and suggestions for score improvement;

[0092] Step S3E: Environment interaction and coordination intelligent agent. The role is the intelligent agent administrator. The main task is to generate reports, collect the analysis results of each intelligent agent, calculate the comprehensive evaluation score of bank contract compliance according to the scoring table calculation rules for the different scores given by each intelligent agent, and finally form a comprehensive evaluation report on bank contract compliance.

[0093] The above are the intelligent agents respectively set for the main evaluation items of contract compliance. The task flow content and the number of intelligent agent items of the intelligent agent can be increased correspondingly according to policy changes.

[0094] In one embodiment, Step S4: Set the scoring rules for each intelligent agent and configure the scoring weights, which specifically include the following steps:

[0095] Score respectively with different evaluation items of bank contracts as scoring items. Among them, the full score of the basic score of a single scoring item is defined as a, a i is the i-th scoring item, 0 < a i < a; the weight of the i-th scoring item is expressed as b i , b 1 + b 2 + b 3 + ··· + b i = 1;

[0096] Table 1 Scoring and weight configuration rules for evaluation items

[0097]

[0098] In one embodiment, the said Step S5: Based on multiple intelligent agents, based on the constructed compliance evaluation large model, combined with the set scoring rules and configured scoring weights of each intelligent agent, obtain the compliance score of the bank contract to be evaluated, which specifically includes the following steps:

[0099] Step S51: For compliance large models with different parameter scales, extract the embedding layers and perform average pooling operations to obtain the embedding models of compliance large models with different parameter scales;

[0100] Step S52: Use the embedding models of compliance large models with different parameter scales to vectorize the content in the bank contract compliance evaluation knowledge base, obtain the vector representations of the knowledge texts in the knowledge base required for bank contract compliance evaluation, and store them in a vector database. The vector database can choose open-source ones such as Milvus, Weaviate, Faiss, Chroma, etc., and select a suitable vectorization model;

[0101] Step S53: According to the task flow in the multi-intelligent agent, based on the vector representations of the knowledge texts in the knowledge base required for bank contract compliance evaluation, obtain the scores of each evaluation item of the bank contract to be evaluated;

[0102] Step S54: Numerically transform the scores of each evaluation item of the bank contract to be evaluated according to the following comprehensive score calculation formula, and calculate to obtain the compliance score S of the bank contract to be evaluated:

[0103] S = s 1 + s 2 + s 2 + s 3 + s 4 + ··· + s i

[0104] In the formula, s i is the compliance score of the i-th evaluation item.

[0105] On the second aspect, such as Figure 2As shown in the figure, the present application provides a bank contract compliance evaluation system based on multi-intelligent agents, including an evaluation knowledge base acquisition module 100, a compliance large model acquisition module 200, a multi-intelligent agent configuration module 300, a scoring rule and weight setting module 400, and a compliance scoring module 500; the evaluation knowledge base acquisition module 100 is used to construct a bank contract compliance evaluation knowledge base; the compliance large model acquisition module 200 is communicatively connected to the evaluation knowledge base acquisition module 100, and is used to train an open-source pre-trained model with the constructed bank contract compliance evaluation knowledge base to obtain a compliance large model; the multi-intelligent agent configuration module 300 is used to configure intelligent agents with corresponding evaluation task flows for each evaluation item of bank contract compliance; the scoring rule and weight setting module 400 is used to set the scoring rules of each intelligent agent and configure the scoring weights; the compliance scoring module 500 is communicatively connected to the compliance large model acquisition module 200, the multi-intelligent agent configuration module 300, and the scoring rule and weight setting module 400, and is used to obtain the compliance score of the bank contract to be evaluated based on the multi-intelligent agents, the constructed compliance evaluation large model, and the combined scoring rules and configured scoring weights of each intelligent agent.

[0106] In one embodiment, the compliance large model acquisition module includes:

[0107] A pre-trained model selection unit, which is used to select a natural language pre-trained large model or a model supporting fine-tuning as the open-source pre-trained model;

[0108] A knowledge text form processing unit, which is used to process the knowledge text in the constructed bank contract compliance evaluation knowledge base into a dataset-form knowledge text;

[0109] A compliance large model acquisition unit, which is communicatively connected to the pre-trained model selection unit and the knowledge text form processing unit, and is used to train and obtain compliance large models with different parameter scales based on the open-source pre-trained model and the dataset-form knowledge text.

[0110] Among them, the function implementation of each module in the above bank contract compliance evaluation system based on multi-intelligent agents corresponds to each step in the above embodiment of the bank contract compliance evaluation method based on multi-intelligent agents, and its functions and implementation processes will not be elaborated here one by one.

[0111] In a third aspect, an embodiment of the present application provides a bank contract compliance evaluation device based on multi-intelligent agents. The bank contract compliance evaluation device based on multi-intelligent agents can be a device with data processing functions such as a personal computer (PC), a laptop computer, a server, etc.

[0112] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces, etc., which are used to implement the interconnection of components inside the multi-agent-based bank contract compliance evaluation device, as well as interfaces for implementing the interconnection between the multi-agent-based bank contract compliance evaluation device and other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber optic interface, an ATM interface, etc.; the user device can be a display, a keyboard, etc.

[0113] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical memory, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0114] The processor can be a general-purpose processor, which can call the multi-agent-based bank contract compliance evaluation program stored in the memory and execute the multi-agent-based bank contract compliance evaluation method provided in the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). Among them, the method executed when the multi-agent-based bank contract compliance evaluation program is called can refer to the various embodiments of the multi-agent-based bank contract compliance evaluation method of the present application, which will not be elaborated here.

[0115] Fourthly, the embodiments of the present application also provide a readable storage medium.

[0116] The readable storage medium of the present application stores a multi-agent-based bank contract compliance evaluation program, and when the multi-agent-based bank contract compliance evaluation program is executed by a processor, it realizes the steps of the multi-agent-based bank contract compliance evaluation method as described above.

[0117] Among them, the method realized when the multi-agent-based bank contract compliance evaluation program is executed can refer to the various embodiments of the multi-agent-based bank contract compliance evaluation method of the present application, which will not be elaborated here.

[0118] It should be noted that the serial numbers of the above embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0119] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above and includes several instructions for causing a terminal device to execute the methods described in various embodiments of the present application.

[0120] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A bank contract compliance assessment method based on multi-intelligent agents, characterized in that: The following steps are involved: Build a knowledge base for bank contract compliance assessment; Use the constructed bank contract compliance assessment knowledge base to train the open source pre-trained model to obtain a large compliance model; Configure intelligent agents with corresponding assessment task flows for each assessment item of bank contract compliance; Set scoring rules for each intelligent agent and configure scoring weights; Based on multiple intelligent agents and the constructed compliance assessment model, the compliance score of the bank contract to be evaluated is obtained by combining the scoring rules of each intelligent agent and configuring the scoring weights.

2. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 1 is characterized in that: The construction of the bank contract compliance assessment knowledge base specifically includes the following steps: Get the knowledge base you need for contract compliance assessments for banks: The knowledge texts in the knowledge base required for the bank contract compliance assessment are processed in segments to obtain the knowledge base after the knowledge texts are processed in segments.

3. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 2 is characterized in that: The knowledge base required for obtaining the bank contract compliance assessment includes an internal business process knowledge base, an internal system knowledge base, a financial regulations knowledge base, an internal contract document knowledge base, an internal violation document knowledge base and a regulatory policy knowledge base.

4. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 2 is characterized in that: The segmented processing of the knowledge text in the knowledge base required for the bank contract compliance assessment, after obtaining the vector database, also includes the following steps: Update the bank contract compliance assessment knowledge base according to changing financial regulatory policies.

5. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 1 is characterized in that: The method of using the constructed bank contract compliance assessment knowledge base to train the open source pre-trained model to obtain the compliance model specifically includes the following steps: Choose a large natural language pre-trained model or a model that supports fine-tuning as an open source pre-trained model; The knowledge text in the constructed bank contract compliance assessment knowledge base is processed into a knowledge text in the form of a data set; Based on the open source pre-trained model, knowledge text training in the form of data sets is used to obtain compliant large models with different parameter scales.

6. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 1 is characterized in that: The intelligent agent in the step of configuring the intelligent agent with the corresponding assessment task flow for each assessment item of the bank contract compliance includes: Intelligent agent for subject qualification verification; Legal content review intelligent agent; Program compliance monitoring intelligent agent; Intelligent agents for regulatory policy evaluation items; Environment interaction and coordination intelligent agents.

7. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 1 is characterized in that: The step of setting scoring rules for each intelligent agent and configuring scoring weights specifically includes the following steps: Scoring is performed based on different evaluation items of bank contracts; The basic score of a single scoring item is defined as a; a i Marked as the i-th scoring item, 0 i <a;​ b i Marked as the weight of the i-th scoring item, b1+b2+b3+···+b i =1.

8. The bank contract compliance assessment method based on multi-intelligent agents as claimed in claim 1 is characterized in that: The method of obtaining the compliance score of the bank contract to be evaluated by using multiple intelligent agents based on the constructed compliance assessment model, combining the scoring rules of each intelligent agent and configuring the scoring weights, specifically includes the following steps: Extract the embedding layer of the compliant large model with different parameter scales and perform average pooling operation to obtain the embedding model of the compliant large model with different parameter scales; Using the embedding model of the compliance model with different parameter scales, the content in the knowledge base of bank contract compliance assessment is vectorized to obtain the vector representation of the knowledge text in the knowledge base required for bank contract compliance assessment; According to the task flow in the multi-intelligent agent, based on the vector representation of the knowledge text in the knowledge base required for the compliance assessment of the bank contract, the score of each assessment item of the bank contract to be assessed is obtained; The scores of each evaluation item of the bank contract to be evaluated are numerically transformed according to the comprehensive score calculation formula to calculate the compliance score of the bank contract to be evaluated.

9. A bank contract compliance assessment system based on multi-intelligent agents, characterized in that: include: Evaluation knowledge base acquisition module, used to build a bank contract compliance evaluation knowledge base; A compliance large model acquisition module is in communication connection with the evaluation knowledge base acquisition module and is used to train an open source pre-trained model using the constructed bank contract compliance evaluation knowledge base to acquire a compliance large model; A multi-intelligent agent configuration module is used to configure intelligent agents with corresponding assessment task flows for each assessment item of bank contract compliance; Scoring rules and weight setting module, used to set scoring rules for each intelligent agent and configure scoring weights; The compliance scoring module is communicated with the compliance big model acquisition module, the multi-intelligent agent configuration module, and the scoring rule and weight setting module, and is used to obtain the compliance score of the bank contract to be evaluated based on the constructed compliance assessment big model based on the multi-intelligent agent, combined with the scoring rules set for each intelligent agent and the configured scoring weights.

10. The bank contract compliance assessment system based on multi-intelligent agents as claimed in claim 9, characterized in that: The compliance large model acquisition module includes: The pre-trained model selection unit is used to select a large natural language pre-trained model or a model that supports fine-tuning as an open source pre-trained model; A knowledge text form processing unit, used for processing the knowledge text in the constructed bank contract compliance assessment knowledge base into a data set form knowledge text; The compliance large model acquisition unit is communicatively connected with the pre-trained model selection unit and the knowledge text form processing unit, and is used to acquire compliance large models of different parameter scales based on the open source pre-trained model and the knowledge text training in the form of a data set.

Images