Article inspection method and system fusing ai inspection and reinspection information

By integrating AI review and re-review information, setting asynchronous review conditions and target review rules, the problems of AI resource waste and process fragmentation were solved, achieving efficient and accurate product data review, and improving the system's automation level and user experience.

CN123118124APending Publication Date: 2026-11-17GUANGZHOU RES INTERESTING INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611426344.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing AI review solutions lack access control, resulting in a large number of simple tasks occupying AI computing resources for complex matters, which reduces the review response speed. Furthermore, the AI ​​review results are disconnected from the manual review process, which reduces the efficiency of manual review.

Method used

The method integrates AI review and re-review information, sets asynchronous review conditions, determines target review rules, executes asynchronous review operations, and conducts re-review under preset conditions to generate compliance marks or rejection requests and provide feedback on the reasons for rejection.

Benefits of technology

It improved the accuracy and efficiency of product data review, reduced the workload of invalid reviews, shortened the review cycle, enhanced the timeliness and resource utilization of the review process, and improved the user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118124A_ABST
    Figure CN123118124A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of data auditing, and discloses a commodity auditing method and system fusing AI auditing and re-auditing information, which comprises the following steps: when it is detected that there is an auditing matter meeting asynchronous auditing conditions, determining a target auditing rule matched with the auditing matter; performing asynchronous auditing operation on commodity data included in the auditing matter according to the target auditing rule to obtain asynchronous auditing results; if the asynchronous auditing is passed and re-auditing conditions are met, or if the current meets emergency re-auditing conditions, performing re-auditing operation on the commodity data to obtain re-auditing information; if the asynchronous auditing or the re-auditing is passed, authorizing the commodity data to be put on the market according to the generated compliance identifier; and when the asynchronous auditing or the re-auditing is not passed, rejecting the request of putting the commodity on the market and feeding back the rejection reason. It can be seen that the application can reduce invalid auditing load, improve the auditing accuracy of commodity data, and shorten the auditing period.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data auditing technology, and in particular to a product auditing method and system that integrates AI auditing and re-auditing information. Background Technology

[0002] In e-commerce and content platform product listing review scenarios, traditional review methods mainly rely on manual verification one by one, which is inefficient and costly. With the development of artificial intelligence technology, some systems have begun to introduce AI models to automate the review of product images, text, attributes, and other information in order to improve processing speed.

[0003] However, existing AI review solutions typically employ a strategy of directly feeding all data into the model for processing, lacking access control over review items. This results in AI computing resources, capable of handling complex matters, being consumed by a large number of simple review tasks, thus reducing the overall product review response speed. Furthermore, in existing AI review solutions, the AI ​​review process and the human review process are disconnected; the review results from different processes are not shared, preventing the AI ​​review results from being effectively utilized in the human review process and reducing the efficiency of human review.

[0004] Therefore, it is particularly important to provide a corresponding solution to the above-mentioned problems in the existing data review process. Summary of the Invention

[0005] This invention provides a product review method and system that integrates AI review and re-review information, which can reduce the load of invalid reviews, improve the accuracy of product data review, and shorten the review cycle.

[0006] The first aspect of this invention discloses a product review method that integrates AI review and re-review information, the method comprising: When an audit item that meets the asynchronous audit conditions is detected, a target audit rule matching the audit item is determined; according to the target audit rule, the asynchronous audit operation is performed on the product data included in the audit item according to the invoked AI interface to obtain the asynchronous audit result; When the asynchronous review result indicates that the review has passed and the preset re-review conditions are currently met, or when the preset emergency re-review conditions are currently met, a re-review operation is performed on the product data to obtain re-review information; When the target review information indicates that the review has passed, a compliance identifier is generated for the product data, and the product corresponding to the product data is authorized to be listed based on the compliance identifier; the target review information includes the asynchronous review result or the re-review information; When the target review information indicates that the review has failed, the product listing request is rejected, and the rejection reason included in the target review information is fed back to the target merchant who submitted the product data.

[0007] As an optional implementation, in the first aspect of the present invention, determining the target audit rule matching the audit item includes: Based on the product data included in the audit items, determine the product category corresponding to the product data, and determine all audit rules that match the product category; all audit rules include public rules, or public rules and category rules; and all audit rules are currently in effect. When the audit rules only include the public rules, the public rules are identified as the target audit rules that match the audit items; When the audit rules include the public rules and the category rules, a merged rule set of the public rules and the category rules is constructed, and all rules in the merged rule set are identified as target audit rules that match the audit items; The category rules have a higher priority than the public rules. When the category rules and the public rules cover the same type of rule and there is a conflict, the category rules shall prevail.

[0008] As an optional implementation, in the first aspect of the present invention, the target review rule includes corresponding rule review parameters; the rule review parameters include at least one of title parameters, image parameters, category attributes, and rich text of product details; The step of performing asynchronous review operations on the product data included in the review items according to the target review rules based on the invoked AI interface, and obtaining asynchronous review results, includes: Based on the invoked AI interface, and according to the title parameters, image parameters, category attributes, and product detail rich text, title review, image review, category attribute review, and product detail rich text review operations are performed on the product data included in the review items, respectively, to obtain asynchronous review results; the asynchronous review results include title review results, image review results, category attribute review results, and product detail rich text review results.

[0009] As an optional implementation, in the first aspect of the present invention, during the asynchronous review operation on the product data, the method further includes: The cumulative operation time for executing the asynchronous review operation is recorded in real time. Determine whether the cumulative operation time exceeds a preset time threshold; When it is determined that the cumulative operation time exceeds the preset time threshold, a re-review identifier is generated based on the cumulative operation time, and the current asynchronous review operation is terminated based on the re-review identifier, and the current review process is switched to the re-review process; the re-review identifier is used to indicate that the current re-review conditions are met.

[0010] As an optional implementation, in the first aspect of the present invention, satisfying the review conditions includes: the asynchronous review result indicating that a certain benchmark comparison data required for reviewing the product data is currently missing, or the product category corresponding to the product data is a targeted review category; the targeted review category is a category that is required to perform a review operation. The conditions for meeting the emergency review include: detecting a temporarily triggered manual review request, or the asynchronous review operation experiencing an operational anomaly; Furthermore, the method further includes: When the target review information includes the re-review information and indicates that the review has passed, the target review rules are optimized based on the re-review information.

[0011] As an optional implementation, in the first aspect of the present invention, the method further includes: When a product review request is detected, the product data to be reviewed is extracted from the submission data corresponding to the product review request; Perform basic rule verification and price verification operations on the product data respectively to obtain basic rule verification results and price verification results; When both the basic rule verification result and the price verification result indicate that they have passed, it is determined that the product data meets the asynchronous review conditions, and review items are generated based on the product data; When the basic rule verification result and / or the price verification result indicate that the operation failed, the failed target information is identified and marked, and the target information is fed back to the submitter who submitted the product review request to prompt the submitter to modify the target information; the target information includes the reason information for the failure of the basic rule verification operation and / or the reason information for the failure of the price verification operation.

[0012] As an optional implementation, in the first aspect of the present invention, performing basic rule verification operations and price verification operations on the product data to obtain basic rule verification results and price verification results includes: The system verifies whether the information corresponding to the product data and multiple required fields is completely filled to obtain a first verification result; it verifies whether the product data matches the preset format requirements to obtain a second verification result; it verifies whether the submission frequency of the product data is higher than a preset frequency threshold to obtain a third verification result; and it determines the first verification result, the second verification result, and the third verification result as the basic rule verification result. Determine the price verification items and their corresponding price verification conditions that match the product category of the product data; and perform a price verification operation on the product data according to the price verification items and their corresponding price verification conditions to obtain the price verification result.

[0013] As an optional implementation, in the first aspect of the present invention, determining the price verification item and its corresponding price verification condition that match the product category of the product data includes: Identify the product category of the product data and determine at least one price verification item that matches the product category; all the price verification items include deposit, purchase price, and rental fee; Determine the price verification conditions corresponding to each of the price verification items; the price verification conditions are dynamically determined based on the market reference price corresponding to the product category and / or the listing time of the specific product model under the product category; The step of performing a price verification operation on the product data based on the price verification items and their corresponding price verification conditions to obtain a price verification result includes: For each of the price verification items, a verification operation is performed on the corresponding price field in the product data according to the price verification conditions corresponding to the price verification item, and a sub-verification result corresponding to the price verification item is obtained. A price verification result is generated based on all the sub-verification results.

[0014] As an optional implementation, in a first aspect of the present invention, identifying the product category of the product data and determining at least one price verification item matching the product category includes: Identify the product category of the product data and the launch time of the specific product model under the product category; Obtain the market reference price corresponding to the specific product model from the preset product price data table; Based on the listing time and the market reference price, and through predefined price rule configuration, deposit price verification items and their corresponding deposit verification conditions, purchase price verification items and their corresponding purchase price verification conditions, and rental price verification items and their corresponding rental verification conditions are generated respectively.

[0015] A second aspect of this invention discloses a product review system that integrates AI review and re-review information, the system comprising: The determination module is used to determine the target audit rule that matches the audit item when an audit item that meets the asynchronous audit conditions is detected. The asynchronous review module is used to perform asynchronous review operations on the product data included in the review items according to the target review rules based on the called AI interface, and obtain asynchronous review results; The review module is used to perform a review operation on the product data to obtain review information when the asynchronous review result indicates that the review has passed and the preset review conditions are met, or when the preset emergency review conditions are met. The compliance processing module is used to generate a compliance identifier for the product data when the target review information indicates that the review has passed, and to authorize the product corresponding to the product data to be listed based on the compliance identifier; the target review information includes the asynchronous review result or the re-review information; The rejection processing module is used to reject the product listing request when the target review information indicates that the review has failed, and to send the rejection reason included in the target review information back to the target merchant who submitted the product data.

[0016] As an optional implementation, in a second aspect of the present invention, the method by which the determining module determines the target audit rule matching the audit item specifically includes: Based on the product data included in the audit items, determine the product category corresponding to the product data, and determine all audit rules that match the product category; all audit rules include public rules, or public rules and category rules; and all audit rules are currently in effect. When the audit rules only include the public rules, the public rules are identified as the target audit rules that match the audit items; When the audit rules include the public rules and the category rules, a merged rule set of the public rules and the category rules is constructed, and all rules in the merged rule set are identified as target audit rules that match the audit items; The category rules have a higher priority than the public rules. When the category rules and the public rules cover the same type of rule and there is a conflict, the category rules shall prevail.

[0017] As an optional implementation, in the second aspect of the present invention, the target review rule includes corresponding rule review parameters; the rule review parameters include at least one of title parameters, image parameters, category attributes, and rich text of product details; The asynchronous review module performs asynchronous review operations on the product data included in the review items according to the target review rules based on the invoked AI interface, and the specific methods for obtaining asynchronous review results include: Based on the invoked AI interface, and according to the title parameters, image parameters, category attributes, and product detail rich text, title review, image review, category attribute review, and product detail rich text review operations are performed on the product data included in the review items, respectively, to obtain asynchronous review results; the asynchronous review results include title review results, image review results, category attribute review results, and product detail rich text review results.

[0018] As an optional implementation, in a second aspect of the invention, the system further includes: The timeout monitoring and processing module is used to record the cumulative operation time of the asynchronous review operation in real time during the process of performing the asynchronous review operation on the product data; The timeout monitoring and processing module is also used to determine whether the cumulative operation time exceeds a preset time threshold. The timeout monitoring and processing module is further configured to generate a re-review identifier based on the cumulative operation time when it is determined that the cumulative operation time exceeds the preset time threshold, and to terminate the current asynchronous review operation based on the re-review identifier and switch the current review process to the re-review process; the re-review identifier is used to indicate that the current re-review conditions are met.

[0019] As an optional implementation, in the second aspect of the present invention, satisfying the review conditions includes: the asynchronous review result indicating that a certain benchmark comparison data required for reviewing the product data is currently missing, or the product category corresponding to the product data is a targeted review category; the targeted review category is a category that is required to perform a review operation. The conditions for meeting the emergency review include: detecting a temporarily triggered manual review request, or the asynchronous review operation experiencing an operational anomaly; The system also includes: An optimization module is used to optimize the target review rules based on the re-review information when the target review information includes the re-review information and indicates that the review has passed.

[0020] As an optional implementation, in a second aspect of the invention, the system further includes: The verification module is used to extract the product data to be reviewed from the submitted data corresponding to the product review request when a product review request is detected. The verification module is also used to perform basic rule verification operations and price verification operations on the product data respectively, and obtain basic rule verification results and price verification results; The determining module is further configured to determine that the product data meets the asynchronous review conditions when both the basic rule verification result and the price verification result indicate that they have passed, and to generate review items based on the product data; The feedback module is used to determine and mark the failed target information when the basic rule verification result and / or the price verification result indicate that the verification failed, and to feed back the target information to the submitter who submitted the product review request, so as to prompt the submitter to modify the target information; the target information includes the reason information for the failure of the basic rule verification operation and / or the reason information for the failure of the price verification operation.

[0021] As an optional implementation, in the second aspect of the present invention, the verification module performs basic rule verification operations and price verification operations on the product data respectively, and the specific methods for obtaining the basic rule verification results and price verification results include: The system verifies whether the information corresponding to the product data and multiple required fields is completely filled to obtain a first verification result; it verifies whether the product data matches the preset format requirements to obtain a second verification result; it verifies whether the submission frequency of the product data is higher than a preset frequency threshold to obtain a third verification result; and it determines the first verification result, the second verification result, and the third verification result as the basic rule verification result. Determine the price verification items and their corresponding price verification conditions that match the product category of the product data; and perform a price verification operation on the product data according to the price verification items and their corresponding price verification conditions to obtain the price verification result.

[0022] As an optional implementation, in a second aspect of the present invention, the method by which the verification module determines the price verification item and its corresponding price verification condition that match the product category of the product data specifically includes: Identify the product category of the product data and determine at least one price verification item that matches the product category; all the price verification items include deposit, purchase price, and rental fee; Determine the price verification conditions corresponding to each of the price verification items; the price verification conditions are dynamically determined based on the market reference price corresponding to the product category and / or the listing time of the specific product model under the product category; The verification module performs a price verification operation on the product data based on the price verification items and their corresponding price verification conditions, and obtains the price verification result in the following specific ways: For each of the price verification items, a verification operation is performed on the corresponding price field in the product data according to the price verification conditions corresponding to the price verification item, and a sub-verification result corresponding to the price verification item is obtained. A price verification result is generated based on all the sub-verification results.

[0023] As an optional implementation, in a second aspect of the present invention, the method by which the verification module identifies the product category of the product data and determines at least one price verification item matching the product category specifically includes: Identify the product category of the product data and the launch time of the specific product model under the product category; Obtain the market reference price corresponding to the specific product model from the preset product price data table; Based on the listing time and the market reference price, and through predefined price rule configuration, deposit price verification items and their corresponding deposit verification conditions, purchase price verification items and their corresponding purchase price verification conditions, and rental price verification items and their corresponding rental verification conditions are generated respectively.

[0024] A third aspect of this invention discloses a product review device that integrates AI review and re-review information, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the product review method that integrates AI review and re-review information as described in any of the first aspects of the present invention.

[0025] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the product review method integrating AI review and re-review information as described in any of the first aspects of the present invention.

[0026] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a product review method that integrates AI review and re-review information. The method includes: when a review item meeting asynchronous review conditions is detected, determining a target review rule matching the review item; performing an asynchronous review operation on the product data included in the review item according to the target review rule via a called AI interface to obtain an asynchronous review result; when the asynchronous review result indicates approval and currently meets preset re-review conditions, or when currently meets preset emergency re-review conditions, performing a re-review operation on the product data to obtain re-review information; when the target review information indicates approval, generating a compliance identifier for the product data, and authorizing the corresponding product to be listed based on the compliance identifier; the target review information includes the asynchronous review result or re-review information; when the target review information indicates approval failure, rejecting the product listing request, and feeding back the rejection reason included in the target review information to the target merchant who submitted the product data. Therefore, by implementing this invention and setting asynchronous review conditions as a prerequisite for performing asynchronous review operations, the frequent triggering of this complex asynchronous review operation to review product data with basic information errors can be effectively reduced, thereby reducing the risk of main program blocking and improving the response speed of product review. After the initial AI review is passed, a secondary review and an emergency review are introduced as supplementary verification layers. This effectively compensates for potential omissions or misjudgments in AI review, enhancing the fault tolerance and security of the review process in special risk scenarios. For approved cases, a compliance label is automatically generated based on the review conclusion, and listing is authorized, reducing the delays and errors of manual authorization. This improves the efficiency of listing compliant products and enhances the automation level of platform operations. For rejected cases, the reasons for rejection are fed back to the corresponding merchant, guiding them to make targeted rectifications and reducing the number of resubmissions. Ultimately, a three-layer architecture is constructed: asynchronous review condition pre-screening → AI asynchronous review → conditional secondary review. This significantly optimizes the timeliness and resource utilization of the review process while improving the accuracy of product data review. Furthermore, through automated authorization and precise feedback mechanisms, it further improves review efficiency and user experience. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating a product review method that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Figure 2This is a flowchart illustrating another product review method that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a product review system that integrates AI review and re-review information, as disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of another product review system that integrates AI review and re-review information, as disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a product review device that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0031] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0032] This invention discloses a product review method and system that integrates AI review and secondary review information. By setting asynchronous review conditions as a prerequisite for executing asynchronous review operations, it effectively reduces the frequent triggering of this complex operation to review product data with basic information errors, thereby reducing the risk of main program blocking and improving the response speed of product review. After the initial AI review is passed, secondary review and emergency review are introduced as supplementary verification layers, effectively compensating for potential omissions or misjudgments in AI review, enhancing the fault tolerance and security of review in special risk scenarios. For approved cases, a compliance label can be automatically generated and authorization for listing can be granted based on the review conclusion, reducing the delay and errors of manual authorization, which is conducive to improving the listing efficiency of compliant products and enhancing the automation level of platform operations. For unsuccessful cases, the reasons for rejection can be fed back to the corresponding merchant, thereby guiding the merchant to make targeted rectifications and reducing the number of times the merchant submits resubmitted. Ultimately, a three-tier architecture was constructed: asynchronous pre-screening of review conditions → AI asynchronous review → conditional re-review. This not only improved the accuracy of product data review but also significantly optimized the timeliness of the review process, increased resource utilization, and further enhanced review efficiency and user experience through automated authorization and precise feedback mechanisms. These will be explained in detail below.

[0033] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a product review method that integrates AI review and secondary review information, as disclosed in an embodiment of the present invention. Figure 1 The described product review method integrating AI review and secondary review information can be applied to product review systems / devices that integrate AI review and secondary review information; however, this invention does not limit its application. Figure 1 As shown, the product review method that integrates AI review and re-review information may include the following operations: 101. When an audit item that meets the asynchronous audit conditions is detected, determine the target audit rule that matches the audit item.

[0034] In this embodiment of the invention, optionally, satisfying the asynchronous review condition can be as follows: for a review request submitted for a certain product data, the system performs preliminary basic information detection on the product data. After the detection passes, the product data is marked as meeting the asynchronous review condition, and review items for the product data are generated simultaneously. Please refer to the description of this basic information detection in Embodiment Two below; it will not be repeated here.

[0035] In this embodiment of the invention, asynchronous review conditions are set as a prerequisite for performing asynchronous review operations. Only when the asynchronous review conditions are met can the determination of the target review rules and subsequent steps be executed. These asynchronous review conditions can also perform basic checks on product data, effectively filtering out a large amount of product data with non-compliant basic information. At the same time, they can effectively reduce the frequent triggering of the complex asynchronous review operation to review product data with errors in basic information.

[0036] In this embodiment of the invention, each audit item includes its corresponding product data to be audited. The audit rules (which can also be regarded as audit requirements, audit regulations, etc.) for each product data are different. Therefore, it is necessary to determine the corresponding target audit rules in a targeted manner, so as to achieve targeted and differentiated audit of product data.

[0037] Optionally, step 101, which determines the target audit rules that match the audit items, may specifically include the following steps: Step A1: Based on the product data included in the audit items, determine the product category corresponding to the product data, and determine all audit rules that match the product category.

[0038] In this embodiment of the invention, the product category can be identified by parsing the title, attributes, qualifications, and other data included in the product data. For example, the product category can be any one of food and beverage, beauty and personal care, electronic products, home furnishings, baby and toys, clothing and shoes, etc. Then, guided by the product category, all currently effective review rules under that product category are retrieved from a preset rule base, providing clear review standards for subsequent reviews. It should be noted that since the retrieved review rules specifically refer to "currently effective," even if the review rules change, it can still ensure that accurate and usable review rules are retrieved.

[0039] In this embodiment of the invention, all review rules include public rules, or public rules and category rules; and all review rules are currently in effect. Public rules can be general rules applicable to the current product category or all product categories (e.g., "absolute terms are prohibited"); category rules can refer to specific rules set for the current product category (e.g., "registration number must be displayed").

[0040] Optionally, the priority of category rules can be set higher than that of public rules; for the same type of rule covered by category rules and public rules, and when rule conflicts occur, the rule of category rules shall be adopted first.

[0041] Step A2: When the audit rules only include public rules, identify the public rules as the target audit rules that match the audit items.

[0042] Step A3: When the audit rules include public rules and category rules, construct a merged rule set of public rules and category rules, and identify all rules in the merged rule set as target audit rules that match the audit items.

[0043] In this optional embodiment, step 1013, which constructs the merged rule set of public rules and category rules, may specifically include the following steps: Iterate through all category rules and compare the current category rule with each public rule using the rule identifier; this rule identifier is used to uniquely identify the type of rule. If the comparison result shows that there is a target public rule with the same rule identifier as the current category rule among all public rules, then the current category rule replaces the target public rule, and the replaced category rule is added to the merged rule set; If the comparison result shows that there is no target public rule with the same rule identifier as the current category rule among all public rules, then the current category rule will be directly added to the merged rule set; After completing the traversal of all category rules, add all remaining common rules that have not been replaced to the merge rule set; Specifically, for the same type of rules and regulations covered by category rules and public rules, conflict identification is performed through rule identifiers, and the rules and regulations of category rules are adopted first, so that each rule in the merged rule set has unique rule and regulation effect.

[0044] As can be seen, in this optional embodiment, categories are automatically identified through product data, and all currently effective public and category rules are retrieved. This ensures that reviews are always based on the latest rules, reducing omissions caused by manual review or misreports due to outdated rules, and improving the timeliness and accuracy of the determined review rules. When only public rules are available, they are directly adopted, simplifying the rule matching process for simple categories. When both types of rules coexist, a merged rule set is constructed, and conflicts are resolved by prioritizing category rules over public rules. While maintaining general regulatory bottom lines, this fully meets the refined management needs of special categories, and a clear conflict resolution mechanism is established to further improve the consistency and accuracy of rule review judgments.

[0045] 102. Based on the AI ​​interface being called and in accordance with the target review rules, perform asynchronous review operations on the product data included in the review items to obtain asynchronous review results.

[0046] In this embodiment of the invention, the target review rule may include corresponding rule review parameters; the rule review parameters may include at least one of title parameters, image parameters, category attributes, and rich text of product details.

[0047] Furthermore, step 102, based on the invoked AI interface and the target review rules, performs asynchronous review operations on the product data included in the review items. The method for obtaining the asynchronous review result may specifically include the following steps: Based on the invoked AI interface, and according to the title parameters, image parameters, category attributes, and rich text of product details, the product data included in the review items are subjected to title review, image review, category attribute review, and rich text of product details review operations respectively, resulting in asynchronous review results. The asynchronous review results include title review results, image review results, category attribute review results, and rich text of product details review results.

[0048] In this optional embodiment, the title review operation is used to review the compliance of the title. Specifically, it can involve setting a prohibited word library and format templates to identify whether the title contains typos, whether the format conforms to specifications, and whether it contains prohibited words. The image review operation is used to review the compliance of the image. Specifically, it can involve setting a library of prohibited pattern examples to identify and verify the image background, watermarks, and whether the image contains prohibited content. The category attribute review operation is used to review the matching degree of category attributes. Specifically, it can involve verifying whether the relevance between the product data and its product category meets the requirements (e.g., relevance higher than XX%), and / or verifying whether the attribute completeness meets the standards (e.g., completeness reaches XX%). The product details rich text review operation is used to review the compliance of the product details rich text. Specifically, it can involve conducting a compliance review on the text and images included in the product data.

[0049] Specifically, after a "baby lotion" product was submitted, the AI ​​detected that the title contained the prohibited word "strongest" (title failed), the image contained a real baby image (image triggered sensitivity), the attribute lacked an expiration date (attribute failed), and the details contained an uncertified "treatment of eczema" efficacy description (rich text failed), and each sub-item was judged independently.

[0050] As can be seen, in this optional embodiment, the review rules are refined into multi-dimensional independent parameters such as title, image, category attributes, and rich text details. This transforms the review process from a simple, single judgment to a multi-dimensional, refined detection, significantly improving the coverage and accuracy of identifying violations. Furthermore, asynchronous review operations are performed on each parameter in parallel and independently via an AI interface, ultimately outputting results for each sub-item. On one hand, parallel processing significantly reduces overall review time and improves efficiency; on the other hand, independently outputted sub-item results provide precise attribution for subsequent rejections (directly pinpointing whether the violation occurred in the title or image), facilitating targeted rectification by merchants and reducing the waste of system resources caused by repeated submissions due to merchants failing to accurately modify their content.

[0051] 103. When the asynchronous review result indicates that the review has passed and the preset re-review conditions are met, or when the preset emergency re-review conditions are met, a re-review operation is performed on the product data to obtain re-review information.

[0052] In this embodiment of the invention, meeting the conditions for review may include: the asynchronous review result indicating that a certain benchmark comparison data required for reviewing product data is currently missing (for example, the record information of a certain product has not been included, and AI cannot make a compliance judgment based on existing rules and information); or, the product category corresponding to the product data is a targeted review category. The targeted review category is a category for which a review operation is mandatory. This targeted review category may be a category with a high customer complaint rate / review rate, which is automatically marked by the backend system.

[0053] In this embodiment of the invention, meeting the conditions for emergency review may include: detecting a temporarily triggered manual review request, or an operational anomaly occurring in the asynchronous review operation. The triggered manual review request may be manually initiated by operations personnel after discovering a public opinion risk, or it may be a review request submitted voluntarily by the merchant. An operational anomaly in the asynchronous review operation may be due to AI interface response timeouts, incorrect model output formats, etc., and this embodiment of the invention does not limit the scope of the anomaly.

[0054] In this embodiment of the invention, the method may optionally include the following steps: When the target review information includes the re-review information and indicates that the review has passed, optimize the target review rules based on the re-review information.

[0055] Furthermore, when the approved information includes optimization information, the target review rules are optimized based on that optimization information, for example, by supplementing new comparison benchmark data or adjusting the sensitive word threshold of the category rules; When the approved information does not include optimization information, retrieve the approved data from the re-approval information. The target review rule hit results for the approved data are compared item by item with the pass labels of the approved data to generate difference information; this difference information is used to characterize the differences between the target review rule hit results and the pass labels. The discrepancy information is structured and parsed to obtain the rule type and discrepancy direction corresponding to each discrepancy item; the discrepancy direction includes the direction of misjudgment and the direction of omission. Based on the rule type and direction of each difference item, optimization information is extracted from the difference information. The optimization information includes at least one of the following: first optimization information for the direction of misjudgment, second optimization information for the direction of missed judgment, and third optimization information for the rule type common to multiple difference items. The first optimization information is used to extract the corresponding rule type and feature information of the misjudged product data; the second optimization information is used to extract the corresponding rule type and feature information of the missed product data; and the third optimization information is used to extract the hit threshold deviation information of this type of rule. Based on the optimization information, determine the optimization strategy for the corresponding rules and regulations in the target review rules, and perform update operations on the corresponding rules and regulations in the target review rules in accordance with the optimization strategy to generate the optimized target review rules; Specifically, when the first optimization information is included, the corresponding optimization strategy is to increase the sensitive word threshold of the corresponding rule or add exemption conditions; when the second optimization information is included, the corresponding optimization strategy is to decrease the sensitive word threshold of the corresponding rule or supplement new comparison benchmark data; when the third optimization information is included indicating a hit threshold deviation, the corresponding optimization strategy is to adjust the review threshold parameter of this type of rule or regulation.

[0056] For example, during the review process, if a human verifies the ingredient ratio of a certain "children's sunscreen" and provides feedback on the safety and compliance of the ingredient ratio, the system can automatically enter the ingredient ratio data into the benchmark comparison library. This allows the AI ​​to directly compare the product when reviewing similar products in the future, without needing to enter the review process.

[0057] As can be seen, in this optional embodiment, using the lack of benchmark data or belonging to a targeted review category as a trigger for review allows for timely introduction of secondary verification in scenarios where AI review confidence is insufficient or in highly sensitive categories. This effectively compensates for the decision-making deficiencies of AI models in data blind spots or high-risk areas, improving the reliability and security of the final review results. Using manual review requests or AI operation anomalies as triggers for emergency reviews not only responds to temporary human intervention needs but also provides contingency strategies when AI services fail, reducing review pauses due to AI review anomalies and improving the stability of the business process. Furthermore, a closed-loop iterative mechanism of review → feedback → optimization is constructed, enabling AI-based asynchronous review to continuously optimize with human review experience, thereby gradually reducing the reliance on manual review and improving the accuracy and practicality of AI review.

[0058] 104. When the target review information indicates that the review has been passed, a compliance identifier is generated for the product data, and the product corresponding to the product data is authorized to be listed based on the compliance identifier.

[0059] In this embodiment of the invention, the target review information may include asynchronous review results or re-review information.

[0060] 105. When the target review information indicates that the review has failed, the product listing request will be rejected, and the rejection reason included in the target review information will be fed back to the target merchant who submitted the product data.

[0061] In this embodiment of the invention, after the rejection reason is fed back to the target merchant (for example, by clicking the "AI Review Result" button in the upper right corner of the product details page to view the AI ​​review details (rule code and review reason)), the target merchant can make targeted modifications based on this information, avoiding blind modifications. Alternatively, if the target merchant does not agree with the feedback rejection reason, they can also file a review request, thereby entering the review process.

[0062] In this embodiment of the invention, the method may optionally include the following steps: Step B1: When performing any audit process or when each audit process is triggered, record the current execution node of the audit process in real time and update the node status of each execution node.

[0063] In this embodiment of the invention, the current execution node may include: an initial node (corresponding to the starting node for merchants to submit product data), an AI review node (corresponding to the node entering the asynchronous review process), a secondary review node (corresponding to the node entering the secondary review process), and a termination node (corresponding to the node for product listing / rejection). The node status of each execution node may include one of the following states: unreviewed, under review (AI), pending review (human review), under review (human review), listed, approved, rejected.

[0064] Specifically, when a merchant creates or modifies a product but hasn't submitted it for review, the current node is the initial node and is in an unreviewed state. After the merchant submits the product, the system calls the AI ​​interface for asynchronous review, at which point the system enters the AI ​​review node, and the node status is "Reviewing (AI)". If the AI ​​review is successful and the product enters the human review queue waiting for a reviewer to accept the order, the system enters the secondary review node, and the node status is "Pending Review (Human Review)". When a reviewer accepts the secondary review task and the product enters the human review process, the current node status changes to "Reviewing (Human Review)". If the product is approved and automatically listed, or if the AI ​​review or human review rejects it, the node flows to the termination node, and the node status changes to "Listed", "Rejected", or "Approved".

[0065] Step B2: Once any audit process is completed, obtain all operation information corresponding to that audit process and generate an audit history record for all operation information.

[0066] This includes, but is not limited to, the condition judgment results of asynchronous review conditions, detected review items, determined target review rules, obtained asynchronous review results, condition judgment results for re-review conditions, re-review information corresponding to the re-review operation, rejection records of rejected listing requests, and feedback on rejection reasons. This review history can be viewed by clicking the "AI Review Results" button in the upper right corner of the product details page, facilitating subsequent process traceability.

[0067] It is evident that implementation Figure 1 The described product review method, which integrates AI review and secondary review information, sets asynchronous review conditions as a prerequisite for executing asynchronous review operations. This effectively reduces the frequent triggering of this complex asynchronous review operation to review product data with basic information errors, thereby reducing the risk of main program blocking and improving the response speed of product review. After the initial AI review is passed, secondary review and emergency review are introduced as supplementary verification layers, which can effectively compensate for possible omissions or misjudgments in AI review, enhancing the fault tolerance and security of review in special risk scenarios. For those that pass the review, compliance labels can be automatically generated and authorization for listing can be granted based on the review conclusion, reducing the delay and errors of manual authorization, which is conducive to improving the listing efficiency of compliant products and improving the automation level of platform operations. For those that fail the review, the reasons for rejection can be fed back to the corresponding merchants, thereby guiding them to make targeted rectifications and reducing the number of times merchants resubmit. Finally, a three-layer architecture of asynchronous review condition pre-screening → AI asynchronous review → conditional secondary review is constructed. While improving the accuracy of product data review, the timeliness and resource utilization of the review process are greatly optimized. Through automated authorization and accurate feedback mechanisms, the review efficiency and user experience of the system are further improved.

[0068] In an optional embodiment, during the asynchronous review operation on product data, the method may further include the following steps: Step C1: Record the cumulative operation time of asynchronous audit operations in real time.

[0069] In this optional embodiment, the system starts timing when the asynchronous review operation is initiated, continuously monitoring the execution time of the current review task. For example, if a product contains a large number of high-definition images and lengthy rich text, the AI ​​interface processing may take a long time, and the system's background real-time timer has recorded 8 seconds of processing time.

[0070] Step C2: Determine whether the cumulative operation time exceeds the preset time threshold.

[0071] In this optional embodiment, the preset duration threshold can be a static threshold, for example, set to 10 seconds, 20 seconds or 30 seconds by default; or it can be a dynamic threshold, for example, a duration threshold can be dynamically determined based on the amount of product data currently being reviewed, historical review status, etc.

[0072] Step C3: When it is determined that the cumulative operation time exceeds the preset time threshold, a re-review identifier is generated based on the cumulative operation time, and the current asynchronous review operation is terminated based on the re-review identifier, and the current review process is switched to the re-review process; the re-review identifier is used to indicate that the current re-review conditions are met.

[0073] As can be seen, in this optional embodiment, a clear time boundary is set for the review process, reducing the long-term occupation of system resources by a single review task due to data complexity or interface performance fluctuations. When a timeout occurs, a re-review flag is generated, the current AI operation is terminated, and the process is switched to the re-review process. This allows the system to automatically perform downgrade processing when the AI ​​processing efficiency does not meet expectations, effectively reducing process blocking caused by AI review lag and improving the responsiveness and stability of the overall review process.

[0074] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating another product review method that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Figure 2 The described product review method integrating AI review and secondary review information can be applied to product review systems / devices that integrate AI review and secondary review information; however, this invention does not limit its application. Figure 2 As shown, the product review method that integrates AI review and re-review information may include the following operations: 201. When a product review request is detected, extract the product data to be reviewed from the submitted data corresponding to the product review request.

[0075] In this embodiment of the invention, the merchant fills in product information on the merchant's end, including marketing statements, product type, category attributes, sales information, sales specifications, product information, payment information, logistics information, etc., and submits the information after completion. At this time, the system detects the submitted product review request, parses the submitted data carried in the product review request, and obtains the product data. For example, if a merchant uploads a listing application for a "Bluetooth headset," the system extracts data such as "product name," "price 299 yuan," "inventory 500 units," and "product image."

[0076] 202. Perform basic rule verification and price verification operations on the product data respectively to obtain the basic rule verification results and price verification results.

[0077] In this embodiment of the invention, optionally, step 202 performs basic rule verification and price verification operations on the product data respectively, and the method for obtaining the basic rule verification results and price verification results may specifically include the following steps: Step D1: Verify whether the product data and the information corresponding to multiple required fields are filled in completely to obtain the first verification result; verify whether the product data matches the preset format requirements to obtain the second verification result; verify whether the submission frequency of the product data is higher than the preset frequency threshold to obtain the third verification result; and determine the first verification result, the second verification result, and the third verification result as the basic rule verification result.

[0078] In this embodiment of the invention, integrity verification is used to reduce data loss, format verification is used to ensure data standardization, and frequency verification is used to reduce interface abuse.

[0079] In this embodiment of the invention, examples of the verification results for each basic rule are as follows: For the first verification result: When a merchant submits a "rice cooker" product, the required attribute "energy efficiency rating" is not filled in. The system marks the first verification result as failing, with the reason being "missing energy efficiency rating". For the second verification result: The product title is "Huge Discount, Super Value Hot Selling Recommendation..." (75 characters in total), exceeding the 60-character limit. The system marks the second verification result as failing, with the reason being "title too long". For the third verification result: If the same "fitness tracker" is submitted repeatedly 6 times within 30 minutes, exceeding the threshold of 5 times / hour, the system marks the third verification result as failing, with the reason being "abnormal submission frequency".

[0080] In this embodiment of the invention, after obtaining multiple independent verification results by executing step D1, all verification results are summarized. If any verification result fails, the overall basic rule verification result is determined to be unsuccessful.

[0081] Step D2: Determine the price verification items and their corresponding price verification conditions that match the product category in the product data; and perform price verification operations on the product data according to the price verification items and their corresponding price verification conditions to obtain the price verification results.

[0082] Optionally, step D2 may specifically include the following steps: Step D21: Identify the product category of the product data, determine at least one price verification item that matches the product category, and determine the price verification condition corresponding to each price verification item.

[0083] In this embodiment of the invention, all price verification items may include deposit, purchase price, and rental fee; optionally, the purchase price may include, but is not limited to, at least one of the following: purchase unit price, combined purchase price, and discounted price.

[0084] In this embodiment of the invention, the price verification condition can be dynamically determined based on the market reference price corresponding to the product category and / or the launch time of the specific product model under the product category.

[0085] Furthermore, step D21 may specifically include the following steps: Identify the product category of the product data and the launch time of the specific product model under that product category; Obtain the market reference price corresponding to the specific product model from the preset product price data table; Based on the listing date and market reference price, and through predefined price rule configuration, deposit price verification items and their corresponding deposit verification conditions, purchase price verification items and their corresponding purchase price verification conditions, and rental price verification items and their corresponding rental verification conditions are generated respectively.

[0086] In this embodiment of the invention, the market reference price may include a deposit reference price; the deposit verification condition may be: the deposit value range is x% to y% of the deposit reference price.

[0087] In this embodiment of the invention, the purchase price verification condition can be: dynamically calculating the upper limit of the purchase price based on the listing time and the lease duration. Specifically, when the lease duration is less than a first preset number of days, the upper limit of the purchase price is the base purchase price; when the lease duration is greater than or equal to the first preset number of days and less than a second preset number of days, the upper limit of the purchase price is the base purchase price minus a first preset percentage; when the lease duration is greater than or equal to the second preset number of days, the upper limit of the purchase price is the base purchase price minus a second preset percentage.

[0088] Wherein, the first preset number of days is less than the second preset number of days; the first preset ratio is less than the second preset ratio. The setting of the number of days and the ratio can be adjusted according to the actual situation, and the embodiments of the present invention do not limit it.

[0089] In this embodiment of the invention, the rent verification condition can be: determining the upper limit of the total rent based on the relationship between the rental duration and a preset duration threshold. Specifically, when the rental duration is less than or equal to the preset duration threshold, the upper limit of the total rent is a first upper limit value; when the rental duration is greater than the preset duration threshold, the upper limit of the total rent is a second upper limit value. The first upper limit value is less than the second upper limit value.

[0090] Step D22: For each price verification item, perform a verification operation on the corresponding price field in the product data according to the price verification conditions corresponding to the price verification item, and obtain the sub-verification result corresponding to the price verification item.

[0091] In this embodiment of the invention, step D22 may optionally include the following steps: The deposit field in the product data is validated according to the deposit validation conditions to obtain the deposit sub-validation result. The purchase price field in the product data is validated according to the purchase price validation conditions to obtain the purchase price validation result. The rental field in the product data is validated based on the rental validation conditions to obtain the rental sub-validation result.

[0092] Step D23: Generate the price verification result based on all sub-verification results.

[0093] In this embodiment of the invention, step D23 may optionally include the following steps: When the deposit verification result, purchase price verification result, and rental fee verification result all indicate that they have passed, the price verification result is determined to be passed. When any sub-verification result indicates failure, the price verification result is determined to be failure, and the failure price verification item and its corresponding failure reason information are marked so that the failure reason information can be fed back to the submitter; among them, all sub-verification results include deposit sub-verification results, purchase price sub-verification results and rental sub-verification results.

[0094] As can be seen, in this embodiment of the invention, the basic rule verification is refined into three dimensions: completeness, format, and frequency, and each dimension outputs an independent sub-result. This achieves fine-grained identification of problem types and facilitates subsequent problem classification feedback and rapid correction. For different product categories, corresponding price verification items and conditions are dynamically matched, and then targeted price verification is performed, reducing false positives or false negatives and improving the accuracy and adaptability of price anomaly detection.

[0095] 203. When both the basic rule verification result and the price verification result indicate that they have passed, the product data is determined to meet the asynchronous review conditions, and review items are generated based on the product data.

[0096] In this embodiment of the invention, if both verification results indicate that the product data meets the asynchronous review conditions (i.e., it passes the basic information detection), the product data can be assembled into standardized structured data, and a review item to be processed can be generated accordingly, and then pushed to the subsequent AI asynchronous review process.

[0097] 204. When the basic rule verification result and / or price verification result indicate that the application fails, identify and mark the target information that fails, and feed back the target information to the submitter who submitted the product review request to prompt the submitter to modify the target information.

[0098] In this embodiment of the invention, the target information includes reasons for the failure of the basic rule verification operation and / or reasons for the failure of the price verification operation. For example, if the system detects that the price of a "down jacket" product is "0.01 yuan" and determines it to be abnormal, and at the same time the basic rule detects "size attribute not filled", both reasons are returned to the merchant. After receiving this target information, the merchant can make targeted corrections accordingly, thereby reducing the situation where merchants blindly modify information and repeatedly submit unqualified information.

[0099] 205. When an audit item that meets the asynchronous audit conditions is detected, determine the target audit rule that matches the audit item.

[0100] 206. Based on the AI ​​interface being called and in accordance with the target review rules, perform asynchronous review operations on the product data included in the review items to obtain asynchronous review results.

[0101] 207. When the asynchronous review result indicates that the review has passed and the preset re-review conditions are met, or when the preset emergency re-review conditions are met, perform a re-review operation on the product data to obtain re-review information.

[0102] 208. When the target review information indicates that the review has been passed, a compliance identifier is generated for the product data, and the product corresponding to the product data is authorized to be listed based on the compliance identifier.

[0103] 209. When the target review information indicates that the review has failed, the product listing request will be rejected, and the rejection reason included in the target review information will be fed back to the target merchant who submitted the product data.

[0104] For further descriptions of steps 205 to 209 in this embodiment of the invention, please refer to the other specific descriptions of steps 101 to 105 in Embodiment 1. These descriptions will not be repeated in this embodiment of the invention.

[0105] It is evident that implementation Figure 2 The described product review method, which integrates AI review and secondary review information, performs preliminary rapid checks on basic rules and prices before entering asynchronous AI review. This intercepts low-value issues such as obvious formatting errors, missing attributes, or abnormal pricing before AI review, establishing an admission filtering mechanism for asynchronous review. This reduces the consumption of computing power for subsequent complex asynchronous reviews by invalid data, and to some extent improves the effective pass rate and processing efficiency of subsequent asynchronous reviews. For cases where checks fail, the method can generate and provide specific reasons for rejection, allowing merchants to clearly identify the direction for modification early in the review process. This reduces the likelihood of merchants being dragged into lengthy review processes due to simple errors and then rejected, effectively shortening the correction and iteration cycle for merchants and reducing the number of times the system processes data repeatedly.

[0106] In an optional embodiment, the method may further include the following steps: Step E1: Perform a full lifecycle inspection on the published and / or listed products according to the preset inspection frequency to obtain the inspection results; wherein, the inspection operation is used to inspect and verify the multiple states of the product from publication to listing.

[0107] In this optional embodiment, the inspection operation may include at least one of the following inspection items: sensitive word detection, product authenticity verification, price compliance verification, special category compliance verification, product out-of-stock verification, and merchant operation anomaly monitoring.

[0108] In this optional embodiment, step E1 may specifically include the following steps: Step E11: Determine the pre-configured inspection frequency, which includes at least one of daily inspection, weekly inspection, or scheduled inspection.

[0109] Step E12: Based on the inspection frequency, poll the target products currently in the target status and perform inspection operations on the polled target products to obtain the inspection results for the target products; wherein, the target status includes any one of the following: publishing status, shelf status, or pending review status.

[0110] Optionally, step E12 performs an inspection operation on the polled target products, and the specific methods for obtaining the inspection results for the target products may include the following operations: Sensitive word detection is performed on the text information of the polled target products to obtain a first result; if any sensitive word in the preset sensitive word library is matched, the first result is used to mark the target product as having a sensitive word violation; Perform an authenticity check on the target product to obtain a second result; this authenticity check is used to verify whether the text and image description of the target product is consistent with its actual attributes, and / or to verify whether the inventory information of the target product matches the actual inventory. A price verification is performed on the target product to obtain a third result; the price verification is described in step 202 of Example 1, and will not be repeated here. A special category compliance verification is performed on the target product to obtain the fourth result. The special category compliance verification is used to identify whether the target product belongs to a special regulatory category. If the target product is identified as belonging to a special regulatory category, it is further verified whether it meets the publication qualifications and / or content specifications of the corresponding special category. Perform a stockout check on the target product to obtain the fifth result; the stockout check is used to detect whether the current inventory of the target product is zero or lower than the preset safety stock threshold. If so, the fifth result is used to mark the target product as having a stockout violation. Perform operational anomaly monitoring on the merchants corresponding to the target products to obtain the sixth result; if it is detected that the merchant has expired qualifications, has been complained about or reported, or has a downgraded credit rating, the sixth result is used to mark the merchant as having operational anomalies and violations; The first, second, third, fourth, fifth, and sixth results are determined as the inspection results for the target product.

[0111] Step E2: When the inspection results indicate that any inspection item is in violation, generate a violation warning message for the violation inspection item, and, based on the violation level included in the violation warning message, remove the product from the shelves or send a manual processing notification to the management terminal.

[0112] In this optional embodiment, the violation level of the violation inspection item can be determined according to a preset violation level mapping relationship; if the violation level is the first level, the product is removed from the shelves; if the violation level is the second level, a manual processing notification containing the violation inspection item and its details is sent to the management terminal.

[0113] As can be seen, in this optional embodiment, by configuring periodic inspection mechanisms such as daily / weekly / scheduled inspections, the entire lifecycle of a product from its release to its listing is continuously monitored. By comprehensively utilizing multi-dimensional inspection methods such as sensitive word detection, authenticity verification, price compliance verification, special category verification, out-of-stock detection, and monitoring of abnormal merchant operations, various violations can be detected and automatically intercepted or warned in a timely manner. At the same time, based on the level of violation, targeted strategies such as product removal or manual intervention are adopted, which effectively improves the comprehensiveness, timeliness, and automation of product compliance monitoring. It also helps to reduce the cost of manual inspections and improve the quality of products and transaction security on the platform.

[0114] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a product review system that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Figure 3 As shown, the product review system integrating AI review and re-review information may include a determination module 301, an asynchronous review module 302, a re-review module 303, a compliance processing module 304, and a rejection processing module 305, wherein: The determination module 301 is used to determine the target audit rule that matches the audit item when an audit item that meets the asynchronous audit conditions is detected.

[0115] The asynchronous audit module 302 is used to perform asynchronous audit operations on the product data included in the audit items according to the target audit rules based on the called AI interface, and obtain asynchronous audit results.

[0116] The review module 303 is used to perform a review operation on the product data and obtain review information when the asynchronous review result indicates that the review has passed and the current preset review conditions are met, or when the current preset emergency review conditions are met.

[0117] The compliance processing module 304 is used to generate a compliance identifier for the product data when the target review information indicates that the review has passed, and to authorize the corresponding product to be listed based on the compliance identifier; the target review information includes asynchronous review results or re-review information.

[0118] The rejection processing module 305 is used to reject the product listing request when the target review information indicates that the review has failed, and to send the rejection reason included in the target review information back to the target merchant who submitted the product data.

[0119] It is evident that implementation Figure 3 The described product review system, which integrates AI review and secondary review information, sets asynchronous review conditions as a prerequisite for executing asynchronous review operations. This effectively reduces the frequent triggering of this complex asynchronous review operation to review product data with basic information errors, thereby reducing the risk of main program blocking and improving the response speed of product review. After the initial AI review is passed, secondary review and emergency review are introduced as supplementary verification layers, which can effectively compensate for possible omissions or misjudgments in AI review, enhancing the fault tolerance and security of review in special risk scenarios. For those that pass the review, a compliance mark can be automatically generated based on the review conclusion, and authorization for listing can be granted, reducing the delay and errors of manual authorization, which is conducive to improving the listing efficiency of compliant products and improving the automation level of platform operations. For those that fail the review, the reasons for rejection can be fed back to the corresponding merchants, thereby guiding them to make targeted rectifications and reducing the number of times merchants resubmit. Finally, a three-layer architecture of asynchronous review condition pre-screening → AI asynchronous review → conditional secondary review is constructed. While improving the accuracy of product data review, the timeliness and resource utilization of the review process are greatly optimized. Through automated authorization and accurate feedback mechanisms, the review efficiency is further improved and the user experience of the system is enhanced.

[0120] In an optional embodiment, the method by which the determining module 301 determines the target audit rule that matches the audit item specifically includes: Based on the product data included in the audit items, determine the product category corresponding to the product data, and determine all audit rules that match the product category; all audit rules include public rules, or public rules and category rules; and all audit rules are currently in effect. When the audit rules only include public rules, the public rules are identified as the target audit rules that match the audit items; When the audit rules include public rules and category rules, construct a merged rule set of public rules and category rules, and identify all rules in the merged rule set as target audit rules that match the audit items; Category rules have higher priority than public rules. When category rules and public rules cover the same type of rules and regulations and there is a conflict, the category rules shall prevail.

[0121] As can be seen, in this optional embodiment, categories are automatically identified through product data, and all currently effective public and category rules are retrieved. This ensures that reviews are always based on the latest rules, reducing omissions caused by manual review or misreports due to outdated rules, and improving the timeliness and accuracy of the determined review rules. When only public rules are available, they are directly adopted, simplifying the rule matching process for simple categories. When both types of rules coexist, a merged rule set is constructed, and conflicts are resolved by prioritizing category rules over public rules. While maintaining general regulatory bottom lines, this fully meets the refined management needs of special categories, and a clear conflict resolution mechanism is established to further improve the consistency and accuracy of rule review judgments.

[0122] In another optional embodiment, the target review rule includes corresponding rule review parameters; the rule review parameters include at least one of title parameters, image parameters, category attributes, and rich text of product details; The asynchronous review module 302 performs asynchronous review operations on the product data included in the review items according to the target review rules based on the called AI interface. The specific methods for obtaining asynchronous review results include: Based on the invoked AI interface, and according to the title parameters, image parameters, category attributes, and rich text of product details, the product data included in the review items are subjected to title review, image review, category attribute review, and rich text of product details review operations respectively, resulting in asynchronous review results. The asynchronous review results include title review results, image review results, category attribute review results, and rich text of product details review results.

[0123] As can be seen, in this optional embodiment, the review rules are refined into multi-dimensional independent parameters such as title, image, category attributes, and rich text details. This transforms the review process from a simple, single judgment to a multi-dimensional, refined detection, significantly improving the coverage and accuracy of identifying violations. Furthermore, by using an AI interface to perform asynchronous review operations on each parameter in parallel and independently, and finally outputting the results for each sub-item, the parallel processing method can significantly shorten the overall review time and improve review efficiency. On the other hand, the independently output sub-item results can provide accurate attribution basis for subsequent rejections (directly pinpointing whether the violation occurred in the title or the image), facilitating targeted rectification by merchants and reducing the waste of system resources caused by merchants failing to accurately modify and submitting repeatedly.

[0124] In yet another alternative embodiment, please refer to Figure 4 , Figure 4 This is a schematic diagram of another product review system that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Figure 4 As shown, the system may also include a timeout monitoring and processing module 306, wherein: The timeout monitoring and processing module 306 is used to record the cumulative operation time of asynchronous review operations in real time during the asynchronous review operation of product data. The timeout monitoring and processing module 306 is also used to determine whether the cumulative operation time exceeds the preset time threshold; The timeout monitoring and processing module 306 is also used to generate a re-review identifier based on the cumulative operation time when it is determined that the cumulative operation time exceeds the preset time threshold, and to terminate the current asynchronous review operation based on the re-review identifier and switch the current review process to the re-review process; the re-review identifier is used to indicate that the current re-review conditions are met.

[0125] As can be seen, in this optional embodiment, a clear time boundary is set for the review process, reducing the long-term occupation of system resources by a single review task due to data complexity or interface performance fluctuations. When a timeout occurs, a re-review flag is generated, the current AI operation is terminated, and the process is switched to the re-review process. This allows the system to automatically perform downgrade processing when the AI ​​processing efficiency does not meet expectations, effectively reducing process blocking caused by AI review lag and improving the responsiveness and stability of the overall review process.

[0126] In another optional embodiment, the conditions for re-review include: the asynchronous review result indicates that a certain benchmark comparison data required for reviewing the product data is currently missing, or the product category corresponding to the product data is a targeted re-review category; the targeted re-review category is a category that is required to perform the re-review operation. The conditions for emergency review include: detecting a temporarily triggered manual review request, or an operational anomaly occurring during asynchronous review operations; And, such as Figure 4 As shown, the system may also include an optimization module 307, wherein: The optimization module 307 is used to optimize the target review rules based on the re-review information when the target review information includes re-review information and indicates that the review has passed.

[0127] As can be seen, in this optional embodiment, using the lack of benchmark data or belonging to a targeted review category as a trigger for review allows for timely introduction of secondary verification in scenarios where AI review confidence is insufficient or in highly sensitive categories. This effectively compensates for the decision-making deficiencies of AI models in data blind spots or high-risk areas, improving the reliability and security of the final review results. Using manual review requests or AI operation anomalies as triggers for emergency reviews not only responds to temporary human intervention needs but also provides contingency strategies when AI services fail, reducing review pauses due to AI review anomalies and improving the stability of the business process. Furthermore, a closed-loop iterative mechanism of review → feedback → optimization is constructed, enabling AI-based asynchronous review to continuously optimize with human review experience, thereby gradually reducing the reliance on manual review and improving the accuracy and practicality of AI review.

[0128] In yet another alternative embodiment, such as Figure 4 As shown, the system may also include a verification module 308 and a feedback module 309, wherein: The verification module 308 is used to extract the product data to be reviewed from the submitted data corresponding to the product review request when a product review request is detected. The verification module 308 is also used to perform basic rule verification operations and price verification operations on the product data respectively, and obtain the basic rule verification results and the price verification results. The determination module 301 is also used to determine that the product data meets the asynchronous review conditions when both the basic rule verification result and the price verification result indicate that they have passed, and to generate review items based on the product data; Feedback module 309 is used to identify and mark the target information that failed when the basic rule verification result and / or price verification result indicate that the verification failed, and to feed back the target information to the submitter who submitted the product review request to prompt the submitter to modify the target information; the target information includes the reason information for the failure of the basic rule verification operation and / or the reason information for the failure of the price verification operation.

[0129] As can be seen, in this optional embodiment, before entering the asynchronous AI review, two preliminary quick checks are performed on basic rules and prices. This intercepts low-value issues such as obvious formatting errors, missing attributes, or abnormal prices before AI review, setting up an admission filtering mechanism for asynchronous review. This reduces the consumption of computing power for subsequent complex asynchronous review by invalid data, and to a certain extent improves the effective pass rate and processing efficiency of subsequent asynchronous reviews. For cases where checks fail, specific reasons for rejection can be generated and fed back, allowing merchants to clearly identify the direction of modification early in the review process. This reduces the likelihood of merchants being dragged into a long review process due to simple errors and then being rejected, effectively shortening the correction and iteration cycle on the merchant's side and reducing the number of times the system processes data repeatedly.

[0130] In another optional embodiment, the verification module 308 performs basic rule verification operations and price verification operations on the product data respectively, and the specific methods for obtaining the basic rule verification results and price verification results include: The system verifies whether the product data and the information corresponding to multiple required fields are filled in completely, and obtains the first verification result; it verifies whether the product data matches the preset format requirements, and obtains the second verification result; it verifies whether the submission frequency of the product data is higher than the preset frequency threshold, and obtains the third verification result; and it determines the first verification result, the second verification result, and the third verification result as the basic rule verification result. Determine the price verification items and their corresponding price verification conditions that match the product categories in the product data; and perform price verification operations on the product data based on the price verification items and their corresponding price verification conditions to obtain the price verification results.

[0131] In this optional embodiment, the verification module 308 determines the price verification item and its corresponding price verification condition that match the product category of the product data in the following specific ways: Identify the product category in the product data and determine at least one price verification item that matches the product category; all price verification items include deposit, purchase price, and rental fee; Determine the price verification conditions corresponding to each price verification item; the price verification conditions are dynamically determined based on the market reference price corresponding to the product category and / or the listing time of the specific product model under the product category. Optionally, the verification module 308 performs a price verification operation on the product data based on the price verification items and their corresponding price verification conditions, and the specific methods for obtaining the price verification result include: For each price verification item, perform a verification operation on the corresponding price field in the product data according to the price verification conditions corresponding to the price verification item, and obtain the sub-verification result corresponding to the price verification item. Generate a price verification result based on all sub-verification results.

[0132] Optionally, the verification module 308 identifies the product category of the product data and determines at least one price verification item that matches the product category in the following ways: Identify the product category of the product data and the launch time of the specific product model under the product category; Obtain the market reference price corresponding to the specific product model from the preset product price data table; Based on the listing time and market reference price, and through predefined price rule configuration, deposit price verification items and their corresponding deposit verification conditions, purchase price verification items and their corresponding purchase price verification conditions, and rental price verification items and their corresponding rental verification conditions are generated respectively.

[0133] As can be seen, in this optional embodiment, the basic rule verification is refined into three dimensions: completeness, format, and frequency, and each dimension outputs an independent sub-result. This achieves fine-grained identification of problem types and facilitates subsequent problem classification feedback and rapid correction. Dynamically matching corresponding price verification items and conditions for different product categories and then performing targeted price verification reduces false positives or false negatives, improving the accuracy and adaptability of price anomaly detection.

[0134] Example 4 Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a product review device that integrates AI review and re-review information, as disclosed in an embodiment of the present invention. Figure 5 As shown, the product verification device that integrates AI verification and re-verification information may include: Memory 401 storing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute some or all of the steps in any of the product review methods that integrate AI review and re-review information as described in Embodiment 1 or Embodiment 2 of the present invention.

[0135] Example 5 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the product review methods that integrate AI review and re-review information as described in Embodiment 1 or Embodiment 2 of this invention.

[0136] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0137] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0138] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention 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 the present invention.

Claims

1. A commodity review method that fuses AI review and re-review information, characterized by, The method includes: When an audit item that meets the asynchronous audit conditions is detected, a target audit rule matching the audit item is determined; according to the target audit rule, the asynchronous audit operation is performed on the product data included in the audit item according to the invoked AI interface to obtain the asynchronous audit result; When the asynchronous review result indicates that the review has passed and the preset re-review conditions are currently met, or when the preset emergency re-review conditions are currently met, a re-review operation is performed on the product data to obtain re-review information; When the target review information indicates that the review has passed, a compliance identifier is generated for the product data, and the product corresponding to the product data is authorized to be listed based on the compliance identifier; the target review information includes the asynchronous review result or the re-review information; When the target review information indicates that the review has failed, the product listing request is rejected, and the rejection reason included in the target review information is fed back to the target merchant who submitted the product data. 2.The merchandise auditing method of fusing AI auditing and re-auditing information according to claim 1, characterized in that, The determination of the target audit rule matching the audit item includes: Based on the product data included in the audit items, determine the product category corresponding to the product data, and determine all audit rules that match the product category; all audit rules include public rules, or public rules and category rules; and all audit rules are currently in effect. When the audit rules only include the public rules, the public rules are identified as the target audit rules that match the audit items; When the audit rules include the public rules and the category rules, a merged rule set of the public rules and the category rules is constructed, and all rules in the merged rule set are identified as target audit rules that match the audit items; The category rules have a higher priority than the public rules. When the category rules and the public rules cover the same type of rule and there is a conflict, the category rules shall prevail. 3.The merchandise auditing method of fusing AI auditing and re-auditing information according to claim 2, characterized in that, The target review rules include corresponding rule review parameters; the rule review parameters include at least one of the following: title parameters, image parameters, category attributes, and rich text of product details; The step of performing asynchronous review operations on the product data included in the review items according to the target review rules based on the invoked AI interface, and obtaining asynchronous review results, includes: Based on the invoked AI interface, and according to the title parameters, image parameters, category attributes, and product detail rich text, title review, image review, category attribute review, and product detail rich text review operations are performed on the product data included in the review items, respectively, to obtain asynchronous review results; the asynchronous review results include title review results, image review results, category attribute review results, and product detail rich text review results. 4.The merchandise auditing method of fusing AI auditing and re-auditing information according to claim 3, characterized in that, During the asynchronous review operation on the product data, the method further includes: The cumulative operation time for executing the asynchronous review operation is recorded in real time. Determine whether the cumulative operation time exceeds a preset time threshold; When it is determined that the cumulative operation time exceeds the preset time threshold, a re-review identifier is generated based on the cumulative operation time, and the current asynchronous review operation is terminated based on the re-review identifier, and the current review process is switched to the re-review process; the re-review identifier is used to indicate that the current re-review conditions are met.

5. The product review method integrating AI review and re-review information according to claim 3, characterized in that, The conditions for re-review include: the asynchronous review result indicates that a certain benchmark comparison data required for reviewing the product data is currently missing, or the product category corresponding to the product data is a targeted re-review category; the targeted re-review category is a category that is required to perform the re-review operation. The conditions for meeting the emergency review include: detecting a temporarily triggered manual review request, or the asynchronous review operation experiencing an operational anomaly; Furthermore, the method further includes: When the target review information includes the re-review information and indicates that the review has passed, the target review rules are optimized based on the re-review information.

6. The product review method integrating AI review and re-review information according to any one of claims 1-5, characterized in that, The method further includes: When a product review request is detected, the product data to be reviewed is extracted from the submission data corresponding to the product review request; Perform basic rule verification and price verification operations on the product data respectively to obtain basic rule verification results and price verification results; When both the basic rule verification result and the price verification result indicate that they have passed, it is determined that the product data meets the asynchronous review conditions, and review items are generated based on the product data; When the basic rule verification result and / or the price verification result indicate that the operation failed, the failed target information is identified and marked, and the target information is fed back to the submitter who submitted the product review request to prompt the submitter to modify the target information; the target information includes the reason information for the failure of the basic rule verification operation and / or the reason information for the failure of the price verification operation. 7.The merchandise auditing method of fusing AI auditing and re-auditing information according to claim 6, characterized in that, The basic rule verification operation and price verification operation are performed on the product data respectively to obtain the basic rule verification result and the price verification result, including: The system verifies whether the information corresponding to the product data and multiple required fields is completely filled to obtain a first verification result; it verifies whether the product data matches the preset format requirements to obtain a second verification result; it verifies whether the submission frequency of the product data is higher than a preset frequency threshold to obtain a third verification result; and it determines the first verification result, the second verification result, and the third verification result as the basic rule verification result. Determine the price verification items and their corresponding price verification conditions that match the product category of the product data; and perform a price verification operation on the product data according to the price verification items and their corresponding price verification conditions to obtain the price verification result. 8.The merchandise auditing method of fusing AI auditing and re-auditing information according to claim 7, characterized in that, The step of determining the price verification item and its corresponding price verification condition that match the product category of the product data includes: Identify the product category of the product data and determine at least one price verification item that matches the product category; all the price verification items include deposit, purchase price, and rental fee; Determine the price verification conditions corresponding to each of the price verification items; the price verification conditions are dynamically determined based on the market reference price corresponding to the product category and / or the listing time of the specific product model under the product category; The step of performing a price verification operation on the product data based on the price verification items and their corresponding price verification conditions to obtain a price verification result includes: For each of the price verification items, a verification operation is performed on the corresponding price field in the product data according to the price verification conditions corresponding to the price verification item, and a sub-verification result corresponding to the price verification item is obtained. A price verification result is generated based on all the sub-verification results. 9.The merchandise auditing method of fusing AI auditing and re-auditing information according to claim 8, characterized in that, The step of identifying the product category of the product data and determining at least one price verification item that matches the product category includes: Identify the product category of the product data and the launch time of the specific product model under the product category; Obtain the market reference price corresponding to the specific product model from the preset product price data table; Based on the listing time and the market reference price, and through predefined price rule configuration, deposit price verification items and their corresponding deposit verification conditions, purchase price verification items and their corresponding purchase price verification conditions, and rental price verification items and their corresponding rental verification conditions are generated respectively.

10. A commodity inspection system that fuses AI inspection and reinspection information, characterized by, The system includes: The determination module is used to determine the target audit rule that matches the audit item when an audit item that meets the asynchronous audit conditions is detected. The asynchronous review module is used to perform asynchronous review operations on the product data included in the review items according to the target review rules based on the called AI interface, and obtain asynchronous review results; The review module is used to perform a review operation on the product data to obtain review information when the asynchronous review result indicates that the review has passed and the preset review conditions are met, or when the preset emergency review conditions are met. The compliance processing module is used to generate a compliance identifier for the product data when the target review information indicates that the review has passed, and to authorize the product corresponding to the product data to be listed based on the compliance identifier; the target review information includes the asynchronous review result or the re-review information; The rejection processing module is used to reject the product listing request when the target review information indicates that the review has failed, and to send the rejection reason included in the target review information back to the target merchant who submitted the product data.