Data processing method and system, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a data processing method and system, electronic equipment, a storage medium and a program product. According to one method, after a first article classification identifier corresponding to a target article is determined according to multi-modal information of the target article, a tax rule suitable for transportation of the target article is obtained from a knowledge graph based on the first article classification identifier; the knowledge graph is dynamically constructed on the basis of import and export article control information, obtained in real time, of each region, and the knowledge graph contains a plurality of nodes and node relations; the nodes have two types, namely article classification identifiers and tax rules; the node relationship reflects the association relationship between the article classification identifier and the tax rule. And further, determining a tax value of the target article based on the applicable tax rule, thereby automatically generating a customs declaration file of the target article according to the tax value. Therefore, automatic generation of the customs declaration file can be realized, so that the customs declaration efficiency is improved, and errors and tedious processes of manual operation are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, and in particular, relates to a data processing method and system, electronic device, storage medium and program product. BACKGROUND

[0002] With the continuous development of globalization trade and the continuous expansion of cross-border e-commerce scale, the tariff policy interpretation and clearance process in international trade has become a key link of enterprise operation. However, the frequent adjustment of tariff policy of each country and the complexity of multi-language and multi-rule make enterprises face more complex tax and clearance challenges. The traditional tariff policy interpretation and clearance process mainly relies on manual processing, which has problems such as complex processing, slow response and easy to make mistakes. Therefore, it is urgent to provide an intelligent means to optimize the cognition of tariff policy of each country and clearance decision. SUMMARY

[0003] The embodiments of the present application provide a data processing method, system, electronic device, storage medium and program product to improve the clearance efficiency of goods and reduce the clearance risk.

[0004] In a first embodiment, the present application provides a data processing method, which comprises:

[0005] According to the multi-modal information of the target object, a first object classification identifier corresponding to the target object is determined;

[0006] Based on the first object classification identifier, a tariff rule applicable to the transportation of the target object is obtained from a knowledge graph;

[0007] Based on the applicable tariff rule, a tariff value of the target object is determined;

[0008] According to the tariff value, a customs declaration file of the target object is generated.

[0009] The knowledge graph is constructed based on the obtained import and export object control information of each region, and contains a plurality of nodes and node relationships; the nodes have two types, which are object classification identifier and tariff rule; the node relationship reflects the association relationship between the object classification identifier and the tariff rule.

[0010] In a second embodiment, the present application further provides a data processing method, which comprises:

[0011] Real-time collection of import and export commodity control information published by each region;

[0012] Using a second preset model, the collected import and export commodity control information is extracted and processed to obtain an extraction result;

[0013] Constructing a knowledge graph based on the extraction results corresponding to each customs area;

[0014] Based on the knowledge graph, tax rate analysis is performed on the target product.

[0015] Among them, the knowledge graph contains multiple nodes and node relationships; there are two types of nodes, namely commodity classification identification and tariff rules, and the node relationship reflects the relationship between commodity classification identification and tariff rules. One commodity classification identification represents a category of commodities.

[0016] In a third embodiment, the present application further provides a data processing method, the method comprising:

[0017] Obtain multimodal information of the target product;

[0018] Based on the multimodal information, a first product category identifier corresponding to the target product is determined.

[0019] Fill the first commodity classification identifier into the customs declaration template applicable to the target commodity to generate a customs declaration document for the target commodity.

[0020] In a fourth embodiment, the present application further provides a data processing method, the method comprising:

[0021] In response to a user operation on a target commodity, obtaining an assumed condition of the target commodity, the assumed condition indicating supply chain change information set for the target commodity, the supply chain change information causing a change in commodity attribute information required for customs declaration of the target commodity;

[0022] Based on the above assumptions, show the extent of the tax rate change for the target product after the supply chain change;

[0023] Among them, the tax rate change range is determined after the commodity attribute information of the target commodity after the change is determined based on the supply chain change information, and the tariff rules applicable to the target commodity after the supply chain change are queried from the knowledge graph based on the changed commodity attribute information. The queried tariff rules are used to trigger the preset tax rate fluctuation simulation analysis model for prediction.

[0024] In a fifth embodiment, the present application further provides a data processing method, the method comprising:

[0025] Determining a first product category identifier corresponding to the target product based on the multimodal information of the target product;

[0026] Extracting and processing the updated import and export commodity control information of the first region to obtain an extraction result;

[0027] Determining, based on the extraction result, a tariff rule after the change in the first region for the first commodity classification identifier;

[0028] Analyze the multi-dimensional trade impact of the changed tariff rules on target commodities;

[0029] Based on the multi-dimensional trade impact, early warning information of the target commodity is output.

[0030] In a sixth embodiment, the present application further provides a data processing system, the system comprising:

[0031] The client is configured to respond to a user's operation on a target item and send multimodal information of the target item to the server;

[0032] A server, configured to execute the method provided in the first embodiment above upon receiving the multimodal information of the target object; and further configured to implement the methods provided in the second, third, and fifth embodiments above;

[0033] Furthermore, the client is further configured to, in response to a user operation on a target commodity, obtain hypothetical conditions for the target commodity and request the server to analyze tax rate changes for the target commodity under the hypothetical conditions; wherein the hypothetical conditions represent supply chain change information set for the target commodity, and the supply chain change information causes changes in commodity attribute information required for customs declaration of the target commodity;

[0034] The server is further configured to: in response to a request from the client, determine the commodity attribute information of the target commodity after the change based on the supply chain change information; based on the changed commodity attribute information, query the tariff rules applicable to the target commodity after the supply chain change from the knowledge graph; and use the queried tariff rules to trigger a preset tax rate fluctuation simulation analysis model to predict the tax rate change range of the target commodity after the supply chain change;

[0035] The client is also used to display the tax rate change range.

[0036] In a seventh aspect, the present application provides an electronic device comprising: a memory and a processor, wherein the memory is configured to store a program; and the processor is coupled to the memory and configured to execute the program stored in the memory to implement any of the above methods.

[0037] In an eighth aspect, the present application provides a computer-readable storage medium storing a computer program, which, when executed by a computer, can implement any of the methods described above.

[0038] In a ninth aspect, the present application provides a computer program product, comprising a computer program, which implements any of the above-described methods when executed by a processor.

[0039] In the technical solution provided in the embodiment of the present application, the first item classification identifier (such as the first commodity classification identifier) ​​corresponding to the target item is determined based on the multimodal information of the target item (such as the target commodity). This can ensure that the target item is classified more accurately, thereby facilitating the subsequent improvement of the accuracy of the determination of the tariff rules corresponding to the target item. Afterwards, based on the first item classification identifier determined above, the tariff rules applicable to the transportation of the target item can be obtained from the knowledge graph. This knowledge graph is dynamically constructed based on the import and export item control information (such as import and export commodity control information) of various regions obtained in real time. The knowledge graph contains multiple nodes and node relationships; the nodes are of two types, namely item classification identifiers and tariff rules; the node relationship reflects the relationship between the item classification identifier and the tariff rules. It can be seen that the present application realizes the use of knowledge graphs to intuitively represent the complex relationship between each item classification identifier (HS code) and its related tariff rules, which is convenient for querying the applicable tariff rules for the target item and simplifying the complexity of the target item declaration. In addition, when the tariff rules of a region change, the updated tariff rules of this region can be quickly reflected in the knowledge graph by updating the knowledge graph. This can help users to timely understand the tariff rule adjustments of various regions through the knowledge graph, so as to make more accurate business decisions. Furthermore, based on the tariff rules applicable to the target items obtained by querying the knowledge graph, the tariff value of the target items can be determined, and the customs declaration documents of the target items can be automatically generated based on this tariff value. In summary, it can be seen that the present application solution can realize the automatic generation of customs declaration documents, which can help improve customs declaration efficiency and reduce errors and cumbersome processes in manual customs declaration. Furthermore, using the target item as an example, this application also supports counterfactual analysis for the target item. Specifically, the counterfactual analysis is performed based on user-entered hypothetical conditions for the target item. These hypothetical conditions represent supply chain changes set for the target item, which cause changes in the commodity attribute information required for customs declaration. Based on this changed commodity attribute information, the tariff rules applicable to the target item after the supply chain change can be retrieved from the knowledge graph. This retrieved tariff rule is then used to trigger a preset tax rate fluctuation model analysis model to predict and output the tax rate change range for the target item after the supply chain change. This counterfactual analysis can help users understand the specific impact of a specific supply chain change on tariffs, thereby making better trade decisions. Furthermore, this application also supports proactive compliance warnings. Specifically, if the trade market involved in the target item includes a first region, after extracting key information from the first region's updated import and export control information and determining the tariff rules for the first commodity classification identifier corresponding to the target item in the first region based on the proposed results, the multi-dimensional trade impact of this changed tariff rule on the target item can be analyzed, and based on this multi-dimensional trade impact, warning information for the target item can be output.This application provides a proactive compliance early warning mechanism that not only helps users (such as enterprises) keep abreast of changes in import and export control information in relevant regions, but also enables in-depth analysis of the specific impacts of these changes on their products (such as costs, origin compliance, etc.), allowing users to better manage compliance risks and optimize trade based on these impacts. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0041] Figure 1 A flowchart of a data processing method provided by an exemplary embodiment of the present application;

[0042] Figure 2a and Figure 2b An interactive diagram of multimodal product information input provided by an exemplary embodiment of the present application;

[0043] Figure 3a and Figure 3b A schematic diagram of a counterfactual analysis trigger provided for an exemplary embodiment of the present application;

[0044] Figure 4a and 4b A schematic diagram of a tax rate retrospective analysis trigger provided for an exemplary embodiment of this application;

[0045] Figure 5 A schematic diagram of the data processing architecture provided for an exemplary embodiment of the present application;

[0046] Figure 6 A schematic structural diagram of an electronic device provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0047] In recent years, the share of cross-border e-commerce in global international trade has continued to rise, with an average annual growth rate of approximately 7%, becoming a core driving force for the digital transformation of the global supply chain. However, technical barriers and compliance complexities in customs clearance have severely constrained the efficiency of cross-border e-commerce commodity trade. Currently, the mainstream customs clearance technology system has the following key flaws:

[0048] 1) Policy update lag (i.e. policy response lag)

[0049] Traditional customs declaration systems rely on manual retrieval of policy documents. When faced with dynamic regulations such as the EU Carbon Border Adjustment Mechanism (CBAM) and the Regional Comprehensive Economic Partnership (RCEP), there are problems such as deviations in multilingual policy interpretation and update delays, which cause companies to incur high storage costs due to compliance errors.

[0050] 2) The accuracy of the commodity tariff code (HS code) matching is insufficient

[0051] The classification of the HS code depends on artificial experience judgment based on a static rule library, so that some enterprises cause clearance delay due to the incorrect matching of the HS code of the commodity.

[0052] 3) Low utilization rate of tariff preference

[0053] Although the coverage of regional trade agreements (such as RCEP) is improved, due to the fragmentation of supply chain data and the high complexity of rules of origin, the actual utilization rate of tariff preference through the use of RCEP and other agreements is only about 57%.

[0054] Among them, RCEP is a regional comprehensive economic partnership agreement, which involves the rules of tariff reduction and exemption of multiple countries.

[0055] In view of the defects existing in the clearance link, although some related technical solutions have given related solutions, there are still some problems. For example, in a current tariff calculation system based on a rule engine, the updating of the tariff policies of each country needs to be manually programmed, which causes a delay of about 3-5 days in updating the tariff policies, and the system cannot analyze unstructured text. In addition, the HS code matching depends on keyword search, and the matching accuracy is low (less than about 85%). For another example, there is a policy analysis system based on natural language processing (NLP) technology, which needs to predefine entity types, and the multi-language support effect is poor (such as only Chinese and English), and cannot understand the conditional logic (such as "if... then..." sentence) in the policy.

[0056] Through the analysis of the above problems, the present application proposes an intelligent solution for the clearance link. The basic idea of the present application is to use the combination of language models and multi-modal technology to optimize the tariff policy cognition and clearance decision through intelligent means, and to improve the efficiency of commodity clearance and reduce risks.

[0057] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application will be described clearly and completely in combination with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0058] It should be noted that, in the case of user information involved in the embodiments of the present application, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) are in compliance with relevant laws and standards.

[0059] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0060] First, the terms used in the embodiments of the present application are explained. It should be understood that this explanation is for a clearer understanding of the embodiments of the present application and does not necessarily constitute a limitation on the embodiments of the present application.

[0061] The HS Code (Harmonization System Code, HSCODE), also known as a tariff number or tariff number, stands for the International Convention on the Harmonized Commodity Description and Coding System. It is an internationally recognized commodity coding and classification system, also known as a product code. Developed by the International Customs Council, the Harmonized Coding System quantifies the tax rates payable and refundable for various import and export products. The fundamental element used by customs and commodity entry-exit management agencies in various countries to confirm commodity categories, manage commodity classifications, review tariff standards, and inspect commodity quality indicators is the universal certification for imported and exported goods: the HS Code. In e-commerce terms, HSCODE is the classification system for customs-managed commodities, with each commodity classified under only one HS code. Each country can define its own HS Code, but the first six digits are internationally standardized, while the remaining digits are defined based on the specific circumstances of each country. Therefore, the HS code (tariff number) can be understood as the commodity code number listed in the Tariff Code. It is used to identify specific categories of goods and help determine the tariff rate, value-added tax rate and other possible regulatory requirements that should be applied to the goods when they are imported or exported, affecting many aspects such as the cost calculation, customs declaration process, and trade statistics of the goods.

[0062] Specifically, the structure of HS code is usually composed of 6-digit numbers, which divides all international trade goods into 22 categories, 98 chapters, and further into headings and subheadings. Among them, the first two digits of HS code (commodity code) represent the "chapter", the third and fourth digits represent the "heading" (also known as title), and the fifth and sixth digits represent the "subheading" (also known as sub-title). The first 6 digits are the HS international standard code, and HS has 1241 four-digit tax items and 5113 six-digit sub-items. Therefore, HS code is mainly divided into the following three levels: the first two digits represent the "chapter" (Section), the middle two digits represent the "heading" (Heading), and the last two digits represent the "subheading" (Subheading). For example, HS code 01.01.01.00 represents live horses. In HS code, "categories" are basically divided according to economic sectors; "chapters" are basically classified in the following two ways: one is to classify by the properties of commodity raw materials, products made from the same raw materials are generally classified into the same chapter, and the chapters are arranged in order from raw materials to finished products according to the degree of processing; or they are classified into different chapters according to their functions or uses, regardless of the raw materials used, and the chapters are further arranged into headings or subheadings according to the raw materials or processing procedures.

[0063] Customs clearance: In international trade, customs clearance refers to a series of legal procedures and procedures that the trade object must go through when entering or leaving a country or region and reporting to the customs. This process aims to ensure the legal and compliant cross-border circulation of trade objects, protect safety, tax benefits, and implement trade policies. The main steps of customs clearance are as follows:

[0064] Counterfactual analysis: refers to the deduction of the impact of tariffs under certain conditions (such as changing the origin of raw materials).

[0065] Knowledge graph: a structured semantic network used to store the association between "HS code (referred to as item classification identifier in this application)-tariff rules (including corresponding regions, tax rates, and item constraints that need to be met to enjoy tax rates, etc.)".

[0066] Pre-set model, a trained machine learning model. In this application, the pre-set model includes a first pre-set model, a second pre-set model, a third pre-set model, etc. The first pre-set model and the second pre-set model are language models (LLM). Language models (LLM) are deep learning models with a large number of parameters, mainly used for natural language processing (NLP) tasks. In addition, the third pre-set model is a multi-modal model. Multi-modal model refers to a model that can handle multiple data types including images, text, etc. The application embodiments do not limit the number of parameters supported by each pre-set model, aiming to meet the actual application needs.

[0067] Classification code matching: Tariff matching refers to the process of accurately classifying items (such as commodities) into corresponding HS codes based on their characteristics, uses, and other information.

[0068] Local Value-Added Rate (LVA): It is an important indicator that measures the proportion of economic value created by the country / region during the production of an item (such as a commodity). It is often used to determine the rules of origin of international trade agreements (such as free trade agreements). Among them, local value-added rate = (local input value / final value of goods) × 100%. Local input value: includes the costs incurred in the country / region for raw materials, labor, technology, energy, etc. Final value of goods: the free on board price (FOB) or total production cost of the goods when they are exported. Different trade agreements may have different definitions of local value-added rate. For example: North American Free Trade Agreement (USMCA): requires that the local value-added rate of core components of automobiles be ≥75%; China-ASEAN Free Trade Agreement: certain goods must meet the "regional value content ≥40%."

[0069] Customs Articles Terminology Database (specifically, Customs Commodity Terminology Database): refers to a standardized database of descriptions and classifications of articles (such as commodities) constructed and maintained by customs or relevant authoritative agencies. Its core function is to unify article naming rules, technical parameters and classification logic, and provide authoritative terminology references for import and export declarations, HS code matching, compliance reviews and other links.

[0070] Modality is a form of representation of an object. Multimodality typically encompasses two or more modalities, describing an object from multiple perspectives. For example, multimodal information for a product might include text descriptions, images, audio and video information, and more. Using multimodal information can make product presentations more comprehensive and holistic.

[0071] Supply chain information of an item: refers to all data and information involved in the process from raw material procurement to the delivery of the final product to the consumer, including but not limited to supplier information, production and processing sites, transportation routes, warehousing conditions, distribution channels, and other data records and information for each link.

[0072] The following method embodiments provided herein are applied to electronic devices with corresponding logic operation functions, which can be either clients or servers. The clients can be any terminal device, such as mobile phones, tablet computers, or smart wearable devices. The servers can be a single server, a service cluster, a virtual server deployed on a server, or a cloud-based server.

[0073] Figure 1 FIG1 shows a flow chart of a data processing method provided by an exemplary embodiment of the present application, which is applied to the server. Figure 1As shown, the data processing method includes the following steps:

[0074] 101. Determine a first item classification identifier corresponding to the target item based on multimodal information of the target item;

[0075] 102. Based on the first item classification identifier, obtain the tariff rules applicable to the transportation of the target item from the knowledge graph;

[0076] 103. Determine the tariff value of the target item using the applicable tariff rules;

[0077] 104. Generate customs declaration documents for the target item based on the tariff value.

[0078] The knowledge graph is dynamically constructed based on the real-time import and export control information of each region. The knowledge graph also contains the relationship between the classification identification of items and tariff rules. Here, "region" refers to a region or a set of countries that implement unified import and export control information. For example, the European Union (EU) is an economic union composed of 27 European countries. In addition, import and export control information is the import and export policy documents (such as PDF files, announcements, etc.) issued by the corresponding regional agencies such as customs agencies, which include tariff policies. In cross-border trade scenarios, import and export control information is also commonly referred to as import and export commodity control information.

[0079] In this embodiment, the target items correspond to different categories in different cross-border item transportation scenarios. For example, in a cross-border trade scenario, the target items can be various commodities traded, such as machinery and equipment, electronic products, textiles, beauty products, etc., which are not limited here. In a cross-border travel scenario, the target items can be personal items carried or hauled by inbound and outbound users, such as clothing, travel goods, personal electronic products, etc. Generally, in a cross-border travel scenario, if the quantity of items such as cigarettes, jewelry, cameras, etc. carried by an individual exceeds the prescribed duty-free carrying amount, the excess amount must be declared and taxed according to the tax rate.

[0080] As mentioned above, target items refer to any inbound or outbound items that require customs declaration and tax payment. They can be for commercial or personal use, and are not specifically limited here. This application mainly introduces this application solution in the context of cross-border trade, with the target items being traded commodities. Among them, cross-border trade scenarios include but are not limited to cross-border e-commerce.

[0081] Taking the cross-border e-commerce scenario as an example, cross-border e-commerce platforms currently usually provide customs declaration functions or integrate with customs declaration services to simplify customs procedures in cross-border transactions. Assume that a cross-border e-commerce platform has an integrated "I want to declare customs" entrance. When merchants need to declare customs for goods orders placed by buyers, they can enter the online customs declaration page through the "I want to declare customs" entrance. Figure 2a The online customs declaration page displayed on the client 100 is shown; further, the merchant can click on the "Import Declaration" control on the online customs declaration page, and a list of commodity orders will be displayed. The merchant can select the commodity order that needs to be declared from the commodity order list, and click the "OK" control after the selection is completed. In response to the "OK" operation triggered by the merchant, the server 200 will start to generate relevant customs declaration documents based on the commodity order information selected by the merchant. For example, the server 200 can obtain the multimodal information of the target commodity from the commodity order information selected by the merchant (including text description information of the target commodity (such as "***Su Yan Cream Whitening and Brightening...", commodity image), and then determine the HS code of the target commodity based on the multimodal information, and then combine the knowledge graph to determine the tariff rules applicable to the target commodity based on the HS code, destination and other information of the target commodity, and then calculate the tariff value of the target commodity based on the tax rate contained in the tariff rules, which is used to participate in the generation of customs declaration documents for the target commodity.

[0082] For example, in other cross-border goods transportation scenarios (such as cross-border trade, etc.), users can also realize the customs declaration of target goods through the public customs declaration service platform. Figure 2b As shown, still taking cross-border trade as an example, after the user enters the online customs declaration page of the public customs declaration service platform through the client 100, he can enter the text description information of the target product to be traded across borders and upload the product image of the target product in the product information input entry provided on the online customs declaration page, and then click the "Submit" control. In response to this submission operation, the server 200 will start to execute the corresponding customs declaration document generation process for the target product. For specific implementation, please refer to the above-mentioned combined Figure 2a Description of relevant content.

[0083] Based on the above example content, the multimodal information of the target item in the above 101 includes but is not limited to: text description information and image information (such as main image and / or sub-image) of the target item. Taking the target item as a commodity as an example: the text description information may include the commodity title, commodity attributes (such as material, color, size, etc.), a detailed description of the commodity (such as name, brand, specifications, purpose, applicable population, freight, place of origin, etc.); and the image information may be uploaded by the user or obtained by converting the video information uploaded by the user into an image format, etc. Specifically, the image information may include: display pictures of the commodity (such as a main picture with a white background, a scene picture, a detail picture, etc.). Among them, this multimodal information can be, for example, input and / or uploaded by the user, or it can also be obtained based on the order information corresponding to the target item, etc. The method of obtaining the multimodal information of the target item is not limited here.

[0084] Furthermore, the third preset model may be used to process the multimodal information of the target object to obtain the text features required for the classification of the target object.

[0085] For example, after the text description information of the target item is input into the third preset model, the processing process performed by the third preset model may include the following steps:

[0086] Step 1: Standardize the text description information of the target item to obtain standardized text description information.

[0087] Here, the reasons why the text description information of the target item needs to be standardized include: the text description information of the target item is usually written and input by the user, and often has but is not limited to the following problems: the presence of colloquial expressions, abbreviations, spelling errors or non-standard terminology. In order to ensure the accuracy of subsequent HS code matching, the text description information entered by the user needs to be converted into a standardized expression that complies with industry standards.

[0088] For example, assuming that the text description information of the target item input by the user is "Apple mobile phone 13th generation uses lithium battery technology", the normalization processing of the text description information may include but is not limited to the following:

[0089] 1) Terminology replacement: For example, you can load the customs article terminology database (specifically, the customs commodity terminology database) and, based on the article terms in the customs article terminology database, replace the non-standard term "Apple phone" in the text description information with "Apple", "13th generation" with "iPhone 13", and "lithium battery" with "lithium-ion battery";

[0090] 2) Supplementing missing key attributes of items: For example, the text description information entered by the user above is missing key attributes such as the network type and product type of the Apple mobile phone. In order to infer that the network type supported by the Apple mobile phone is 5G and the product type is a smartphone based on the model information of the Apple mobile phone in the text description information, "5G smartphone" can be supplemented in the text description information.

[0091] After the above standardization process, the resulting standardized text description is "Apple iPhone 13 is a 5G smartphone that uses lithium-ion battery technology."

[0092] Step 2: Extract keywords from the standardized text description to obtain the first key text for determining the classification of the target item.

[0093] Continuing with the above example, for example, the following structured first key text {"Brand: Apple", "Model: iPhone 13", network type "5G", "Product type: smartphone" can be extracted from the above standardized text description "Apple iPhone 13 is a 5G smartphone that uses lithium-ion battery technology" based on regular expressions.

[0094] Based on the above-mentioned first key text, the text features of the target item can be determined. For example, the fourth model can be used to vectorize the first key text to obtain the corresponding first vector; and this first vector and the above-mentioned first key text in text form are determined as the text features of the target object. Among them, in this application, the reasons for determining the first key text in text form and the first vector representation in vector form as the text features of the target item include: it is convenient to use a hybrid matching algorithm (including text retrieval and vector retrieval) to determine the first item classification identifier corresponding to the target item based on the text features of the target object. The hybrid matching algorithm will be described in detail below and will not be described in detail here.

[0095] Among them, the above-mentioned third preset model is a multimodal model that can process multimodal information such as text, images, and videos. Specifically, the third preset model can be, for example, a GPT-4V model. In order to ensure that the third preset model has a high accuracy of image recognition and analysis, this application uses an image dataset (containing hundreds of thousands of images) specifically for customs items (such as customs commodities) as a training sample set, and fine-tunes a model based on a Transformer architecture (such as a GPT-4V model) to obtain a third preset model. Here, the fine-tuning training method includes but is not limited to: LoRA fine-tuning method. In addition, the third preset model can focus on the key parts or characteristics of the objects in the picture through the attention mechanism adopted therein, such as: for mobile phones, pay special attention to the type of charging interface at the bottom of the mobile phone, the network standard logo supported by the device, the shell material, etc.; for electrical plugs, pay special attention to the specific standard of the electrical plug (such as shape, size, rated voltage, etc.); for textiles, pay special attention to texture. Furthermore, the fourth preset model is a model for text vectorization, such as a Sentence-BER model, and the output dimension of the Sentence-BER model can be a 768-dimensional vector.

[0096] In addition, in order to ensure that more comprehensive text features of the target object can be obtained, a third preset model can be used to identify and analyze the image information of the target object to generate a structured text description to supplement the text features of the target object determined above.

[0097] For example, continuing with the above example, the target item is the 13th generation of Apple mobile phone. After obtaining one or more images uploaded by the user for the target item, these images are identified and analyzed by using the third preset model, and structured text descriptions are generated, such as "color: starlight color", "material glass back panel and aluminum metal frame". These generated structured text descriptions are called second key text information. The attributes of the target item reflected by this second key text are not covered in the first key text extracted based on the text description information of the target item. For this reason, the second key text can be supplemented to the text features of the target item. In addition, the second key text can be vectorized and the corresponding second vector representation can be supplemented to the text features of the target item. For the reasons for determining this second key text and its corresponding second vector representation as the text features of the target item, please refer to the reasons for determining the first key text and its corresponding first vector representation as the text features of the target item.

[0098] As mentioned above, the target item's text features include: a first key text and its corresponding first vector representation, and a second key text and its corresponding second vector representation. HS code matching can then be performed using the target item's text features to determine the target item's corresponding HS code. This determined HS code serves as the target item's first item classification identifier.

[0099] Based on this, in one possible implementation, the above-mentioned step 101 of “determining a first item classification identifier corresponding to the target item based on the multimodal information of the target item” may include:

[0100] 1011. Standardize the text description information of the target item to obtain a standardized text description; wherein the standardization includes at least one of the following: replacing non-standard descriptive words in the text description information according to the customs item terminology database; and supplementing missing key attributes of the item.

[0101] 1012. Extracting a first key text for determining a classification of a target item from the standardized text description information;

[0102] 1013. Perform recognition analysis on the image information of the target object to obtain a second key text for determining the classification of the target object;

[0103] 1014. Determine text features of the target item based on the first key text and the second key text;

[0104] 1015. Determine the first item classification identifier based on the text feature.

[0105] For the specific implementation of steps 1011 to 1014, please refer to the relevant content in the above other embodiments. The text features of the target item include: a first key text and a first vector representation obtained by vectorizing the first key text, and a second key text and a second vector representation obtained by vectorizing the second key text.

[0106] In step 1015, in addition to combining the first and second vector representations in vector form, the first and second key texts in text form may also be used to determine the first item classification identifier corresponding to the target item. Specifically, a hybrid matching approach may be employed to perform HS code matching based on the text features of the target item to obtain a matching HS code for the target item. This matched HS code is then determined as the first item classification identifier corresponding to the target item. Hybrid matching approaches include, but are not limited to, text search and vector search.

[0107] In specific implementations, a text search, such as fuzzy matching using Elasticsearch (an open-source distributed search engine), is first performed to screen a set of candidate HS codes from a vast HS code database based on the first and second key texts in text form. A vector search is then performed on the candidate HS codes, specifically using FAISS index acceleration, to perform a similarity analysis between the vector representations corresponding to each candidate HS code in the candidate HS code set and the first and second vector representations in vector form. Based on the similarity analysis results, a relatively matching HS code is identified as the first item classification identifier for the target item.

[0108] The aforementioned FAISS index acceleration primarily refers to utilizing the various index structures and algorithms provided by FAISS to speed up vector searches on large-scale datasets. This allows, for example, returning the top three HS codes with the highest similarity scores within 10 milliseconds. Furthermore, during the vector retrieval phase, the similarity score calculation can be performed using the following formula: final_score = α*text_similarity + β*image_similarity. final_score represents the similarity score. Furthermore, in the vector retrieval scenario, text_similarity represents the similarity between the first vector representation and the vector representation corresponding to the corresponding HS code (obtained by vectorizing the encoding description information corresponding to the HS code); image_similarity represents the similarity between the second vector representation and the vector representation corresponding to the corresponding HS code (obtained by vectorizing the encoding description information corresponding to the HS code). Furthermore, the aforementioned α and β are weight coefficients that can be adjusted dynamically, for example, α = 0.7 and β = 0.3.

[0109] It should be noted that during the text retrieval phase, the candidate HS code set is primarily screened through text matching. During the text matching process, if text similarity analysis is used to screen the candidate HS code set, the similarity score corresponding to each HS code in the HS code database can also be calculated based on the similarity score calculation formula given above, thereby recalling HS codes with similarity scores greater than or equal to a set threshold as candidate HS codes. In the text retrieval phase, in the similarity score calculation formula given above, text_similarity represents the similarity between the first key element information and the coding description information (in text form) corresponding to the corresponding HS code; and mage_similarity represents the similarity between the second key element information and the coding description information (in text form) corresponding to the corresponding HS code.

[0110] For example, assuming that a hybrid matching method can be used to perform HS code matching based on the text features of the target item, three relatively relevant HS codes and their similarity scores are obtained as follows:

[0111] 8517.1200 (for wireless telephone equipment): score of 0.95

[0112] 8517.6200 (for other digital units used to transmit or receive sound, images, or other data): score 0.85

[0113] 8471.4140 (for portable digital devices): score 0.80

[0114] Based on the above, we can determine that the HS code that best matches the target item is 8517.1200. In other words, the HS code 8517.1200 is the first item classification identifier corresponding to the target item.

[0115] In summary, this application uses the multimodal information of the target items to match HS codes, which can ensure more accurate classification of the target items, thereby facilitating the subsequent improvement of the accuracy of the determination of tariff rules corresponding to the target items.

[0116] After determining the first item classification identifier of the target item, the tariff rules applicable to the transportation of the target item can be obtained in combination with the dynamically constructed knowledge graph.

[0117] In specific implementations, the knowledge graph is constructed by real-time acquisition and parsing of import and export control information published by various regions around the world. Import and export control information, also known as unstructured policy information or unstructured policy text, includes tariff policies and various other import and export management measures. Acquiring import and export control information from various regions can be accomplished through, but is not limited to, automated acquisition from multiple data sources using appropriate data acquisition technologies. It should be noted that the data acquired using these data acquisition technologies is public and complies with relevant laws and regulations. Examples of these data sources include official customs and agency websites, the official World Trade Organization (WTO) website, websites of free trade zones or economic cooperation organizations, and the website of the United Nations Conference on Trade and Development (UNCTAD). Furthermore, acquired import and export control information can include, for example, announcements (such as those published on the WTO official website) and the original text of trade agreements (typically provided in PDF or HTML format or via email). The aforementioned data acquisition technologies include, but are not limited to, distributed data acquisition technologies. For example, a policy acquisition module can be developed based on the Scrapy framework to automatically acquire import and export control information from various data sources. Scrapy is an open source Python web data acquisition framework. Furthermore, developing a policy acquisition module based on the Scrapy framework also supports CAPTCHA cracking and dynamic rendering. CAPTCHA cracking can be achieved by integrating third-party services (such as 2Captcha or DeathByCaptcha) to resolve CAPTCHAs on web pages, or by training machine learning models to recognize simple CAPTCHAs. Dynamic rendering can be achieved by combining Scrapy with Selenium or Puppeteer to simulate browser behavior and retrieve the complete page content for pages that rely on JavaScript to dynamically load content.

[0118] Furthermore, given that the formats of the imported and exported goods control information obtained from various regions may vary and the languages ​​used may also differ, preprocessing is typically performed before parsing the obtained imported and exported goods control information, allowing subsequent parsing and processing of the preprocessed imported and exported goods control information using a corresponding second preset model. Preprocessing includes, but is not limited to, file format conversion and multilingual encoding. File format conversion involves selecting appropriate tools for content extraction and conversion based on different file formats (PDF, HTML, etc.). For example, if a PDF file needs to be converted, Apache PDFBox (an open source Java library) can be used to extract text from the PDF file, converting it to plain text for subsequent analysis and processing. Alternatively, if an HTML file needs to be converted, BeautifulSoup or LXML libraries in Python can be used. Alternatively, if the content of an email or body needs to be converted, the Python email library can be used to read .eml files and further process the PDF or HTML attachments contained therein using the aforementioned methods. Furthermore, the aforementioned multilingual encoding can be implemented, for example, using UTF-8 standardization. UTF-8 standardization means that when processing information from different countries and regions, all text data is stored in UTF-8 encoding. After the preprocessing is completed, the second preset model can be used to process the original import and export control information after preprocessing. The processing flow includes: inputting the original import and export control information after preprocessing into the second preset model. The second preset model will extract key information (i.e. key entities) from this import and export control information and output the extraction results. In addition, the extraction results output by the second preset model will also be verified using a rule engine. After the verification is passed, a knowledge graph will be constructed based on the corresponding extraction results of each region. Among them, during extraction, the second preset model can extract specific key information from the given import and export control information based on the prompt template.

[0119] Exemplarily, for the second preset model, a prompt template (prompt_template) is designed in advance. The prompt template (prompt_template) may be, for example, "You are a customs policy analysis expert. Please extract the HS code (6-10 digits), tax rate changes (original tax rate, new tax rate, effective date), and rules of origin (such as "value added ≥ 40%) from the following text, and output it in JSON format, with missing fields marked as null." This prompt template is used to guide the first preset model to accurately extract the corresponding key information from the given import and export control information (policy_text) and output the extraction results in JSON format. Among them, the key information that needs to be extracted in the prompt template includes: HS code, tax rate change details (such as original tax rate, new tax rate, effective date, etc.) and rules of origin, and it is required that if some key information is missing due to not being extracted, the field value of the corresponding field will be marked with null when output. Null is a null pointer, which is usually used to indicate that a variable field does not point to any valid object or value.

[0120] Furthermore, the error correction mechanism can be used to perform logical verification on the key information output by the second preset model (such as verifying whether the output HS code format meets the requirements) to ensure the accuracy and standardization of the output results. For example, assuming that the field corresponding to the HS code is hs_code, when performing logical verification on HS editing, you can first check whether the field value of the hs_code field output by the second preset model exists and is not empty; if it exists and is not empty, then trigger verification. Specifically, use the regular expression \d{6,10}" to check whether the field value of the hs_code field is a pure number of 6 to 10 digits. If not, a custom exception such as HSFormatError (representing a suspected HS code format error) can be thrown, and an error message can be clearly prompted such as "Invalid HS code structure" to facilitate manual debugging and error correction. After the logical verification is passed, the knowledge graph can be constructed based on the key information output by the second preset model.

[0121] In the above, the second preset model is a language model, and specifically, for example, can be a GPT-4 model (a language model developed based on OpenAI). The constructed knowledge graph structure includes multiple nodes and node relationships, and the nodes in the multiple nodes have two categories, classified as HS code (also referred to as article classification identifier in the present application, such as can be represented by HS:Code) and tariff rule (such as can be represented by Policy:Rule), and the node relationship reflects the association relationship between the HS code and the tariff rule. The tariff rule includes: the corresponding region (such as the corresponding country, which can be represented by the country field), the tax rate (rate) and the article constraint condition (condition) that needs to be met to enjoy the tax rate. And, the corresponding graph database (such as Neo4j) can be used to construct and manage the knowledge graph.

[0122] Exemplarily, in Neo4j, the storage structure of a certain node relationship contained in the knowledge graph is as follows:

[0123] (HS:Code{value:"8541.40"})-[:HAS_POLICY]->(Policy:Rule{country:"EU",rate:18.0,condition:"origin=CN,LVA≥40%"})

[0124] In this example, the HS:Code node represents a specific HS code (i.e., an article classification identifier), and its value is 8541.40. The Policy:Rule node represents the tariff rule (also referred to as policy rule) related to the HS code, including the tax rate (rate, such as the value of 18.0), the applicable country or region (country, such as the value of EU (European Union)), and the article constraint condition (condition, such as "origin=CN" (indicating the place of origin is China)) that needs to be met. The HAS_POLICY relationship: used to connect the HS code and its applicable tariff rule, indicating that this tariff rule applies to the specified HS code. One HS:Code node may generally have an association relationship with multiple Policy:Rule nodes.

[0125] In addition, the update strategy adopted by this application for the knowledge graph is: incremental update and version control. Incremental update means that when the tariff policy of a new region is monitored or the tariff policy of an existing region changes, the knowledge graph is updated by adding a corresponding tariff rule node to the knowledge graph, thereby updating the difference part. Version control adopts the method of retaining historical strategy snapshots to support tax rate retrospective analysis. Specifically, version control can be understood as: when the tariff policy of an existing region changes, instead of directly modifying the corresponding tariff rule node in the knowledge graph, a corresponding new tariff rule node is added to the knowledge graph, and the previous corresponding tariff rule node is retained as the historical version of the tariff rule, which is convenient for supporting tax rate retrospective analysis.

[0126] For the convenience of understanding the update strategy adopted for the knowledge graph, the following will continue the foregoing example, record the current HS code "8541.40" in the knowledge graph as the first HS code, and record the current applicable EU (European Union) tariff rule of the first HS code (i.e. Policy:Rule{country: "EU", rate: 18.0, condition: "origin = CN"}) as the first version of the EU (European Union) tariff rule. Subsequently, through periodic acquisition, a new import and export commodity regulation information is acquired in the EU (European Union) region, and through analysis of the new import and export commodity regulation information, it is determined that the EU (European Union) tariff rule of the first HS code has changed, such as the change in the tariff rule including: the tax rate is raised from the original 18.0 to 20.0, and the effective date of the new tax rate is: 2025-06-01; in addition, the corresponding tax rate enjoying constraint condition of the goods has also changed, such as the local value-added rate of the original place of origin rule being lowered from greater than or equal to 40% to greater than or equal to 35%. Then, for the changed tariff rule, a second version of the EU (European Union) tariff rule is added in the knowledge graph, which includes: country: "EU", rate: 20.0, condition: "origin = CN, LVA > 35%" in addition to these contents, it can also include effective_start_date: "2025-06-01" (which indicates the effective date of the new tax rate), version: "2" (which indicates the version number of the second tariff rule), etc. An association will be established between the first HS code and the second version of the EU (European Union) tariff rule, and the association between the first HS code and the first version of the EU (European Union) tariff rule will also be retained in the knowledge graph, and some contents will be added to the first version of the tariff rule adaptively, such as version: "1" (which indicates the version number of the first version of the tariff rule) and effective_end_date: "2025-05-31" (which indicates the expiration date of the first version of the tariff rule) can be added to the first version of the tariff rule. With the above update of the knowledge graph, in the updated knowledge graph, the first HS code is associated with two versions of the EU (European Union) tariff rule, and the corresponding node relationship storage structure includes:

[0127] (HS:Code{value:"8541.40"})-[:HAS_POLICY]->(Policy:Rule{country:"EU",rate:18.0,condition:"origin=CN, LVA≥40%"},effective_end_date:"2025-05-31",version:"1");

[0128] (HS:Code{value:"8541.40"})-[:HAS_POLICY]->(Policy:Rule{country:"EU",rate:20.0,condition:"origin=CN, LVA≥35%"},effective_start_date:"2025-06-01",version:"2").

[0129] This application uses intelligent means to obtain import and export control information from various regions in real time and to generate knowledge graphs. This allows the use of knowledge graphs to intuitively represent the complex relationship between the classification identifiers (HS codes) of various items and their associated tariff rules, making it easier to query applicable tariff rules for the corresponding items and simplifying the complexity of customs declarations. In addition, when a region's tariff rules change, the knowledge graph can be quickly updated and the updated tariff rules reflected in the knowledge graph. This can help users understand tariff rule adjustments in various regions in a timely manner through the knowledge graph, allowing them to make more accurate business decisions.

[0130] Based on the above description of the knowledge graph, in step 102, when determining the tariff rules applicable to the transportation of the target item, in addition to combining the first item classification identifier corresponding to the target item, other information is often required, such as the destination, local value-added rate, and place of origin of the target item. This other information is the relevant attribute information required for customs declaration of the target item. Based on this, in one possible implementation, step 102, "obtaining the tariff rules applicable to the transportation of the target item from the knowledge graph based on the first item classification identifier," may include:

[0131] 1021. Determine the item attribute information required for customs declaration of the target item.

[0132] In this step, item attribute information refers to relevant information about the target item required to complete the import / export declaration process. This information can assist in assessing tariffs and taxes, and verifying compliance with import / export regulations. Specifically, item attribute information includes, but is not limited to, the target item's name, destination, specifications, origin, local value-added rate, primary material source, assembly location, intended use, and export location. This item attribute information can be automatically acquired by parsing the target item's text description or image information, querying a database, or integrating the target item's supply chain records. Alternatively, it can be user-entered, for example, by simultaneously entering the target item's multimodal information and the item attribute information. Some attributes included in this item attribute information are elements of the target item's customs declaration, such as origin, destination, item name, destination, specifications, and model. Other attributes serve as key data supporting the identification of declaration elements, such as the local value-added rate, primary material source, and assembly location. This information is crucial for identifying the target item's origin.

[0133] 1022. According to the item attribute information and the first item classification identifier of the target item, obtain from the knowledge graph a first version of tariff rules that is related to the first item classification identifier and is in effect.

[0134] Here, the first version of the tariff rule includes a first region, a first tax rate, and an item constraint. If the target item's customs declaration type is an import declaration, the first region includes the target item's destination, and accordingly, the first tax rate is the import tax rate. Furthermore, if the target item's customs declaration type is an export declaration, the first region includes the target item's export destination, and accordingly, the first tax rate is the export tax rate.

[0135] 1023. Determine the first version of the tariff rules as the tariff rules applicable to the transportation of the target item.

[0136] To facilitate understanding of the above 1021 to 1023, the following example takes the import declaration of target items as an example to illustrate the process of determining the tariff rules applicable to the transportation of target items.

[0137] Example scenario: The target item needs to be transported to Thailand, and the item information of the target item includes the first item classification identifier (HS code, 8541.40) and the corresponding item attribute information (item attributes required for customs declaration, such as origin of China, local value added rate of 45%, and destination of Thailand). Based on the knowledge graph and the item information of the target item, the corresponding decision engine can be used to perform reasoning on the tariff rules applicable to the target item. The decision engine includes a rule reasoning module, which may include: a rule engine (such as Drools) and a first preset model (such as a language model). The rule engine is used to process the display rule logic. Specifically, it determines whether the target item complies with a specific tariff preferential policy based on the corresponding rule chain, such as whether it complies with the RCEP tariff rules for 8541.40. The rule chain can be generated based on the item constraints contained in the corresponding tariff rules. For example, since the destination "Thailand" belongs to the countries on the RCEP effective list, it uses the tariff rules specified in the RCEP accordingly. Based on this, based on the first item classification identifier "8541.40" of the target item, the RCEP tariff rules associated with "8541.40" are queried in the knowledge graph, and then a rule chain is generated based on the item constraints contained in the queried RCEP tariff rules. For example, item constraints include checking that both the item's origin and destination must be RCEP member countries and that the local value-added rate must be greater than or equal to 40%. These constraints are relatively simple. Based on these relatively simple constraints, the rule engine can generate a rule chain: If the item's origin ∈ RCEP member country, the local value-added rate ≥ 40%, and the destination ∈ a country on the RCEP effective list (i.e., the destination ∈ RCEP member country), then the item is determined to be subject to the RCEP tariff rules (specifically, the preferential tariff rate within the RCEP tariff rules). Based on this rule chain, combined with the item attribute information of the target item, rule matching can be performed to determine that the target item's origin "China" is an RCEP member country, the local value-added rate "45%" is greater than 40%, and the destination is also on the RCEP effective list. Therefore, it can be determined that the target item is subject to the RCEP tariff rules for 8541.40, specifically, the preferential tariff rate within the RCEP for 8541.40. The first preset model mentioned above is used for implicit reasoning to handle some more complex conditions.For example, the item attribute information of the target item also includes some declaration elements that affect the determination of the origin of the item, such as the main raw materials coming from China and the assembly place being Vietnam. The previously given origin of "China" is inconsistent with the main raw material place and assembly place given here, and they belong to different countries. This may lead to a substantial change in the determination of the origin of the item. For this reason, the element of "China" as the origin included in the item attribute information of the target item may be inaccurate. It is necessary to use the first preset model to re-analyze the specific country of "origin qualification" of the target item based on the main raw material source, assembly place and other elements of the target item. Among them, the analysis process is based on the "substantial change" in the RCEP rules of origin. Based on this, the first pre-set model's inputs might include: the target item's primary raw materials originate in China, its final assembly location is Vietnam, and the target agreement is RCEP. The model then determines whether the target item meets the "substantial change" requirement under the RCEP rules of origin. Based on this input analysis, the third pre-set model's output might be that the target item is considered Vietnamese-origin because, while the primary raw materials originate in China, the final assembly takes place in Vietnam. This assembly process results in a change in the product's HS code (tariff classification), meeting the RCEP rules of origin's substantial change requirement. Therefore, the target item can be considered Vietnamese-origin. Based on this output, the target item's actual origin is determined to be Vietnam. To this end, "Origin is Vietnam" can be used as one of the input parameters of the rule engine to match the "Origin of the item ∈ RCEP member country" rule in the rule chain. Since Vietnam is also an RCEP member country, the target item can be determined to meet the "Origin of the item ∈ RCEP member country" rule.

[0138] Based on the above example, the target item's item attribute information includes, in addition to the destination, a first attribute and a second attribute, and the second attribute affects the determination of the first attribute. The first attribute is one of the attributes required to determine whether the target item satisfies the item constraint. Furthermore, between steps 1022 and 1023, the provided method may further include the following steps:

[0139] S11. If it is determined based on the second attribute that the first attribute needs to be re-identified, triggering a first preset model to perform a re-identification analysis on the first attribute using the second attribute to obtain an analysis result;

[0140] S12. Update the first attribute based on the analysis result;

[0141] S13. Using the updated first attribute as one of the input parameters of a rule engine, triggering the rule engine to determine whether the target item satisfies the item constraint based on the input parameter, and obtaining a determination result;

[0142] S14. If the judgment result is satisfied, the first version of the tariff rule is determined to be the tariff rule applicable to the transportation of the target item.

[0143] In S11-12 above, the first attribute is, for example, the country of origin. There can be one or more (two or more) second attributes, such as the country of origin of the main materials or the country of assembly. For example, if either the country of origin of the main materials or the country of assembly is a different country from the country of origin, the requalification criteria may be met, necessitating a requalification analysis of the country of origin. During the requalification analysis, in addition to the second attributes of the target item, such as the country of origin of the main materials or the country of assembly, other information may also be used, such as the "substantial change rule" under the corresponding agreement, such as the "substantial change" in the RCEP rules of origin in the example above. The requalification analysis may result in a "substantial change" in the first attribute. In this case, the first attribute needs to be updated based on the analysis results, such as changing the country of origin from "China" to "Vietnam" in the example above. Alternatively, the requalification analysis may result in no "substantial change" in the first attribute. In this case, no update or revision of the first attribute is required, and the first attribute remains unchanged.

[0144] In the above S13-S14, the rule engine will first generate a rule chain based on the item constraints, and then match and analyze the input parameters based on the rule chain to determine whether the target item meets the item constraints. If so, the first version of the tariff rules will be determined as the tariff rules applicable to the transportation of the target item. Among them, the input parameters of the rule engine are determined based on the factors required to determine whether the item constraints are met. Specifically, in addition to the updated first attribute (such as origin), the input parameters may also include other attributes of the target item such as destination, local value-added rate, etc.

[0145] Furthermore, the above-mentioned 103 “determining the tariff value of the target article based on the applicable tariff rules” may include:

[0146] 1031. Calculate the tariff value of the target item based on the first tax rate included in the first version of the tariff rules.

[0147] If the target item is declared for import, the calculated tariff value is the import tariff payable to the destination customs. If the target item is declared for export, the calculated tariff value is the export tariff payable to the local customs (the customs at the place of export).

[0148] For the first version of the tariff rules, please refer to the first version of the European Union (EU) tariff rules given for the HS code "8541.40" (recorded as the first HS code) in other embodiments mentioned above.

[0149] After calculating the tariff value of the target item, the customs declaration process can be entered to automatically generate customs declaration documents that meet the requirements of the corresponding region. This automatic generation of customs declaration documents can be achieved by automatically filling in the corresponding customs declaration template. The customs declaration module is a standardized document framework used to declare item information to customs in import and export trade.

[0150] Based on this, in one possible technical solution, the above-mentioned step 104 of “generating customs declaration documents for the target article based on the tariff value” may include:

[0151] 1041. Obtain a customs declaration template suitable for the target article;

[0152] 1042. Fill the tariff value and the first item classification identifier of the target item into the customs declaration template to obtain the customs declaration document of the target item.

[0153] In step 1041 above, the customs declaration template can be loaded and retrieved from the corresponding template library based on the export location and / or destination of the target item. For example, if the customs declaration scenario for the target item is import declaration, the customs declaration template for customs transactions between the export location and the destination will be loaded from the template library based on the export location and destination of the target item. If the customs declaration scenario for the target item is export declaration, the customs declaration template for export customs transactions at the export location will be loaded from the template library based on the export location of the target item.

[0154] In step 1042, in addition to filling in the customs declaration template based on the tariff value and first article classification identifier of the target item, other information may also be required, such as the net weight of the target item. The net weight of the target item is also a customs declaration element and can be entered by the user. If the user does not enter the net weight of the target item, the server can automatically obtain it. Specifically, but not limited to, automatically obtaining this missing net weight data can be done by calling the logistics API or other methods.

[0155] For example, for a target item, when automatically filling in a customs declaration template, if it is found that the net weight of the target item is missing, the net weight can be obtained by querying the user's warehouse management system or accessing the system of a transportation company applicable to the transportation of the target item.

[0156] Once all the information items in the customs declaration template are filled in, a customs declaration document will be automatically generated. The format of the customs declaration document can be PDF, and this customs declaration document contains the necessary information required for customs declaration, such as the first classification identification (HS code), net weight, tariff value, etc. of the target item.

[0157] The following is an example of a customs declaration template for trade between China (CN) and the European Union (EU). The data structure of the customs declaration template is as follows:

[0158] {

[0159] "template_id":"CN_EU_CUSTOMS_001",

[0160] "required_fields":["hs_code","net_weight"],

[0161] "import duties":"tariff_value",

[0162] "format":{"type":"PDF","schema":"UDEF-2.1"}

[0163] }

[0164] The template_id field represents the unique identifier of this customs declaration template. In this example, its value is CN_EU_CUSTOMS_001, indicating that this is the standardized customs declaration template applicable to customs declarations between China and the EU. "required_fields" is an array that specifies the fields that must be filled in when using this customs declaration template. "hs_code" represents the HS code (i.e., the corresponding item classification identifier) ​​of the item, and "net_weight" represents the net weight of the item. "format" defines the format requirements of this customs declaration template. "import duties" represents import duties. "type":"PDF" indicates that the generated file type is PDF format. "schema":"UDEF-2.1" indicates that the generated PDF file must comply with the DEF-2.1 standard.

[0165] It should be noted here that the customs declaration templates corresponding to different regions are different, and the field content in the customs declaration template is not limited to the content contained in the sample customs declaration template given above.

[0166] Through the above, the generated customs declaration document is specifically an import declaration document or an export declaration document, which depends on the customs declaration scenario corresponding to the target item. After the customs declaration document is generated, the server can first send it to the client for display and review. After the user triggers the confirmation operation after review, the server can automatically submit it to the corresponding customs system in response to this confirmation operation. For example, if it is an import declaration, it is submitted to the corresponding destination customs system; if it is an export declaration, it is submitted to the corresponding export customs system. If the user discovers errors during review, they can proactively modify the errors in the customs declaration document and trigger the confirmation operation again after the modification is completed. In response to this confirmation operation, the server can submit the user's modified customs declaration document to the corresponding customs system. Of course, with user authorization, the server can also directly send the generated customs declaration document to the corresponding customs system without user review.

[0167] After the customs declaration documents reach the relevant customs system, they will undergo automatic and / or manual review. Upon completion, the server will return the review results to the user. The server can also automatically update the knowledge graph based on the review results. For example, if the review result indicates a miscalculation of the tariff value, the cause of the miscalculation may be the invalidation of the tariff rules previously used. Based on this review result, the server can also mark the corresponding tariff rules in the knowledge graph used as invalid.

[0168] In addition, in addition to submitting the generated customs declaration documents to the corresponding customs system to complete customs clearance, the customs declaration documents can also be submitted to the blockchain for evidence storage.

[0169] This application can reduce the possibility of errors in tariff calculation and manual filling of customs declaration documents through automatic calculation of tariffs, automatic generation and submission of customs declaration documents, improve the compliance of customs declaration documents, and enhance the automation of customs clearance, thereby speeding up the entire customs clearance process.

[0170] Furthermore, this application also supports counterfactual analysis for target items, helping users make better trade decisions in trade scenarios. Counterfactual analysis simulates the impact of changes in the item's supply chain (such as changing suppliers, adjusting production and assembly locations, etc.) on taxes and fees.

[0171] For example, a user is considering changing the battery supplier of a target product from China to Vietnam and wants to understand the impact of this change on the tariffs on the target product exported to Thailand. The following is the counterfactual analysis process for this change:

[0172] 1) The user enters hypothetical conditions for the target item

[0173] The current supply chain configuration for the target item is as follows:

[0174] Battery supplier: China (a major source of materials for the target item)

[0175] Other major parts suppliers: remain unchanged (all RCEP member countries)

[0176] The target finished product is exported to Australia

[0177] Current applicable tax rate: Based on the existing supply chain, finished products are subject to the RCEP preferential tax rate of 5%.

[0178] Hypothetical change: Changing the battery supplier from China to Vietnam (also a RCEP member)

[0179] 2) Traverse the knowledge graph to find association rules

[0180] First, the knowledge graph is used to search for tariff rules that are compatible with the change, and the origin and local value-added rate are reconfirmed. Since the destination has changed, after searching, the adapted tariff rules are still the RCEP tariff rules for smartphones. For example, after reconfirmation, Vietnam is the origin, and the local value-added rate (LVC) is raised from the original 45% to 47%. Therefore, based on the item constraints contained in the found tariff rules, as well as the origin, local added value, destination and other information of the target items after the change, after rule matching analysis, it is determined that the requirements for enjoying the RCEP preferential tariff rate of 5% are still met after the change.

[0181] 3) Forecast the fluctuation range of tax rates after the change

[0182] Prediction implementation method: Monte Carlo simulation can be used, but is not limited to, to predict the tax rate fluctuations after the above change (battery supplier changes from China to Vietnam). The specific prediction steps can be as follows:

[0183] ① Identify sources of uncertainty: such as exchange rate fluctuations, changes in trade agreements between RCEP member countries, etc.

[0184] ② Establish a probability model: assume the possibility of changes in different factors and the extent of their impact (such as the probability of increased costs due to rising exchange rates).

[0185] ③Simulation process:

[0186] Set the number of simulations (e.g. 10,000).

[0187] In each simulation, the values ​​of each variable are randomly selected based on the set probability distribution.

[0188] Calculates expected tax rates and other expenses based on selected variable values.

[0189] ④Result analysis: Determine the tax rate fluctuation range based on the results of all simulations.

[0190] For example, after 10,000 simulation runs, the tax rate fluctuation range (also known as the tax rate fluctuation range) is roughly between 4.8% and 5.2%. This tax rate fluctuation range is provided to users to assist them in optimizing their purchases.

[0191] This application uses Monte Carlo simulation to help users understand the specific impact of specific supply chain changes on tariffs, which is particularly important for making strategic decisions.

[0192] Based on the above content, the provided method may further include the following steps:

[0193] 105. Receive a first analysis request from the client, the first analysis request being used to analyze tax rate fluctuations of a target item under an assumed condition. The assumed condition indicates supply chain change information set for the target item, the supply chain change information causing a change in item attribute information required for customs declaration of the target item.

[0194] The assumptions in this step are entered by the user through the client.

[0195] For example, see Figure 3a , inherit and combine Figure 2b In the example given, if the user first performs export declaration on the target item and then triggers the counterfactual analysis function on the customs declaration page, a prompt message such as "Do you want to perform counterfactual analysis on the ** item that was just exported?" can be displayed here first. If the user confirms yes, a hypothesis condition input entry (such as Figure 3a As shown in the figure on the right, the user can enter the corresponding assumptions.

[0196] For example, see Figure 3b If the user directly triggers the counterfactual analysis function in the customs declaration page and has not performed any other operations such as import declaration on the target item before, the user will be required to enter the basic information of the target item, the current actual supply chain information, the hypothetical conditions of the supply chain information, etc.

[0197] For example, in the hypothetical condition given in the previous example, the battery supplier of the target item is changed from China to Vietnam. In this hypothetical condition, the change of the supplier node in the supply chain causes a change in the main material source attribute in the item attribute information of the target item.

[0198] 106. Based on the assumptions, determine the changed item attribute information. For example, if the battery supplier changes, the original main material source, local value-added rate, and place of origin in the item attribute information need to be reconfirmed to obtain the changed item attribute information.

[0199] 107. Based on the changed item attribute information and the knowledge graph, predict the tax rate change range of the target item. The specific implementation of this step can be found in 2) and 3) of the counterfactual analysis process given above.

[0200] 108. Send the tax rate change range to the client for display.

[0201] In the above, the target items are mainly described as follows: based on the tariff rules (called the first version of the tariff rules) that are currently in effect for the first item classification identifier (HS code corresponding to the target object) in the first region found from the knowledge graph, the tariff value of the target item is calculated and the customs declaration documents are automatically generated. When the import and export control information of the first region is updated, the tariff rules of the first region for the first item classification identifier or other item classification identifiers may change, and the knowledge graph needs to be updated at this time. Combined with the construction and update related content of the knowledge graph described in the other embodiments mentioned above, it can be seen that the update strategy adopted by this application for the knowledge graph is incremental update and version control. Based on the relevant content of this update strategy, this application also provides an example of updating the knowledge graph when it is determined that the tariff rules for the first item classification identifier in the first region have changed. Specifically, the example of updating the knowledge graph may include the following steps:

[0202] 109. When an update of the import and export control information of the first region is detected, extract key information from the updated import and export control information of the first region using a second preset model, and output the extraction result;

[0203] 110. If it is determined based on the extraction result that the tariff rules for the first article classification identifier in the first region have changed, updating the knowledge graph;

[0204] Among them, the extraction result includes the tax rate change information corresponding to the first item classification identifier; and updating the knowledge graph includes: according to the tax rate change information, adding the second version of the tariff rules corresponding to the first region to the knowledge graph, and the second version of the tariff rules includes the first region, the second tax rate and the effective date corresponding to the second tax rate. In addition, the second version of the tariff rules may also include item constraints (which may be different from or the same as the item constraints in the previous first version of the tariff rules), etc.; establishing an association relationship between the first item classification identifier and the second version of the tariff rules, and maintaining the previous association relationship between the first item classification identifier and the second version of the tariff rules, and marking the expiration date of the first version of the tariff rules according to the effective date corresponding to the second tax rate.

[0205] For the specific implementation description of the content here, please refer to the relevant example content given in the aforementioned other embodiments when describing the construction and update of the knowledge graph using the item classification code "8541.40" as an example, which will not be described in detail here.

[0206] Based on this, it can be seen that in this application, each item classification identifier in the knowledge graph may be associated with one or more versions of tariff rules corresponding to each customs area, and each version of the tariff rules is marked with a corresponding invalid status or effective status.

[0207] Furthermore, the provided method further comprises the following steps:

[0208] 111. Receive a second analysis request sent by a client, where the second analysis request is used to instruct analysis of changes in tariff rules for the target item in a second region;

[0209] 112. In response to the second analysis request, generating a tariff rule change visualization graph based on at least one version of the tariff rule corresponding to the first region that has a relationship with the first item classification identifier in the knowledge graph; the tariff rule change graph including a trend of tariff rate changes;

[0210] 113. Send the tariff rule change view to the client for display.

[0211] In the above 111, the second analysis request is generated by the client based on the tax rate retrospective analysis operation triggered by the user. In addition, the second region may be the same as or different from the above-mentioned first region.

[0212] For example, if the user has previously made an export declaration for the target object before performing the tax rate retrospective analysis operation, a prompt message may be output to prompt whether to perform a tax rate retrospective analysis corresponding to the destination for the target item previously exported. If the user clicks Yes (e.g. Figure 4a As shown), a second analysis request can be generated for the target item directly based on the first region to which the destination belongs (that is, the second region, which is related to the aforementioned first region). If the user clicks "No", a corresponding input page can be displayed for the user to enter the second region (which may be a different region from the aforementioned first region) that requires tax rate retrospective analysis for the target item on this input page, thereby generating a second analysis request using this second region. Alternatively, the user has not performed any operation on the target item before and directly clicks on tax rate retrospective analysis. In this case, an input page can also be displayed. On this input page, in addition to entering the second region that requires tax rate retrospective analysis for the target item, multimodal information of the target object (such as Figure 4bAs shown), the information entered by the user will participate in the generation of the second analysis information; further, after the second analysis request reaches the server, the server can first determine the first item classification identifier corresponding to the target item based on the multimodal information carried in the request, and then perform a tax rate retrospective analysis on the second region for the target item based on this first item classification identifier.

[0213] In the above 112 to 113, the tariff rule change view may include the tax rate change trend (such as can be reflected by the corresponding curve), and / or may also include other change trends, such as the local value-added rate change trend.

[0214] Based on the view of tariff rule changes displayed on the client, users can, but are not limited to, optimize the supply chain of target items (such as re-evaluating supplier selection and logistics routes, and seeking more cost-effective solutions) and strengthen risk management of target items (gaining early insights into the changing trends of tariff policies in the region and preparing for possible risks).

[0215] Furthermore, this application also provides a proactive compliance early warning mechanism. This proactive compliance early warning mechanism is used to implement automatic impact analysis based on policy (import and export control information) changes. Automated impact analysis can refer to the use of technical means (such as natural language processing and machine learning) to analyze the content of policy changes in real time, assess the specific impact of these changes on the existing operations of enterprises' items, and issue warnings when changes that may affect the compliance of enterprises' items are discovered.

[0216] For example, consider a multinational company engaged in the import and export of electronics, with its primary markets encompassing multiple countries and regions. The company has deployed a proactive compliance alert mechanism to monitor changes in import and export control information in various countries. Specifically, this mechanism obtains new import and export control information in real time from, but is not limited to, official announcements issued by relevant agencies in these countries and regions. A second pre-set model (a large language model) is then used to parse and analyze this acquired import and export control information. For example, by analyzing a notice regarding an impending increase in import tariffs on certain electronic products in one region, the company can extract the product categories involved (e.g., portable computers, HS code 8471.30) and the effective date (six months from now). Comparing this analysis with the company's product supply chain data reveals that some products currently fall into the "8471.30" category, primarily entering the market through a single country. Further analysis indicates that if no action is taken, the new tariff rate will increase the cost of these products by approximately 5%, impacting profit margins. Based on the results of this analysis, alerts are automatically generated and sent to team members in the company responsible for supply chain management. In addition, a detailed report can be provided, suggesting that you consider adjusting procurement strategies (such as looking for suppliers in other low-tax regions) or optimizing logistics routes to reduce overall costs.

[0217] This application provides a proactive compliance alert mechanism that not only helps users (such as businesses) stay informed of changes to import and export control information in relevant regions, but also provides in-depth analysis of the specific impact of these changes on their products and provides feasible response plans. In this way, especially in cross-border commodity trade scenarios, users can better manage compliance risks, maintain competitive advantages, and ensure stable operations in a complex and ever-changing global import and export control information environment.

[0218] The solution provided by the above application can be applied to the server-side deployment such as Figure 5 In the exemplary data processing architecture, taking the customs declaration scenario of cross-border trade goods as an example, the main interactions involved in the data processing architecture include:

[0219] 1) Policy update trigger: When the policy database (i.e., the customs policy database, also known as import and export commodity control information) monitors new policy documents (new import and export commodity control information) issued by various regions, the parsing pipeline is automatically started.

[0220] 2) Product matching asynchronous queue: Highly concurrent requests (such as customs declaration requests for products) enter the Kafka queue and are processed in batches by the Spark cluster.

[0221] 3) Feedback loop: The customs declaration document review results returned by the customs automatically update the knowledge graph in reverse (such as marking invalid tariff rules).

[0222] Specifically, see Figure 5 As shown, the main components of this data processing architecture include the processing layer and the output layer. The processing layer includes the cognitive engine module, matching, and decision engines. The output layer includes the customs clearance material generator. The following briefly describes the modules in the processing layer and the output layer.

[0223] 1. Cognitive Engine

[0224] The function of the cognitive engine is to parse unstructured policy texts (also known as unstructured import and export commodity control information) in various regions around the world in real time and dynamically construct a knowledge graph.

[0225] Furthermore, the functional technical implementation of this cognitive engine includes the following:

[0226] 11) Multi-source policy acquisition. For example, distributed data acquisition technology can be used to automatically obtain policy documents published on the official customs websites of various regions, as well as WTO notices and the original text of trade agreements (in PDF, HTML, or email format). Distributed data acquisition can be implemented using the Scrapy framework, which supports verification code cracking and dynamic rendering. Preprocessing of acquired policy documents includes file format conversion (e.g., using Apache PDFBox to extract text from PDF files) and unified multilingual encoding (e.g., using UTF-8 standardization).

[0227] 12) Use a second preset model (a language model, specifically, a GPT-4 model) to parse the obtained policy document and form structured information.

[0228] Exemplarily, the processing flow of a policy document may be as follows: the original policy document obtained is used as the input of the second preset model -> the second preset model parses the policy document to extract key entities, and outputs the extracted key entities -> uses the rule engine to verify the output key entities -> after verification, the knowledge graph is updated.

[0229] The extraction of key entities is implemented by the second preset model based on the corresponding prompt template. For a detailed description of the prompt template, please refer to the relevant content in other embodiments. The extracted key entities include, but are not limited to, HS codes, information indicating tax rate changes (such as the original tax rate, the new tax rate, and the effective date), and information used for determining rules of origin (such as a local value-added rate ≥ 40%). Furthermore, a rule engine is used to implement an error correction mechanism that performs logical verification on the key entities output by the second preset model. For example, the extracted HS codes can be verified. Furthermore, the knowledge graph is constructed and managed in a graph database such as Neo4j. For details on the knowledge graph's storage structure (such as the node relationship storage contained therein), please refer to the relevant content in other embodiments. The update strategy adopted for the knowledge graph is: incremental updates (such as detecting policy changes every hour and updating only the differences) and version control (retaining historical policy snapshots to support tax rate retrospective analysis). For a detailed description of the knowledge graph's update strategy, please refer to the relevant content in other embodiments.

[0230] 2. Matcher

[0231] The function of the matcher is to match the HS code based on the multimodal information of the product (such as text description information and image information) to determine the corresponding HS code of the product (that is, the corresponding product classification identifier).

[0232] Furthermore, the functional technical implementation of the matcher includes: utilizing a multimodal model (which is a third preset model) to process the multimodal information of the commodity.

[0233] For example, the processing flow of text description information may be as follows:

[0234] The user enters a product description, which is then normalized using the multimodal model. Keyword extraction is then performed, which then generates a vector representation of the extracted key text. For example, if the user enters the product description as "Apple iPhone 13th Generation," the multimodal model outputs "Apple iPhone 13 5G Smartphone."

[0235] During standardization, domain dictionaries are injected. Specifically, the customs commodity terminology database is loaded to replace non-standard terms in text descriptions. For example, the standard term for "lithium battery" is "lithium-ion battery." Furthermore, when vectorizing text, the Sentence-BERT model can be used for vectorization, outputting a corresponding 768-dimensional vector.

[0236] The image information processing flow can be as follows:

[0237] The user uploads a product image, which is then analyzed by the multimodal model (GPT-4V). This image is then output as a structured text description, which is then supplemented with the product's text features. For example, if a "metal zipper pull" is identified in an image, the output might be "Material: Zinc Alloy."

[0238] In the above, when training the multimodal model, the model optimization training includes: fine-tuning the training strategy (such as using the customs commodity image dataset (about 500,000) for LoRA fine-tuning). In addition, the built-in attention mechanism can focus on the key parts of the commodity in the image (such as the type of electrical plug and the texture of textiles).

[0239] Furthermore, based on the above-obtained text features of the commodities, hybrid matching is used to perform HS code matching analysis. The implementation of hybrid matching includes text search (e.g., using Elasticsearch for fuzzy matching, which supports spelling correction) and vector search (e.g., using FAISS index acceleration, which can return the top three results within 10 milliseconds). For a detailed description of hybrid matching, please refer to the relevant content in other embodiments.

[0240] 3. Decision Engine

[0241] The function of the decision engine is to perform tariff rule reasoning and compliance decisions based on the knowledge graph.

[0242] Furthermore, the functional technical implementation of this decision engine includes:

[0243] 31) Rule reasoning module (also known as causal reasoning engine), which uses the corresponding rule chain to determine whether the goods comply with the corresponding tariff preferential policies (specifically, whether they comply with the corresponding tariff rules in the knowledge graph).

[0244] For example, the rule chain is as follows:

[0245] If the origin of goods is an RCEP member country and the local value added rate is ≥40%, and the country of destination is a country on the RCEP effective list, the goods will be subject to the RCEP preferential tax rate.

[0246] In the above, the framework of the rule reasoning module includes: a rule engine (such as Drools) (for processing explicit logic), a first preset model (such as a language model, which is used for implicit reasoning to process complex conditions (such as "the main raw materials come from ASEAN"). For detailed descriptions of the specific functions of the above rule engine and the first preset model, please refer to the relevant content of other embodiments.

[0247] 32) Counterfactual Analysis Module: Its function is to simulate the impact of supply chain changes on taxes and fees

[0248] The functional implementation steps of the counterfactual analysis module may include: the user inputs a hypothesis (such as "changing the motor supplier from China to Vietnam"); ​​based on the hypothesis, traversing the knowledge graph to find related tariff rules; calling Monte Carlo simulation to predict the tariff fluctuation range. For the specific implementation of the function of this counterfactual analysis model, please refer to the relevant content in other embodiments.

[0249] 32) Real-time computing services.

[0250] The structural design of this real-time computing service includes: API request layer (access layer) -> message queue layer (such as Kafka queue) -> distributed computing layer (such as Spark distributed computing) -> result cache layer (such as Redis cache results).

[0251] The API request layer is used to receive real-time external requests, such as requests for product tariff inquiries, counterfactual analysis of a product, and retroactive tax rate analysis for a specific region. These requests are queued in Kafka and executed in parallel using distributed engines such as Spark. The results are cached in databases such as Redis. Kafka is a highly reliable, highly concurrent, and persistent messaging middleware service. Spark is a distributed computing engine, and Redis is an in-memory data structure storage system that can be used as a database.

[0252] The performance metrics of this real-time computing service include: ensuring a single request response latency of less than 500ms in scenarios with a large number of product libraries; and achieving a concurrency of approximately 10,000 TPS (processing approximately 10,000 requests per second). This concurrency relies on features such as the Spark distributed engine and Redis cache.

[0253] 3. Customs Clearance Material Generator

[0254] The function of this customs clearance material generator is to automatically generate customs declaration documents that meet the requirements of various regions.

[0255] Furthermore, the functional technology of the customs clearance material generator is implemented as follows: dynamically loading the corresponding customs declaration module, then intelligently filling the customs declaration template to generate the customs declaration documents for the goods. During the intelligent filling process, missing field processing can be performed, which means automatically triggering data completion. For example, if necessary information is missing during the customs declaration template filling process, such as the net weight of the goods, it can automatically attempt to call the relevant API to complete this information (such as calling the logistics API to obtain the net weight).

[0256] In addition, it is necessary to supplement the present application that: in addition to adopting the solution provided in the present application, other solutions can of course be adopted to implement it. For example, an expert system + manual review solution can be adopted (such as establishing a customs expert knowledge base, and then combining it with manual review of key decisions). However, this solution has the following defects: long response time (such as response time greater than 24 hours) and high labor costs (such as labor costs will increase by about 40%). For another example, a federated learning cross-platform model (multiple logistics companies jointly train models), but this solution has problems such as data privacy leakage risks and policy update synchronization delays.

[0257] In addition, in the cross-border trade scenario of commodities, this application also provides several data processing method embodiments, as follows:

[0258] This application also provides a data processing method, which is applied to the server. Specifically, the method includes the following steps:

[0259] 201. Real-time collection of import and export commodity control information released by various regions;

[0260] 202. Using a second preset model, extract and process the collected import and export commodity control information to obtain an extraction result;

[0261] 203. Construct a knowledge graph based on the extraction results corresponding to each customs area;

[0262] 204. Based on the knowledge graph, perform tax rate analysis on the target product.

[0263] Among them, the knowledge graph contains multiple nodes and node relationships; there are two types of nodes, namely commodity classification identification and tariff rules, and the node relationship reflects the relationship between commodity classification identification and tariff rules. One commodity classification identification represents a category of commodities.

[0264] In the above, the tax rate analysis of the target commodity may include but is not limited to: retrospective analysis of tax rates in a certain region (such as analyzing the trend changes in tax rates for the target commodity in a certain region based on multiple versions of tariff rules related to the target commodity in the region), counterfactual analysis (such as analyzing tax rate fluctuations caused by changes in the supply chain of the target commodity), etc.

[0265] For the specific implementation of the above steps 201 to 204, please refer to the relevant content in other embodiments, and will not be described in detail here. In addition, the method provided in this embodiment may include other steps in addition to the steps shown above. For other steps that may be included and related specific implementations, please refer to the relevant content in other embodiments.

[0266] This application also provides a data processing method, which is applied to the server. Specifically, the method includes the following steps:

[0267] 301. Obtain multimodal information of the target product;

[0268] 302. Determine a first product category identifier corresponding to the target product based on the multimodal information.

[0269] 303. Fill the first commodity classification identifier into a customs declaration template applicable to the target commodity to generate a customs declaration document for the target commodity.

[0270] For the specific implementation of the above steps 301 to 303, please refer to the relevant content in other embodiments, and will not be described in detail here. In addition, the method provided in this embodiment may include other steps in addition to the steps shown above. For other steps that may be included and related specific implementations, please refer to the relevant content in other embodiments.

[0271] This application also provides a data processing method, which is applied to a client. Specifically, the method includes the following steps:

[0272] 401. In response to a user operation on a target commodity, obtaining an assumed condition of the target commodity, wherein the assumed condition indicates supply chain change information set for the target commodity, and the supply chain change information causes a change in commodity attribute information required for customs declaration of the target commodity;

[0273] 402 Based on the assumptions, the tax rate change range of the target product after the supply chain change is displayed;

[0274] Among them, the tax rate change range is determined after the commodity attribute information of the target commodity after the change is determined based on the supply chain change information, and the tariff rules applicable to the target commodity after the supply chain change are queried from the knowledge graph based on the changed commodity attribute information. The queried tariff rules are used to trigger the preset tax rate fluctuation simulation analysis model for prediction.

[0275] In specific implementations, when querying the knowledge graph for tariff rules applicable to target commodities, the query can also be combined with the first commodity classification identifier corresponding to the target commodity. The first commodity classification identifier is the HS code, which can be determined based on the multimodal information of the target commodity. Furthermore, the tax rate fluctuation simulation analysis model can be, for example, but not limited to, a Monte Carlo simulation model.

[0276] For the specific implementation of steps 401 to 402, please refer to the relevant content in other embodiments, and will not be described in detail here. In addition, the method provided in this embodiment may include other steps in addition to the steps shown above. For other steps that may be included and their specific implementation, please refer to the relevant content in other embodiments.

[0277] This application also provides a data processing method, which is applied to the server. Specifically, the method includes the following steps:

[0278] 501. Determine a first product category identifier corresponding to the target product based on multimodal information of the target product;

[0279] 502. Extract and process the updated import and export commodity control information of the first region to obtain an extraction result;

[0280] 503. Determine, based on the extraction result, the tariff rules of the first region after the change in the classification identifier of the first commodity;

[0281] 504. Analyze the multi-dimensional trade impact of the changed tariff rules on the target commodities;

[0282] 505. Output warning information of the target commodity based on the multi-dimensional trade impact.

[0283] In the above, the first region is one of the current trading markets of the target commodity. Multidimensional trade impacts may include but are not limited to: cost impacts (such as increased trade costs caused by increased tax rates), customs declaration compliance impacts (such as compliance impacts on rules of origin, such as the increase in local value-added rate in the changed tariff rules, which may cause the current local value-added rate of the target commodity to not meet this requirement). In addition, the output warning information may include but is not limited to warnings of increased trade costs, warnings that the target commodity does not comply with the tariff rules of the first region, etc. In addition, the warning information may also include other information, such as supply chain adjustment suggestions (such as suggestions for adjusting suppliers, optimizing logistics routes, etc.). The implementation of this content can also refer to the relevant content described in the other aforementioned embodiments for the active compliance warning mechanism.

[0284] For the specific implementation of steps 501 to 502, please refer to the relevant content in other embodiments, and will not be described in detail here. In addition, the method provided in this embodiment may include other steps in addition to the steps shown above. For other steps that may be included and their specific implementation, please refer to the relevant content in other embodiments.

[0285] It should be noted here that in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0286] This application also provides a data processing system, comprising a server and a client. The client is configured to respond to a user's operation on a target item by sending multimodal information about the target item to the server. Upon receiving the multimodal information about the target item, the server is configured to execute the steps of the corresponding method described above.

[0287] Furthermore, the client is further configured to, in response to a user operation on a target commodity, obtain hypothetical conditions for the target commodity and request the server to analyze tax rate changes for the target commodity under the hypothetical conditions; wherein the hypothetical conditions represent supply chain change information set for the target commodity, and the supply chain change information causes changes in commodity attribute information required for customs declaration of the target commodity;

[0288] The server is further configured to: in response to a request from the client, determine the commodity attribute information of the target commodity after the change based on the supply chain change information; based on the changed commodity attribute information, query the tariff rules applicable to the target commodity after the supply chain change from the knowledge graph; use the queried tariff rules to trigger a preset tax rate fluctuation simulation analysis model to predict the tax rate change range of the target commodity after the supply chain change; the client is further configured to display the tax rate change range.

[0289] For the specific forms of the client and server and the specific implementation of their respective functions, please refer to the relevant content in other embodiments and will not be described in detail here.

[0290] This application also provides device embodiments corresponding to each of the above methods. The details are as follows:

[0291] The present application provides a data processing device, which includes: a determination module, an acquisition module, and a generation module. The determination module is used to determine the first item classification identifier corresponding to the target item based on the multimodal information of the target item. The acquisition module is used to obtain the tariff rules applicable to the transportation of the target item from the knowledge graph based on the first item classification identifier. The determination module is also used to determine the tariff value of the target item using the applicable tariff rules. The generation module is used to generate the customs declaration documents for the target item based on the tariff value. The knowledge graph is constructed based on the acquired import and export control information of various regions, and the knowledge graph contains multiple nodes and node relationships; the nodes are of two types, namely item classification identifiers and tariff rules; the node relationship reflects the relationship between the item classification identifier and the tariff rules.

[0292] Optionally, the acquisition module, when used to obtain the tariff rules applicable to the transportation of the target item from the knowledge graph based on the first item classification identifier corresponding to the target item, is specifically used to: determine the item attribute information required for customs declaration of the target item; based on the item attribute information, obtain from the knowledge graph a first version of the tariff rules that has a relationship with the first item classification identifier and is in effect; determine the first version of the tariff rules as the tariff rules applicable to the transportation of the target item; wherein the item attribute information includes the destination; the first version of the tariff rules includes a first region, a first tax rate, and item constraints; when the target item is declared for import, the first region includes the destination.

[0293] Optionally, the item attribute information further includes a first attribute and a second attribute, and the second attribute affects the identification of the first attribute; the first attribute is one of the attributes required to determine whether the target item satisfies the item constraint condition. Furthermore, the device further includes: a trigger module and an update module. The trigger module is configured to, if it is determined based on the second attribute that the first attribute needs to be re-identified, trigger the first preset model to perform a re-identification analysis on the first attribute element using the second attribute to obtain an analysis result. The update module is configured to update the first attribute based on the analysis result. The trigger module is further configured to use the updated first attribute as one of the input parameters of the rule engine, triggering the rule engine to determine whether the target item satisfies the item constraint condition based on the input parameters. Furthermore, the determination module is further configured to, if satisfied, determine that the first version of the tariff rule is the tariff rule applicable to the transportation of the target item.

[0294] Optionally, the device further includes: a receiving module, a prediction module, and a sending module. The receiving module is configured to receive a first analysis request sent by a client, the first analysis request being used to instruct analysis of tax rate changes for the target item under hypothetical conditions; the hypothetical conditions indicating supply chain change information set for the target item, the supply chain change information causing changes in the item attribute information. The determination module is further configured to determine the changed item attribute information based on the hypothetical conditions. The prediction module is configured to predict the extent of tax rate changes for the target item based on the knowledge graph and the changed item attribute information. The sending module is configured to send the tax rate change extent to the client for display.

[0295] Optionally, the device further includes: an extraction module. The extraction module is used to extract and process the updated import and export control information of the first region using a second preset model, and output the extraction result. And, the above-mentioned update module is used to update the knowledge graph if it is determined based on the extraction result that the tariff rules of the first region for the first item classification identifier have changed. Wherein, the extraction result includes tax rate change information; and updating the knowledge graph includes: according to the tax rate change information, adding the second version of the tariff rules corresponding to the first region to the knowledge graph, the second version of the tariff rules including the first region, the second tax rate and the effective date of the second tax rate; establishing an association relationship between the first item classification identifier and the second version of the tariff rules, maintaining the association relationship between the first item classification identifier and the first version of the tariff rules, and marking the expiration date of the first version of the tariff rules according to the effective date of the second tax rate.

[0296] Optionally, the receiving module is further configured to receive a second analysis request from a client, the second analysis request instructing an analysis of changes in tariff rules for the target item in a second region. The generating module is further configured to respond to the second analysis request and generate a tariff rule change view based on at least one version of tariff rules corresponding to the second region that has a relationship with the first item classification identifier and is contained in the knowledge graph; the tariff rule change view includes a trend in tax rate changes. The sending module is further configured to send the tariff rule change view to the client for display.

[0297] Optionally, the multimodal information includes textual description information and image information of the target item. Furthermore, the determination module, when used to determine the item classification identifier corresponding to the target item based on the multimodal information of the target item, includes: standardizing the textual description information to obtain standardized textual description information; extracting a first key text for determining the target item classification from the standardized textual description information; performing recognition analysis on the image information to obtain a second key text for determining the target item classification; and determining the first item classification identifier based on the first key text and the second key text. The standardization process includes at least one of the following: replacing non-standard descriptive terms in the textual description information according to the customs item terminology database; and supplementing missing key item attributes.

[0298] The present application also provides a data processing device, which includes: an acquisition module, an extraction module, a construction module, and an analysis module. The acquisition module is used to collect import and export commodity control information released by various regions in real time. The extraction module is used to extract and process the collected import and export commodity control information using a second preset model to obtain an extraction result. The construction module is used to construct a knowledge graph based on the extraction results corresponding to each customs area. The analysis module is used to perform tax rate analysis on the target commodity based on the knowledge graph. The knowledge graph contains multiple nodes and node relationships; the nodes are of two types, namely commodity classification identifiers and tariff rules, and the node relationship reflects the relationship between commodity classification identifiers and tariff rules. One commodity classification identifier represents a category of commodities.

[0299] The present application also provides a data processing device comprising: an acquisition module, a determination module, and a filling module. The acquisition module is configured to acquire multimodal information of a target commodity. The determination module is configured to determine, based on the multimodal information, a first commodity classification identifier corresponding to the target commodity. The filling module is configured to fill the first commodity classification identifier into a customs declaration template applicable to the target commodity to generate a customs declaration document for the target commodity.

[0300] The present application also provides a data processing device, which includes: an acquisition module and a display module. The acquisition module is used to obtain the hypothetical conditions of the target commodity in response to the user's operation on the target commodity, and the hypothetical conditions represent the supply chain change information set for the target commodity, and the supply chain change information causes the commodity attribute information required for the target commodity to be declared for customs to change. The display module is used to display the tax rate change range of the target commodity after the supply chain change based on the hypothetical conditions. The tax rate change range is determined based on the supply chain change information to determine the commodity attribute information of the target commodity after the change, and the tariff rules applicable to the target commodity after the supply chain change are queried from the knowledge graph based on the changed commodity attribute information. The queried tariff rules are used to trigger the preset tax rate fluctuation simulation analysis model for prediction.

[0301] The present application also provides a data processing device, which includes: a determination module, an extraction module, an analysis module, and an output module. The determination module is used to determine the first commodity classification identifier corresponding to the target commodity based on the multimodal information of the target commodity. The extraction module is used to extract and process the updated import and export commodity control information of the first region to obtain an extraction result. The determination module is also used to determine the tariff rules of the first region after the change for the first commodity classification identifier based on the extraction result. The analysis module is used to analyze the multidimensional trade impact of the changed tariff rules on the target commodity. The output module is used to output early warning information of the target commodity based on the multidimensional trade impact.

[0302] What needs to be explained here about the above-mentioned devices is that the above-mentioned devices can implement the technical solutions described in the above-mentioned corresponding method embodiments. The specific implementation principles of the above-mentioned modules or units can be found in the relevant contents of the above-mentioned corresponding method embodiments, and will not be described in detail here.

[0303] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 6 As shown, in practice, the electronic device includes: a memory 54 and a processor 55.

[0304] The memory 54 is used to store computer programs and may be configured to store various other data to support operations on the computing platform. Examples of such data include instructions for any application or method operating on the electronic device, data structures, contact data, phone book data, messages, images, videos, etc.

[0305] The processor 55 is coupled to the memory 54 and is configured to execute the computer program in the memory 54 to implement the steps in the various method embodiments of the present application.

[0306] Further, if Figure 6 As shown, the electronic device also includes: a communication component 56, a display 57, a power component 58, an audio component 59 and other components. Figure 6 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 6 In addition, Figure 6 The components in the dotted box are optional components, not mandatory components, and the specific components may depend on the product form of the electronic device. The working node of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone or an IOT device, or a server device such as a conventional server, a cloud server or a server array. If the electronic device of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, etc., it can include Figure 6 If the working node of this embodiment is implemented as a server device such as a conventional server, a cloud server or a server array, it may not include Figure 6 Components within the dotted box.

[0307] The above-mentioned memory can be implemented by any type of volatile or non-volatile memory 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 memory, flash memory, magnetic disk or optical disk.

[0308] The communication component is configured to facilitate wired or wireless communication between the device in which the communication component resides and other devices. The device in which the communication component resides can access a wireless network based on a communication standard, such as a 2G, 3G, 4G / LTE, 5G, or other mobile communication network, or a combination thereof. In an exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel.

[0309] The above-mentioned display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundary of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0310] The power supply assembly provides power to various components of the device in which the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which the power supply assembly is located.

[0311] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), and when the device where the audio component is located is in an operating mode, such as call mode, recording mode, and voice recognition mode, the microphone is configured to receive external audio signals. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0312] Accordingly, an embodiment of the present application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to implement the steps in the above method embodiment. The computer-readable storage medium includes volatile or non-volatile or a combination thereof, and may be removable or non-removable. Examples of computer-readable storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic cassette, tape disk storage or other magnetic storage device or any other non-transmission medium.

[0313] Accordingly, an embodiment of the present application further provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, the processor is enabled to implement the steps in the above-mentioned method embodiment. It should be understood that each process or a combination of multiple processes in the above-mentioned method flow can be implemented by a computer program or instruction. In addition, these computer programs or instructions can be applied to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device, so that the processor of the general-purpose computer, the special-purpose computer, the embedded processor or other programmable data processing device can be implemented as a device for implementing the corresponding functions in the above-mentioned method embodiment.

[0314] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0315] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A data processing method, characterized in that: include: Determining a first item classification identifier corresponding to the target item based on the multimodal information of the target item; Based on the first item classification identifier, obtaining tariff rules applicable to the transportation of the target item from the knowledge graph; Determining the tariff value of the target item using the applicable tariff rules; generating customs declaration documents for the target item according to the tariff value; Among them, the knowledge graph is constructed based on the import and export control information obtained from various regions, and the knowledge graph contains multiple nodes and node relationships; there are two types of nodes, namely item classification identification and tariff rules; the node relationship reflects the relationship between item classification identification and tariff rules.

2. The method according to claim 1, characterized in that Based on the first item classification identifier corresponding to the target item, obtaining tariff rules applicable to the transportation of the target item from the knowledge graph includes: Determining the item attribute information required for customs declaration of the target item; Based on the item attribute information, obtaining from the knowledge graph a first version of tariff rules that is related to the first item classification identifier and is in effect; Determine the first version of the tariff rules as the tariff rules applicable to the transportation of the target item; Among them, the item attribute information includes the destination; the first version of the tariff rules includes the first region, the first tax rate, and the item constraint conditions; when the target item is imported, the first region includes the destination.

3. The method according to claim 2, characterized in that The item attribute information also includes a first attribute and a second attribute, and the second attribute affects the determination of the first attribute; the first attribute is one of the attributes required to determine whether the target item meets the item constraint condition; The method further comprises: If it is determined based on the second attribute that the first attribute needs to be re-identified, the second attribute is used to trigger the first preset model to perform a re-identification analysis on the first attribute element to obtain an analysis result; Based on the analysis result, updating the first attribute; Using the updated first attribute as one of the input parameters of a rule engine, triggering the rule engine to determine whether the target item satisfies the item constraint based on the input parameter; If the conditions are met, the first version of the tariff rules is determined to be the tariff rules applicable to the transportation of the target item.

4. The method according to claim 2 or 3, characterized in that Also includes: receiving a first analysis request sent by a client, where the first analysis request is used to instruct analysis of tax rate changes of the target item under an assumed condition; The assumption condition indicates supply chain change information set for the target item, and the supply chain change information causes the item attribute information to change; Based on the assumption, determining the changed attribute information of the item; Predicting the tax rate change range of the target item based on the knowledge graph and the changed item attribute information; The tax rate change range is sent to the client for display.

5. The method according to claim 2 or 3, characterized in that Also includes: Using a second preset model, extracting and processing the updated import and export control information of the first region, and outputting an extraction result; If it is determined according to the extraction result that the tariff rules of the first region for the first item classification identifier have changed, updating the knowledge graph; In which, the extraction result includes tax rate change information; and updating the knowledge graph includes: adding a second version of the tariff rules corresponding to the first region to the knowledge graph based on the tax rate change information, the second version of the tariff rules including the first region, the second tax rate and the effective date of the second tax rate; establishing an association between the first item classification identifier and the second version of the tariff rules, maintaining the association between the first item classification identifier and the first version of the tariff rules, and marking the expiration date of the first version of the tariff rules according to the effective date of the second tax rate.

6. The method according to claim 5, characterized in that Also includes: receiving a second analysis request sent by a client, where the second analysis request is used to instruct analysis of changes in tariff rules for the target item in a second region; In response to the second analysis request, generating a tariff rule change view based on at least one version of the tariff rule corresponding to the second region that has a relationship with the first item classification identifier and is contained in the knowledge graph; The tariff rule change view includes the trend of tariff rate changes; The tariff rule change view is sent to the client for display.

7. The method according to any one of claims 1 to 3, characterized in that The multimodal information includes text description information and image information of the target object; and Determining an item classification identifier corresponding to the target item based on the multimodal information of the target item includes: Standardizing the text description information to obtain standardized text description information; Extracting a first key text for determining a classification of a target item from the standardized text description information; performing recognition analysis on the image information to obtain a second key text for determining the classification of the target object; Determining the first item classification identifier based on the first key text and the second key text; The standardization process includes at least one of the following: replacing non-standard descriptive words in the text description information according to the customs article terminology database; and supplementing missing key attributes of the article.

8. A data processing method, characterized in that: include: Real-time collection of import and export commodity control information released by various regions; Using the second preset model, extracting and processing the collected import and export commodity control information to obtain an extraction result; Constructing a knowledge graph based on the extraction results corresponding to each customs area; Based on the knowledge graph, perform tax rate analysis on the target product; Among them, the knowledge graph contains multiple nodes and node relationships; there are two types of nodes, namely commodity classification identification and tariff rules, and the node relationship reflects the relationship between commodity classification identification and tariff rules. One commodity classification identification represents a category of commodities.

9. A data processing method, characterized in that: include: Obtain multimodal information of the target product; Determining a first product category identifier corresponding to the target product based on the multimodal information; Fill the first commodity classification identifier into the customs declaration template applicable to the target commodity to generate a customs declaration document for the target commodity.

10. A data processing method, characterized in that: include: In response to a user operation on a target commodity, obtaining an assumed condition of the target commodity, the assumed condition indicating supply chain change information set for the target commodity, the supply chain change information causing a change in commodity attribute information required for customs declaration of the target commodity; Based on the above assumptions, show the extent of the tax rate change for the target product after the supply chain change; Among them, the tax rate change range is determined after the commodity attribute information of the target commodity after the change is determined based on the supply chain change information, and the tariff rules applicable to the target commodity after the supply chain change are queried from the knowledge graph based on the changed commodity attribute information. The queried tariff rules are used to trigger the preset tax rate fluctuation simulation analysis model for prediction.

11. A data processing method, characterized in that: include: Determining a first product category identifier corresponding to the target product based on the multimodal information of the target product; Extracting and processing the updated import and export commodity control information of the first region to obtain an extraction result; Determining, based on the extraction result, a tariff rule after the change in the first region for the first commodity classification identifier; Analyze the multi-dimensional trade impact of the changed tariff rules on target commodities; Based on the multi-dimensional trade impact, early warning information of the target commodity is output.

12. A data processing system, characterized in that: include: The client is configured to respond to a user's operation on a target item and send multimodal information of the target item to the server; A server, configured to execute the steps of any one of the methods of claims 1 to 8 upon receiving the multimodal information of the target object; Furthermore, the server is further configured to implement the steps of the method described in claim 8, 9 or 11 above; Furthermore, the client is further configured to, in response to a user operation on a target commodity, obtain hypothetical conditions for the target commodity and request the server to analyze tax rate changes for the target commodity under the hypothetical conditions; wherein the hypothetical conditions represent supply chain change information set for the target commodity, and the supply chain change information causes changes in commodity attribute information required for customs declaration of the target commodity; The server is further configured to: in response to a request from the client, determine the commodity attribute information of the target commodity after the change based on the supply chain change information; Based on the changed commodity attribute information, the tariff rules applicable to the target commodity after the supply chain change are queried from the knowledge graph; using the queried tariff rules, a preset tax rate fluctuation simulation analysis model is triggered to predict the tax rate change range of the target commodity after the supply chain change; The client is also used to display the tax rate change range.

13. An electronic device, characterized in that: include: memory and a processor, wherein The memory is used to store programs; The processor is coupled to the memory, and is configured to execute the program stored in the memory to implement the method according to any one of claims 1 to 11.

14. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a computer, the method according to any one of claims 1 to 11 can be implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images