Material review processing platform based on AI model

Abstract

The invention relates to the technical field of material review, and discloses a material review processing platform based on an AI model, and the platform comprises a front-end interaction and display module which is used for receiving application information containing one or more application materials from a user terminal; the AI cognition and processing center is used for receiving the application information and carrying out first-stage dominant error check and second-stage implicit logic risk study and judgment on the application material; and the dominant error verification aims at determining whether the application material has preset dominant errors in the aspects of integrity, format normalization and cross-material data dominant consistency. According to the method and the device, the multi-modal material analysis and dominant error verification functions are set, so that the technical problem that applications are frequently rejected and repeatedly submitted due to the fact that simple form errors exist in applied materials is solved, and the effect of remarkably improving the one-time passing rate of material submission is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of material review, and particularly to a material review processing platform based on an AI model. Background Art

[0002] In the wave of informatization and digital transformation, the material review link in government services and enterprise operations is gradually evolving from the traditional manual mode to the direction of automation and intelligence. The existing automated auxiliary audit systems have improved the efficiency of material circulation and processing to a certain extent.

[0003] However, the current pre-audit capabilities of most automated systems for application materials are very limited, and these systems usually only enable the receipt and archiving of documents. For issues such as whether the materials are complete, whether the key fields conform to a specific format, and whether the core information among multiple materials is consistent, the system cannot perform pre-emptive and effective automatic verification. This results in a large number of applications being rejected due to some simple formal errors. Applicants only receive feedback from manual review several days or even weeks after submission, and then need to re-prepare and submit again. This process not only lengthens the overall cycle of business handling but also greatly consumes the energy of applicants and reviewers. Summary of the Invention

[0004] To make up for the above deficiencies, the present invention provides a material review processing platform based on an AI model, aiming to improve the technical problems in the existing material review process, such as low audit efficiency, lagging feedback caused by relying on manual or simple rules for verification, and the inability to effectively identify the deep logical risks among materials.

[0005] In the first aspect, the present invention provides the following technical solution: A material review processing platform based on an AI model, comprising:

[0006] A front-end interaction and presentation module, configured to receive application information including one or more application materials from a user terminal;

[0007] An AI cognition and processing center, configured to receive the application information and perform a first-stage explicit error verification and a second-stage implicit logical risk analysis on the application materials;

[0008] The explicit error verification is intended to determine whether there are preset explicit errors in the integrity, format standardization, and cross-material data explicit consistency of the application materials;

[0009] The implicit logical risk analysis is intended to construct a global context for the multiple application materials and comprehensively analyze whether there are preset implicit logical risks among the multiple application materials based on the global context;

[0010] An intelligent feedback generation module that generates corresponding review feedback information based on the results of the explicit error verification and / or the results of the implicit logical risk assessment;

[0011] Among them, the review feedback information is sent to the front-end interaction and presentation module for presentation to the user terminal.

[0012] Preferably, the AI cognition and processing center includes:

[0013] A multi-modal material analysis sub-module that, before performing the explicit error verification in the first stage, analyzes the application materials in various different formats uploaded by the user terminal into unified structured data containing text content, layout structure information, and semantic type information for use by the explicit error verification and the implicit logical risk assessment.

[0014] Preferably, when the AI cognition and processing center performs the explicit error verification in the first stage, it is used to:

[0015] Compare the list of submitted materials with a preset list of necessary materials for integrity verification;

[0016] Match the key fields in the application materials with preset verification rules for format standardization verification;

[0017] Cross-compare the same core entity information between different application materials for cross-material data explicit consistency verification.

[0018] Preferably, when the AI cognition and processing center performs the implicit logical risk assessment in the second stage, it is used to:

[0019] Aggregate all or part of the text content of the multiple application materials to construct the global context;

[0020] Based on the global context, perform reasoning analysis on whether there is a logical matching conflict between the core capabilities reflected in the application materials and the core goals to be achieved by the application matters or whether there is a logical feasibility conflict between the resource scale reflected in the application materials and the resource scale required by the application matters to determine whether there is the implicit logical risk.

[0021] Preferably, the platform further includes:

[0022] A government affairs knowledge vector database for storing background knowledge files related to government affairs reviews;

[0023] When the AI cognition and processing center performs the implicit logical risk assessment, it is also used to:

[0024] Retrieve one or more relevant background knowledge files from the government affairs knowledge vector database based on the global context, and use the background knowledge files and the global context together as the basis for comprehensive judgment.

[0025] Preferably, the intelligent feedback generation module is used for:

[0026] When the result of the explicit error check is that there are explicit errors, generate interactive guidance information aimed at guiding the user to make real-time modifications as the review feedback information;

[0027] When the result of the implicit logical risk judgment is that there are implicit logical risks, generate a risk summary report for the reference of the back-end reviewers as the review feedback information.

[0028] Preferably, the platform further includes:

[0029] A business process and scheduling module, used to control the execution process of the AI cognition and processing center, and this process includes:

[0030] After receiving the application information, first trigger the explicit error check in the first stage, and feedback the generated interactive guidance information to the user;

[0031] After determining that there are no explicit errors in the application materials, then trigger the implicit logical risk judgment in the second stage to complete the complete review of the application information.

[0032] In a second aspect, the present invention provides the following technical solution, a method for reviewing materials based on an AI model, including the following steps:

[0033] Receive application information submitted by a user terminal and containing one or more application materials;

[0034] Conduct a first-stage explicit error check on the application materials. The explicit error check aims to determine whether there are preset explicit errors in the integrity, format standardization, and cross-material data explicit consistency of the application materials, and execute the corresponding judgment process;

[0035] If there are the explicit errors, generate and feedback interactive guidance information aimed at guiding the user to make real-time modifications;

[0036] If there are no such explicit errors, conduct a second-stage implicit logical risk judgment on the application materials. The implicit logical risk judgment aims to construct the multiple application materials into a global context, and comprehensively judge whether there are preset implicit logical risks among the multiple application materials based on the global context;

[0037] Based on the result of the implicit logical risk judgment, generate and output a decision-making assistance summary for use in subsequent review processes.

[0038] In a third aspect, the invention provides the following technical solution: a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned material review method based on an AI model.

[0039] In a fourth aspect, the invention provides the following technical solution: a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned material review method based on an AI model.

[0040] The invention has the following beneficial effects:

[0041] 1. In the present invention, by setting up a multi-modal material parsing and explicit error checking function, real-time structured parsing and multi-dimensional form checking are carried out on various formats of application materials submitted by users, realizing the automation of the lagging and manually judged form review link in the traditional process and bringing it forward. The technical problem of frequent rejection and repeated submission of applications due to simple form errors in application materials is solved, and the effect of significantly improving the one-time passing rate of material submission is achieved.

[0042] 2. In the present invention, by aggregating the structured data of multiple application materials to construct a global context and using the comprehensive logical reasoning ability of the large language model for in-depth judgment, the accurate identification of non-explicit internal logical contradictions hidden among multiple materials is realized, such as the mismatch between capabilities and goals, and resources and scales. The technical problem that the prior art can only perform surface rule checking and cannot discover deep logical defects is solved, so that high-risk applications can be screened and warned in advance, and the effect of avoiding the investment of ineffective review resources is achieved.

[0043] 3. In the present invention, by introducing a vector database storing professional background knowledge, assisting the AI center for reasoning through retrieval-enhanced generation technology, and generating the judged implicit risks into a structured decision-making assistance summary, the efficient combination of the in-depth logical judgment result of AI and the final decision-making ability of human experts is realized. The technical problems of information overload and difficult detection of key risk points in traditional manual review are solved, a new review paradigm of human-machine collaboration is constructed, and the overall quality and final decision-making efficiency of the review work are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a system architecture diagram of the material review processing platform based on an AI model proposed by the present invention;

[0045] Figure 2 This is a method flowchart of a material review method based on an AI model proposed by the present invention. Detailed implementation manners

[0046] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0047] Embodiment 1:

[0048] Referring to Figure 1 , in the first embodiment of the present invention, the present invention provides a material review processing platform based on an AI model, including:

[0049] A front-end interaction and presentation module, configured to receive application information including one or more application materials from a user terminal;

[0050] An AI cognition and processing center, configured to receive the application information and perform a first-stage explicit error check and a second-stage implicit logical risk judgment on the application materials; including:

[0051] A multimodal material analysis sub-module, configured to parse various application materials uploaded by the user terminal into unified structured data including text content, layout structure information, and semantic type information before performing the first-stage explicit error check, for use in the explicit error check and implicit logical risk judgment. When the AI cognition and processing center performs the first-stage explicit error check, it is configured to:

[0052] Compare the submitted material list with a preset necessary material list for integrity check;

[0053] Match the keyword fields in the application materials with preset check rules for format standardization check;

[0054] Cross-compare the same core entity information between different application materials for cross-material data explicit consistency check.

[0055] When the AI cognition and processing center performs the second-stage implicit logical risk judgment, it is configured to:

[0056] Aggregate all or part of the text content of multiple application materials to construct a global context;

[0057] Based on the global context, conduct reasoning and analysis on whether there is a logical matching conflict between the core capabilities reflected in the application materials and the core objectives to be achieved by the application matters, or whether there is a logical feasibility conflict between the resource scale reflected in the application materials and the resource scale required for the application matters, so as to determine whether there is a hidden logical risk.

[0058] Explicit error checking aims to determine whether there are preset explicit errors in the integrity, format standardization, and explicit data consistency across materials of the application materials;

[0059] Hidden logical risk assessment aims to construct multiple application materials into a global context and comprehensively assess whether there are preset hidden logical risks among multiple application materials based on the global context;

[0060] The intelligent feedback generation module generates corresponding review feedback information based on the results of explicit error checking and / or the results of hidden logical risk assessment; the intelligent feedback generation module is used for:

[0061] When the result of explicit error checking shows that there are explicit errors, generate interactive guidance information aimed at guiding users to make real-time modifications as the review feedback information;

[0062] When the result of hidden logical risk assessment shows that there are hidden logical risks, generate a risk summary report for the reference of backend reviewers as the review feedback information.

[0063] Among them, the review feedback information is sent to the front-end interaction and presentation module for presentation to the user terminal.

[0064] The government affairs knowledge vector database is used to store background knowledge files related to government affairs reviews;

[0065] When the AI cognitive and processing center executes the hidden logical risk assessment, it is also used for:

[0066] Retrieve one or more relevant background knowledge files from the government affairs knowledge vector database based on the global context, and use the background knowledge files and the global context together as the basis for comprehensive assessment.

[0067] The business process and scheduling module is used to control the execution process of the AI cognitive and processing center, and this process includes:

[0068] After receiving the application information, first trigger the explicit error checking in the first stage, and feedback the generated interactive guidance information to the user;

[0069] After determining that there are no explicit errors in the application materials, then trigger the hidden logical risk assessment in the second stage to complete the complete review of the application information.

[0070] Specifically, this platform is deployed in a server cluster. The server is configured with a processor (CPU / GPU), a memory, and a network interface required to run the large model. The memory stores the computer program involved in the present invention, which is executed by the processor to implement the functions of the following modules and the data flow between them.

[0071] First, the front-end interaction and presentation module receives application information containing one or more application materials from the user terminal. This information flow triggers the business process and scheduling module, which precisely controls the execution order of the AI cognition and processing center according to the preset two-stage process logic.

[0072] In the first stage of the review process, the aim is to achieve accurate cognition of the application materials. The multi-modal material analysis sub-module within the AI cognition and processing center is first called. This sub-module uses the multi-modal processing ability of the core large model to parse the original files in formats such as PDF and JPG uploaded by the user into a structured document object D.

[0073] For a document containing N logical blocks, its mathematical representation is , where each logical block consists of a triple which respectively represent the text content, the position coordinates, and the semantic type.

[0074] Subsequently, based on the parsed high-quality structured data, the system performs explicit error checking. This step includes: Integrity check: By comparing the list of required materials with the list of submitted materials , the integrity score is calculated, and if the score is not 1, it is identified as an error.

[0075] Content normality check: For the specified fields , the predefined rules are called to execute the verification function , and its return value is of boolean type, indicating whether it conforms to the norm.

[0076] Explicit consistency check: For the core entity , the set of its values is extracted from all materials, and it is determined by the consistency function whether is greater than 1, and if it is greater than 1, it is identified as an error.

[0077] When any explicit error is detected, the intelligent feedback generation module converts the structured error information into an interactive guidance text for the applicant to achieve instant feedback and correction.

[0078] After the accurate recognition and verification in the first stage, the process enters the second stage, which embodies the design concept of intelligent decision-making. The AI recognition and processing center starts the implicit logical risk judgment. In this process, all text information from various sources is first aggregated to construct a global context , where represents aggregating all text information from various sources for aggregation, represents additional form information or other information in a specific format.

[0079] Next, the system executes the Retrieval-Augmented Generation (RAG) process, encodes into a query vector , and performs semantic retrieval in the government knowledge vector database. This knowledge base can store background knowledge files across domains to support cross-domain collaboration at the knowledge level. By calculating the cosine similarity , where represents the dot product of vectors and , represents the Euclidean norm of vector , represents the Euclidean norm of vector , and thus retrieves the top-K relevant knowledge entries .

[0080] Subsequently, the enhanced context is fed into the DeepSeek-V2671B large model for in-depth reasoning. The model is used to analyze whether there are deep logical matching conflicts or feasibility contradictions between the core capabilities reflected in the application materials and the application goals, and between the resource scale and the matter requirements. These contradictions are identified as implicit logical risks and form a risk set .

[0081] The system can further quantify the risks, for example, by calculating the total risk score through weighted summation , where is the weight, is the severity assessment score of each risk point.

[0082] Finally, the intelligent feedback generation module combines the judged risk set with the score to generate a decision-making assistance summary for the backend reviewers. This summary, together with the application material package, is pushed to the manual review link, realizing the effective combination of the intelligent decision-making results of Al and the final adjudication power of human experts, constituting cross-domain collaboration at the human-machine level, and completing the full-process enhancement of "accurate recognition-intelligent decision-making-cross-domain collaboration".

[0083] Example 2:

[0084] Referring to Figure 2 , in the second embodiment of the present invention, the present invention provides a material review method based on an AI model, including the following steps:

[0085] Receiving application information submitted by a user terminal and including one or more application materials;

[0086] Performing a first-stage explicit error check on the application materials. The explicit error check aims to determine whether there are preset explicit errors in the integrity, format standardization, and cross-material data explicit consistency of the application materials, and execute the corresponding judgment process;

[0087] If there are explicit errors, generating and feedbacking interactive guidance information aimed at guiding the user to make real-time modifications;

[0088] If there are no explicit errors, performing a second-stage implicit logic risk judgment on the application materials. The implicit logic risk judgment aims to construct multiple application materials into a global context and comprehensively judge whether there are preset implicit logic risks between the multiple application materials based on the global context;

[0089] Based on the result of the implicit logic risk judgment, generating and outputting a decision-making assistance summary for use in subsequent review processes.

[0090] Specifically, to further clarify the technical solution of the present invention, a specific application scenario is now used as an example for illustration. This embodiment aims to solve the technical problem in the prior art that applicants for government affairs services need to make multiple round trips to the administrative service center due to cumbersome preparation of application materials, missing materials, or filling errors, resulting in a long processing time and poor experience.

[0091] Scenario setting:

[0092] A certain technology enterprise (hereinafter referred to as the "applicant") intends to apply for a "High-tech Enterprise R & D Project Support Fund" online through the review platform described in the present invention. The application process for this fund is complex and requires submission of multiple application materials including, but not limited to, the "Project Application Form", scanned copy of the enterprise legal person business license, financial audit reports for the past three years, existing intellectual property certification documents, resumes of R & D team members, etc.

[0093] The implementation steps are as follows:

[0094] Step 1: Receive application information. The applicant accesses the interactive interface of this platform through the user terminal. First, fill in all the electronic fields of the "Project Application Form" online; then, batch upload the above-mentioned business license, audit report, intellectual property certificate and other documents in PDF or JPG format. The front-end module of the platform receives all the application information and packages it into an independent application event, triggering the background review process.

[0095] Step 2: Perform the first-phase explicit error check. At the moment when the applicant clicks "Submit", the platform immediately performs the first-phase explicit error check on this application event. This process requires no manual intervention and aims to discover and feedback all form-level errors in real time to avoid the application being rejected due to low-level errors.

[0096] Integrity check: Based on the application item of "High-tech Enterprise R & D Project Support Fund", the platform retrieves the set of required material lists from the knowledge base , when it includes the "Financial Audit Report for the Past Three Years", but the applicant only uploads the reports for the most recent two years.

[0097] The system compares the list of materials already submitted with through set operations , and identifies that the "Financial Audit Report for 2022" is missing. At this time, the integrity score .

[0098] Format standardization check: The platform checks the "Unified Social Credit Code" field filled in the "Project Application Form". When the applicant mistakenly fills in an 18-digit code as 17 digits. The system applies the preset regular expression rules to this field value , and its verification function returns a value of 0, determining that the format is not standard.

[0099] Cross-material data explicit consistency check: The platform extracts the core entity of "Enterprise Legal Representative" . In the "Project Application Form" filled in by the applicant, the value of this entity is "Li Ming"; however, in the scanned copy of the business license uploaded, the value of this entity is parsed as "Li Ming". However, in a scanned copy of the board resolution attached as an annex, due to the scribbled signature, it is recognized as "Li Peng" by the multi-modal parsing module. At this time, the value set of this entity is . The system executes the consistency function , and because , it determines that the information of this core entity is inconsistent among different materials.

[0100] Step 3: Generate and provide interactive guidance information based on explicit errors:

[0101] Since multiple explicit errors were identified in Step 2, the result of the judgment process is "Explicit errors exist". Instead of submitting this application to the background, the platform immediately invokes the intelligent feedback generation module. This module converts the above-structured error list (missing materials, field format errors, information inconsistencies) into natural language and presents it in the form of a pop-up window or highlighting on the user interface: "Hello, the system has detected the following issues that need to be corrected:

[0102] 1. Missing materials: Please supplement and upload the '2022 Annual Financial Audit Report';

[0103] 2. Filling error: The 'Unified Social Credit Code' should be 18 digits. Please check;

[0104] 3. Information inconsistency: The name of the 'Enterprise Legal Representative' is identified as 'Li Peng' in the board resolution, which does not match 'Li Ming' in other materials. Please check and confirm."

[0105] Through this step, the applicant does not need to wait for several days or even weeks for a manual review rejection notice. Instead, at the moment of submission, they obtain accurate and actionable modification guidelines and can immediately complete the supplementary upload and information correction. This directly addresses the technical pain points of "multiple submissions" and "cumbersome processing" due to material issues, reducing the communication cost that could have taken weeks to just a few minutes.

[0106] Step 4: If there are no explicit errors, perform the implicit logical risk assessment in the second stage:

[0107] After the applicant corrects all errors based on the real-time feedback and resubmits, the platform performs the first-stage verification again to confirm that there are no explicit errors.

[0108] At this time, the result of the judgment process is "No explicit errors", and the platform automatically triggers the implicit logical risk assessment in the second stage.

[0109] 1. Construct the global context: The platform aggregates the text content of all the application materials submitted by the applicant after correction and the online form data to construct a global context that contains the complete information of this application event . This context comprehensively describes the applicant's qualifications, financial status, technical strength, and the specific content of the current application project.

[0110] 2. Perform comprehensive assessment: The platform will Encode it as a query vector, and retrieve relevant policy documents and past successful cases from the government affairs knowledge vector database as background knowledge. Subsequently, send the enhanced context into the large language model for in-depth reasoning. The model discovers that the applicant claims in the "Project Application Form" that the total project investment is five million yuan, but the financial audit reports uploaded in the past three years show that the company's average annual net profit is only three hundred thousand yuan, and the asset-liability ratio is relatively high. This constitutes an implicit logical risk between the "scale of applied resources" and the "resources required to achieve the goal". In addition, the model also discovers that all the intellectual property certificates submitted by the applicant are "design patents", while the declared project is "underlying algorithm research and development", and the relevance between the two in the technical field is weak, constituting a risk of mismatch between the "core capabilities" and the "application objectives".

[0111] Step 5: Generate and output a decision-making assistance summary:

[0112] Based on the above research and judgment results, the platform generates a decision-making assistance summary for the subsequent manual review process. This summary clearly lists:

[0113] 1. Financial feasibility risk: There is a significant gap between the applicant's own profitability and the high investment required for the project, and the source of project funds is uncertain. It is recommended to pay attention to the effectiveness of its financing plan, and the risk assessment score is relatively high.

[0114] 2. Technical ability matching risk: The applicant's past technical accumulation (design patent) has a low correlation with the hardcore technical R & D direction declared this time, and there are doubts about the project's technical implementation ability.

[0115] Embodiment 3:

[0116] In the third embodiment of the present invention, based on the same inventive concept, a computer-readable storage medium proposed by the present invention stores a computer program, and when the computer program is executed by a processor, it realizes the steps of a material review method based on an AI model in the above embodiment.

[0117] Embodiment 4:

[0118] In the fourth embodiment of the present invention, based on the same inventive concept, a computer terminal proposed by the present invention includes: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute a material review method based on an AI model in the above embodiment.

[0119] It should be understood that each part of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following techniques well known in the art or a combination thereof can be used: discrete logic circuits having logic gate circuits for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0120] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An AI model-based material review processing platform, characterized in that, Including: A front-end interaction and presentation module, which is used to receive application information containing one or more application materials from a user terminal; An AI cognition and processing center, which is used to receive the application information and perform a first-stage explicit error check and a second-stage implicit logical risk judgment on the application materials; The explicit error check aims to determine whether there are preset explicit errors in the integrity, format standardization, and cross-material data explicit consistency of the application materials; The implicit logical risk judgment aims to construct the multiple application materials into a global context and comprehensively judge whether there are preset implicit logical risks between the multiple application materials based on the global context; An intelligent feedback generation module, which generates corresponding review feedback information based on the result of the explicit error check and / or the result of the implicit logical risk judgment; Among them, the review feedback information is sent to the front-end interaction and presentation module for presentation to the user terminal.

2. The material review processing platform based on the AI model according to claim 1, wherein The AI cognition and processing center includes: A multi-modal material analysis sub-module, which is used to parse the application materials in various different formats uploaded by the user terminal into unified structured data containing text content, layout structure information, and semantic type information before performing the first-stage explicit error check, for use in the explicit error check and the implicit logical risk judgment.

3. The material review processing platform based on the AI model according to claim 1, characterized in that When the AI cognition and processing center performs the first-stage explicit error check, it is used to: Compare the list of submitted materials with a preset list of necessary materials for integrity check; Match the keyword fields in the application materials with preset check rules for format standardization check; Cross-compare the same core entity information between different application materials for cross-material data explicit consistency check.

4. The material review processing platform based on the AI model according to claim 1, characterized in that, When the AI cognition and processing center performs the second-stage implicit logical risk judgment, it is used to: [[ID= ​ 5. The material review processing platform based on the AI model according to claim 1, characterized in that, ​ ​ ​ ​ 6. The material review processing platform based on the AI model according to claim 1, characterized in that ​ ​ When the result of the implicit logical risk judgment indicates the existence of an implicit logical risk, a risk summary report for the reference of backend reviewers' decision-making is generated as the review feedback information.

7. The material review processing platform based on the AI model according to claim 1, characterized in that The platform further includes: A business process and scheduling module for controlling the execution process of the AI cognition and processing center, and the process includes: After receiving the application information, first trigger the explicit error check in the first stage, and feedback the generated interactive guidance information to the user; After determining that there are no explicit errors in the application materials, trigger the implicit logical risk judgment in the second stage to complete the complete review of the application information.

8. A material review method based on an AI model, characterized in that, For the material review processing platform based on the AI model according to any one of claims 1-7, the method includes the following steps: Receive application information submitted by a user terminal and including one or more application materials; Perform an explicit error check on the application materials in the first stage, and the explicit error check is intended to determine whether there are preset explicit errors in the integrity, format standardization, and explicit cross-material data consistency of the application materials, and execute the corresponding judgment process; If there are such explicit errors, generate and feedback interactive guidance information for guiding the user to make real-time modifications; If there are no such explicit errors, perform an implicit logical risk judgment on the application materials in the second stage, and the implicit logical risk judgment is intended to construct the multiple application materials into a global context, and comprehensively judge whether there are preset implicit logical risks among the multiple application materials based on the global context; Generate and output a decision-making assistance summary based on the result of the implicit logical risk judgment for use in subsequent review processes.

9. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a material review method based on the AI model as described in claim 8.

10. A readable storage medium, characterized in that, A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, it implements a material review method based on the AI model as described in claim 8.

Images