An ai-based complex table automatic generation method and system

Abstract

This invention discloses an AI-based method and system for automatically generating complex tables, belonging to the fields of artificial intelligence and low-code development technology. The method includes: establishing a hierarchical and categorized prompt word rule base; building a model recognition knowledge base and inputting table architecture metadata; receiving user natural language requirement descriptions; a large model performing semantic parsing based on the prompt word rule base and knowledge base to generate advanced intelligent table basic JSON data; converting the JSON data into a canvas-recognizable format and rendering it as a real table component; receiving user manual adjustment operations and updating the configuration in real time; and saving the final configuration to generate the target intelligent table. This invention, through prompt word management and knowledge base retrieval, BERT intent recognition, bidirectional conversion between JSON and AST, and visual bidirectional binding, achieves fully automated generation from natural language requirements to editable intelligent tables, solving the problems of cumbersome generation, limited functionality, and poor adaptability in existing technologies. It has the advantages of high efficiency, flexibility, and low barrier to entry.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and low-code development technology, and more specifically, to an AI-based method and system for automatically generating complex tables. Background Technology

[0002] Current intelligent table generation technologies mainly rely on various development tools or simple generation platforms. Their core goal is to achieve rapid table generation and basic display. Their mainstream generation logic is relatively simple: developers often need to manually configure the table structure (such as the number of columns, rows, and column names) or generate a basic table framework through simple text commands. After generation, the table data and related configurations need to be manually added. The table generation functions of some low-code platforms have not broken through the limitations of the traditional generation mode.

[0003] The core capabilities of existing intelligent table generation technologies are concentrated on basic table construction, and their specific shortcomings include:

[0004] (1) The generation method is cumbersome and inefficient. It requires manually defining the column attributes and row data of the table one by one, or repeatedly operating the basic components to splice the table. Even if it supports simple command generation, it can only generate the basic framework and lacks batch generation and batch configuration capabilities.

[0005] (2) Insufficient intelligence, unable to accurately understand users' personalized needs for tables (such as table item types, data associations, and filtering rules) through natural language.

[0006] (3) The functions are limited and the adaptability is poor. After generation, it only supports basic text display functions and lacks diverse table items such as images, files, progress bars, and tags. It also does not have the function of custom filtering and mapping of table data (such as mapping the code 0 / 1 to female / male). The generated table cannot directly adapt to the usage requirements of complex business scenarios. It requires a lot of manual optimization before it can be used, and the adjustment is cumbersome and prone to errors. Summary of the Invention

[0007] The present invention aims to provide an AI-based method and system for automatically generating complex tables, in order to solve the problems of cumbersome and inefficient table generation methods, insufficient intelligence, limited functionality, and poor adaptability in the existing technology.

[0008] The technical solution of this invention is as follows:

[0009] To achieve the above objectives, this invention provides an AI-based method for automatically generating complex tables, comprising the following steps:

[0010] Step S1: Establish a prompt word management system, which includes a hierarchical and categorized prompt word rule library. The prompt word rule library covers general requirement description specifications, field definition prompt words, layout style prompt words, and business rule prompt words.

[0011] Step S2: Build a model recognition knowledge base and input table architecture metadata. The table architecture metadata includes technical configuration items such as table container, column configuration, row configuration, cell style, data binding, interaction rules, filtering and sorting, access control and linkage logic. Configure the business meaning, value range, constraints and related dependencies for each configuration item.

[0012] Step S3: Receive the user's natural language description of their requirements;

[0013] Step S4: Based on the prompt word management system and the model recognition knowledge base, the large model performs semantic parsing, intent recognition and demand decomposition on the natural language demand description, extracts the core elements of the table, and generates advanced intelligent table basic JSON data that conforms to the platform's technical standards.

[0014] Step S5: Convert the basic JSON data into a component configuration format that the canvas can recognize, and render and generate an interactive smart table component in the canvas;

[0015] Step S6: Receive manual adjustment operations from the user on the smart table component in the canvas, and update the table configuration in real time.

[0016] Step S7: Save the final table configuration and generate the target smart table.

[0017] Furthermore, in step S4, a fine-tuned BERT intent recognition model is used, combined with semantic vector retrieval from the knowledge base, to parse entity parameters in user requirements. The prompt word engineering guides the large model to generate intelligent table basic JSON data with compliant structure and complete parameters.

[0018] Furthermore, in step S5, converting the basic JSON data into a canvas-recognizable format includes: using the Esprima parser to convert the JSON data into an abstract syntax tree (AST), extracting the tag names, attributes, child nodes, and core nodes of the table component, performing format mapping and validation according to the component library JSON Schema specification, and outputting standard JSON recognizable by the canvas.

[0019] Furthermore, in step S6, a two-way data binding mechanism is adopted to support adaptive adjustment of table layout, online debugging of component styles, and real-time modification of configuration parameters. All operations are updated in real time through the data synchronization engine to update table JSON data and DOM rendering.

[0020] Further, in step S4, the basic JSON data generated by the large model includes table attribute configuration and table item configuration attributes. The table attribute configuration includes border attributes, pagination attributes, header display attributes, component size attributes, pagination configuration attributes, zebra stripe attributes, row height attributes, and table content height attributes. The table item configuration attributes include whether to display overflow prompts, whether to sort, minimum width, tag name, alignment, header color, whether to merge headers, header font size, dynamic data binding fields, filter type, width, content font size, whether to fix, static data binding fields, content color, content binding field type, content filter field configuration set, and content field configuration set. The content field configuration set includes image width, image height, image URL, whether to display document preview, whether to display document download, icon configuration, icon color, icon font size, icon insertion position, tag configuration name, tag size, tag type, progress bar height, progress bar minimum value, progress bar maximum value, dynamic image height, dynamic image width, and whether to display default content.

[0021] This invention also provides an AI-based system for automatically generating complex tables, including: a prompt word and knowledge base management module, used to establish a standardized prompt word management system and build a large-scale model-specific recognition knowledge base, providing support for table architecture JSON Schema, component metadata, and semantic mapping of configuration items;

[0022] The requirement parsing and JSON generation module is used to receive users' natural language requirement descriptions, and through intent recognition and entity extraction, parse the table structure and table item type requirements to generate advanced intelligent table basic JSON data that conforms to the knowledge base specification.

[0023] The JSON conversion module is used to parse the basic JSON data and convert it into a standard component JSON format that the visualization canvas can recognize, thereby realizing a two-way mapping between JSON data and visualization components.

[0024] The visual canvas rendering module is used to render interactive, realistic table components based on the converted standard JSON data, providing visual interactive support for canvas dragging and component preview.

[0025] The table editing module provides visual editing capabilities for table component properties, styles, and layouts, enabling real-time synchronization of operations with JSON data.

[0026] The technical effects and advantages of this invention are as follows:

[0027] (1) Efficient and convenient generation: No need to manually configure the table structure and fill in the data, it can quickly generate a table with complete static data, complete the basic configuration of the table in batches, avoid the redundancy and errors of manual operation, and greatly shorten the generation cycle;

[0028] (2) Diverse table items: It can automatically generate various personalized table items such as images, files, progress bars, tags, file previews, and downloads, without the need for manual addition, and adapts to diverse usage scenarios;

[0029] (3) Strong data adaptability: It can be directly and flexibly bound to the real business data model to realize dynamic synchronization of table data. It also supports the automatic completion of secondary data filtering and custom mapping (such as mapping code 0 to female and code 1 to male) through the configuration of filter items in the table. After generation, it can be directly adapted to complex business scenarios without manual optimization.

[0030] (4) Flexible and accessible editing: The table layout can be adjusted intuitively, the table item type can be replaced and the original configuration can be inherited. The table style can be modified in real time and the effect can be previewed. The operation is simple and no professional technical skills are required.

[0031] (5) End-to-end full-process automation: It integrates five major modules: knowledge base management, requirement analysis, JSON conversion, visualization rendering, and real-time editing. After the user inputs natural language commands, no human intervention is required to complete the full-process automation from requirement to deployable and editable smart table. Attached Figure Description

[0032] Figure 1 This is a flowchart of an AI-based method for automatically generating complex tables in an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example 1

[0035] This embodiment provides an AI-based method for automatically generating complex tables, such as... Figure 1 As shown.

[0036] (1) Establish a prompt word management system

[0037] Design a hierarchical and categorized prompt word rule library to cover all scenarios of table generation: including general requirement description specifications, field definition prompt words, layout style prompt words, business rule prompt words, and fallback prompt words for abnormal scenarios, etc.

[0038] Clearly define the grammatical structure, parameter constraints, and output format requirements of the prompt words to ensure that the large model can generate a JSON structure that conforms to the system specifications, regardless of the user requirements it receives. At the same time, it supports iterative optimization of prompt words to adapt to the needs of different business scenarios.

[0039] (2) Build a model recognition knowledge base and input table structure metadata.

[0040] Completely input the full architecture JSON data of the smart tables in the existing platform, including all technical configuration items such as table container, column configuration, row configuration, cell style, data binding, interaction rules, filtering and sorting, access control, and linkage logic;

[0041] For each configuration item, provide detailed business meaning, value range, constraints, and dependencies, such as "the table's filtering configuration condition is matching the array in filterData".

[0042] (3) User Input Smart Form Requirements Description

[0043] Users do not need to understand the technical details of the platform. They can describe the business needs of the table completely using only natural language. For example: "Generate a monthly performance report for the sales department, including employee name, region, sales this month, target completion rate, year-on-year growth rate, and ranking. It supports filtering by region and sorting by sales, and the table header is fixed."

[0044] It supports multi-dimensional requirement input, covering all requirements such as table business purpose, field structure, data type, layout style, interactive functions, and permission requirements.

[0045] (4) The large model parses user descriptions and generates advanced intelligent table base JSON data.

[0046] The large model, based on pre-defined prompt word rules and a knowledge base, performs semantic parsing, intent recognition, and requirement decomposition on users' natural language needs, extracting the core elements of the table: field list, data type, business rules, style requirements, interaction configuration, etc.

[0047] By combining the table architecture specifications in the knowledge base, it automatically generates advanced intelligent table base JSON data that conforms to the platform's technical standards, fully including all the technical configurations of the table, without the need for manual coding.

[0048] (5) Convert the generated basic JSON data into a canvas-recognizable format and render it as a real table component.

[0049] The system performs format conversion and structure adaptation on the basic JSON data generated by the large model, converting it into a component configuration format that the canvas can recognize and adapting it to the canvas's rendering engine.

[0050] The system automatically renders realistic and interactive smart table components on the canvas, fully replicating all configurations defined in the JSON, including fields, layout, style, and interaction rules, allowing users to intuitively see the final effect of the table.

[0051] (6) Users can manually adjust the tables in the canvas.

[0052] Users can manually adjust AI-generated tables in all dimensions within the low-code canvas, including layout adjustments, style enhancements, configuration optimizations, and interactive enhancements.

[0053] The table configuration is updated in real time during the adjustment process, and users can preview the adjustment effect instantly, achieving "what you see is what you get";

[0054] Version rollback is supported. If you are not satisfied with the adjustments, you can revert to the initial AI-generated version and re-optimize it.

[0055] (7) Generation complete, the target intelligent table is obtained.

[0056] After the user completes all adjustments, the system automatically saves the final table configuration and generates the target intelligent table that meets the business requirements.

[0057] The generated tables retain complete configuration data and can be edited, iterated, and optimized at any time to adapt to changes in business needs.

[0058] The core of this invention lies in:

[0059] 1.1 Table-specific knowledge base and prompt word association mechanism

[0060] To address the shortcomings of existing technologies, such as the disconnect between knowledge base and table configuration and the inability of the system to accurately identify table requirements, a solution of "structured input of table configuration rules + standardized management of prompt words" is adopted. This solution builds a table-specific knowledge base and constructs semantic associations, standardizes prompt word templates, enables the system to quickly and accurately identify table requirements, solves the problem of requirement identification bias, and provides basic support for rapid table generation.

[0061] 1.2 Intelligent Table Requirement Analysis and Basic JSON Generation Solution

[0062] To address the shortcomings of existing technologies, such as inaccurate understanding of requirements, non-standardized generated table data, and the need for repeated fine-tuning, a BERT fine-tuning model combined with table requirement entity extraction logic is adopted. This model accurately parses key parameters such as table structure and table item types from the user's natural language, automatically generating standardized and complete basic JSON data for the table, reducing manual fine-tuning and achieving efficient conversion of requirements into table data.

[0063] 1.3 Two-way conversion architecture between table JSON and visualization components

[0064] To address the shortcomings of existing technologies where tabular data and visual tables are disconnected and require manual adaptation, a two-stage transformation logic of "Esprima parsing AST + table component schema verification" is designed. This logic automatically converts basic JSON data into a format that can be recognized by the visualization, while simultaneously updating the JSON in reverse synchronization with table editing operations, thus achieving seamless integration between data and visualization components without manual intervention.

[0065] 1.4 Intelligent Table Visualization Rendering and Real-time Editing Synchronization Mechanism

[0066] To address the shortcomings of existing technologies, such as cumbersome table editing, asynchronous modifications, and susceptibility to errors, a two-way binding mechanism of "JSON-virtual DOM rendering-visual editing" is implemented. This mechanism quickly renders standard JSON into operable table components, supporting drag-and-drop layout, style modification, parameter adjustment, and other operations. All editing actions are synchronized with the JSON in real time and persistently stored, eliminating the need to regenerate code.

[0067] 1.5 Intelligent Form End-to-End Collaborative Architecture

[0068] To address the shortcomings of existing technologies in table generation processes, such as fragmentation, disconnect between stages, and low efficiency, this paper integrates five major modules: knowledge base management, requirement analysis, JSON conversion, visualization rendering, and real-time editing. This creates an end-to-end collaborative system where users can input natural language commands to automate the entire process from requirement to deployable and editable smart table without human intervention.

[0069] Example 2

[0070] Function Description

[0071] 2.1.1 Prompt Word and Knowledge Base Management Module

[0072] (1) Core functions: Establish a standardized prompt word management system, build a large model-specific recognition knowledge base, and provide support for table-based JSON Schema, component metadata, and semantic mapping of configuration items;

[0073] (2) Technical implementation: A standardized prompt word template engine is adopted, and the existing table architecture JSON data structure, component configuration item dictionary and semantic relationship are entered to build a knowledge base vector index for semantic matching and accurate recognition when the large model is called.

[0074] 2.1.2 Requirements Analysis and JSON Generation Module

[0075] (1) Core function: Receive the user's natural language requirement description, and through intent recognition and entity extraction, parse the requirements such as table structure and table item type to generate advanced intelligent table basic JSON data that conforms to the knowledge base specification;

[0076] (2) Technical implementation: Based on the fine-tuned BERT intent recognition model, combined with knowledge base semantic vector retrieval, the entity parameters in the user's needs are parsed, and the large model is guided by prompt word engineering to generate intelligent table basic JSON data with compliant structure and complete parameters.

[0077] 2.1.3 JSON Conversion Module

[0078] (1) Core function: Parse the basic JSON data generated by the large model and convert it into the standard component JSON format that the visualization canvas can recognize, so as to realize the two-way mapping connection between JSON data and visualization components;

[0079] (2) Technical implementation: Esprima parser is used to convert JSON data into an abstract syntax tree (AST), extract core nodes such as tagName, props, and children of the table component, perform format mapping and validation according to the JSON Schema specification of the component library, and output standard JSON that the canvas can recognize.

[0080] 2.1.4 Visual Canvas Rendering Module

[0081] (1) Core function: Based on the converted standard JSON data, render and generate an interactive real table component, and provide visual interaction support such as canvas drag and drop and component preview;

[0082] (2) Technical implementation: Based on the HTML5 native drag-and-drop API and virtual DOM rendering mechanism, the standard JSON data after conversion is read, and the corresponding table DOM structure is loaded and rendered through the component rendering engine to realize the real-time preview of the component and the interactive response of the canvas.

[0083] 2.1.5 Table Editing Module

[0084] (1) Core functions: Provides visual editing capabilities for table component attributes, styles, and layouts, enables real-time synchronization of operations and JSON data, and optimizes the final presentation of the table;

[0085] (2) Technical implementation: A two-way data binding mechanism is adopted to support adaptive adjustment of table layout, online debugging of component style, and real-time modification of configuration parameters. All operations are updated in real time through the data synchronization engine to update the table JSON data and DOM rendering, and finally generate the target intelligent table that meets the requirements and persist it.

[0086] 2.2 Table JSON Configuration

[0087] 2.2.1 Attribute Configuration

[0088] "attribute": {

[0089] "border": false, / / Border property

[0090] "pagination": true, / / Pagination attribute

[0091] "showHeader": true, / / Display header attributes

[0092] "compSize": "superLarge", / / Component size property

[0093] "paginationValue": "[10, 20, 50, 100]", / / Pagination configuration property

[0094] "stripe": false, / / Zebra stripe attribute

[0095] "lineHeight": "35", / / Line height attribute

[0096] "height": "400" / / Table content height attribute

[0097] }

[0098] 2.2.3 Table Item Configuration Properties

[0099] "tableColumnData": [

[0100] {

[0101] "showOverflowTooltip": true, / / Whether to display overflow warnings for table items

[0102] "isSortable": false, / / Whether table items are sorted

[0103] "minWidth": "", / / Minimum width of table items

[0104] "label": "Date", / / Table item label name

[0105] "align": "left", / / Table item alignment

[0106] "headerColor": "", / / Table header color

[0107] "mergeable": false, / / Whether to merge table headers

[0108] "headerFontSize": "", / / Table header font size

[0109] "prop": "", / / Dynamic data binding field for table items

[0110] "valueType": "", / / Table item filter type

[0111] "width": "", / / Table item width

[0112] "columnFontSize": "", / / Font size of table item content

[0113] "fixed": false, / / Whether table items are fixed

[0114] "staticProp": "date", / / Static data binding field for table item

[0115] "columnColor": "", / / Color of table column content

[0116] "filedType": "", / / Field type bound to table item content

[0117] "filterData": [], / / Collection of table item content filtering fields

[0118] "configData": { / / Collection of table item content field configurations

[0119] "imageWidth": "", / / Width of table item images

[0120] "imageHeight": "", / / Image height of table item content

[0121] "imageUrl": "", / / Image content for table items

[0122] "showFilePreview": false, / / Whether to display a document preview of the table field content.

[0123] "showFileDownload": false, / / Whether to display document downloads in the table field

[0124] "icon": "", / / Icon configuration for table item content

[0125] "iconColor": "", / / Color of table item icons

[0126] "iconFontSize": "", / / Font size of table item icons

[0127] "insertType": "before", / / Position where table item icons are inserted

[0128] "tagName": "tag", / / The name of the tag configuration for the table item content

[0129] "tagSize": "mini", / / Configure the size of the table item content label

[0130] "tagType": "primary", / / Table item content tag type

[0131] "processStroke": "", / / Configure the height of the progress bar for the table item content

[0132] "processMinValue": "", / / Configure the minimum value for the progress bar of the table item content

[0133] "processMaxValue": "", / / Configure the maximum value for the progress bar of the table item content

[0134] "dynamicImageHeight": "", / / Height of the dynamic image for the table item content

[0135] "dynamicImageWidth": "", / / Width of dynamic image for table items

[0136] "showDefaultValue": true, / / Whether to display default content for table items

[0137] }

[0138] }

[0139] ].

[0140] Example 3

[0141] This embodiment provides an AI-based system for automatically generating complex tables, corresponding to the method described above. The system includes:

[0142] Prompt word and knowledge base management module: It adopts a standardized prompt word template engine, inputs the existing table-based JSON data structure, component configuration item dictionary and semantic relationship, and builds a knowledge base vector index.

[0143] The requirement parsing and JSON generation module: Based on the fine-tuned BERT intent recognition model and combined with semantic vector retrieval from the knowledge base, it parses the entity parameters in the user's requirements and guides the large model to generate intelligent table-based JSON data with compliant structure and complete parameters through prompt word engineering.

[0144] JSON conversion module: Using the Esprima parser, JSON data is converted into AST, core nodes such as tagName, props, and children of the table component are extracted, and format mapping and validation are performed according to the component library's JSON Schema specification to output standard JSON that the canvas can recognize.

[0145] Visual canvas rendering module: Based on the HTML5 native drag-and-drop API and virtual DOM rendering mechanism, it reads the converted standard JSON data, loads and renders the corresponding table DOM structure through the component rendering engine, and realizes real-time component preview and canvas interactive response.

[0146] Table editing module: It adopts a two-way data binding mechanism, supports adaptive adjustment of table layout, online debugging of component styles, and real-time modification of configuration parameters. All operations are updated in real time through the data synchronization engine to update the table JSON data and DOM rendering, and finally generate the target intelligent table that meets the requirements and store it persistently.

[0147] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatically generating complex tables based on AI, characterized in that, Includes the following steps: Step S1: Establish a prompt word management system, which includes a hierarchical and categorized prompt word rule library. The prompt word rule library covers general requirement description specifications, field definition prompt words, layout style prompt words, and business rule prompt words. Step S2: Build a model recognition knowledge base and input table architecture metadata. The table architecture metadata includes technical configuration items such as table container, column configuration, row configuration, cell style, data binding, interaction rules, filtering and sorting, access control and linkage logic. Configure the business meaning, value range, constraints and related dependencies for each configuration item. Step S3: Receive the user's natural language description of their requirements; Step S4: Based on the prompt word management system and the model recognition knowledge base, the large model performs semantic parsing, intent recognition and demand decomposition on the natural language demand description, extracts the core elements of the table, and generates advanced intelligent table basic JSON data that conforms to the platform's technical standards. Step S5: Convert the basic JSON data into a component configuration format that the canvas can recognize, and render and generate an interactive smart table component in the canvas; Step S6: Receive manual adjustment operations from the user on the smart table component in the canvas, and update the table configuration in real time. Step S7: Save the final table configuration and generate the target smart table.

2. The method according to claim 1, characterized in that, In step S4, a fine-tuned BERT intent recognition model is used, combined with semantic vector retrieval from the knowledge base, to parse entity parameters in user needs. The prompt word engineering guides the large model to generate intelligent table basic JSON data with compliant structure and complete parameters.

3. The method according to claim 1, characterized in that, In step S5, converting the basic JSON data into a canvas-recognizable format includes: using the Esprima parser to convert the JSON data into an abstract syntax tree (AST), extracting the label names, attributes, child nodes, and core nodes of the table component, performing format mapping and validation according to the component library JSON Schema specification, and outputting standard JSON that the canvas can recognize.

4. The method according to claim 1, characterized in that, In step S6, a two-way data binding mechanism is adopted to support adaptive adjustment of table layout, online debugging of component styles, and real-time modification of configuration parameters. All operations are updated in real time through the data synchronization engine to update table JSON data and DOM rendering.

5. The method according to claim 1, characterized in that, In step S4, the basic JSON data generated by the large model includes table attribute configuration and table item configuration attributes. The table attribute configuration includes border attributes, pagination attributes, header display attributes, component size attributes, pagination configuration attributes, zebra stripe attributes, row height attributes, and table content height attributes. The table item configuration attributes include whether to display overflow prompts, whether to sort, minimum width, tag name, alignment, header color, whether to merge headers, header font size, dynamic data binding fields, filter type, width, content font size, whether to fix, static data binding fields, content color, content binding field type, content filter field configuration set, and content field configuration set.

6. The method according to claim 5, characterized in that, The content field configuration set includes image width, image height, image URL, whether to display document preview, whether to display document download, icon configuration, icon color, icon font size, icon insertion position, tag configuration name, tag size, tag type, progress bar height, minimum progress bar value, maximum progress bar value, animated image height, animated image width, and whether to display default content.

7. An AI-based system for automatically generating complex tables, characterized in that, include: The prompt word and knowledge base management module is used to establish a standardized prompt word management system and build a dedicated recognition knowledge base for large models, providing support for table-based JSON Schema, component metadata, and semantic mapping of configuration items; The requirement parsing and JSON generation module is used to receive users' natural language requirement descriptions, and through intent recognition and entity extraction, parse the table structure and table item type requirements to generate advanced intelligent table basic JSON data that conforms to the knowledge base specification. The JSON conversion module is used to parse the basic JSON data and convert it into a standard component JSON format that the visualization canvas can recognize, thereby realizing a two-way mapping between JSON data and visualization components. The visual canvas rendering module is used to render interactive, realistic table components based on the converted standard JSON data, providing visual interactive support for canvas dragging and component preview. The table editing module provides visual editing capabilities for table component properties, styles, and layouts, enabling real-time synchronization of operations with JSON data.

8. The system according to claim 7, characterized in that, The requirement parsing and JSON generation module is based on a fine-tuned BERT intent recognition model, combined with semantic vector retrieval from the knowledge base, to parse entity parameters in user requirements. Through prompt word engineering, it guides the large model to generate intelligent table-based JSON data with compliant structure and complete parameters.

9. The system according to claim 7, characterized in that, The JSON conversion module uses the Esprima parser to convert JSON data into an abstract syntax tree (AST), extracts the tag names, attributes, child nodes, and core nodes of the table component, performs format mapping and validation according to the component library's JSON Schema specification, and outputs standard JSON that the canvas can recognize.

10. The system according to claim 7, characterized in that, The table editing module adopts a two-way data binding mechanism, which supports adaptive adjustment of table layout, online debugging of component styles, and real-time modification of configuration parameters. All operations are updated in real time through the data synchronization engine to update the table JSON data and DOM rendering, and finally generate the target intelligent table that meets the requirements and store it persistently.

Images