Dual-purpose item-based controlled item identification method, device, equipment, medium and product

Abstract

The application provides a dual-purpose item-based controlled article identification method, device, equipment, medium and product. It relates to the technical field of artificial intelligence. The method comprises: obtaining an article to be identified; determining whether the article to be identified belongs to a non-dual-purpose item category based on a preset category library; in the case that the article to be identified belongs to a dual-purpose item category, generating a safety score of the article to be identified according to a preset keyword library and a preset reasoning logic rule; in response to the safety score being greater than or equal to a preset safety threshold, generating a relevance score of the article to be identified based on a vector retrieval result and a preset scoring rule; in the case that the relevance score is equal to a preset relevance threshold, identifying the article to be identified using a preset multi-dimensional judgment strategy to generate an identification result. The method of the application significantly improves the review efficiency, reduces the misjudgment rate, and meets the needs of intelligent, explainability and scalability in the export business scenario while ensuring compliance.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, equipment, medium and product for identifying controlled articles based on dual-use items. Background Technology

[0002] In international trade and cross-border export scenarios, dual-use items are subject to strict export controls by various countries because they can be used for both civilian and military purposes, as well as in sensitive fields. my country requires exporters to conduct compliance reviews of goods to determine whether they fall within the scope of the controlled goods list. The international express delivery, cross-border trade, and e-commerce industries process massive amounts of declarations daily, involving various categories such as industrial equipment, precision components, special materials, and chemicals. The review process is characterized by its large volume, high timeliness requirements, complex rules, and heavy compliance responsibilities.

[0003] Currently, the review of dual-use items mainly relies on human expert judgment, with some solutions incorporating large language models for auxiliary identification. Human review depends on professionals' understanding of the control list, technical parameters, and usage restrictions; the large language model-based solution inputs the item description into the model, which directly provides a conclusion on whether it falls within the control scope. Some solutions combine keyword matching and text retrieval to achieve comparative analysis between controlled items and declared items.

[0004] However, existing solutions lack a tiered risk filtering mechanism, consume significant computational resources, have a high false positive rate, and are prone to over-interception or under-interception. Furthermore, the judgment process is not standardized and cannot meet the requirements of compliance auditing. Summary of the Invention

[0005] The methods, apparatus, equipment, media, and products for identifying controlled articles based on dual-use items provided in this application are used to achieve rapid and accurate determination of dual-use items.

[0006] In a first aspect, embodiments of this application provide a method for identifying controlled articles based on dual-use items, the method comprising:

[0007] Acquire the item to be identified, the item to be identified including item description information;

[0008] Determine whether the item to be identified belongs to the category of non-dual-use items based on a pre-defined category library;

[0009] If the item to be identified belongs to the dual-use item category, a security score for the item to be identified is generated based on a preset keyword library and preset reasoning logic rules.

[0010] In response to the security score being greater than or equal to a preset security threshold, a relevance score for the item to be identified is generated based on the vector retrieval results and preset scoring rules;

[0011] When the correlation score is equal to a preset correlation threshold, a preset multi-dimensional judgment strategy is used to identify the item to be identified in order to generate an identification result, which is used to indicate whether the item to be identified is a controlled item.

[0012] In one possible implementation, the preset keyword library includes keywords of multiple categories and corresponding weights for each category. The step of generating a security score for the item to be identified based on the preset keyword library and preset inference logic rules includes:

[0013] The item description information is matched against a preset keyword library to obtain keyword recognition results;

[0014] The initial security score of the item to be identified is generated by combining and weighting the keyword categories and locations in the keyword recognition results.

[0015] The keyword recognition results are inferred using preset inference logic rules to generate inference results; the preset inference logic rules include at least one of category matching, usage constraint judgment, technical feature matching, and item conformity inference.

[0016] The initial security score is adjusted based on the reasoning results to obtain a security score for the item to be identified.

[0017] In one possible implementation, generating a relevance score for the item to be identified based on the vector retrieval results and a preset scoring rule includes:

[0018] Based on the item description information, a similarity search is performed from a preset vector database to obtain vector search results. The preset vector database stores dual-purpose item control entry vectors that have undergone structured segmentation and parameter extraction.

[0019] Based on the vector retrieval results, a weighted score is applied to multiple preset dimensions using preset scoring rules. The preset dimensions include at least two of the following: item use, technical parameters, material type, and item form.

[0020] A relevance score for the item to be identified is generated based on the scores corresponding to multiple preset dimensions.

[0021] In one possible implementation, the article form includes raw materials, semi-finished products, and finished products, and the weighted score includes:

[0022] Increase the weight score for items in raw material form and decrease the weight score for items in finished product form.

[0023] Items intended for civilian use will have their weight score reduced, while items made of pre-defined sensitive materials or with pre-defined control components will have their weight score increased.

[0024] In one possible implementation, the step of identifying the item to be identified using a preset multi-dimensional determination strategy includes:

[0025] Perform a preset verification process on the item to be identified and the vector retrieval result. The preset verification process includes at least one of the following: category matching verification, application scenario verification, combination characteristic verification, and typical example verification.

[0026] If any verification fails, the item to be identified is determined not to be a dual-use item;

[0027] If all verifications pass, the item to be identified is determined to be or may be a dual-purpose item.

[0028] In one possible implementation, the typical example verification includes:

[0029] Determine whether the item to be identified meets any of the preset examples. The preset examples include items with similar keywords but different categories, items containing preset sensitive materials but which are finished products, items where the testing equipment is confused with the tested object, and items where a single parameter matches but the combined conditions are not met.

[0030] In one possible implementation, it also includes:

[0031] If the security score is less than the preset security threshold, then the item to be identified is determined not to be a dual-use item.

[0032] If the correlation score is less than the preset correlation threshold, then the item to be identified is determined not to be a dual-purpose item.

[0033] If the correlation score is greater than the preset correlation threshold, then the item to be identified is determined to be or may be a dual-purpose item.

[0034] Secondly, embodiments of this application provide a controlled article identification device based on dual-use items, comprising:

[0035] The acquisition module is used to acquire the item to be identified, the item to be identified including item description information;

[0036] The determination module is used to determine whether the item to be identified belongs to the category of non-dual-use items based on a preset category library.

[0037] The generation module is used to generate a security score for the item to be identified based on a preset keyword library and preset reasoning logic rules when the item to be identified belongs to the dual-use item category.

[0038] The generation module is also used to generate a relevance score for the item to be identified based on the vector retrieval results and the preset scoring rules in response to the security score being greater than or equal to a preset security threshold.

[0039] The identification module is used to identify the item to be identified by adopting a preset multi-dimensional judgment strategy when the relevance score is equal to a preset relevance threshold, so as to generate an identification result, which is used to indicate whether the item to be identified is a controlled item.

[0040] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0041] The memory stores computer-executed instructions;

[0042] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0043] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0044] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0045] The method, apparatus, equipment, medium, and product for identifying controlled items based on dual-use items provided in this application first perform simple matching based on a preset category library to quickly screen out ordinary civilian items that obviously pose no control risk, significantly reducing the computational load of subsequent deep identification. This avoids unnecessary complex analysis of non-sensitive items, reducing false positive rates and resource consumption. A coarse-grained risk classification is achieved through a dual-track approach of keyword identification and rule-based reasoning. Risk levels are represented by quantitative scoring, standardizing and objectifying risk judgment and eliminating subjective human bias. This provides clear triage criteria for subsequent processes, enabling tiered processing where low-risk items are quickly processed and high-risk items are more finely assessed. When the security score is greater than or equal to a preset security threshold, the item to be identified is considered high-risk and requires further identification. Vector retrieval is used to achieve accurate matching of control items, solving the problems of knowledge lag and inaccurate understanding in large models. Multi-dimensional scoring quantifies the correlation between items and control items, resulting in more refined and stable identification. Using scores as triage thresholds makes the system's identification logic explainable, verifiable, and auditable. For borderline ambiguous items—those with a relevance score equal to a preset relevance threshold—a refined secondary verification is performed, significantly improving the accuracy of identifying boundary scenarios. Common misjudgments are systematically eliminated through multi-dimensional verification. Structured conclusions and supporting evidence are output, meeting export control compliance audit requirements. While ensuring compliance, review efficiency is significantly improved, and the false judgment rate is reduced. Attached Figure Description

[0046] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0047] Figure 1 An application scenario diagram of the method for determining dual-use items provided in this application;

[0048] Figure 2 A schematic flowchart illustrating a controlled article identification method based on dual-use items provided in an embodiment of this application;

[0049] Figure 3 A flowchart illustrating a controlled article identification method based on dual-use items, provided as another embodiment of this application;

[0050] Figure 4 A schematic diagram of the structure of a controlled article identification device based on dual-use items provided in an embodiment of this application;

[0051] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0052] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0053] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0054] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0055] To clearly understand the technical solution of this application, the solutions of the prior art will be described in detail first.

[0056] In international trade and cross-border export, dual-use items can be used simultaneously for civilian and military purposes, and are sensitive areas, making them a key target of export controls in various countries. Relevant Chinese laws and regulations explicitly require exporters to conduct compliance reviews of declared items to determine whether they fall within the scope of controlled items. The international express delivery, cross-border trade, and e-commerce industries process massive amounts of export declarations daily, covering many categories such as industrial equipment, precision components, special materials, and chemicals. Related reviews are characterized by large volumes of business, strict time requirements, complex control rules, and significant compliance responsibilities. Currently, compliance reviews of dual-use items mainly rely on human experts, with some existing technologies attempting to introduce large language models for auxiliary identification. Human review depends on the reviewers' experience in judging controlled lists, technical parameters, and usage restrictions; large language model-based solutions directly input item descriptions into the model, which outputs control determination results. Some solutions also combine keyword matching and text retrieval methods to complete the comparative analysis between controlled items and declared items. However, existing review schemes generally lack a tiered risk filtering mechanism, resulting in high computational resource consumption and problems such as high false positive rates, excessive interception, and missed detection. At the same time, the existing scheme lacks a unified standard for the judgment process, and the judgment basis and process are difficult to trace, which cannot meet the rigid requirements of export control compliance audit.

[0057] Therefore, to address technical issues in existing technologies and improve judgment efficiency and resource utilization, a rapid pre-screening step is implemented at the very beginning of the process. By pre-setting a non-dual-use item category library, obviously risk-free items are directly excluded, significantly reducing subsequent computational pressure and increasing overall review speed. To reduce subjective judgment, a security scoring system combining keyword recognition and rule-based reasoning is introduced, quantifying risk levels into scores to achieve coarse-grained risk grading and providing clear triage criteria for subsequent processes. To address the issue of inaccurate model understanding, a relevance scoring method combining vector retrieval and multi-dimensional scoring is employed. Retrieval ensures regulatory accuracy, while scoring rules achieve precise quantification, avoiding both misjudgments and omissions of sensitive items. To further reduce the misjudgment rate, for items at the threshold, a multi-dimensional judgment strategy is implemented. Typical misjudgments are eliminated through item-by-item verification, outputting standard and traceable judgment results, thereby comprehensively improving accuracy and compliance.

[0058] Figure 1 This is an application scenario diagram illustrating the controlled article identification method based on dual-use items provided in this application, such as... Figure 1 As shown in the diagram, the scenario diagram corresponding to the controlled article identification method based on dual-use items provided in this application includes: a declaration terminal device 101, an intelligent review server 102, a preset category library 103, a preset vector database 104, and an identification rule base 105. It can be understood that the controlled article identification device based on dual-use items is integrated into the intelligent review server 102.

[0059] Specifically, the user submits item declaration information to the intelligent review server 102 through the declaration terminal device 101. After obtaining the item to be identified, which includes item description information, the intelligent review server 102 calls the preset category library 103 to determine whether the item to be identified belongs to the category of non-dual-use items. If it does not belong to the category, a security score is generated based on the keyword library and reasoning logic rules in the identification rule library 105. When the security score is greater than or equal to the preset security threshold, a similarity search is performed from the preset vector database 104, and a relevance score is generated based on the preset scoring rules. If the relevance score is equal to the preset relevance threshold, the intelligent review server 102 performs refined identification of the item to be identified according to the multi-dimensional judgment strategy in the identification rule library 105, and generates the final identification result. The processing result and identification basis are synchronized to the declaration terminal device 101 in real time and displayed to the user.

[0060] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0061] Figure 2 A schematic flowchart illustrating a controlled article identification method based on dual-use items provided in an embodiment of this application is shown below. Figure 2 As shown, the executing entity in this embodiment is a controlled article identification device based on dual-use items. This device can be implemented through a computer program, or through a medium storing the relevant computer program, such as a USB flash drive and / or optical disc, or through a physical device integrating or installing the relevant computer program, such as a chip, server, or server cluster. The controlled article identification method based on dual-use items provided in this embodiment includes the following steps:

[0062] S201. Obtain the item to be identified, which includes item description information.

[0063] Among them, items to be identified refer to items that need to be identified for compliance in export declaration, cross-border logistics or trade compliance review, including their name, material, specifications, uses, model and other relevant descriptive information.

[0064] Specifically, the user inputs the item to be identified, including its name, material, specifications, purpose, model, and other relevant descriptive information. The system collects the target objects for compliance review of dual-use items from the cross-border declaration business portal, logistics declaration entry process, or business data flow link, thus completing the acquisition of the item to be identified.

[0065] S202. Determine whether the item to be identified belongs to the category of non-dual-use items based on the preset category library.

[0066] Among them, the preset category library refers to the set of standard categories of non-dual-use items that are built and stored in advance, including item categories that clearly belong to ordinary food, civilian clothing, daily stationery, civilian kitchenware, and regular daily necessities.

[0067] Among them, the category of non-dual-use items refers to the category of items that are clearly not subject to dual-use item control according to relevant export control regulations and control catalogs. These items only have conventional civilian uses and do not involve sensitive technologies, special materials, high-precision equipment, or characteristics that can be used for military purposes.

[0068] Specifically, the name, material, purpose, and description of the item to be identified are obtained, and this information is compared with the categories of non-dual-use items stored in a pre-defined category library. During the matching process, semantic parsing is performed on the description information of the item to be identified to extract its category, purpose attributes, and general civilian characteristics. The extracted results are then compared with the categories in the pre-defined category library for consistency judgment. If the item to be identified does not belong to any non-dual-use item category in the pre-defined category library, or contains characteristics that may be related to the scope of regulation, then it is determined that the item to be identified does not belong to a non-dual-use item category, and the identification process continues in the subsequent security scoring steps.

[0069] Optionally, if the category, purpose, and attributes of the item to be identified are consistent with any non-dual-use item category in the preset category library, and it does not contain any characteristics related to sensitive materials, sensitive equipment, or sensitive uses, then the item to be identified is determined to belong to the non-dual-use item category. The identification result is then directly output, and the identification process ends.

[0070] S203. When the item to be identified belongs to the dual-use item category, a security score for the item to be identified is generated based on the preset keyword library and preset reasoning logic rules.

[0071] The pre-set keyword library refers to a set of keywords pre-constructed based on export control regulations and control catalogs for dual-use items. These keywords are categorized according to their function and risk attributes, including materials (high strength, high temperature resistance, composite materials, special alloys, ceramics, carbon fiber, etc.), chemicals (specific CAS numbers, precursors, phosphides, fluorides, etc.), equipment (CNC, controllers, high precision, positioning, vacuum, centrifuges, lasers, radars, sensors, etc.), software and technology (design software, simulation software, source code, etc.), and end-use (aerospace, nuclear energy, missiles, drones, military, etc.).

[0072] Among them, the pre-set reasoning logic rules refer to the standardized judgment logic formulated based on the item structure, category classification, usage restrictions, and technical parameter requirements of the dual-use item control catalog, which is used to transform the keyword identification results into risk level judgments.

[0073] Among them, the safety score refers to the score used to quantitatively characterize the degree of association between the identified item and the scope of dual-use item control. The higher the score, the higher the control risk of the item.

[0074] Specifically, when the item to be identified does not belong to the category of non-dual-use items, the process proceeds to the security scoring stage. The description information of the item, including its name, material, specifications, and purpose, is fully analyzed. The analyzed content is then matched one by one with various keywords in a pre-set keyword database to identify sensitive keywords contained in the item description and determine the category to which each keyword belongs. Following pre-set reasoning logic rules, structured reasoning is performed on the keyword identification results. First, the technical field and control category of the item are determined based on the matched keywords. Then, it is determined whether the item involves sensitive uses as defined in the control catalog, and further, whether the item possesses the technical characteristics and functional attributes required by the control catalog. Finally, combining the keyword matching results and the rule-based reasoning conclusions, a comprehensive security score for the item to be identified is calculated.

[0075] S204. In response to a security score greater than or equal to a preset security threshold, generate a relevance score for the item to be identified based on the vector retrieval results and preset scoring rules.

[0076] The preset safety threshold refers to the pre-set critical score used to determine whether to proceed with deep correlation analysis.

[0077] The vector retrieval result refers to the relevant regulatory item information recalled after semantic similarity matching between the item to be identified and the pre-built dual-use item control item vector library. This includes the control clauses, category descriptions, technical parameters, and control codes that are closest to the item description.

[0078] Among them, the preset scoring rules refer to the quantitative scoring standards that are pre-established based on export control compliance requirements, and the comprehensive scoring is carried out by combining the degree of matching between the item and the control item, the purpose attribute, the material sensitivity, the form of the item, and other dimensions.

[0079] The relevance score is a quantitative measure of the degree of association between the item to be identified and the dual-use item control item. The higher the score, the more likely the item is to fall within the control scope.

[0080] Specifically, when the security score of the item to be identified is greater than or equal to a preset security threshold, the relevance scoring process begins. The descriptive information of the item, such as its name, material, specifications, and purpose, is converted into a semantic vector, and a similarity search is performed in a pre-built vector library of dual-use item control entries. The control entry content most closely matching the item description is obtained, forming the vector search result. Then, according to preset scoring rules, a multi-dimensional comparative analysis is performed between the item to be identified and the recalled control entries, evaluating the degree of matching in terms of technical category, material properties, technical parameters, functional purpose, and item form, and weighting the scores based on the matching results of each dimension. The scores from each dimension are then summed and normalized according to a preset range to ultimately form a relevance score that objectively reflects the degree of association between the item to be identified and the control entry.

[0081] For example, the pre-defined scoring rules reflect the principles of prioritizing civilian use and making sensitive exceptions: 1 point (unrelated): ordinary civilian general-purpose goods with no technical / use connection to the controlled scope. 2-5 points (weakly related): weak technical connection but intended for ordinary civilian use, with no controlled features. 6-8 points (potentially related): civilian products made of sensitive materials, such as civilian decorative parts made of carbon fiber or titanium alloy, or core control components, such as PLC modules or industrial controllers; even if labeled as having civilian use, their dual-use potential must be assessed. 9-10 points (strongly related): the item is highly consistent with controlled items in terms of core technical parameters, materials, and uses.

[0082] S205. When the relevance score is equal to the preset relevance threshold, a preset multi-dimensional judgment strategy is used to identify the item to be identified in order to generate an identification result. The identification result is used to indicate whether the item to be identified is a controlled item.

[0083] Among them, the preset relevance threshold refers to the pre-set critical score used to distinguish between clearly irrelevant, suspectedly relevant, and clearly relevant.

[0084] Among them, the pre-set multi-dimensional judgment strategy refers to a set of multi-level, progressive comprehensive judgment rules pre-established based on dual-use item control regulations and review logic, which is used to verify borderline suspected items item by item.

[0085] The identification result refers to the final compliance conclusion output after a complete review process.

[0086] Specifically, when the relevance score of the item to be identified equals a preset relevance threshold, the item is identified as a borderline suspected item, and a preset multi-dimensional judgment strategy is initiated for final refined judgment. The item to be identified is matched with the control entries obtained from vector retrieval to determine whether they belong to the same material category, equipment category, or technical field. If the categories are significantly inconsistent, the control risk is directly excluded.

[0087] Furthermore, the application scenarios of the identified items are verified for clarity, determining whether the items have clear and conventional civilian or industrial uses. For items with clear uses and not involving sensitive areas, their regulatory risk level is reduced. Next, a combination characteristic integrity verification is performed to determine whether the item simultaneously meets all the limiting conditions required by the regulatory item; items that only meet some conditions are not considered compliant with regulatory requirements. Finally, typical misjudgment traps are identified, eliminating common misjudgment situations such as similar keywords but different categories, containing regulated materials but being a final product, confusion between testing equipment and the tested object, and matching a single parameter but not meeting the overall conditions.

[0088] Optionally, after completing the verification of all dimensions, the final identification result is formed by combining all judgment criteria, and the reasons for the conclusion are clearly marked for subsequent review and auditing.

[0089] The controlled item identification method based on dual-use items provided in this application first performs simple matching based on a preset category library to quickly screen out ordinary civilian items that obviously pose no control risk, significantly reducing the computational load of subsequent deep identification. It avoids unnecessary complex analysis of non-sensitive items, reducing false positive rates and resource consumption. A coarse-grained risk classification is achieved through a dual-track approach of keyword identification and rule-based reasoning. Risk levels are represented by quantitative scoring, standardizing and objectifying risk judgment and eliminating subjective human bias. This provides clear triage criteria for subsequent processes, enabling tiered processing where low-risk items are quickly processed and high-risk items are more finely assessed. When the security score is greater than or equal to a preset security threshold, the item to be identified is considered high-risk and requires further identification. Vector retrieval is used to achieve accurate matching of control items, solving the problems of knowledge lag and inaccurate understanding in large models. Multi-dimensional scoring quantifies the correlation between items and control items, resulting in more refined and stable identification. Using scores as triage thresholds makes the system's identification logic explainable, verifiable, and auditable. For borderline ambiguous items, i.e., items with a correlation score equal to a preset correlation threshold, a refined secondary verification is performed, significantly improving the accuracy of identification in boundary scenarios. The system systematically eliminates common misjudgments through multi-dimensional verification. It outputs structured conclusions and supporting evidence, meeting export control compliance audit requirements. While ensuring compliance, it significantly improves review efficiency and reduces the misjudgment rate.

[0090] As an optional implementation, based on the above embodiments, a preset keyword library includes keywords of multiple categories and corresponding weights for each category. A security score for the item to be identified is generated based on the preset keyword library and preset inference logic rules, including:

[0091] The item description information is matched against a preset keyword library to obtain keyword recognition results;

[0092] The initial security score of the item to be identified is generated by combining and weighting the keyword categories and locations in the keyword recognition results.

[0093] The keyword recognition results are inferred using preset inference logic rules to generate inference results; the preset inference logic rules include at least one of category matching, usage constraint judgment, technical feature matching, and item conformity inference.

[0094] The initial security score is adjusted based on the reasoning results to obtain the security score of the item to be identified.

[0095] The keyword category refers to the type of keywords classified according to the controlled objects of dual-use items, including materials, equipment, chemicals, software technology, end use, etc.

[0096] Among them, category weight refers to the score coefficient pre-set for keyword categories with different risk levels.

[0097] Among them, combined weighting refers to a calculation method that combines multiple factors such as the keyword category, the number of hits, and the position of appearance, and adds and adjusts the score according to preset rules.

[0098] The initial security score refers to the preliminary risk score obtained solely through keyword matching and weighting.

[0099] Category matching refers to matching the category corresponding to the keyword with the major and intermediate categories in the control catalog to determine the control scope to which the item may belong.

[0100] Among them, the use constraint judgment refers to determining whether the use information in the item description falls within the scope of sensitive uses limited by the control catalog.

[0101] Technical feature matching refers to comparing the structure, performance, parameters, and other characteristics of an item with the technical requirements of the control catalog.

[0102] Among them, item compliance reasoning refers to judging whether an item fully meets all the conditions stipulated in a certain control item according to the item logic of the control catalog.

[0103] The reasoning result refers to the conclusion that the items are related to the control catalog obtained through rule-based reasoning.

[0104] Specifically, the process involves obtaining complete descriptive information about the item to be identified, including its name, material, specifications, purpose, and technical parameters. The text content is then globally matched against various keywords in a pre-defined keyword library to identify all matched keywords. The category of each keyword and its position within the description are recorded, forming the keyword identification result. A basic score is assigned based on the category weight of each matched keyword, and then further weighted according to the keyword's position. Keywords appearing in key positions such as the item name or core parameters receive higher scores, while those appearing in secondary positions such as notes or supplementary descriptions receive lower scores. Additionally, cross-category keyword combinations are given extra weight. All weighted results are then summed to obtain an initial security score.

[0105] For example, based on keyword recognition results, items are first categorized into corresponding major and minor control categories to complete category matching. Next, it is determined whether the item's purpose falls under sensitive scenarios prohibited or restricted by the control catalog, completing purpose constraint judgment. Then, the item's technical specifications, structural functions, and control requirements are compared with those of the control entries to complete technical feature matching. Finally, following the logic of simultaneously satisfying multiple conditions of a control entry, it is determined whether the item fully complies with a certain control requirement, forming an inference result. The initial score is calibrated based on the inference result; a high match rate between the item and the control catalog increases the score, while a significant non-compliance with control conditions decreases the score; and a partial match but not a complete fulfillment of the requirement maintains a moderate score, ultimately yielding the security score of the item to be identified.

[0106] The controlled article identification method based on dual-use items provided in this application achieves rapid location of sensitive features of the article through keyword matching, transforming unstructured text into structured risk signals, providing standardized input for subsequent scoring, and improving the efficiency and consistency of risk identification. It distinguishes the strength of risk by using category weight and position weight, avoiding misjudgments caused by simple keyword counting, making the initial score more closely reflect the actual risk level, and achieving coarse-grained risk quantification. It transforms the logic of regulatory regulations into an executable judgment process, compensating for the lack of logical depth in keyword matching, making risk judgments more aligned with the original intent of regulations, and reducing misjudgments caused by simple keyword matching. Through dual calibration of keyword weighting and rule reasoning, the score is more accurate, stable, and closely aligned with regulatory requirements, providing a reliable basis for subsequent processes, balancing review efficiency and identification accuracy.

[0107] As an optional implementation, based on the above embodiments, a relevance score for the item to be identified is generated based on the vector retrieval results and preset scoring rules, including:

[0108] Similarity searches are performed from a preset vector database based on item description information to obtain vector search results. The preset vector database stores dual-purpose item control entry vectors that have undergone structured segmentation and parameter extraction.

[0109] Based on the vector retrieval results, a weighted score is applied to multiple preset dimensions using preset scoring rules. The preset dimensions include at least two of the following: item use, technical parameters, material type, and item form.

[0110] A relevance score for the item to be identified is generated based on the scores corresponding to multiple preset dimensions.

[0111] Among them, the preset vector database refers to a professional database that pre-builds and stores vector data of dual-use item control entries. For example, the control list document is split into second-level headings using MarkdownHeaderTextSplitter. Each document block contains complete control codes, original description information, structured rewritten descriptions and key technical parameters. Each document block is converted into a text embedding vector for storage.

[0112] Among them, weighted scoring refers to assigning corresponding weights to each preset dimension based on its influence on control identification, and then converting the scores of each dimension based on the weights.

[0113] Specifically, the descriptive information of the item to be identified is converted into vector form. A similarity search is performed in a pre-set vector database. The degree of similarity is determined by calculating the distance between the item vector and the vector of the controlled item in the database. Controlled items whose similarity distance meets the preset conditions are selected and returned, forming vector search results. Evaluation is conducted according to at least two of the four dimensions of item use, technical parameters, material type, and item form. Scores are assigned based on the degree of matching between each dimension and the control requirements for dual-use items. At the same time, corresponding weights are assigned to each dimension according to the importance of control identification. The original scores of each dimension are converted into weighted scores for each dimension.

[0114] For example, if it is a sensitive material, add 2 points; if the intended use is clearly defined as civilian / sports / home / automotive / decoration, add 2 points; if the technical parameters do not match, add 0 points; if the item is a raw material / semi-finished product, add 2 points.

[0115] Furthermore, the weighted scores of each dimension are arithmetically summed to obtain the original total score. The original total score is then normalized to a preset score range, such as 1-10, through a linear mapping method, and finally the relevance score of the item to be identified is generated.

[0116] The controlled article identification method based on dual-use items provided in this application achieves accurate similarity matching between the item to be identified and the controlled item by using a pre-set vector database of controlled item vectors that have undergone structured segmentation and parameter extraction. This effectively solves the problems of semantic bias and keyword omission that are prone to occur in simple text matching, and significantly improves the accuracy and efficiency of retrieval. By comprehensively covering four core dimensions—item purpose, technical parameters, material type, and item form—it achieves a multi-dimensional and all-round assessment of the compliance risk of the item, avoiding the one-sidedness caused by single-dimensional scoring. Through a weighted scoring method, it highlights the differentiated importance of different dimensions in controlled article identification, making the scoring more in line with the actual logic of dual-use item control, effectively distinguishing items of different risk levels, and reducing misjudgments and omissions.

[0117] As an optional implementation, based on the above embodiments, the product form includes raw materials, semi-finished products, and finished products, and the weighted score includes:

[0118] Increase the weight score for items in raw material form and decrease the weight score for items in finished product form.

[0119] Items intended for civilian use will have their weight score reduced, while items made of pre-defined sensitive materials or with pre-defined control components will have their weight score increased.

[0120] Among them, pre-defined sensitive materials refer to high-risk materials pre-designated according to the dual-use item control catalog, such as high-strength alloys, carbon fibers, special ceramics, and high-temperature resistant composite materials.

[0121] Among them, the preset control components refer to the core functional components that are of key concern in the regulatory catalog, such as high-precision controllers, CNC modules, servo drives, sensors and other key components.

[0122] Specifically, when conducting multi-dimensional weighted scoring, the first step is to identify and determine the form of the item to be identified based on its description information, distinguishing whether it belongs to raw materials, semi-finished products, or finished products. For items in the form of raw materials, due to their higher risk of proliferation and modifiability, the corresponding weight score is increased according to preset rules. For items in the form of finished products, due to their fixed use and clear civilian attributes, the corresponding weight score is decreased according to preset rules.

[0123] Furthermore, the actual purpose and key component information of the item are identified to determine whether it belongs to a clearly defined civilian purpose and whether it contains pre-set sensitive materials or pre-set control components. For items with a clearly defined civilian purpose, the corresponding weight score is reduced to avoid over-sensitivity to ordinary civilian items. For items containing pre-set sensitive materials or belonging to pre-set control components, the corresponding weight score is increased to strengthen the identification of high-risk objects. The final weighted score for this dimension is formed by combining the above weighted adjustment results.

[0124] The controlled article identification method based on dual-use items provided in this application can accurately distinguish the level of control risk based on the physical form of the article by adjusting the weights according to the differences in article form. It increases the risk weight for raw materials that are easily repurposed or reprocessed, and decreases the weight for finished products with fixed uses and controllable risks. This makes the scoring more consistent with the actual regulatory logic of export control and reduces misjudgments of end-use civilian products. Adjusting the weights based on use, sensitive materials, and control components adheres to the identification principle of prioritizing civilian use and exempting sensitive items. This ensures that ordinary civilian articles receive reasonable score reductions and achieve rapid release, while accurately increasing the scores for high-risk materials and core components, avoiding the underestimation of the risk of sensitive articles, and significantly improving the rationality and accuracy of the scoring.

[0125] As an optional implementation, based on the above embodiments, a preset multi-dimensional judgment strategy is used to identify the item to be identified, including:

[0126] Perform preset verification processing on the item to be identified and the vector retrieval results. The preset verification processing includes at least one of the following: category matching verification, application scenario verification, combination characteristic verification, and typical example verification.

[0127] If any verification fails, the item to be identified is determined not to be a dual-use item;

[0128] If all verifications pass, the item to be identified is determined to be or may be a dual-purpose item.

[0129] Among them, category matching verification refers to comparing the technical category, material category, or equipment category to which the item to be identified belongs with the category of controlled items. If the categories do not match, the item is directly identified, and dual-use items are not involved.

[0130] Application scenario verification refers to analyzing the actual usage scenario and purpose description of the item to be identified, and determining whether it is a regular civilian scenario, a general industrial scenario, or a sensitive scenario such as aerospace, military, or nuclear energy.

[0131] Among them, the combined characteristic verification refers to verifying whether the item to be identified meets multiple limiting conditions simultaneously, rather than only meeting a single scattered characteristic, in accordance with the requirements of the dual-use item control items.

[0132] Among them, typical example verification refers to comparing and verifying based on common historical misjudgment situations to exclude typical easily confused situations such as similar keywords but irrelevant in substance, and similar parameters but inconsistent overall.

[0133] Specifically, the items to be identified undergo four progressive checks in sequence. First, a category matching check is performed, comparing the item's material, equipment, or technology category with the relevant control item to determine if they belong to the same category. If the categories are clearly inconsistent, the check fails. Next, an application scenario check is performed, analyzing the purpose information in the item's description to confirm whether its application scenario is clearly civilian or conventional industrial. Items with clear purposes and no sensitive indications can be directly excluded from control risks.

[0134] Furthermore, a combined characteristic verification is performed to determine whether the item simultaneously meets multiple technical, parameter, and usage constraints stipulated in the control item. If only some conditions are met, it is considered not to meet the control requirements. Finally, a typical example verification is performed to compare the item with common misjudged samples to eliminate situations such as similar keywords but different categories, containing sensitive materials but being a finished product, or confusion between the test equipment and the tested object.

[0135] Understandably, if any check fails during the entire verification process, the item is deemed not to be a dual-use item. Only when all checks pass is the item identified as either a dual-use item or potentially a dual-use item. Finally, the verification conclusions are compiled into the identification criteria and output along with the identification result.

[0136] The identification basis refers to the structured description used to support the final identification conclusion, including the verification results, the characteristics of the matched control items, and the reasons for risk exclusion.

[0137] The controlled article identification method based on dual-use items provided in this application uses category matching verification to quickly distinguish whether an article is related to a controlled item from a holistic perspective. This can eliminate a large number of irrelevant categories of articles in the early stages, reducing subsequent invalid verifications and improving overall identification efficiency. Application scenario verification prioritizes civilian use, promptly eliminating risks for articles with clear civilian applications and effectively reducing the false positive rate. Combination characteristic verification aligns with the regulatory logic that multiple conditions must be met simultaneously for a dual-use item to constitute a controlled item, avoiding false positives triggered by a single feature or scattered keywords, significantly improving identification accuracy and regulatory consistency. Typical example verification specifically addresses boundary scenarios where the model is prone to confusion and misjudgment, further filtering incorrect matches through empirical verification rules, making the identification more stable and closer to the judgment logic of human experts. Simultaneously, it outputs traceable identification evidence, making the results interpretable and meeting export control compliance audit requirements.

[0138] As an optional implementation, based on the above embodiments, typical example verification includes:

[0139] Determine whether the item to be identified meets any of the preset examples. The preset examples include items with similar keywords but different categories, items containing preset sensitive materials but which are finished products, items where the test equipment is confused with the test object, and items where a single parameter matches but the combined conditions are not met.

[0140] Specifically, for items with similar keywords but different categories, the core category and keywords of the item to be identified are extracted and precisely compared with the controlled item categories retrieved by vector retrieval. If only the keywords are similar in literal sense, but the categories are completely different, such as CNC lathe parts and CNC missile launchers, it is identified as a false positive scenario and excluded. Secondly, for items containing preset sensitive materials but which are finished products, it is first identified whether the item contains preset sensitive materials, and then it is determined whether the item form is a finished product and whether the purpose is clearly civilian use. If it is confirmed to be a civilian finished product containing sensitive materials, such as carbon fiber bicycle frames or titanium alloy civilian tableware, it is identified as a false positive scenario and excluded.

[0141] Furthermore, for items where the testing equipment and the object under test are confused, the functional description of the item is analyzed to distinguish whether the item itself is a testing device or the object under test. If the item is only used for testing or inspection of sensitive objects and does not have sensitive functions or uses itself, such as a drone battery testing instrument, it is identified as a misjudgment scenario and excluded. Finally, for items where a single parameter matches but the combined conditions are not met, the parameters, uses, and morphological characteristics of the item are checked against all the combined conditions specified in the control item. If only a single parameter matches, but other conditions are not met, such as a device whose rotation speed matches the control parameter, but whose use and material do not match, it is identified as a misjudgment scenario and excluded.

[0142] The controlled article identification method based on dual-use items provided in this application specifically addresses four types of high-frequency misjudgment scenarios, compensates for minor loopholes in the preceding verification process, makes the system's identification logic closer to the judgment thinking of human experts, significantly reduces the misjudgment rate, improves identification stability, and makes the identification results more interpretable, providing a clear basis for misjudgment exclusion for compliance audits.

[0143] As an optional implementation, based on the above embodiments, it further includes:

[0144] If the security score is less than the preset security threshold, the item to be identified is determined not to be a dual-use item.

[0145] If the correlation score is less than the preset correlation threshold, the item to be identified is determined not to be a dual-use item.

[0146] If the correlation score is greater than the preset correlation threshold, the item to be identified is determined to be or may be a dual-purpose item.

[0147] Specifically, after completing the security scoring, the score is compared with a preset security threshold. If the security score is lower than the preset security threshold, it indicates that the sensitive features of the item to be identified are weak and the regulatory risk is low. The item is then directly identified as not being a dual-use item, and the subsequent process is terminated.

[0148] Furthermore, for items entering the relevance scoring stage, a further triage judgment is performed after obtaining the relevance score. If the relevance score is less than a preset relevance threshold, it indicates that the item has a low correlation with the regulated item, and the item is directly identified as not belonging to the dual-use category. If the relevance score is greater than the preset relevance threshold, it indicates that the item is highly matched with the regulated item and has significant regulated risk, and the item is directly identified as belonging to or potentially belonging to the dual-use category. Only borderline items whose relevance score equals the preset relevance threshold will proceed to the subsequent multi-dimensional fine-grained identification.

[0149] The controlled article identification method based on dual-use items provided in this application adopts a multi-level threshold triage mechanism to achieve tiered processing of items with different risk levels. This ensures accurate identification of high-risk items while allowing low-risk items to be released quickly, significantly improving overall review efficiency. By using clear threshold judgment rules and unified identification standards, subjective differences and model illusions are reduced, lowering the probability of false positives and false negatives. Multi-level threshold filtering effectively reduces the computational cost of deep verification, while making the entire identification process clear and traceable, meeting export control compliance audit requirements, and balancing review efficiency, identification accuracy, and business compliance.

[0150] Figure 3 A flowchart illustrating a controlled article identification method based on dual-use items, as provided in another embodiment of this application, is shown below. Figure 3 As shown, the controlled article identification method based on dual-use items provided in this embodiment includes the following steps:

[0151] S301. Obtain the item to be identified, which includes item description information.

[0152] S302. Determine whether the item to be identified belongs to the category of non-dual-use items based on the preset category library.

[0153] S303. If the item to be identified belongs to the category of non-dual-use items, output that the item to be identified belongs to the category of non-dual-use items and end the process.

[0154] S304. If the item to be identified belongs to the dual-use item category, the description information of the item to be identified is matched in the preset keyword library to obtain the keyword identification result.

[0155] S305. Based on the keyword category and keyword location in the keyword recognition results, a weighted combination is performed to generate an initial security score for the item to be identified.

[0156] S306. Use preset reasoning logic rules to reason about the keyword recognition results to generate reasoning results; the preset reasoning logic rules include at least one of category matching, usage constraint judgment, technical feature matching and item conformity reasoning.

[0157] S307. Adjust the initial security score based on the reasoning results to obtain the security score of the item to be identified.

[0158] S308. Compare the security score of the item to be identified with the preset security threshold.

[0159] S309. If the security score is less than the preset security threshold, then execute S303.

[0160] S310. In response to a security score greater than or equal to a preset security threshold, a similarity search is performed from a preset vector database based on the description information of the item to be identified to obtain vector search results. The preset vector database stores dual-purpose item control entry vectors that have undergone structured segmentation and parameter extraction.

[0161] S311. Based on the vector retrieval results, a weighted score is applied to multiple preset dimensions using preset scoring rules. The preset dimensions include at least two of the following: item use, technical parameters, material type, and item form.

[0162] S312. Generate a relevance score for the item to be identified based on the scores corresponding to multiple preset dimensions.

[0163] S313. Compare the relevance score of the item to be identified with the preset relevance threshold.

[0164] S314. If the correlation score is less than the preset correlation threshold, then execute S303.

[0165] S315. If the correlation score is greater than the preset correlation threshold, the identified item is determined to be or may be a dual-purpose item, the identification result containing the identification basis is output, and the process ends.

[0166] S316. If the relevance score is equal to the preset relevance threshold, then the preset verification process is performed on the item to be identified and the vector retrieval result. The preset verification process includes at least one of the following: category matching verification, application scenario verification, combination characteristic verification, and typical example verification.

[0167] S317. If any check fails, then execute S303.

[0168] S318. If all checks pass, then execute S315.

[0169] In this embodiment, the implementation method and technical effect of S301-S318 are similar to those of the corresponding solutions in the above embodiments, and will not be repeated here.

[0170] Figure 4 A schematic diagram of the structure of the controlled article identification device based on dual-use items provided in this application is shown below. Figure 4 As shown, the controlled article identification device 40 based on dual-use items provided in this embodiment includes: a determination module 41, a generation module 42, and an identification module 43.

[0171] The system includes: an acquisition module 41 for acquiring the item to be identified, which includes item description information; a determination module 42 for determining whether the item to be identified belongs to a non-dual-use item category based on a preset category library; a generation module 43 for generating a security score for the item to be identified based on a preset keyword library and preset reasoning logic rules when the item to be identified belongs to a dual-use item category; a generation module 43 for generating a relevance score for the item to be identified based on vector retrieval results and preset scoring rules when the security score is greater than or equal to a preset security threshold; and an identification module 44 for identifying the item to be identified using a preset multi-dimensional judgment strategy when the relevance score is equal to a preset relevance threshold, thereby generating an identification result, which indicates whether the item to be identified is a controlled item.

[0172] The controlled article identification device based on dual-use items provided in this embodiment can perform... Figure 2 The implementation principles and technical effects of the methods shown are similar, and will not be repeated here.

[0173] Optionally, the preset keyword library includes keywords of multiple categories and the corresponding weights for each category. The generation module 42, when generating a security score for the item to be identified based on the preset keyword library and preset inference logic rules, specifically performs the following: matching the item description information in the preset keyword library to obtain keyword recognition results; combining and weighting the keyword categories and positions in the keyword recognition results to generate an initial security score for the item to be identified; inferring the keyword recognition results using preset inference logic rules to generate an inference result; the preset inference logic rules include at least one of category matching, usage constraint judgment, technical feature matching, and item conformity inference; and adjusting the initial security score based on the inference result to obtain a security score for the item to be identified.

[0174] Optionally, the generation module 42, when generating a relevance score for the item to be identified based on the vector retrieval results and preset scoring rules, specifically performs the following: a similarity retrieval is performed from a preset vector database based on the item description information to obtain vector retrieval results. The preset vector database stores dual-purpose item control entry vectors that have undergone structured segmentation and parameter extraction; a weighted score is applied to multiple preset dimensions based on the vector retrieval results using preset scoring rules. The preset dimensions include at least two of the following: item use, technical parameters, material type, and item form; and a relevance score for the item to be identified is generated based on the scores corresponding to the multiple preset dimensions.

[0175] Optionally, the item form includes raw materials, semi-finished products, and finished products. The generation module 42 is specifically used in weighted scoring to: increase the weight score of items in the form of raw materials, decrease the weight score of items in the form of finished products; decrease the weight score of items for civilian use, and increase the weight score of items that belong to preset sensitive materials or preset control components.

[0176] Optionally, when the identification module 43 identifies the item to be identified using a preset multi-dimensional judgment strategy, it specifically performs the following: performs preset verification processing on the item to be identified and the vector retrieval result. The preset verification processing includes at least one of category matching verification, application scenario verification, combination characteristic verification, and typical example verification. If any verification fails, it is determined that the item to be identified is not a dual-purpose item. If all verifications pass, it is determined that the item to be identified is or may be a dual-purpose item.

[0177] Optionally, during typical example verification, the identification module 43 is specifically used to: determine whether the item to be identified meets any of the preset examples. The preset examples include items with similar keywords but different categories, items containing preset sensitive materials but which are finished products, items where the test equipment and the test object are confused, and items where a single parameter matches but the combined conditions are not met.

[0178] Optionally, the determining module 41 is further configured to determine that the item to be judged is not a dual-use item if the security score is less than a preset security threshold; determine that the item to be judged is not a dual-use item if the correlation score is less than a preset correlation threshold; and determine that the item to be judged is or may be a dual-use item if the correlation score is greater than a preset correlation threshold.

[0179] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Figure 5 As shown, the electronic device 50 provided in this embodiment includes a processor 51 and a memory 52. ​​The processor 51 and the memory 52 are connected via a bus and communicate with each other.

[0180] In the specific implementation process, the processor 51 executes the computer execution instructions stored in the memory 52, causing the processor 51 to perform the above-described method.

[0181] The specific implementation process of processor 51 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0182] In the above embodiments, it should be understood that the processor 51 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0183] The memory 52 may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0184] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0185] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0186] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0187] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0188] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0189] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0190] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0191] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0192] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0193] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0194] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for identifying controlled articles based on dual-use items, characterized in that, The method includes: Acquire the item to be identified, the item to be identified including item description information; Determine whether the item to be identified belongs to the category of non-dual-use items based on a pre-defined category library; If the item to be identified belongs to the dual-use item category, a security score for the item to be identified is generated based on a preset keyword library and preset reasoning logic rules. In response to the security score being greater than or equal to a preset security threshold, a relevance score for the item to be identified is generated based on the vector retrieval results and preset scoring rules; When the correlation score is equal to a preset correlation threshold, a preset multi-dimensional judgment strategy is used to identify the item to be identified in order to generate an identification result, which is used to indicate whether the item to be identified is a controlled item.

2. The method according to claim 1, characterized in that, The preset keyword library includes keywords of multiple categories and corresponding weights for each category. The step of generating a security score for the item to be identified based on the preset keyword library and preset inference logic rules includes: The item description information is matched against a preset keyword library to obtain keyword recognition results; The initial security score of the item to be identified is generated by combining and weighting the keyword categories and locations in the keyword recognition results. The keyword recognition results are inferred using preset inference logic rules to generate inference results; the preset inference logic rules include at least one of category matching, usage constraint judgment, technical feature matching, and item conformity inference. The initial security score is adjusted based on the reasoning results to obtain a security score for the item to be identified.

3. The method according to claim 1, characterized in that, The process of generating a relevance score for the item to be identified based on vector retrieval results and preset scoring rules includes: Based on the item description information, a similarity search is performed from a preset vector database to obtain vector search results. The preset vector database stores dual-purpose item control entry vectors that have undergone structured segmentation and parameter extraction. Based on the vector retrieval results, a weighted score is applied to multiple preset dimensions using preset scoring rules. The preset dimensions include at least two of the following: item use, technical parameters, material type, and item form. A relevance score for the item to be identified is generated based on the scores corresponding to multiple preset dimensions.

4. The method according to claim 3, characterized in that, The product forms include raw materials, semi-finished products, and finished products, and the weighted score includes: Increase the weight score for items in raw material form and decrease the weight score for items in finished product form. Items intended for civilian use will have their weight score reduced, while items made of pre-defined sensitive materials or with pre-defined control components will have their weight score increased.

5. The method according to claim 1, characterized in that, The step of using a preset multi-dimensional judgment strategy to identify the item to be identified includes: Perform a preset verification process on the item to be identified and the vector retrieval result. The preset verification process includes at least one of the following: category matching verification, application scenario verification, combination characteristic verification, and typical example verification. If any verification fails, the item to be identified is determined not to be a dual-use item; If all verifications pass, the item to be identified is determined to be or may be a dual-purpose item.

6. The method according to claim 5, characterized in that, The typical example verification includes: Determine whether the item to be identified meets any of the preset examples. The preset examples include items with similar keywords but different categories, items containing preset sensitive materials but which are finished products, items where the testing equipment is confused with the tested object, and items where a single parameter matches but the combined conditions are not met.

7. The method according to claim 1, characterized in that, Also includes: If the security score is less than the preset security threshold, then the item to be identified is determined not to be a dual-use item. If the correlation score is less than the preset correlation threshold, then the item to be identified is determined not to be a dual-purpose item. If the correlation score is greater than the preset correlation threshold, then the item to be identified is determined to be or may be a dual-purpose item.

8. A controlled article identification device based on dual-purpose items, characterized in that, include: The acquisition module is used to acquire the item to be identified, the item to be identified including item description information; The determination module is used to determine whether the item to be identified belongs to the category of non-dual-use items based on a preset category library. The generation module is used to generate a security score for the item to be identified based on a preset keyword library and preset reasoning logic rules when the item to be identified belongs to the dual-use item category. The generation module is also used to generate a relevance score for the item to be identified based on the vector retrieval results and the preset scoring rules in response to the security score being greater than or equal to a preset security threshold. The identification module is used to identify the item to be identified by adopting a preset multi-dimensional judgment strategy when the relevance score is equal to a preset relevance threshold, so as to generate an identification result, which is used to indicate whether the item to be identified is a controlled item.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

11. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-7.

Images