A qualification auditing method, system, device and medium

Abstract

The application is suitable for the field of computer technology, and provides a qualification auditing method, which comprises the following steps: obtaining application metadata of an application to be audited, generating a qualification auditing list through a dynamic demand perception model; receiving a qualification file uploaded by a user, extracting information from the qualification file in combination with the qualification auditing list, obtaining qualification auditing content, and identifying the authenticity of the qualification auditing content; based on the qualification auditing content, performing authorization chain analysis from a global authorization network graph to obtain a target authorization subgraph, and performing topology health degree evaluation on the target authorization subgraph; and based on the target authorization subgraph, performing comprehensive auditing on the qualification auditing content. The qualification auditing list generated by the application avoids the preparation and inspection of irrelevant qualifications, and the auditing of qualifications is adapted to different business modes and market demands, so as to ensure the comprehensiveness of the auditing.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and in particular relates to a qualification verification method, system, device and medium. Background Technology

[0002] When submitting an application, such as a game, to the App Store or other Android app markets, it usually requires an approval process. The operator needs to provide a large number of qualification documents, such as a business license, ICP license, software copyright certificate, ISBN number, and a complex authorization chain for qualification review.

[0003] Existing qualification verification methods typically involve obtaining qualification materials provided by users, then using OCR technology to identify and extract information from these materials for subsequent review. However, OCR technology has limited recognition accuracy and a single recognition dimension, making it unable to accurately identify various types of qualification materials. Furthermore, the review checklists used for verification are predefined and cannot be flexibly adapted to different application types, making it difficult to meet the needs of flexible business models and market demands.

[0004] Therefore, this application proposes a qualification review method that can dynamically perceive and identify the review list from multiple dimensions, in order to improve the review effect. Summary of the Invention

[0005] This application provides a qualification review method, system, device, and medium that can solve one of the aforementioned problems in the prior art.

[0006] In a first aspect, embodiments of this application provide a qualification verification method, including: Obtain application metadata of the applications to be reviewed, and generate a qualification review list through a dynamic demand awareness model; Receive the qualification documents uploaded by the user, extract information from the qualification documents in conjunction with the qualification review checklist, obtain the qualification review content, and verify the authenticity of the qualification review content; Based on the qualification review content, authorization chain analysis is performed from the global authorization network graph to obtain the target authorization subgraph, and the topology health of the target authorization subgraph is evaluated. Based on the target authorization subgraph, a comprehensive review of the qualification review content is conducted.

[0007] Furthermore, the step of obtaining the application metadata of the application to be reviewed, and generating a qualification review list through a dynamic demand-aware model, includes: The application metadata of the application to be reviewed is parsed and quantified to obtain the application feature vector. The application metadata includes application category, content description text, operation mode description and target market list. The application feature vectors are matched with the qualification rule knowledge base through multiple channels to obtain the qualification review list; Each qualification in the qualification review list is labeled with a qualification attribute, which includes the level of necessity, expected document type, and authorization chain depth requirement.

[0008] Furthermore, the qualification rule knowledge base is a qualification rule knowledge graph, specifically including qualification type nodes, regulatory field nodes, operating model nodes, target market nodes, and keyword nodes; Establish association attributes between the qualification type node and other node types to form qualification trigger rule paths, wherein the association attribute types include requirement type, trigger type, location type, and belonging type.

[0009] Furthermore, the multi-channel matching includes semantic channel matching and logical channel matching; The step of performing multi-channel matching between the application feature vector and the qualification rule knowledge base to obtain the qualification review list includes: In the semantic channel matching, the vectorized content description text is matched with the keyword node for similarity, so as to trigger the keyword node and the corresponding qualification triggering rule path; In the logical channel matching, the vectorized application category, operation mode description and target market list are sequentially matched or judged with the regulatory field node, the operation mode node and the target market node in a symbolic way to trigger the corresponding node and the corresponding qualification trigger rule path. A qualification type node is added to the qualification review list only when all necessary condition nodes leading to it are triggered.

[0010] Furthermore, the process of receiving the qualification documents uploaded by the user, combining them with the qualification review checklist, extracting information from the qualification documents to obtain qualification review content, and verifying the authenticity of the qualification review content includes: Obtain the expected document type for each qualification in the qualification review list, and assign a multimodal processing pipeline for each expected document type from the predefined processing strategy library; Each of the qualification documents is input into the corresponding multimodal processing pipeline to extract the qualification review content and verify its authenticity. The processing results of each of the multimodal processing pipelines are aligned with the qualification review list to generate a review list item that corresponds one-to-one with each qualification in the qualification review list.

[0011] Furthermore, based on the qualification review content, the step of performing authorization chain analysis from the global authorization network graph to obtain the target authorization subgraph, and then performing a topology health assessment on the target authorization subgraph, includes: Based on historical audit data, a global authorization network graph is constructed with entities and qualification assets as nodes. Entities are companies, individuals, or organizations, and qualification assets are specific intellectual property rights or administrative licenses. Starting with the submitter of the application to be reviewed, and combining the authorization chain depth requirements corresponding to each item in the review list, the target authorization subgraph is constructed by traversing the global authorization network graph. Calculate the topology health index of the target authorization subgraph, and measure the stability of the authorization chain of the qualification review content through the topology health index, wherein the topology health index includes path redundancy and critical node dependency.

[0012] Furthermore, the comprehensive review of the qualification verification content based on the target authorization subgraph includes: Align and review the qualification review content with the target authorization subgraph to generate a composite feature representation for each review list item; The composite feature representation is input into a multi-task learning model, and the multi-task learning model is used to verify the qualification validity, authorization logic consistency and commercial rationality of each audit list item to obtain preliminary verification results; Based on the necessity level of each item on the audit checklist, a dynamic weighted comprehensive decision is made on the preliminary verification results to generate a final audit conclusion.

[0013] Secondly, embodiments of this application provide a qualification verification system, including: The first processing module is used to obtain the application metadata of the application to be reviewed and generate a qualification review list through a dynamic demand awareness model. The second processing module is used to receive the qualification documents uploaded by the user, extract information from the qualification documents in conjunction with the qualification review list, obtain the qualification review content, and verify the authenticity of the qualification review content. The third processing module is used to perform authorization chain analysis from the global authorization network graph based on the qualification review content, obtain the target authorization subgraph, and perform topology health assessment on the target authorization subgraph. The fourth processing module is used to conduct a comprehensive review of the qualification review content based on the target authorization sub-graph.

[0014] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned qualification review method.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium, including a computer program stored in the computer-readable storage medium, which, when executed by a processor, implements the qualification review method described above.

[0016] The beneficial effects of the embodiments in this application compared with the prior art are: This application discloses a qualification review method that automatically generates a qualification review list precisely matched to the current application through a dynamic demand-aware model, avoiding the preparation and inspection of irrelevant qualifications. Simultaneously, the qualification review is adapted to different business models and market demands, ensuring comprehensiveness. Furthermore, based on the qualification review list, corresponding multimodal processing pipelines are invoked, avoiding performance losses caused by using a general model to process various qualification documents, ensuring the accuracy of extracted content. Further, by constructing a global authorization network graph, complex authorization relationships can be obtained, and based on this, target authorization subgraphs related to the qualification review content can be obtained, ensuring the integrity and effectiveness of the authorization relationships of the application to be reviewed. This allows for a comprehensive review of the qualification review content, improving the review efficiency of the application to be reviewed and increasing the pass rate of the application. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a qualification review method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a qualification review system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0019] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0020] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0021] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0023] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0025] Please see Figure 1 As shown, this invention is a qualification verification method, comprising the following steps: S100. Obtain the application metadata of the application to be reviewed, and generate a qualification review list through the dynamic demand awareness model. This application uses a dynamic demand awareness model to automatically generate a qualification review checklist that is precisely matched to the current application, avoiding the preparation and inspection of irrelevant qualifications. At the same time, the qualification review is adapted to different business models and market demands, thereby ensuring the comprehensiveness of the review.

[0026] In some embodiments, step S100 above includes: The application metadata of the application to be reviewed is parsed and quantified to obtain the application feature vector. The application metadata includes application category, content description text, operation mode description and target market list. The application feature vectors are matched with the qualification rule knowledge base through multiple channels to obtain the qualification review list; Each qualification in the qualification review list is labeled with a qualification attribute, which includes the level of necessity, expected document type, and authorization chain depth requirement.

[0027] In this embodiment, based on the application metadata of the application to be reviewed and combined with the dynamic demand awareness model, all possible qualifications for reviewing the application are listed. The generated qualification review list is different for different applications to be reviewed, so that the qualification review method of this application can meet the review requirements of different applications and adapt to different business models and market demands. For example, if an application description contains keywords such as "online live broadcast" and "virtual gifts", the dynamic demand awareness model will trigger the review requirements for the "Internet Culture Business License" and the "Radio and Television Program Production and Operation License". If the application is released in both "China" and "North America", the qualification review list must require qualifications in both China and North America, and indicate that there may be differentiated authorization requirements, so as to ensure the comprehensiveness of the review.

[0028] Specifically, the system receives application metadata submitted by application developers, including application category, content description text, operation model description, and target market list. The application category is the application type, such as "game-role-playing" or "tool-productivity". The content description text is a detailed textual description of the application's functions, content, or gameplay. The operation model description is the specific operation method of the application, such as "free download, in-app purchase", "subscription", "buyout", or "includes ads". The target market list is a list of countries or regions where the application will be launched.

[0029] More specifically, the metadata of each application undergoes standardized preprocessing. For example, for content description text, a pre-trained natural language processing model is used for named entity recognition and keyword extraction. For instance, sensitive words such as "live streaming," "social networking," "financial management," and "medical consultation" are identified in the text, which are key signals that trigger specific qualification requirements. For descriptions of operating models, a text classification model is used to map them to several predefined standard models, such as "in-app purchase" and "subscription." Understandably, different operating models correspond to different business qualifications and authorization requirements. Finally, the metadata of each application is converted into standard feature vectors and concatenated to generate application feature vectors, which are used to generate the subsequent qualification review list.

[0030] More specifically, the dynamic demand perception model can perform multi-channel matching between application feature vectors and qualification rule knowledge base. Specifically, the dynamic demand perception model includes an application feature extraction module, a qualification rule knowledge base management module, and a multi-channel matching engine. The modules work together to achieve a complete process from application feature capture to accurate matching with the rule knowledge base.

[0031] In addition, for each qualification in the qualification review list, its qualification attributes need to be marked to provide a basis for subsequent qualification content review, so that each review conclusion can be traced back to the specific attribute definition and verification rules, ensuring the transparency and explainability of the entire review process.

[0032] Specifically, the qualification attributes include the necessity level, expected file type, and authorization chain depth requirements. The necessity level includes mandatory, conditional, and recommended. Mandatory means that the absence or invalidity of the qualification will directly affect the final review conclusion, i.e., the review will fail directly. Conditional means depends on the specific function or mode of the application. If the authorization chain logic pointed to by the qualification fails, it may be judged as a failure in subsequent steps, depending on the conditions set for this qualification. Recommended means that the absence or defect of the qualification will not lead to a failure, but it will affect its overall credibility in the app store, and ultimately affect its search ranking or recommendation weight in the app store.

[0033] More specifically, the expected document type defines the specific document format and content specifications required to prove the qualification. It is a composite attribute including a document template identifier, a set of core information fields, and document format requirements. The document template identifier points to a predefined document template ID, such as "Biz_License_CN" for a Chinese business license and "Copyright_Auth_Letter" for a copyright authorization letter. This document template indicates the identification method required to identify the qualification type, enabling information extraction from various qualification documents. The set of core information fields represents the key data fields to be extracted from this type of qualification document. For example, for "Biz_License_CN," the field set includes: ["Company Name," "Unified Social Credit Code," "Legal Representative," "Registered Capital," "Date of Establishment," "Validity Period"]. This serves as the basis for extracting qualification review content. The document format requirements specify the acceptable digital file formats for this type of qualification document, such as [".pdf", ".jpg", ".png"].

[0034] More specifically, the authorization chain depth requirement defines the qualifications requiring an authorization chain, its traceability verification targets and boundaries, and provides the data foundation for the generation of subsequent target authorization subgraphs. It typically indicates the target node to be traced in the authorization chain. For example, for copyright, the target is usually the original copyright holder, which can be an individual or a company; for ISBN numbers, the target is usually the general publishing unit. The authorization chain depth requirement allows traversal within a pre-defined global authorization network graph, specifying traversal methods such as recursive reverse tracing. It also sets verification rules for logical checks after constructing the target authorization subgraph, typically including authorization scope inclusion checks and time continuity checks. The authorization scope inclusion check specifically states that the scope of rights of a lower-level licensee must not exceed the scope of rights of its higher-level licensor, while the time continuity check states that the validity period of a lower-level authorization must be entirely within the validity period of its higher-level authorization. Understandably, for qualifications that do not require authorization chain verification, the corresponding authorization chain depth requirement in the qualification attribute is empty, indicating that authorization analysis of this qualification is not required in subsequent authorization chain analysis.

[0035] In some embodiments, the qualification rule knowledge base is a qualification rule knowledge graph, specifically including qualification type nodes, regulatory field nodes, operating model nodes, target market nodes, and keyword nodes; Establish association attributes between the qualification type node and other node types to form qualification trigger rule paths, wherein the association attribute types include requirement type, trigger type, location type, and belonging type.

[0036] In this embodiment, the qualification rule knowledge base is a graph-based rule system, specifically a qualification rule knowledge graph. Its nodes represent qualification types and the conditions that trigger the qualification type, and the edges represent the conditional logic that triggers the qualification requirement. The matching process of the application feature vector in the qualification rule knowledge base is a reasoning process on the graph that activates the corresponding node path based on the application feature vector.

[0037] Specifically, the node types of the aforementioned qualification rule knowledge graph include main nodes and condition nodes. The association attribute types between main nodes and condition nodes include requirement type, trigger type, location type, and belonging type. The main node is specifically a qualification type node, while the condition nodes include regulatory field nodes, operating model nodes, target market nodes, and keyword nodes. The main node and various condition nodes are connected through association attributes, ultimately forming the qualification trigger rule path.

[0038] Specifically, a unique node identifier is assigned to each node instance within each type of node, and nodes are associated with each other through edges. For example, in one embodiment, qualification type nodes include ["Node ID: A_001, Name: 'Network Culture Business License'", "Node ID: A_002, Name: ISBN Number", ...], regulatory field nodes include ["Node ID: B_101, Name: Online Game", "Node ID: B_102, Name: Online Live Streaming"], operation mode nodes include ["Node ID: C_201, Name: In-App Purchase"], target market nodes include ["Node ID: D_301, Name: China"], and keyword nodes include ["Node ID: E_401, Name: "Live Streaming", vector:[0.12, -0.45, ..., ...]]. [0.67]”], It is understandable that each keyword node has a pre-stored corresponding word vector. By comparing the word vectors, it is determined whether the keyword node is triggered. Furthermore, regarding the definition of edge relationships, as in one embodiment, (Regulatory domain: online games) - [Requirements] -> (Qualification type: ISBN number), (Keyword: live streaming) - [Trigger] -> (Regulatory domain: online performance), (Regulatory domain: online performance) - [Requirements] -> (Qualification type: "Internet Culture Business License"), (Qualification type: "Internet Culture Business License") - [Located in] -> (Target market: China), (Operation model: in-app purchase) - [Trigger] -> (Qualification type: ICP license). Thus, by associating nodes with each other through edges, qualification trigger rule paths are generated.

[0039] More specifically, the qualification rule knowledge base in this embodiment uses natural language processing technology to automatically parse newly released regulations and policy documents from official sources, and then adds, deletes, and modifies them into the nodes and edges of the corresponding qualification rule knowledge graph. This allows the qualification rule knowledge base to keep up with changes in policies and regulations. At the same time, it also receives rules manually added or adjusted by platform operators through a graphical interface, thereby dynamically adjusting the knowledge base to ensure the accuracy and completeness of the rules.

[0040] In some embodiments, the multi-channel matching includes semantic channel matching and logical channel matching; The step of performing multi-channel matching between the application feature vector and the qualification rule knowledge base to obtain the qualification review list includes: In the semantic channel matching, the vectorized content description text is matched with the keyword node for similarity, so as to trigger the keyword node and the corresponding qualification triggering rule path; In the logical channel matching, the vectorized application category, operation mode description and target market list are sequentially matched or judged with the regulatory field node, the operation mode node and the target market node in a symbolic way to trigger the corresponding node and the corresponding qualification trigger rule path. A qualification type node is added to the qualification review list only when all necessary condition nodes leading to it are triggered.

[0041] In this embodiment, the vectorized application feature vectors are classified into semantic channels or logical channels for matching according to their data attributes to ensure matching accuracy. The semantic channel is used to process unstructured content description text. The similarity of the vectorized content description text with the word vectors pre-stored in each keyword node of the qualification rule knowledge base is calculated. If the similarity exceeds the threshold, the keyword node and its associated rule path are activated. The logical channel is used to process structured metadata, namely the vectorized application category, operation mode description, and target market list. The vectorized metadata is matched or the relationship is judged with the corresponding regulatory field node, operation mode node, and target market node in the qualification rule knowledge base. If the match is successful, the corresponding node is activated.

[0042] In this embodiment, the multi-channel matching process is a forward reasoning process on the qualification rule knowledge graph. Specifically, when the keyword node and the target market node are triggered, the trigger signal will be propagated upstream along the related attributes such as requirements or triggers. Finally, the qualification type node is determined to be required and added to the qualification review list only when all the necessary condition nodes leading to a certain qualification type node are activated.

[0043] In one embodiment, the content description text is "an application that provides online voice live streaming social networking". The sentence vector of this application has a word vector similarity of 0.92 with the word vector of the "live streaming" node in the keyword node. At this time, the "keyword: live streaming" node is triggered. This node propagates the trigger signal to the "regulatory field: online performance" node through the trigger type edge. The "regulatory field: online performance" node triggers the "qualification type: Internet Culture Business License" node by requesting the type edge. At the same time, the "qualification type: Internet Culture Business License" triggers the "target market: China" node through the type edge. Finally, the qualification type: Internet Culture Business License is added to the qualification review list. The qualification type: Internet Culture Business License is marked with the corresponding qualification attributes.

[0044] S200: Receive the qualification documents uploaded by the user, extract information from the qualification documents in conjunction with the qualification review list, obtain the qualification review content, and verify the authenticity of the qualification review content. Based on the qualification review checklist, this application calls the corresponding multimodal processing pipeline, avoiding the performance loss caused by using a general model to process various qualification documents, and ensuring the accuracy of the extracted content.

[0045] In some embodiments, step S200 above includes: Obtain the expected document type for each qualification in the qualification review list, and assign a multimodal processing pipeline for each expected document type from the predefined processing strategy library; Each of the qualification documents is input into the corresponding multimodal processing pipeline to extract the qualification review content and verify its authenticity. The processing results of each of the multimodal processing pipelines are aligned with the qualification review list to generate a review list item that corresponds one-to-one with each qualification in the qualification review list.

[0046] In this embodiment, based on the expected file type of the qualification attributes carried by each qualification type in the qualification review list, the method for information extraction is determined, thereby selecting the corresponding multimodal processing pipeline to extract content and verify the authenticity of the qualification files corresponding to the corresponding qualification types. At the same time, the extracted content is aligned with the qualification type to generate the corresponding review list items, and the authenticity of each review list item is verified to ensure the accuracy of the extracted content.

[0047] Specifically, the process involves obtaining user-uploaded qualification documents, categorizing them according to filename or format, and combining this with the qualification review checklist to their corresponding qualification types. Based on the file template identifier corresponding to each qualification type, a predefined multimodal processing pipeline is used to extract the qualification review content from a predefined processing strategy library. For example, in one embodiment, a user uploads a business license, which is categorized under the "Enterprise Legal Person Qualification" qualification type. The corresponding qualification attribute for this type has a file template identifier of "Biz_License_CN," and its field set includes: ["Enterprise Name," "Unified Social Credit Code," "Legal Representative"]. The data includes fields such as "Person", "Registered Capital", "Date of Establishment", and "Validity Period". The file format is [".pdf", "".jpg", "".png"]. Based on this field set, the multimodal processing pipeline is determined from the processing strategy library as structure flow -> text flow -> visual flow -> special identification. Further, the structure flow locates the area where each field in the field set of the business license is located, and then the image of this area is input into the text flow for recognition, thereby obtaining the specific review content of each field. At the same time, the visual flow loads the official seal template features of the administrative department and inputs them into the special identification for strict comparison with the detected seal, thereby performing authenticity verification.

[0048] Understandably, the predefined processing strategy library stores multimodal processing pipelines for qualification documents corresponding to different qualification types. Each multimodal processing pipeline defines the set of key information fields that need to be extracted from the qualification document, as well as the pipeline process and special authenticity verification process that need to be enabled. For example, for "business license" documents, the pipeline relies on structured flow to locate fields such as "unified social credit code" and "company name" and enables the official seal template comparison function of visual flow. For "authorization letter" documents, the pipeline relies on the named entity recognition function of text flow to extract fields such as "authorizer", "authorized party", "authorized subject", and "validity period" and enables the signature handwriting consistency analysis function of visual flow. For "license" documents, the pipeline needs to call text flow to extract the license number and validity period and call visual flow to identify its official watermark or anti-counterfeiting QR code. That is, different qualification documents have different processing pipelines.

[0049] Furthermore, the processing results of each multimodal processing pipeline are aligned with each qualification in the qualification review list to check whether the user-uploaded file set covers all qualifications in the qualification review list. If the qualification file corresponding to a mandatory qualification in the qualification review list is not in the user-uploaded file set, the user is prompted to upload the file to ensure the completeness of the review results. For the extracted relevant qualification review content, a review list item is generated by matching it one-to-one with each qualification for subsequent comprehensive review.

[0050] S300. Based on the qualification review content, perform authorization chain analysis from the global authorization network graph to obtain the target authorization subgraph, and conduct topology health assessment on the target authorization subgraph; In this application, by constructing a global authorization network graph, complex authorization relationships can be obtained, and based on this, target authorization subgraphs related to the qualification review content can be obtained, ensuring the integrity and effectiveness of the authorization relationships of the application to be reviewed, and using this to conduct a comprehensive review of the qualification review content, thereby improving the review efficiency of the application to be reviewed and also increasing the pass rate of the application to be reviewed.

[0051] In some embodiments, step S300 above includes: Based on historical audit data, a global authorization network graph is constructed with entities and qualification assets as nodes. Entities are companies, individuals, or organizations, and qualification assets are specific intellectual property rights or administrative licenses. Starting with the submitter of the application to be reviewed, and combining the authorization chain depth requirements corresponding to each item in the review list, the target authorization subgraph is constructed by traversing the global authorization network graph. Calculate the topology health index of the target authorization subgraph, and measure the stability of the authorization chain of the qualification review content through the topology health index, wherein the topology health index includes path redundancy and critical node dependency.

[0052] In this embodiment, it is necessary not only to review the qualification content, but also to verify whether the authorization chain of each qualification is complete and valid, thereby improving the effectiveness of qualification review.

[0053] In this embodiment, a global authorization network graph is constructed based on historical audit data as the foundation for the authorization chain analysis of qualification audit content. Specifically, the global authorization network graph includes entity nodes and qualification asset nodes. Entity nodes represent companies, individuals, or organizations, and their key attributes include entity_id, name, entity_type, credit_code, and status. Specifically, entity_id is a unique identifier generated based on the entity name and / or credit code; name represents the name; entity_type represents the entity type, such as Company or Individual; credit_code represents the unified social credit code (which can be empty when type is Individual); and status represents the entity status, such as normal or abnormal. Qualification asset nodes represent specific intellectual property rights or administrative licenses, and their key attributes include asset_id, asset_type, registration_number, and validity_period. Specifically, asset_id represents a unique hash; asset_type represents the asset type, such as Software... The system uses data such as reCopyright, ISBN, ICP, registration_number (registration number), and validity_period (validity period) to connect the nodes. Specifically, HOLDS represents the holding relationship, connecting the principal node and the qualification asset node, indicating that the principal is the legal holder of the qualification. Other relationship attributes include acquisition_date (acquisition date) and GRANTS (authorization relationship), connecting two qualification asset nodes and associating them with a principal node, representing the authorization behavior between the principal and the qualification asset. GRANTS attributes include grant_id, scope, territory, start_date, end_date, and source_file. grant_id represents a unique ID, scope represents the authorization scope (e.g., "global issuance"), territory represents the authorization region, start_date represents the authorization start date, end_date represents the authorization end date, and source_file represents the source file hash for traceability.

[0054] In this embodiment, when traversing the global authorization network graph, the submitter of the application to be reviewed is taken as a main node. Simultaneously, based on the authorization chain depth requirements in the review list items, a correlation query is performed from the global authorization network graph. The query process employs a bounded recursive query method. Specifically, in one embodiment, starting with "Company Z," the global authorization network graph is traced backward along the GRANTS edges, using the authorization chain depth requirements of each review list item as the boundary. For example, for the "Software Copyright" qualification, the node authorized to Company Z is traced backward along the GRANTS edges. At each node, the node type is judged to determine if it meets the target node set in the authorization chain depth requirements. If it does, the query is considered successful, and an authorization chain is constructed with the target node as the starting point and Company Z as the ending point. For each review list item in the qualification review list, the same bounded recursive query method is used to obtain the corresponding authorization chain. Finally, one or more target authorization subgraphs are constructed, with Company Z as the ending point and the target nodes of each authorization chain depth requirement as the starting point.

[0055] In some embodiments, during the traversal of the global authorization network graph, in order to prevent loops or excessive searching, a maximum number of recursive steps, such as 10 steps, is set for each type of qualification. If the target node is not found after reaching the maximum number of steps, it is determined to be a tracing failure, thereby avoiding infinite exploration of the global authorization network graph and thus wasting computing resources.

[0056] Furthermore, in the target authorization subgraph, its topology health index is calculated to measure the stability of the authorization chain in the qualification review content. The topology health index includes at least path redundancy and critical node dependency. Path redundancy specifically checks whether there are multiple independent authorization paths from any qualification node to its end user. If the redundancy is low, it indicates that the authorization chain of the qualification is fragile. For example, in one embodiment, in its target authorization subgraph, in addition to the authorization chain path "Author -> Company A -> Company Z", there is also an authorization chain path "Author -> Company B -> Company Z". Even if Company A has a problem, Company Z's authorization will still remain valid through Company B. The critical node dependency index is calculated by counting the node with the highest intermediary centrality in the target authorization subgraph. If the node fails, it will cause the proportion of broken paths in the subgraph. This index reflects the fragility of the authorization chain that the current review application depends on. For example, in one embodiment, 90% of the paths in the target authorization subgraph must pass through "Copyright Agency Company C". The dependency of the target authorization subgraph on "Copyright Agency Company C" is 90%, thereby quantifying the risk of authorization chain breakage faced by the application.

[0057] S400. Based on the target authorization sub-graph, a comprehensive review of the qualification review content is conducted.

[0058] In some embodiments, step S400 above includes: Align and review the qualification review content with the target authorization subgraph to generate a composite feature representation for each review list item; The composite feature representation is input into a multi-task learning model, and the multi-task learning model is used to verify the qualification validity, authorization logic consistency and commercial rationality of each audit list item to obtain preliminary verification results; Based on the necessity level of each item in the audit checklist, a dynamic decision is made on the preliminary verification results to generate a final audit conclusion.

[0059] In this embodiment, the text information and authenticity verification results in the qualification review content are concatenated with the graph neural network embedding vector formed by the corresponding nodes and paths in the target authorization subgraph to generate a composite feature representation. Thus, the composite feature representation of each qualification being reviewed simultaneously includes its own document evidence features and its authorization features in the global authorization network, avoiding review bias that may be caused by a single feature source, and providing a data foundation for accurately judging the validity, consistency and rationality of the qualification in the future.

[0060] Furthermore, the composite feature representation is input into a multi-task learning model, which executes three verification tasks in parallel: qualification validity verification, authorization logic consistency verification, and commercial reasonableness inference. Specifically, the multi-task learning model includes three dedicated task modules: a validity verification module, a consistency verification module, and a reasonableness inference module. Specifically, the composite feature representation is input into this multi-task learning model; the validity verification module outputs the validity status probability distribution of the qualification documents corresponding to each qualification; the consistency verification module outputs the logical consistency score of each qualification in the authorization chain; and the reasonableness inference module outputs the commercial reasonableness score of the authorization behavior involved in that qualification.

[0061] Furthermore, based on the necessity level of each item on the audit checklist, dynamic decisions are made regarding the preliminary verification results. Specifically, for "mandatory necessary" qualifications, if the "validity verification" result fails, the overall audit conclusion is failure. For "conditionally necessary" qualifications, the pass status of the qualification is determined by the verification results of "validity verification" and "consistency verification." For example, in one embodiment, for "software copyright license," the audit conditions are that the license itself must be valid and the authorization chain must be complete and compliant. During the judgment process, if both conditions are met, the qualification passes; if one or two conditions are not met, the audit fails. For "recommended" qualifications, the verification results of the three aspects generate a comprehensive risk score, which serves as the risk rating for the overall application.

[0062] Please see Figure 2 As shown, the present invention also provides a qualification review system, the system comprising: First processing module 201: Used to obtain application metadata of the application to be reviewed, and generate a qualification review list through a dynamic demand awareness model; The second processing module 202 is used to receive the qualification documents uploaded by the user, extract information from the qualification documents in conjunction with the qualification review list, obtain the qualification review content, and verify the authenticity of the qualification review content. The third processing module 203 is used to perform authorization chain analysis from the global authorization network graph based on the qualification review content, obtain the target authorization subgraph, and perform topology health assessment on the target authorization subgraph. The fourth processing module 204 is used to conduct a comprehensive review of the qualification review content based on the target authorization subgraph.

[0063] It is understandable that, such as Figure 1 The content of the qualification review method embodiments shown is applicable to the qualification review system embodiments. The specific functions implemented by the qualification review system embodiments are the same as those shown in the examples. Figure 1 The qualification verification method shown is the same as the embodiment, and the beneficial effects achieved are the same as those shown. Figure 1 The beneficial effects achieved by the qualification review method shown in the embodiment are also the same.

[0064] It should be noted that the information interaction and execution process between the above systems are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0065] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0066] Please see Figure 3As shown, this embodiment of the invention also provides a computer device 3, including: a memory 302 and a processor 301, and a computer program 303 stored on the memory 302. When the computer program 303 is executed on the processor 301, it implements the qualification review method as described in any of the above methods.

[0067] The computer device 3 may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that... Figure 3 The computer device 3 is merely an example and does not constitute a limitation on the computer device 3. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0068] The processor 301 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0069] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may be an external storage device of the computer device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 3. Furthermore, the memory 302 may include both internal and external storage units of the computer device 3. The memory 302 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0070] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the qualification review method as described in any of the above methods.

[0071] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographic device / computer device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0072] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A qualification verification method, characterized in that, include: Obtain application metadata of the applications to be reviewed, and generate a qualification review list through a dynamic demand awareness model; Receive the qualification documents uploaded by the user, extract information from the qualification documents in conjunction with the qualification review checklist, obtain the qualification review content, and verify the authenticity of the qualification review content; Based on the qualification review content, authorization chain analysis is performed from the global authorization network graph to obtain the target authorization subgraph, and the topology health of the target authorization subgraph is evaluated. Based on the target authorization subgraph, a comprehensive review of the qualification review content is conducted.

2. The method as described in claim 1, characterized in that, The process of obtaining application metadata of the application to be reviewed, and generating a qualification review list through a dynamic demand-aware model, includes: The application metadata of the application to be reviewed is parsed and quantified to obtain the application feature vector. The application metadata includes application category, content description text, operation mode description and target market list. The application feature vectors are matched with the qualification rule knowledge base through multiple channels to obtain the qualification review list; Each qualification in the qualification review list is labeled with a qualification attribute, which includes the level of necessity, expected document type, and authorization chain depth requirement.

3. The method as described in claim 2, characterized in that, The qualification rule knowledge base is a qualification rule knowledge graph, specifically including qualification type nodes, regulatory field nodes, operating model nodes, target market nodes, and keyword nodes; Establish association attributes between the qualification type node and other node types to form qualification trigger rule paths, wherein the association attribute types include requirement type, trigger type, location type, and belonging type.

4. The method as described in claim 3, characterized in that, The multi-channel matching includes semantic channel matching and logical channel matching; The step of performing multi-channel matching between the application feature vector and the qualification rule knowledge base to obtain the qualification review list includes: In the semantic channel matching, the vectorized content description text is matched with the keyword node for similarity, so as to trigger the keyword node and the corresponding qualification triggering rule path; In the logical channel matching, the vectorized application category, operation mode description and target market list are sequentially matched or judged with the regulatory field node, the operation mode node and the target market node in a symbolic way to trigger the corresponding node and the corresponding qualification trigger rule path. A qualification type node is added to the qualification review list only when all necessary condition nodes leading to it are triggered.

5. The method as described in claim 1, characterized in that, The process of receiving the qualification documents uploaded by the user, combining them with the qualification review checklist, extracting information from the qualification documents to obtain qualification review content, and verifying the authenticity of the qualification review content includes: Obtain the expected document type for each qualification in the qualification review list, and assign a multimodal processing pipeline for each expected document type from the predefined processing strategy library; Each of the qualification documents is input into the corresponding multimodal processing pipeline to extract the qualification review content and verify its authenticity. The processing results of each of the multimodal processing pipelines are aligned with the qualification review list to generate a review list item that corresponds one-to-one with each qualification in the qualification review list.

6. The method as described in claim 5, characterized in that, Based on the qualification review content, the process involves performing authorization chain analysis from the global authorization network graph to obtain a target authorization subgraph, and then conducting a topology health assessment of the target authorization subgraph, including: Based on historical audit data, a global authorization network graph is constructed with entities and qualification assets as nodes. Entities are companies, individuals, or organizations, and qualification assets are specific intellectual property rights or administrative licenses. Starting with the submitter of the application to be reviewed, and combining the authorization chain depth requirements corresponding to each item in the review list, the target authorization subgraph is constructed by traversing the global authorization network graph. Calculate the topology health index of the target authorization subgraph, and measure the stability of the authorization chain of the qualification review content through the topology health index, wherein the topology health index includes path redundancy and critical node dependency.

7. The method as described in claim 6, characterized in that, The comprehensive review of the qualification verification content based on the target authorization subgraph includes: Align and review the qualification review content with the target authorization subgraph to generate a composite feature representation for each review list item; The composite feature representation is input into a multi-task learning model, and the multi-task learning model is used to verify the qualification validity, authorization logic consistency and commercial rationality of each audit list item to obtain preliminary verification results; Based on the necessity level of each item on the audit checklist, a dynamic weighted comprehensive decision is made on the preliminary verification results to generate a final audit conclusion.

8. A qualification verification system, characterized in that, include: The first processing module is used to obtain the application metadata of the application to be reviewed and generate a qualification review list through a dynamic demand awareness model. The second processing module is used to receive the qualification documents uploaded by the user, extract information from the qualification documents in conjunction with the qualification review list, obtain the qualification review content, and verify the authenticity of the qualification review content. The third processing module is used to perform authorization chain analysis from the global authorization network graph based on the qualification review content, obtain the target authorization subgraph, and perform topology health assessment on the target authorization subgraph. The fourth processing module is used to conduct a comprehensive review of the qualification review content based on the target authorization sub-graph.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images