Intelligent order generation and management system

Abstract

The invention discloses an intelligent order generation and management system, and relates to the technical field of information processing. The system comprises a data acquisition module; the multi-modal information processing module is used for calling the data in the database and performing character extraction through unified analysis and structured processing; the knowledge graph construction module is used for establishing a knowledge graph of product entities and entity relationships according to the products, mapping the extracted text information to entity nodes of the knowledge graph, and performing semantic association of different source information to form preprocessed order information; and the intelligent generation module is used for generating standardized JSON structured order information from the preprocessed order information and automatically inputting the standardized JSON structured order information into a database. According to the invention, through a unified data acquisition and storage mechanism, automatic identification and collection of various information formats are realized, manual arrangement is not needed, the work processing efficiency is greatly improved, and the complexity of manual input and proofreading is reduced.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of information processing, and particularly relates to an intelligent order generation and management system. BACKGROUND

[0002] In a private domain operation scenario such as a retail industry, users are accustomed to submitting orders containing key information such as the names and quantities of purchased items through non-standardized means such as handwritten documents, group chat messages, and picture attachments. However, relying on manual extraction, organization, and entry of order data from these fragmented, multi-modal (text, image, table) information sources is not only extremely inefficient, but also prone to human error, which can seriously affect subsequent processes.

[0003] Currently, due to the limitations of the technical solutions for structured order automatic generation, it is difficult to effectively meet the needs of the above scenarios: (1) Template / form-based entry systems (such as ERP, CRM): Users are forced to manually enter information through standardized interfaces, which is a cumbersome process that is severely disconnected from users' natural ordering habits (such as group chat communication and uploading handwritten orders by taking pictures), resulting in low user acceptance and difficulty in promotion.

[0004] (2) Optical character recognition (OCR) technology: Although it can handle image information, the recognition accuracy significantly decreases when faced with complex and variable handwriting, blurred images, or non-standard forms, making it difficult to independently ensure the reliability of order key information extraction (such as item name, specification, and quantity).

[0005] (3) Natural language processing (NLP) technology: It is good at processing text chat content, but it is usually limited to analyzing sentences with relatively clear structures. For user's free expression of colloquial descriptions, ambiguous references, industry jargon abbreviations, and table information embedded in pictures / attachments, its understanding and information extraction capabilities are severely insufficient.

[0006] For example, Chinese patent CN119006094A discloses an ERP-based order management method and system, which includes verifying order data transmitted from a CRM system to an ERP system when the ERP system or CRM system enters a warning state, and evaluating the transmission process of the order data in combination with the load states of the ERP system and CRM system to effectively control the transmission reliability of raw data between different systems. Also, Chinese patent CN117217876A discloses an order preprocessing method and device based on OCR technology, equipment, and medium, and Chinese patent CN114677080A proposes a logistics export order processing method combining RPA and AI, and the like, all of which provide an order processing technology. However, the core deficiencies of the prior art are: Modal processing unification: Existing solutions focus on single data type (pure text or pure image), lack of collaborative perception, correlation and unified understanding ability of multi-modal information such as group chat text, handwritten pictures and scanned forms; Insufficient automation: The processing process highly depends on manual intervention for information screening, proofreading and system input, with low automation level, which cannot meet the high-frequency, fragmented and real-time order processing demand of private domain operation; Lack of deep semantic understanding: The identification accuracy of user's unstructured and free expression order intention (such as ambiguous description "old look a portion", "the last red"), complex context (such as multiple modifications and multi-person discussions in chat records) and non-standard format information (such as handwritten body and randomly arranged tables) is not high, which is easy to miss or misjudge the key order elements; Lack of scene adaptability: Especially lack of optimization design for "private domain group chat" which is a typical scene. This scene has the characteristics of multi-user participation, information cross-channel (text, picture, voice to text, file), high fragmentation, unstructured and other characteristics. The existing technology has not provided an effective end-to-end intelligent solution to integrate and process these information and accurately generate structured orders.

[0007] Therefore, an intelligent order generation and management system is needed, which can accurately adapt to the intelligent order generation technology of complex scenes such as private domain group chat, to completely solve the core pain points of low efficiency and easy to make mistakes of manual processing. SUMMARY

[0008] The purpose of the present application is to provide an intelligent order generation and management system, which realizes the automatic identification and collection of various information formats through a unified data collection and storage mechanism, without manual sorting, greatly improving the work processing efficiency and solving the existing problems.

[0009] To solve the above technical problems, the present application is realized by the following technical scheme: The present application is an intelligent order generation and management system, comprising: Data acquisition module: it is used to acquire multi-modal data and store it in the database; Multi-modal information processing module: it is used to call the data in the database and extract the text through unified analysis and structured processing, automatically identify the key information such as goods, quantity, customer information in the order; The knowledge graph construction module is configured to establish a knowledge graph of product entities and entity relationships according to products, and map the extracted text information to the entity nodes of the knowledge graph through basic mapping, fuzzy mapping and completion mapping, associate the semantic information of different sources, form preprocessed order information, and automatically convert non-standard, fuzzy, alias and other expressions into standardized structured order data through semantic understanding and information extraction technology, thereby greatly reducing the complexity of manual input and correction. The intelligent generation module is configured to generate standardized JSON structured order information from the preprocessed order information based on a prompt and a large language model, and automatically input the information into a database.

[0010] Further, the data collection module periodically collects chat records in the chat tool, and the chat records include natural language text, Excel tables and handwritten order pictures.

[0011] Further, the multi-modal information processing module extracts text through an optical character recognition algorithm or a semantic reasoning matching algorithm or a large language model.

[0012] Further, the multi-modal information processing module performs intent recognition and key information extraction on the extracted text, forms order information, and converts the order information into structured JSON format.

[0013] Further, the knowledge graph includes product, alias, attribute and specification information, and a multi-dimensional knowledge graph covering products, aliases, attributes and specifications is constructed, and semantic matching and vector retrieval technologies are combined to realize automatic error correction, completion and standardization of order information.

[0014] Further, the knowledge graph construction module performs completion mapping according to the entity relationships and context reasoning algorithm of the knowledge graph, and automatically corrects and completes fragmented expressions.

[0015] Further, the context reasoning algorithm tracks historical session IDs associated with historical orders based on a conversation memory network (CMN) to automatically correct and complete fragmented expressions.

[0016] Further, the knowledge graph construction module performs basic mapping according to the precise string matching of the database, and automatically corrects and completes fragmented expressions.

[0017] Further, the knowledge graph construction module performs fuzzy mapping according to the alias library and vector similarity calculation of the database, and automatically corrects and completes fuzzy expressions.

[0018] According to the knowledge graph, the semantic association and reasoning of information of different sources such as pictures, texts and tables are realized, compared with the current single identification technology, and the problem that the existing technology has insufficient processing capacity for fuzzy and alias expressions can be effectively solved.

[0019] Further, a verification module is further included, and the verification module comprises: a syntax verification submodule which performs integrity verification of JSON format and fields corresponding to order information based on a JSON Schema Validator; a business verification submodule which performs inventory and price validity verification of order information based on an ERP API real-time calling tool; a semantic verification submodule which verifies whether the order information conforms to the original intention of the user based on an LLM similarity comparison tool.

[0020] The present application has the following beneficial effects: The present application automatically collects, fuses and processes order information in group chats and the like, realizes automatic identification and collection of various information formats through a unified data collection and storage mechanism, and greatly improves work processing efficiency without manual manual arrangement. The present application integrates large language models, semantic reasoning matching algorithms and OCR recognition technologies, uniformly analyzes and structures multiple source order information such as chat texts, picture OCR results and table data, can automatically identify key information such as goods, quantity and customer information in orders, and automatically converts non-standard, fuzzy, alias and the like expressions into standardized structured order data through semantic understanding and information extraction technology, greatly reducing the complexity of manual input and correction.

[0021] Of course, any product implementing the present application does not necessarily need to achieve all the advantages described above at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed for the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0023] Figure 1 The figure is a schematic diagram of an intelligent order generation and management system corresponding to the first embodiment of the present application. Figure 2 The figure is a schematic diagram of an intelligent order generation and management system corresponding to the second embodiment of the present application. Figure 3 The figure is a schematic diagram of a verification module in the intelligent order generation and management system corresponding to the second embodiment of the present application. Figure 4 This is a schematic diagram of verification optimization in the intelligent order generation and management system corresponding to the second embodiment of the present invention. DETAILED DESCRIPTION

[0024] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0025] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0026] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0027] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0028] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0029] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0030] Example 1: See also Figure 1 As shown, the present invention is an intelligent order generation and management system, comprising: a data acquisition module, a multimodal information processing module, a knowledge graph construction module, and an intelligent generation module, wherein each module is communicatively connected; The data collection module is used to obtain multimodal data and store it in the database, enabling channels to obtain original order information, such as connecting to WeChat / DingTalk APIs to capture group chat text, images, and files; As an embodiment provided by the present invention, preferably, the system automatically obtains chat logs (chatlogs) from chat tools such as WeChat and QQ groups on a regular basis, including various data types such as text, pictures, and documents, and stores them uniformly in a database to ensure data integrity and traceability; The multimodal information processing module is used to retrieve data from the database and perform text extraction through unified parsing and structured processing, automatically identifying key information such as goods, quantities, and customer information in orders; As an embodiment of the present invention, preferably, images in chat logs (such as handwritten orders and product photos) are first subjected to text extraction using OCR (Optical Character Recognition) technology. For handwritten text and complex backgrounds, semantic matching algorithms (such as vector retrieval and knowledge base comparison) are used to retrieve the most similar standard product names from a pre-built product knowledge base.

[0031] For example: If the user inputs "Tu Shouyi Qiansanxiang", the system can automatically match it to the standard product "Wang Shouyi Shisanxiang".

[0032] As an embodiment provided by the present invention, preferably, for natural language text and documents (such as Excel) in the chat, a large language model (such as ChatGPT, etc.) is used in combination with a pre-designed prompt template to perform intent recognition and key information extraction on the text, and organize the order content (such as product, quantity, customer information, etc.) into a structured JSON format; As an embodiment provided by the present application, preferably, the knowledge graph construction module is used to establish a knowledge graph of product entities and entity relationships according to products, and map the extracted text information to entity nodes of the knowledge graph through basic mapping, fuzzy mapping and completion mapping, so as to perform semantic association of information of different sources and form preprocessed order information. As an embodiment provided by the present application, preferably, through semantic understanding and information extraction technology, non-standard, fuzzy, alias and other expressions are automatically converted into standardized structured order data, greatly reducing the complexity of manual input and correction. As an embodiment provided by the present application, preferably, a knowledge graph covering entities such as products, aliases, attributes and specifications and their relationships is constructed to support dynamic expansion and maintenance.

[0033] The OCR recognition result, text content and table field are mapped to entity nodes in the knowledge graph to realize semantic association of information of different sources. The entity relationships and context reasoning capability of the knowledge graph are used to automatically correct and complete fuzzy, alias and fragmented expressions.

[0034] For example: The picture recognizes “shisanxiang” and the text mentions “Wang Shouyi”, and the system can associate them as the same product through the knowledge graph; The intelligent generation module is used to generate standardized JSON structured order information based on the prompt and the large language model, and automatically input the database, so that the whole system forms a closed loop feedback update of management and generation. Through the prompt and the large language model, the preprocessed order information is uniformly generated into standardized JSON structured data, and is automatically input into the database, providing support for subsequent order management, statistical analysis and system docking.

[0035] For natural language text information, Excel data and handwritten order photograph pictures provided by the user, a large language model is used for recognition, and fuzzy matching is realized in combination with information retrieval technology, and then the information is uniformly generated into structured order information, which is suitable for private domain operation scenarios such as QQ groups and WeChat groups that lack unified information management systems.

[0036] The embodiment has the following advantages: (1) Automatically collect and analyze various formats of order information, without manual input, greatly improving work efficiency; (2) It can be connected with existing order systems to realize end-to-end business process automation; (3) It can reduce the workload of manual order information input by chat information, improve work efficiency, and avoid errors in collaboration.

[0037] Example 2: See also Figures 2-4 As shown, the present invention is an intelligent order generation and management system. It proposes an intelligent order automatic generation system based on multimodal information. It aims to solve the problems of information dispersion, non-standard expression, and low manual processing efficiency when users collect order statistics in multiple forms (text, pictures, tables, etc.) through multiple channels such as group chats in private domain operation environments. The system includes: Data acquisition module, multimodal information processing module, knowledge graph construction module, intelligent generation module, verification module, and communication connections between modules; Based on the first embodiment, the verification module includes: Syntax verification submodule, which performs integrity verification of the JSON format and fields corresponding to order information based on JSON Schema Validator; The business verification submodule verifies the inventory and price validity of order information based on the ERP API real-time call tool; The semantic verification submodule verifies whether the order information conforms to the user's original intention based on the LLM similarity comparison tool.

[0038] As an embodiment provided by the present invention, preferably, after a verification error is found, the verification module delineates an error node. When delineating an error node, the verification module performs the following steps: S1: Obtain the error node and preliminarily define the radiation range, which includes: S11: Use the error node as the root node of the binary tree; S12: Use the node directly associated with the error node as the first-level child node of the root node of the binary tree; S13: Using the node directly associated with the first-level child node as the second-level child node of the root node of the binary tree; S14: The radiation range includes the error node, first-level child nodes, and second-level child nodes; S2: Within the radiation range, select any node as a trial node and determine whether deleting the trial node will affect the mapping of the error node. If so, the trial node is used as the direct radiation range of the error node; otherwise, the trial node is used as the indirect radiation range of the error node. S3: Node self-healing: Select the shortest binary branch from the direct radiation range and perform self-healing repair. Self-healing repair includes: S31: Remapping: For nodes with mapping errors, re-execute basic mapping, fuzzy mapping, and complete mapping; S32: Data correction: If the node data itself is incorrect (such as incorrect inventory quantity), the data is updated by calling the external API; S33: After adjusting the relationship between nodes, check through the verification module until there is no error, and then self-repair is successful, if the self-repair fails, select the next shortest binary tree branch, and perform self-repair; S34: If all the branches in the direct radiation range are not successfully self-repaired, the radiation range is expanded, and the third-level child nodes are included for self-repair again; S35: If self-repair is finally impossible, manual intervention process is triggered.

[0039] As an embodiment provided by the application, preferably, when selecting a shortest binary tree branch from the direct radiation range, the following algorithm is executed: Optionally, a branch is selected; The node type, node association strength, and historical repair success rate of the branch are respectively weighted; The weighted binary tree branch is updated and generated; The weight of each branch is calculated; A shortest and highest-weighted binary tree branch is selected.

[0040] For example:

[0041] As an embodiment provided by the application, preferably, according to the verification result, a continuous optimization closed-loop order processing mode is realized, and the steps are as follows: After finding the verification error, manually correct the order, and analyze the error type; If it is an LLM understanding error, add a PROMPT example, fine-tune the LLM, and regenerate the order; If it is a knowledge gap, update the knowledge graph, optimize entity mapping, and regenerate the order.

[0042] Embodiment three: Based on embodiment one or two, as an embodiment provided by the application, preferably, the data collection module periodically collects chat records in the chat tool, and the chat records include natural language text, Excel table, and handwritten order picture, such as pulling new messages every 5 minutes or parsing MIME email attachments based on SMTP protocol every day.

[0043] As an embodiment provided by the application, preferably, it further includes metadata storage, and the metadata includes source channel, timestamp, and user anonymous ID, and the metadata output is: a unified format of original data package, represented as text+picture URL+file link.

[0044] As an embodiment provided by the present application, preferably, the multi-modal information processing module extracts text through an optical character recognition algorithm or a semantic reasoning matching algorithm or a large language model, and the picture confidence is > text > history record.

[0045] As an embodiment provided by the present application, preferably, the multi-modal information processing module performs intent recognition and key information extraction on the extracted text to form order information, and converts the order information into a structured JSON format, such as a fresh food retail order, which requires input conversion to JSON: Unit conversion: box -> kg (1 box = 10 kg); Add preservation requirements; Missing fields are completed with default values; <<Knowledge graph>> Product default value: {"preservation requirement": "cold chain transportation"} User preference: {"delivery time": "14:00-18:00"} <<Input>> {items: [{product: "red fuji apple", quantity: 5, unit: "box"}]} <<Output Schema>>.

[0046] As an embodiment provided by the present application, preferably, the knowledge graph includes product, alias, attribute, and specification information, and a multi-dimensional knowledge graph covering product, alias, attribute, and specification is constructed, and combined with semantic matching and vector retrieval technologies, automatic error correction, completion, and standardization of order information are realized.

[0047] As an embodiment provided by the present application, preferably, the knowledge graph construction module performs mapping based on the entity relationship of the knowledge graph and the context reasoning algorithm, automatically corrects and completes fragmented expressions, constructs a multi-dimensional knowledge graph covering product, alias, attribute, and specification, and combines semantic matching and vector retrieval technologies to realize automatic error correction, completion, and standardization of order information. The system can realize semantic association and reasoning of information from different sources such as pictures, text, and tables based on the knowledge graph, and compared with the current single recognition technology, it can effectively solve the problem of insufficient processing capacity for fuzzy and alias expressions in the prior art.

[0048] As an embodiment provided by the present application, preferably, the context reasoning algorithm automatically corrects and completes fragmented expressions based on a conversation memory network (CMN) that tracks historical conversation ID and associates historical orders.

[0049] As an embodiment provided by the present application, preferably, the knowledge graph construction module performs basic mapping according to accurate string matching of the database, and automatically corrects and completes the expression of the fragments, such as accurate matching: "Apple Red Fuji" → basic mapping to "Apple Red Fuji", such as: "Mr. Li ordered a set of white flagship PAD, requiring a fast charging kit and a screen breakage insurance, and the delivery address is Suzhou, Jiangsu Province" "Customer": ["Mr. Li"]; "Product": ["White flagship PAD"]; "Accessories": ["Fast charging kit", "Screen breakage insurance"]; "Address": ["Suzhou, Jiangsu Province"]; Then: basic mapping (direct matching)

[0050] Direct matching success: customer ID, quantity, and other standardized fields are directly mapped.

[0051] As an embodiment provided by the present application, preferably, the knowledge graph construction module performs fuzzy mapping according to the alias library and vector similarity calculation of the database, and automatically corrects and completes the expression of the fragments, such as fuzzy matching: "large fruit" → vector similarity search → associated L specification; "high sweetness" → add attribute [sugar content ≥ 18%].

[0052] For example: "Mr. Li ordered a set of white flagship PAD, requiring a fast charging kit and a screen breakage insurance, and the delivery address is Suzhou, Jiangsu Province"; then the fuzzy mapping is:

[0053] Fuzzy mapping logic: IF similarity > 0.91 THEN directly associated ELSE transfer to manual review.

[0054] According to the semantic association and reasoning of information from different sources such as pictures, texts, and tables realized by the knowledge graph, compared with the current single recognition technology, the problem of insufficient processing capacity of the prior art for fuzzy and alias expressions can be effectively solved, and the order collection, analysis, standardization, and structured entry can be automatically completed, the order management system can be connected, the end-to-end business process automation can be supported, and the automation and intelligent level of order processing can be improved.

[0055] An intelligent order generation and management system realizes automatic recognition and collection of multiple information formats through automatic collection, fusion and processing of order information in group chats and other channels, and a unified data collection and storage mechanism, without manual manual arrangement, greatly improving work processing efficiency; the large language model, semantic reasoning matching algorithm and OCR recognition technology are integrated, the chat text, picture OCR result, table data and other multi-source order information are uniformly analyzed and structured, the key information such as goods, quantity and customer information in the order can be automatically recognized, and through semantic understanding and information extraction technology, non-standard, fuzzy, alias and other expressions are automatically converted into standardized structured order data, greatly reducing the complexity of manual input and correction.

[0056] In the description of the present specification, the description referring to the terms "one embodiment", "an example", "a specific example" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0057] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details and limit the application to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of the present specification. The present specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent order generation and management system, characterized in that: include: Data acquisition module: used to acquire multimodal data and store it in the database; Multimodal information processing module: It is used to retrieve data from the database and extract text through unified parsing and structured processing; Knowledge graph construction module: This module is used to build a knowledge graph of product entities and entity relationships based on products. It also maps the extracted text information to the entity nodes of the knowledge graph through basic mapping, fuzzy mapping, and completion mapping, and performs semantic association of information from different sources to form pre-processed order information. Intelligent generation module: It is used to generate standardized JSON structured order information from pre-processed order information based on Prompt and a large language model, and automatically enter it into the database.

2. The intelligent order generation and management system according to claim 1, characterized in that: The data collection module regularly collects chat records in the chat tool, and the chat records include natural language text, Excel tables, and handwritten order pictures.

3. The intelligent order generation and management system according to claim 1, characterized in that: The multimodal information processing module extracts text using an optical character recognition algorithm, a semantic reasoning matching algorithm, or a large language model.

4. The intelligent order generation and management system according to claim 3, characterized in that: The multimodal information processing module performs intent recognition and key information extraction on the extracted text to form order information, and converts the order information into a structured JSON format.

5. The intelligent order generation and management system according to claim 1, characterized in that: The knowledge graph includes product, alias, attribute, and specification information.

6. The intelligent order generation and management system according to claim 1, characterized in that: The knowledge graph construction module performs completion mapping based on the entity relationships and contextual reasoning algorithm of the knowledge graph, and automatically corrects and completes fragmented expressions.

7. The intelligent order generation and management system according to claim 6, characterized in that: The contextual reasoning algorithm automatically corrects and completes fragmented expressions by tracking historical session IDs and associating historical orders based on a conversation memory network.

8. The intelligent order generation and management system according to claim 1, characterized in that: The knowledge graph construction module performs basic mapping based on the precise string matching of the database, and automatically corrects and completes the expression of the fragments.

9. The intelligent order generation and management system according to claim 1, characterized in that: The knowledge graph construction module performs fuzzy mapping based on the database's alias library and vector similarity calculation, and automatically corrects and completes fuzzy expressions.

10. The intelligent order generation and management system according to claim 1, characterized in that: It also includes a verification module, which includes: Syntax verification submodule, which performs integrity verification of the JSON format and fields corresponding to order information based on JSON Schema Validator; The business verification submodule verifies the inventory and price validity of order information based on the ERP API real-time call tool; The semantic verification submodule verifies whether the order information conforms to the user's original intention based on the LLM similarity comparison tool.

Images