An e-commerce commodity information dynamic structured processing method and system

Abstract

The application provides an e-commerce commodity information dynamic structured processing method and system, and breaks through the general workflow technical limitation: unlike general arrangement tools such as Dify / Coze / Airflow, the application supports a structured reasoning chain for commodity theme fields; the application supports cross-theme dependent derivation: there is a natural dependence between commodity themes, and the application first incorporates the configurable logic chain to form automatic derivation; the application has highly flexible theme reasoning capability: different categories and different industries can quickly build differentiated theme chains through combined rule steps; and the application has very high robustness, and a step-level bottom-up mechanism avoids single-point failure.

Description

Technical Field

[0001] This invention relates to the field of television data processing technology, and more specifically, to a method and system for dynamic structured processing of e-commerce product information. Background Technology

[0002] In e-commerce and multi-platform product data governance scenarios, product themes (such as brand, model, capacity, series, category, price range, and feature version) are the core of product structured representation. However, current theme maintenance methods mainly rely on manual or single-rule processing, which has the following main problems: 1. Complex dependencies exist between subject fields but cannot be expressed. For example, the series can only be determined after the model is determined; the function version depends on the model or capacity range; and the parsing method of the capacity or specification field can only be determined after the category is determined. Existing systems usually only provide single-field processing logic and cannot describe these "cross-subject dependencies". Second, the rule execution method is fixed and does not support flexible combination by project. Most existing systems use fixed order or independent rules, which cannot allow users to freely combine multiple rule steps according to the characteristics of the category, nor can they customize the theme reasoning chain. Third, it cannot support collaborative processing of AI and rules. In unstructured text, AI (LLM / classification models) is more advantageous for topic extraction; however, AI output often requires further cleaning by rules. Current technology does not have the ability to insert AI nodes into the rule chain. Fourth, there is a lack of step-level fallback mechanisms. For example, if model identification fails, it is necessary to fall back to title word splitting; if capacity is missing, it is necessary to supplement it from the details parameters; if series is missing, it is necessary to deduce it from the model. The existing system only provides global fallback and cannot define different fallback strategies for each step. Fifth, the correct execution order cannot be guaranteed when referencing across topics. For example, if the user configures the system to execute the series first and then the model, it will lead to calculation errors. The existing system does not have a mechanism to automatically deduce the topic dependency topology.

[0003] Therefore, we have made improvements to this by proposing a dynamic structured processing method and system for e-commerce product information. Summary of the Invention

[0004] This invention provides a method for dynamic structuring of e-commerce product information, characterized by the following steps: S1. Receive multiple product themes configured by the user in the visual operation interface; S2. For each product theme, receive multiple rule steps selected by the user from the rule library corresponding to the product theme, and configure the execution order of each rule step; S3. Based on the user's configuration of each rule step, identify and construct cross-topic reference relationships between product themes; S4. Based on the cross-topic reference relationship, parse and generate the dependency topology graph between product topics, and determine the execution order of each product topic and its rule steps based on the dependency topology graph; S5. When executing the rule steps of any product topic, the output of each rule step is automatically used as the input of the next step within the same product topic, and is simultaneously written into the result set of that product topic for reference by other product topics. S6. For each rule step, select the execution method according to the user configuration: rule execution, AI execution, AI first then rule, rule first then AI, or mixed execution of rule and AI, and merge the results through a strategy. S7. Configure an independent step-level fallback strategy for each rule step. When a rule step fails to execute or outputs an abnormality, it will be handled according to the fallback strategy. S8. After completing the execution of all product themes and their rule steps, output structured product information data and provide it to the outside world in JSON format.

[0005] As a preferred technical solution of this application, in S1, the product theme includes at least one or more of the following: brand, model, series, capacity, and functional version.

[0006] As a preferred technical solution of this application, in S2, the rule base contains no less than 15 rule types, covering at least: basic field standardization, field concatenation and combination, keyword matching and extraction, conditional logic judgment, mapping and renaming, interval, numerical and statistical classification, field extraction priority and fallback rules, compound logic, formatted output standardization, text cleaning and language processing, field template reference and inheritance, multi-value output and concatenation, cross-platform field correction, rule priority and fallback strategy, and inter-topic dependency and inference. Users can arbitrarily adjust the order of steps to achieve project-level differentiated topic processing. As a preferred technical solution of this application, in S3, the cross-topic reference relationship includes using the output results of one or more other product topics as the basis for condition judgment, filtering conditions, value derivation basis, or execution triggering conditions of the current topic.

[0007] As a preferred technical solution of this application, in S4, the process of generating the dependent topology graph specifically includes: Iterate through all product themes using the rule steps to extract cross-theme references; Each product theme is abstracted as a node in a directed graph, and cross-theme references are abstracted as directed edges; Perform topological sorting on a directed graph to generate an acyclic execution order; When a circular dependency is detected, the user is prompted to correct the configuration.

[0008] As a preferred technical solution in this application, in S5, the step-level input / output binding mechanism enables: Within the same topic, the output fields of the previous rule step are automatically mapped to the input fields of the next rule step; Different topics can be linked through topic result sets to achieve structured references at the field level.

[0009] As a preferred technical solution in this application, in S7, the fallback strategy includes at least using the results of the preceding steps, using the default value, calling the backup rule, calling the artificial intelligence model to complete the step, skipping the step, or triggering an abnormal interruption.

[0010] As a preferred technical solution in this application, in S8, the structured data includes at least one or more of the following fields: brand, model, series, capacity, function version, price range, size range, and platform normalization field.

[0011] A dynamic structured processing system for e-commerce product information, used to implement a method for dynamic structured processing of e-commerce product information, includes: The theme configuration module is used to receive and manage the product themes configured by the user; The rule chain building module is used to configure rule steps and execution order for each product theme; The cross-topic dependency analysis module is used to identify cross-topic reference relationships and generate a dependency topology graph; The execution engine module is used to schedule and execute each rule step according to the dependency topology graph, and supports the mixed execution of rules and artificial intelligence models; The fallback strategy management module is used to configure and invoke step-level fallback strategies for each rule step; The results set management module is used to store and manage intermediate and final results for each topic; The output interface module is used to output structured JSON format product information to external systems.

[0012] As a preferred technical solution of this application, it also includes a visual operation interface for displaying product themes, rule steps, cross-theme dependencies and execution status in a graphical environment.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This application breaks through the limitations of general workflow technology: unlike general orchestration tools such as Dify / Coze / Airflow, it supports structured inference chains "oriented towards product subject fields"; 2. This application supports cross-topic dependency derivation: there are natural dependencies between product topics, and this invention is the first to incorporate them into a configurable logic chain to form automatic derivation; 3. This application has highly flexible topic reasoning capabilities: different categories and industries can quickly construct differentiated topic chains by combining rule steps; 4. It has extremely high robustness, and the step-by-step fallback mechanism avoids single points of failure; 5. This application supports collaborative processing of AI and rules to improve the accuracy of structured extraction.

[0014] 6. This application can achieve structured output, which is multi-field structured data rather than text fragments, making it suitable for large-scale governance. Attached Figure Description

[0015] Figure 1 A flowchart of the method for dynamic structured processing of e-commerce product information provided in this application; Figure 2 The catch-all flowchart provided for this application; Figure 3 The cross-topic dependency flowchart provided for this application; Figure 4 Example diagram of the output JSON provided for this application. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.

[0018] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0019] Example 1, please refer to Figures 1-3 The method for dynamic structuring of e-commerce product information includes the following steps: S1. Receives multiple product themes configured by the user in the visual operation interface; adopts a visual custom configuration mode, which allows for flexible addition, removal, and adjustment of product parsing themes without modifying the underlying code, thereby improving scenario adaptability and versatility. S2. For each product theme, receive multiple rule steps selected by the user from the rule library corresponding to the product theme, and configure the execution order of each rule step; relying on the independent rule library and the customizable step order mechanism, break the limitation of fixed rule execution chain, and users can customize exclusive processing flow according to product data characteristics to adapt to the field extraction, cleaning and standardization needs of different products. S3. Based on the user's configuration of each rule step, identify and construct cross-topic reference relationships between product themes; realize data linkage processing of multiple product themes; S4. Based on cross-topic reference relationships, parse and generate a dependency topology graph between product topics, and determine the execution order of each product topic and its rule steps based on the dependency topology graph; for example: if "Series" references "Model", the system automatically places the "Model" node before "Series"; if "Feature Version" references "Brand" and "Model", the system automatically deduces the execution order as: Brand → Model → Feature Version, ensuring that the execution chain always conforms to the topic dependency relationship; S5. When executing the rule steps of any product theme, the output of each rule step is automatically used as the input of the next step within the same product theme, and is simultaneously written into the result set of that product theme for reference by other product themes; this achieves seamless connection of step data within a single theme and efficient reuse of data between multiple themes, improving data processing efficiency and consistency. S6. For each rule step, select the execution method according to the user configuration: rule execution, AI execution, AI first then rule, rule first then AI, or mixed execution of rules and AI. And merge the results through strategies to ensure higher accuracy in unstructured data. S7. Configure an independent step-level fallback strategy for each rule step. When a rule step fails to execute or outputs an abnormality, it is handled according to the fallback strategy to reduce the problem of interruption of the overall parsing process and task failure caused by the abnormal execution of a single step. S8. After completing the execution of all product themes and their rule steps, output structured product information data and provide it to the outside world in JSON format, which can be directly used for e-commerce data analysis scenarios.

[0020] Furthermore, in S1, the product theme includes at least one or more of the following: brand, model, series, capacity, and functional version, covering the core attribute dimensions of e-commerce products, meeting the basic needs of structured governance of mainstream e-commerce products, and adapting to multi-category product analysis scenarios.

[0021] Furthermore, in S2, the rule base contains no fewer than 15 rule types, covering at least: basic field standardization, field concatenation and combination, keyword matching and extraction, conditional logic judgment, mapping and renaming, interval, numerical and statistical classification, field extraction priority and fallback rules, compound logic (multi-level conditions), formatted output standardization, text cleaning and language processing, field template reference and inheritance, multi-value output and concatenation, cross-platform field correction, rule priority and fallback strategies, and inter-topic dependency and inference (including dependency inference, filtering relationships, cross-topic concatenation, condition triggering, and contextual semantic reasoning). Users can arbitrarily adjust the order of steps to achieve project-level differentiated topic processing; constructing a comprehensive, multi-level rule system can solve various complex commodity data processing problems and thoroughly meet the differentiated data governance needs of different projects and platforms.

[0022] Furthermore, in S3, cross-topic referencing relationships include using the output results of one or more other product topics as the basis for conditional judgment, filtering conditions, value derivation, or execution triggering conditions for the current topic; for example: IF (brand = "Xiaomi") AND (model contains "Pro") THEN Function Version = "Pro Version". This referencing relationship is explicitly configured by the user, forming a cross-topic dependency. Relying on user-defined explicit cross-topic referencing logic, single-field parsing is upgraded to multi-field association reasoning. Based on the parsed basic attributes, derived attributes are deduced, solving the problem that traditional technologies cannot determine derived fields such as version and series based on basic information such as brand and model, making data parsing more in line with actual e-commerce business logic.

[0023] Furthermore, in S4, the process of generating the dependent topology graph specifically includes: Iterate through all product themes using the rule steps to extract cross-theme references; Each product theme is abstracted as a node in a directed graph, and cross-theme references are abstracted as directed edges; Perform topological sorting on a directed graph to generate an acyclic execution order; When a circular dependency is detected, the user is prompted to correct the configuration.

[0024] Furthermore, such as Figure 3 As shown, it includes: Users can configure rules for each topic and reference results from other topics within those rules. Parse rule configurations to identify reference relationships between topics, such as: series dependent on model / feature version dependent on brand + model; Construct a topic-dependent directed graph where nodes = topic fields and edges = reference relationships; Determine if the dependency topology graph contains cycles; If so, a configuration error message indicating a circular dependency will be displayed, prompting the user to adjust the rules. If not, proceed with the following steps: Perform topological sorting on a directed graph to generate the execution order of topics; The rule chains for each topic are executed in the following topological order: brand → model → series → feature version. During execution, each topic can read the results of its dependent topics; Complete all topic calculations to obtain a complete set of structured topics.

[0025] Furthermore, in S5, the step-level input / output binding mechanism enables: Within the same topic, the output fields of the previous rule step are automatically mapped to the input fields of the next rule step; Different topics are referenced at the field level through topic result sets, rather than through text streams.

[0026] Furthermore, in S7, fallback strategies include at least using the results of previous steps, using default values, calling alternative rules, calling an AI model to complete the steps, skipping the steps, or triggering an abnormal interruption. Different topics can have different fallback mechanisms to improve robustness.

[0027] The process of providing a safety net is as follows: Figure 2 As shown, it includes: Enter a specific topic and execute it, such as: Model Topic; Obtain the rule step chain configured for this topic; Take the first step: read the input fields and the results of the preceding topic; Determine if the execution method is rule execution, AI execution, AI first then rule, rule first then AI, or a mixture of rule and AI execution; if yes, write the output of this step to update the intermediate result of the current topic; otherwise, trigger the fallback strategy of this step; determine if the fallback is successful; if yes, write the output of this step to update the intermediate result of the current topic; otherwise, record the exception or mark it as null, and optionally interrupt the execution of this topic. After updating the intermediate results of the current topic in this step, check if there is a next step. If not, output the final result of the topic for reference by subsequent topics or for final output. If so, repeat the process of "take the first step, read the input fields and the results of the previous topic".

[0028] Furthermore, in S8, structured data must include at least one or more of the following fields: brand, model, series, capacity, function version, price range, size range, and platform normalization fields. It outputs multi-dimensional and comprehensive structured data for products, covering the core field requirements for e-commerce product governance, analysis, and operation, and is compatible with e-commerce data applications across all scenarios.

[0029] Example 2. A dynamic structured processing system for e-commerce product information, used to implement a method for dynamic structured processing of e-commerce product information, including: The theme configuration module is used to receive and manage the product themes configured by the user; The rule chain building module is used to configure rule steps and execution order for each product theme; The cross-topic dependency analysis module is used to identify cross-topic reference relationships and generate a dependency topology graph; The execution engine module is used to schedule and execute each rule step according to the dependency topology graph, and supports the mixed execution of rules and artificial intelligence models; The fallback strategy management module is used to configure and invoke step-level fallback strategies for each rule step; The results set management module is used to store and manage intermediate and final results for each topic; The output interface module is used to output structured JSON format product information to external systems.

[0030] The execution engine module further includes: The rule execution submodule is used to execute pure rule-based steps; The model execution submodule is used to call the artificial intelligence model for inference; The hybrid execution scheduling submodule is used to coordinate the multi-mode collaborative execution of rules and models.

[0031] Furthermore, it also includes a visual user interface for displaying product themes, rule steps, cross-theme dependencies, and execution status in a graphical environment.

[0032] This invention can be widely applied to: e-commerce platform product governance, e-commerce platform product data management (MDM), e-commerce data analysis, product structuring in the retail industry, multi-platform product alignment and matching, e-commerce product data analysis, cross-platform e-commerce product database establishment, AI structured data extraction system, data cleaning / ETL / data standardization system, e-commerce platform market competition intelligence monitoring, cross-platform e-commerce operation system (ERP), e-commerce & cross-border product selection system, and e-commerce marketing data analysis system.

[0033] Example 3, the following description illustrates a preferred embodiment of the present invention: First, the user creates a "Model" theme in the system and configures it in three steps: Step 1: Keyword extraction; Step 2: AI text parsing; Step 3: Normalize mapping rules; II. User configuration of the "Series" theme, and settings: Subject="Model"; Step 1: Derive the series based on the model number; Step 2: Interval-based judgment; Step 3: Align the mapping table; 3. The system automatically analyzes cross-topic reference relationships and establishes execution dependencies: Brand → Model → Series → Feature Version; IV. During execution: Obtain the original product data; Execute each topic according to the dependency chain; Each topic generates results step-by-step according to a chain of steps. If any step fails, a fallback will be executed. Ultimately, this results in a structured thematic output; Example of outputting JSON, such as Figure 4 As shown.

[0034] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a connection that allows communication between them; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0035] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A method for dynamic structuring of e-commerce product information, characterized in that, Includes the following steps: S1. Receive multiple product themes configured by the user in the visual operation interface; S2. For each product theme, receive multiple rule steps selected by the user from the rule library corresponding to the product theme, and configure the execution order of each rule step; S3. Based on the user's configuration of each rule step, identify and construct cross-topic reference relationships between product themes; S4. Based on the cross-topic reference relationship, parse and generate the dependency topology graph between product topics, and determine the execution order of each product topic and its rule steps based on the dependency topology graph; S5. When executing the rule steps of any product topic, the output of each rule step is automatically used as the input of the next step within the same product topic, and is simultaneously written into the result set of that product topic for reference by other product topics. S6. For each rule step, select the execution method according to the user configuration: rule execution, AI execution, AI first then rule, rule first then AI, or mixed execution of rule and AI, and merge the results through a strategy. S7. Configure an independent step-level fallback strategy for each rule step. When a rule step fails to execute or outputs an abnormality, it will be handled according to the fallback strategy. S8. After completing the execution of all product themes and their rule steps, output structured product information data and provide it to the outside world in JSON format.

2. The method for dynamic structured processing of e-commerce product information according to claim 1, characterized in that, In S1, the product theme must include at least one or more of the following: brand, model, series, capacity, and functional version.

3. The method for dynamic structured processing of e-commerce product information according to claim 1, characterized in that, In S2, the rule base contains no fewer than 15 rule types, covering at least: basic field standardization, field concatenation and combination, keyword matching and extraction, conditional logic judgment, mapping and renaming, interval, numerical and statistical classification, field extraction priority and fallback rules, compound logic, formatted output standardization, text cleaning and language processing, field template reference and inheritance, multi-value output and concatenation, cross-platform field correction, rule priority and fallback strategy, and inter-topic dependency and inference. Users can arbitrarily adjust the order of steps to achieve project-level differentiated topic processing.

4. The method for dynamic structuring of e-commerce product information according to claim 1, characterized in that, In S3, the cross-topic reference relationship includes using the output results of one or more other product topics as the basis for condition judgment, filtering conditions, value derivation basis, or execution triggering conditions of the current topic.

5. The method for dynamic structured processing of e-commerce product information according to claim 1, characterized in that, In S4, the process of generating the dependent topology graph specifically includes: Iterate through all product themes using the rules and steps to extract cross-theme references; Each product theme is abstracted as a node in a directed graph, and cross-theme references are abstracted as directed edges; Perform topological sorting on a directed graph to generate an acyclic execution order; When a circular dependency is detected, the user is prompted to correct the configuration.

6. The method for dynamic structured processing of e-commerce product information according to claim 1, characterized in that, In S5, the step-level input / output binding mechanism enables: Within the same topic, the output fields of the previous rule step are automatically mapped to the input fields of the next rule step; Different topics can be linked through topic result sets to achieve structured references at the field level.

7. The method for dynamic structured processing of e-commerce product information according to claim 1, characterized in that, In S7, fallback strategies include at least using the results of previous steps, using default values, calling alternative rules, calling an AI model to complete the steps, skipping the steps, or triggering an abnormal interruption.

8. The method for dynamic structured processing of e-commerce product information according to claim 1, characterized in that, In S8, structured data must include at least one or more of the following fields: brand, model, series, capacity, feature version, price range, size range, and platform normalization field.

9. A dynamic structured processing system for e-commerce product information, used to implement the dynamic structured processing method for e-commerce product information as described in any one of claims 1 to 7, characterized in that, include: The theme configuration module is used to receive and manage the product themes configured by the user; The rule chain building module is used to configure rule steps and execution order for each product theme; The cross-topic dependency analysis module is used to identify cross-topic reference relationships and generate a dependency topology graph; The execution engine module is used to schedule and execute each rule step according to the dependency topology graph, and supports the mixed execution of rules and artificial intelligence models; The fallback strategy management module is used to configure and invoke step-level fallback strategies for each rule step; The results set management module is used to store and manage intermediate and final results for each topic; The output interface module is used to output structured JSON format product information to external systems.

10. The e-commerce product information dynamic structured processing system according to claim 9, characterized in that, It also includes a visual interface for displaying product themes, rule steps, cross-theme dependencies, and execution status in a graphical environment.

Images