A method, apparatus, device and medium for intelligent automatic form creation

By combining natural language processing and multi-objective optimization algorithms with semantic parsing and knowledge graphs, forms for low-code platforms are automatically generated, solving the problems of low form creation efficiency, reliance on experience, and lack of standardization in existing technologies, and achieving efficient and accurate form generation.

CN123114466APending Publication Date: 2026-11-17BEIJING ANZHENGTONG INFORMATION TECH HLDG CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611414010.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-11
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing low-code platforms suffer from inefficiency, reliance on user experience, lack of standardization and semantic understanding capabilities during form creation, and an inability to automatically infer form structure and logical relationships, resulting in time-consuming form creation and the potential for unreasonable data structures.

Method used

By integrating natural language processing, business domain knowledge graphs, multi-objective evolutionary optimization algorithms, and semantic parsing models, the system achieves fully automated generation from user request text to a working form system. This includes the application of semantic parsing, graph attention networks, multi-objective optimization, and reinforcement learning models, automatically constructing form layouts and configuring access permissions.

Benefits of technology

It improves the accuracy and efficiency of form generation, realizes the fully automated generation from user requirement text to a working form system, reduces manual operation time, and improves the standardization and accuracy of form design.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114466A_ABST
    Figure CN123114466A_ABST
Patent Text Reader

Abstract

This application provides a method, apparatus, device, and medium for intelligent automatic form creation, comprising: processing requirement description text based on a semantic parsing model; constructing a business semantic graph based on a set of business entities and a set of business relationships; calculating the semantic similarity between nodes in the business semantic graph using a graph attention network; generating a mapping scheme based on the semantic similarity matching results; constructing a multi-objective optimization function for form layout based on the mapping scheme; solving the multi-objective optimization function using an improved algorithm to determine the optimal layout scheme; generating a logical table structure based on the optimal layout scheme; serializing the logical table structure into a standardized form description file; calling a low-code platform interface to complete form creation; and automatically configuring multi-level access permissions based on role information identified in the business requirement text. This achieves fully automated generation from user requirement text to a runnable form system, improving the accuracy and efficiency of form generation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of computer processing technology, and in particular to a method, apparatus, device and medium for intelligent automatic form creation. Background Technology

[0002] Currently, low-code development platforms (LCDPs) have become an important technological tool for enterprise digital transformation. In existing solutions, form creation on LCDPs typically relies on users manually creating forms through a visual designer interface. Users need to select the required control types from the control library one by one, drag and drop the controls onto the design canvas, and configure the control's attribute parameters, validation rules, relationships, and display conditions one by one. However, the above-mentioned existing technologies have the following significant technical problems and defects: First, form creation is inefficient. For complex business scenarios, users need to manually select, drag, and configure all controls, which can take hours or even days to build a form. Second, control selection is highly dependent on user experience. Different controls differ significantly in data binding methods, storage structures (single-value storage vs. multi-value storage), interaction modes, and applicable scenarios, making it difficult to accurately select the optimal control type and easily leading to unreasonable data structures due to incorrect control selection. Third, form structure design lacks standardization. Different users have significantly different form design solutions for similar business scenarios. Fourth, it lacks semantic understanding and intelligent reasoning capabilities. Existing low-code platforms cannot understand users' business requirements expressed in natural language, nor can they automatically infer the set of fields that a form should include, the logical relationships between fields, data validation rules, and business constraints. Therefore, there is an urgent need for an intelligent form creation method that can understand users' natural language requirements and automatically complete the entire process from semantic parsing to form persistence, in order to solve the technical problems of inefficiency, reliance on experience, and lack of standardization in existing solutions. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method, apparatus, device and medium for intelligent automatic form creation, which integrates natural language processing, business domain knowledge graph, multi-objective evolutionary optimization algorithm and semantic parsing model to achieve fully automated generation from user requirement text to a working form system, thereby improving the accuracy and efficiency of form generation.

[0004] This application provides a method for intelligent automatic form creation, the method comprising: Receive user input of requirement description text, extract information from the requirement description text based on a semantic parsing model, and construct a business semantic graph based on the extracted set of business entities and set of business relationships. Based on the graph attention network, a pre-constructed domain knowledge graph and the semantic similarity between nodes of the business semantic graph are calculated. Based on the semantic similarity matching results, a mapping scheme from business entities to form controls is generated. The domain knowledge graph includes business domain type, form control type, and layout mode type. Based on the mapping scheme, a multi-objective optimization function for form layout is constructed, and the improved NSGA-II algorithm is used to solve the multi-objective optimization function to determine the optimal layout scheme. A logical table structure is generated based on the optimal layout scheme. The logical table structure is serialized into a standardized form description file. The low-code platform interface is called to complete the form creation. Multi-level access permissions are automatically configured based on the role information identified in the business requirement text.

[0005] In one possible implementation, the intelligent automatic form creation method further includes: The system continuously collects user modification records of the generated forms, dynamically adjusts the weight parameters of the semantic parsing model and domain knowledge graph using an online learning mechanism, and introduces a reinforcement learning model with user feedback as a reward signal to iteratively optimize the form generation strategy.

[0006] In one possible implementation, the information extraction from the requirement description text based on the semantic parsing model includes: The BERT pre-trained encoder based on the semantic parsing model obtains the context representation vector of each token in the requirement description text through a self-attention mechanism; The context representation vector is input into the BiLSTM sequence encoding layer based on the semantic parsing model. The forward LSTM propagates forward along the sequence to capture the preceding context dependency information, and the backward LSTM propagates backward along the sequence to capture the following context dependency information. The forward and backward hidden states are concatenated to output the feature representation of each position. The feature representation is input into the CRF conditional random field decoding layer based on the semantic parsing model. The legality of adjacent labels is constrained by the transition probability matrix between labels, and the Viterbi algorithm is used to dynamically search for the globally optimal label sequence. Extract the set of business entities based on the token fragments marked as entity type in the globally optimal tag sequence, and extract the set of business relationships based on the dependency paths between token fragments or entity pairs marked as relationship type in the globally optimal tag sequence.

[0007] In one possible implementation, the multi-objective optimization function is: F(X)= w1·U(X)+ w2·C(X)+ w3·S(X); Where F(X) is the multi-objective optimization function, U(X) is the user experience score, C(X) is the control compatibility score, S(X) is the structural standardization score, w1, w2, and w3 are the weight coefficients of each objective, and X is the form layout scheme to be evaluated.

[0008] In one possible implementation, the improved NSGA-II algorithm includes: A hybrid encoding strategy is adopted, which combines integer encoding of control type, real number encoding of layout position, and tree-like encoding of nested relationships. A structure-preserving crossover operator based on subtree swapping while maintaining the integrity of the hierarchical structure; An adaptive mutation probability strategy is adopted to make the mutation probability decrease with the number of generations. Furthermore, an elite retention strategy is adopted to preserve all Pareto front non-dominated solutions, and the TOPSIS method is finally used to select the optimal layout scheme from the Pareto solution set through weighted Euclidean distance.

[0009] In one possible implementation, generating the logical table structure based on the optimal layout scheme includes: The optimal layout scheme is parsed to determine the main table, sub-tables, and related sub-tables; Map the main form fields to the physical columns of the main form logical table, and automatically infer the field data type based on the control type; Map subforms to independent subform logical tables and automatically create parent fields to establish parent-child table relationships; Map associated subforms to associated subform logical tables, and implement many-to-many relationships through intermediate association tables; For multi-value control types, automatically create a multi-value association table containing three columns: record ID, field identifier, and value from the main table, and automatically create a B-Tree index for fields that need to support searching.

[0010] In one possible implementation, the reward function of the reinforcement learning model is: R=α·R_accuracy+β·R_efficiency+γ·R_satisfaction; Where R_accuracy is the field matching accuracy reward, R_efficiency is the creation efficiency improvement reward, R is the reward signal, R_satisfaction is the user satisfaction reward, and α, β, and γ are reward weight coefficients.

[0011] This application embodiment also provides a form intelligent automatic creation device, the form intelligent automatic creation device comprising: The semantic parsing module is used to receive user input requirement description text, extract information from the requirement description text based on the semantic parsing model, and construct a business semantic graph based on the extracted set of business entities and set of business relationships. The knowledge graph reasoning module is used to calculate the semantic similarity between nodes of a pre-built domain knowledge graph and the business semantic graph based on a graph attention network, and to generate a mapping scheme from business entities to form controls based on the semantic similarity matching results; wherein, the domain knowledge graph includes business domain type, form control type and layout mode type; The layout optimization module is used to construct a multi-objective optimization function for the form layout based on the mapping scheme, and to solve the multi-objective optimization function using the improved NSGA-II algorithm to determine the optimal layout scheme. The form generation module is used to generate a logical table structure based on the optimal layout scheme, serialize the logical table structure into a standardized form description file, call the low-code platform interface to complete the form creation, and automatically configure multi-level access permissions based on the role information identified in the business requirement text.

[0012] This application also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the form intelligent automatic creation method described above are performed.

[0013] This application also provides a computer-readable storage medium storing a computer program, which, when run by a processor, executes the steps of the form intelligent automatic creation method described above.

[0014] This application provides a method, apparatus, device, and medium for intelligent automatic form creation. The method includes: receiving user-inputted requirement description text; extracting information from the requirement description text based on a semantic parsing model; constructing a business semantic graph based on the extracted set of business entities and set of business relationships; calculating the semantic similarity between nodes in a pre-constructed domain knowledge graph and the business semantic graph based on a graph attention network; generating a mapping scheme from business entities to form controls based on the semantic similarity matching results; wherein the domain knowledge graph includes business domain type, form control type, and layout mode type; constructing a multi-objective optimization function for form layout based on the mapping scheme; solving the multi-objective optimization function using an improved NSGA-II algorithm to determine the optimal layout scheme; generating a logical table structure based on the optimal layout scheme; serializing the logical table structure into a standardized form description file; calling a low-code platform interface to complete form creation; and automatically configuring multi-level access permissions based on role information identified in the business requirement text. By integrating natural language processing, business domain knowledge graphs, multi-objective evolutionary optimization algorithms, and semantic parsing models, we can achieve fully automated generation from user requirement text to a working form system, thereby improving the accuracy and efficiency of form generation.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating an intelligent automatic form creation method provided in this application embodiment; Figure 2 This is one of the structural schematic diagrams of an intelligent automatic form creation device provided in an embodiment of this application; Figure 3 This is a second schematic diagram of the structure of a form intelligent automatic creation device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0019] First, the applicable scenarios for this application will be introduced. This application can be applied to the field of computer processing technology.

[0020] Research has revealed the following significant technical problems and shortcomings in existing technologies: First, form creation is inefficient. For complex business scenarios, users need to manually select, drag, and configure all controls, which can take hours or even days to build a form. Second, control selection is highly dependent on user experience. Different controls differ significantly in data binding methods, storage structures (single-value vs. multi-value), interaction modes, and applicable scenarios, making it difficult to accurately select the optimal control type and prone to data structure inconsistencies due to incorrect control selection. Third, form structure design lacks standardization. Different users have significantly different form design solutions for similar business scenarios. Fourth, there is a lack of semantic understanding and intelligent reasoning capabilities. Existing low-code platforms cannot understand users' business requirements expressed in natural language, nor can they automatically infer the set of fields that a form should include, the logical relationships between fields, data validation rules, and business constraints. In conclusion, there is an urgent need for an intelligent form creation method that can understand users' natural language requirements and automatically complete the entire process from semantic parsing to form persistence to solve the technical problems of inefficiency, reliance on experience, and lack of standardization in existing solutions.

[0021] Based on this, embodiments of this application provide a method, apparatus, device, and medium for intelligent automatic form creation. By integrating natural language processing, business domain knowledge graphs, multi-objective evolutionary optimization algorithms, and semantic parsing models, it achieves fully automated generation from user requirement text to a working form system, thereby improving the accuracy and efficiency of form generation.

[0022] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for intelligent automatic form creation provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the intelligent automatic form creation method includes: S101: Receive the user input requirement description text, extract information from the requirement description text based on the semantic parsing model, and construct a business semantic graph based on the extracted set of business entities and set of business relationships.

[0023] In this step, a semantic parsing model is used to extract the set of business entities and the set of business relationships from the requirement description text, and a business semantic graph is constructed based on the set of business entities and the set of business relationships.

[0024] Here, users can enter a description of their requirements for a target form in natural language through the platform's intelligent form creation portal. This description includes information such as the required business fields, the logical relationships between fields, and business process requirements. The description text undergoes preprocessing, including: Chinese word segmentation based on a word segmentation algorithm; identification of key business entities in the text using Named Entity Recognition (NER) technology; and parsing of the sentence's syntactic structure and semantic relationships using dependency parsing.

[0025] It should be noted that the semantic parsing model is the BERT-BiLSTM-CRF architecture.

[0026] In one possible implementation, the information extraction from the requirement description text based on the semantic parsing model includes: A: The BERT pre-trained encoder based on the semantic parsing model obtains the context representation vector of each token in the requirement description text through a self-attention mechanism.

[0027] Here, the BERT pre-trained encoder layer uses a BERT model pre-trained on a large-scale Chinese corpus as the text encoder, and obtains the context representation vector of each token in the input text through a self-attention mechanism.

[0028] B: Input the context representation vector into the BiLSTM sequence encoding layer based on the semantic parsing model. The forward LSTM propagates forward along the sequence to capture the preceding context dependency information, and the backward LSTM propagates backward along the sequence to capture the following context dependency information. The forward and backward hidden states are concatenated to output the feature representation of each position.

[0029] Here, the BiLSTM sequence encoding layer stacks a bidirectional long short-term memory network on top of the BERT encoding output. The forward LSTM captures the positive contextual dependencies of the sequence, and the backward LSTM captures the negative contextual dependencies. The forward and backward hidden states are concatenated to output the feature representation of each position.

[0030] C: Input the feature representation into the CRF conditional random field decoding layer based on the semantic parsing model, use the transition probability matrix between labels to constrain the legality of adjacent labels, and use the Viterbi algorithm to dynamically search for the globally optimal label sequence.

[0031] Here, the CRF (Conditional Random Field) decoding layer models the transition probability matrix between labels and uses the Viterbi decoding algorithm to ensure the global optimality of the entity label sequence and the relationship label sequence, thus avoiding the label inconsistency problem caused by independent classification.

[0032] D: Extract the set of business entities based on the token fragments marked as entity type in the global optimal tag sequence, and extract the set of business relationships based on the dependency paths between token fragments or entity pairs marked as relationship type in the global optimal tag sequence.

[0033] Here, based on the extracted set of business entities and set of business relationships, a business semantic graph G=(V,E,R) is constructed, where V is the set of business entity nodes, E is the set of edges between entities, and R is the set of business relationships.

[0034] This application employs a three-layer semantic parsing model based on BERT-BiLSTM-CRF architecture. Through the deep context representation of the BERT pre-trained encoder layer, the bidirectional context dependency capture of the BiLSTM sequence encoding layer, and the globally optimal label sequence decoding of the CRF conditional random field decoding layer, these three elements work synergistically to significantly improve the understanding of business domain terms and complex semantic structures. The CRF layer, by modeling the transition probability matrix between labels and employing the Viterbi decoding algorithm, avoids the label inconsistency problem caused by independent classification, ensuring the global optimality of entity label sequences and relation label sequences.

[0035] S102: Calculate the semantic similarity between nodes of the pre-built domain knowledge graph and the business semantic graph based on the graph attention network, and generate a mapping scheme from business entities to form controls based on the semantic similarity matching results; wherein, the domain knowledge graph includes business domain type, form control type and layout mode type.

[0036] In this step, the semantic similarity between nodes of the pre-built domain knowledge graph and the business semantic graph is calculated based on the graph attention network, and a mapping scheme from business entities to form controls is generated based on the semantic similarity matching results.

[0037] Here, a domain knowledge graph for low-code forms is constructed using a three-layer architecture: The first layer is the business domain ontology layer, which defines the standard entities and relationship patterns for each business domain. For example, the ontology for contract management includes core concepts such as contract subject, signatory, amount, and term, and their relationships; the ontology for human resources management includes core concepts such as employee, department, position, and job level, and their relationships. The business domain ontology layer is constructed and continuously updated through domain expert annotation and historical form data mining. The second layer is the form control mapping layer, which establishes a semantic mapping rule library from business entity types to form control types. Core mapping relationships include, but are not limited to: personnel information entity → member single-select / member control; department information entity → department single-select / department control; amount / quantity entity → numeric control; date / time entity → date / date range / time control; address information entity → address control; description / remark entity → multi-line text / rich text control; status / type entity → single-select / multi-select control; image / file entity → image / attachment control; nested data entity → sub-form control, etc. The mapping rule library also includes recommended rules for control attributes, such as decimal precision for numeric controls, format templates for date controls, and option enumerations for radio buttons. The third layer, the layout pattern library, stores validated and commonly used form layout patterns, including single-column standard layout, two-column grid layout, multi-column grid layout, tab grouping layout, nested subform layout, and partitioned layout. Each layout pattern comes with a description of applicable scenarios and control compatibility constraints.

[0038] It should be noted that the control types of the domain knowledge graph include: single-line text, multi-line text, numeric, date, date range, multi-select, single-select, slider, text, time, time range, switch, image, attachment, address, rich text, separator, grid container, grid cell, table container, table cell, table row, tab, tab label, member single-select, member, department single-select, department, collaborator, collaborating enterprise, related data, related query, subform, template control, plug-in, parent control, cascading control, role control, job control, automatic counting, and related subform.

[0039] Specifically, a TransE-based knowledge graph embedding algorithm is employed to map the business semantic graph and the domain knowledge graph to a unified low-dimensional vector space. A graph attention network is used to calculate the semantic similarity between nodes in the business semantic graph and nodes in the domain knowledge graph. The core of the graph attention network lies in the calculation of multi-head attention weights, using the following formula: α_ij=softmax(LeakyReLU(a^T[W·h_i‖W·h_j])); The formula for multi-head output splicing is: h'_i=‖_{k=1}^{K}σ(Σ_{j∈N(i)}α_ij^kW^kh_j); Where α_ij is the attention weight of node i to node j, W is the learnable linear transformation weight matrix, a is the attention parameter vector, h_i and h_j are the feature vectors of nodes i and j respectively, K is the number of attention heads, ‖ denotes vector concatenation operation, and σ is a non-linear activation function. h'_i is the final feature vector output by node i after the multi-head attention mechanism. Based on the attention weight calculation results, a mapping scheme from business entities to form controls is obtained through threshold filtering (setting a minimum similarity threshold θ) and a Top-K ranking strategy.

[0040] This application constructs a three-layer architecture for a low-code form domain knowledge graph. It maps the business semantic graph and the domain knowledge graph to a unified low-dimensional vector space using the TransE knowledge graph embedding algorithm. Then, it calculates the semantic similarity between nodes using a graph attention network. Finally, it obtains the optimal mapping scheme through threshold filtering and a Top-K ranking strategy. The TransE algorithm models the relationships in the knowledge graph as translation operations between entity vectors, and the graph attention network captures complex semantic relationships between nodes through a multi-head attention mechanism. The organic combination of these two approaches enables highly accurate and generalizable control mapping across domains and scenarios.

[0041] S103: Construct a multi-objective optimization function for the form layout based on the mapping scheme, solve the multi-objective optimization function using the improved NSGA-II algorithm, and determine the optimal layout scheme.

[0042] In this step, a multi-objective optimization function for the form layout is constructed based on the mapping scheme, and the improved NSGA-II algorithm is used to solve the multi-objective optimization function. The comprehensive optimal layout scheme is selected from the Pareto optimal solution set obtained by the NSGA-II algorithm.

[0043] In one possible implementation, the multi-objective optimization function is: F(X)= w1·U(X)+ w2·C(X)+ w3·S(X); Where F(X) is the multi-objective optimization function, U(X) is the user experience score, C(X) is the control compatibility score, S(X) is the structural standardization score, w1, w2, and w3 are the weight coefficients of each objective, and X is the form layout scheme to be evaluated.

[0044] It should be noted that the user experience score U(X) is composed of three sub-indicators: visual density, information hierarchy, and operation process length; the control compatibility score C(X) is evaluated based on whether the selected control type meets its corresponding data storage mechanism and association constraints; and the structural standardization score S(X) is calculated by comparing the field set, control type distribution, and layout structure matching degree of the generated form with the preset domain standard template.

[0045] In one possible implementation, the improved NSGA-II algorithm includes: a: A hybrid encoding strategy is adopted, which combines integer encoding of control type, real number encoding of layout position, and tree-like encoding of nested relationship.

[0046] Here, the control type uses integer encoding, the layout position uses real number encoding, and the nesting relationship uses tree encoding (representing the hierarchical structure of parent and child controls).

[0047] b: Structure-preserving crossover operator based on subtree swapping while maintaining the integrity of the hierarchical structure.

[0048] Here, a cross operation based on subtree swapping is designed to swap the child forms or layout blocks of two parent individuals while maintaining the integrity of the form hierarchy.

[0049] c: Adopt an adaptive mutation probability strategy so that the mutation probability decreases with the number of generations.

[0050] Here, two mutation operations are designed: control type mutation and layout position mutation. The mutation probability adopts an adaptive strategy, decreasing with the number of generations to balance global exploration and local development capabilities.

[0051] d: In addition, an elite retention strategy is adopted to preserve all Pareto front non-dominated solutions, and the TOPSIS method is finally used to select the optimal layout scheme from the Pareto solution set through weighted Euclidean distance.

[0052] Here, in each generation of evolution, all non-dominated solutions on the Pareto front are retained to ensure that the optimal solution is not lost during the evolution process. From the Pareto optimal solution set obtained by the improved NSGA-II algorithm, the TOPSIS (Topology for Approximating Ideal Solutions) method is used to select the overall optimal layout scheme. The TOPSIS method calculates the weighted Euclidean distance from each Pareto optimal solution to the positive and negative ideal solutions, and selects the solution with the highest relative proximity as the optimal layout scheme.

[0053] This application models the form layout problem as a multi-objective optimization problem, constructs a comprehensive optimization function that includes user experience score, control compatibility score, and structural standardization score, and uses an improved NSGA-II algorithm for solving it. The improved NSGA-II algorithm effectively balances global exploration and local development capabilities while maintaining the integrity of the form's hierarchical structure, and ensures that the optimal solution is not lost during the evolution process through an elite retention strategy. Finally, the TOPSIS method is used to select the comprehensive optimal layout scheme from the Pareto optimal solution set, overcoming the shortcomings of traditional weighted methods, such as difficulty in determining weighting coefficients and inability to guarantee simultaneous optimization of multiple objectives.

[0054] S104: Generate a logical table structure based on the optimal layout scheme, serialize the logical table structure into a standardized form description file, call the low-code platform interface to complete the form creation, and automatically configure multi-level access permissions based on the role information identified in the business requirement text.

[0055] In this step, a logical table structure is automatically generated based on the optimal layout scheme. The logical table structure is then converted into a JSON-formatted form description file. The form engine API of the low-code platform is called to complete the persistent storage of the form. Based on the role information identified in the business requirement text, application-level permissions, view-level permissions, and field-level permissions under the multi-tenant permission system are automatically configured.

[0056] In one possible implementation, generating the logical table structure based on the optimal layout scheme includes: The optimal layout scheme is parsed to determine the main table, sub-tables, and related sub-tables. Main table fields are mapped to physical columns of the main table's logical table, and the data type of the fields is automatically inferred based on the control type. Sub-tables are mapped to independent logical tables, and parent fields are automatically created to establish parent-child table relationships. Related sub-tables are mapped to related logical tables, and many-to-many relationships are achieved through intermediate related tables. For multi-value control types, a multi-value related table containing three columns—main table record ID, field identifier, and value—is automatically created, and B-Tree indexes are automatically created for fields that need to support searching.

[0057] Here, the main form field mapping maps all fields at the main form level to physical columns in the main form logical table. Each field corresponds to a physical column, and the column name is automatically generated by the system based on the field identifier. Subform control mapping maps subform controls to independent subform logical tables, automatically creating parent-related fields and establishing parent-child relationships between sub-table records and main table records. Related subform mapping maps related subform controls to related subform logical tables, achieving many-to-many data associations through intermediate related tables. The data type of the field is automatically inferred based on the control type, with the following inference rules: text controls correspond to STRING type, numeric controls (number) correspond to DECIMAL type, with a default precision of 2 decimal places. Auto-counting controls correspond to INTEGER type, automatically incrementing. Date controls (date / daterange) correspond to DATETIME type; time controls correspond to TIME type; personnel controls correspond to USER type; department controls (department / departmentSingle) correspond to DEPT type; multiple / linked data (multiple / linkData) correspond to JSON array type. Multi-value field handling: For multi-value control types, a multi-value association table is automatically created. This table contains three core columns: the main table record ID, the field identifier, and the value, supporting one-to-many data storage requirements. Automatic index strategy generation: B-Tree indexes are automatically created for fields that need to support searching.

[0058] The process of calling the low-code platform interface to create and persist the form includes: (a) creating form entries, associating the form with the specified application, and setting the form name and description; (b) initializing the default view, setting the view name and sorting number, and configuring the view display settings (including column visibility, sorting rules, and filtering conditions); (c) creating logical tables and logical columns, and establishing the mapping relationship between form fields and database physical columns; (d) configuring the data source and dataset to support CRUD operations on form data; and (e) completing the initial configuration of the dynamic form, including data validation rules, field linkage relationships, and form submission processing logic.

[0059] Here, based on the platform's multi-tenant permission system, access permissions for forms are automatically configured. Permission configuration includes: application-level permissions: configuring access permissions for different roles to the form application; view-level permissions: controlling the view scope visible to different user groups; field-level permissions: setting read and write permissions for sensitive fields.

[0060] In one possible implementation, the intelligent automatic form creation method further includes: The system continuously collects user modification records of the generated forms, dynamically adjusts the weight parameters of the semantic parsing model and domain knowledge graph using an online learning mechanism, and introduces a reinforcement learning model with user feedback as a reward signal to iteratively optimize the form generation strategy.

[0061] Here, subsequent user modifications to automatically generated forms are recorded, constructing a modification log dataset. The log content includes: field information added / deleted / modified by the user, control type changes, layout adjustments, attribute parameter modifications, etc. Each log record includes a timestamp, user ID, operation type, and operation details. An online learning strategy is employed, continuously updating the mapping weights of the semantic parsing model and knowledge graph using the modification log dataset. Specifically, when the mapping from a business entity to a control type is frequently modified by the user, the system automatically reduces the confidence weight of that mapping and learns new mapping rules based on the user's modification behavior.

[0062] In one possible implementation, the reward function of the reinforcement learning model is: R=α·R_accuracy+β·R_efficiency+γ·R_satisfaction; Where R_accuracy is the field matching accuracy reward, R_efficiency is the creation efficiency improvement reward, R is the reward signal, R_satisfaction is the user satisfaction reward, and α, β, and γ are reward weight coefficients.

[0063] It should be noted that R_accuracy measures the similarity between the automatically generated set of fields and the set of fields finally confirmed by the user; R_efficiency measures the reduction in time taken by automatic generation compared to manual creation; and R_satisfaction is indirectly obtained through user ratings or modifications after the form is generated.

[0064] This application provides an intelligent automatic form creation method, which includes: receiving user-input requirement description text; extracting information from the requirement description text based on a semantic parsing model; constructing a business semantic graph based on the extracted set of business entities and set of business relationships; calculating the semantic similarity between nodes in a pre-built domain knowledge graph and the business semantic graph based on a graph attention network; generating a mapping scheme from business entities to form controls based on the semantic similarity matching results; wherein the domain knowledge graph includes business domain type, form control type, and layout mode type; constructing a multi-objective optimization function for form layout based on the mapping scheme; solving the multi-objective optimization function using an improved NSGA-II algorithm to determine the optimal layout scheme; generating a logical table structure based on the optimal layout scheme; serializing the logical table structure into a standardized form description file; calling a low-code platform interface to complete the form creation; and automatically configuring multi-level access permissions based on role information identified in the business requirement text. By integrating natural language processing, business domain knowledge graph, multi-objective evolutionary optimization algorithm, and semantic parsing model, fully automated generation from user requirement text to a working form system is achieved, improving the accuracy and efficiency of form generation.

[0065] Please see Figure 2 , Figure 3 , Figure 2 This is one of the structural schematic diagrams of an intelligent automatic form creation device provided in an embodiment of this application; Figure 3 This is a second schematic diagram of a form intelligent automatic creation device provided in an embodiment of this application. Figure 2 As shown, the intelligent automatic form creation device 200 includes: Semantic parsing module 210 is used to receive user input requirement description text, extract information from the requirement description text based on the semantic parsing model, and construct a business semantic graph based on the extracted set of business entities and set of business relationships. The knowledge graph reasoning module 220 is used to calculate the semantic similarity between nodes of a pre-built domain knowledge graph and the business semantic graph based on a graph attention network, and to generate a mapping scheme from business entities to form controls based on the matching results of semantic similarity; wherein, the domain knowledge graph includes business domain type, form control type and layout mode type; The layout optimization module 230 is used to construct a multi-objective optimization function for the form layout based on the mapping scheme, solve the multi-objective optimization function using the improved NSGA-II algorithm, and determine the optimal layout scheme. The form generation module 240 is used to generate a logical table structure based on the optimal layout scheme, serialize the logical table structure into a standardized form description file, call the low-code platform interface to complete the form creation, and automatically configure multi-level access permissions based on the role information identified in the business requirement text.

[0066] Furthermore, such as Figure 3 As shown, the intelligent automatic form creation device 200 also includes a feedback learning module 250, which is used for: The system continuously collects user modification records of the generated forms, dynamically adjusts the weight parameters of the semantic parsing model and domain knowledge graph using an online learning mechanism, and introduces a reinforcement learning model with user feedback as a reward signal to iteratively optimize the form generation strategy.

[0067] Furthermore, the semantic parsing module 210 is used to extract information from the requirement description text based on the semantic parsing model: The BERT pre-trained encoder based on the semantic parsing model obtains the context representation vector of each token in the requirement description text through a self-attention mechanism; The context representation vector is input into the BiLSTM sequence encoding layer based on the semantic parsing model. The forward LSTM propagates forward along the sequence to capture the preceding context dependency information, and the backward LSTM propagates backward along the sequence to capture the following context dependency information. The forward and backward hidden states are concatenated to output the feature representation of each position. The feature representation is input into the CRF conditional random field decoding layer based on the semantic parsing model. The legality of adjacent labels is constrained by the transition probability matrix between labels, and the Viterbi algorithm is used to dynamically search for the globally optimal label sequence. Extract the set of business entities based on the token fragments marked as entity type in the globally optimal tag sequence, and extract the set of business relationships based on the dependency paths between token fragments or entity pairs marked as relationship type in the globally optimal tag sequence.

[0068] Furthermore, the layout optimization module 230 uses the following multi-objective optimization function: F(X)= w1·U(X)+ w2·C(X)+ w3·S(X); Where F(X) is the multi-objective optimization function, U(X) is the user experience score, C(X) is the control compatibility score, S(X) is the structural standardization score, w1, w2, and w3 are the weight coefficients of each objective, and X is the form layout scheme to be evaluated.

[0069] Furthermore, the layout optimization module 230 for the improved NSGA-II algorithm includes: A hybrid encoding strategy is adopted, which combines integer encoding of control type, real number encoding of layout position, and tree-like encoding of nested relationships. A structure-preserving crossover operator based on subtree swapping while maintaining the integrity of the hierarchical structure; An adaptive mutation probability strategy is adopted to make the mutation probability decrease with the number of generations. Furthermore, an elite retention strategy is adopted to preserve all Pareto front non-dominated solutions, and the TOPSIS method is finally used to select the optimal layout scheme from the Pareto solution set through weighted Euclidean distance.

[0070] Furthermore, the form generation module 240 is used to generate the logical table structure based on the optimal layout scheme: The optimal layout scheme is parsed to determine the main table, sub-tables, and related sub-tables; Map the main form fields to the physical columns of the main form logical table, and automatically infer the field data type based on the control type; Map subforms to independent subform logical tables and automatically create parent fields to establish parent-child table relationships; Map associated subforms to associated subform logical tables, and implement many-to-many relationships through intermediate association tables; For multi-value control types, automatically create a multi-value association table containing three columns: record ID, field identifier, and value from the main table, and automatically create a B-Tree index for fields that need to support searching.

[0071] Furthermore, the reward function for the reinforcement learning model used by the feedback learning module 250 is: R=α·R_accuracy+β·R_efficiency+γ·R_satisfaction; Where R_accuracy is the field matching accuracy reward, R_efficiency is the creation efficiency improvement reward, R is the reward signal, R_satisfaction is the user satisfaction reward, and α, β, and γ are reward weight coefficients.

[0072] This application provides an intelligent automatic form creation device, comprising: a semantic parsing module for receiving user-input requirement description text, extracting information from the requirement description text based on a semantic parsing model, and constructing a business semantic graph based on the extracted set of business entities and set of business relationships; a knowledge graph reasoning module for calculating the semantic similarity between nodes of a pre-constructed domain knowledge graph and the business semantic graph based on a graph attention network, and generating a mapping scheme from business entities to form controls based on the semantic similarity matching results; wherein the domain knowledge graph includes business domain type, form control type, and layout mode type; a layout optimization module for constructing a multi-objective optimization function for form layout based on the mapping scheme, solving the multi-objective optimization function using an improved NSGA-II algorithm, and determining the optimal layout scheme; and a form generation module for generating a logical table structure based on the optimal layout scheme, serializing the logical table structure into a standardized form description file, calling a low-code platform interface to complete the form creation, and automatically configuring multi-level access permissions based on role information identified in the business requirement text. By integrating natural language processing, business domain knowledge graphs, multi-objective evolutionary optimization algorithms, and semantic parsing models, we can achieve fully automated generation from user requirement text to a working form system, thereby improving the accuracy and efficiency of form generation.

[0073] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.

[0074] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, they can perform the operations described above. Figure 1 The steps of the form intelligent automatic creation method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0075] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 The steps of the form intelligent automatic creation method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0076] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0077] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0078] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0080] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0081] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent automatic form creation, characterized in that, The intelligent automatic form creation method includes: Receive user input of requirement description text, extract information from the requirement description text based on a semantic parsing model, and construct a business semantic graph based on the extracted set of business entities and set of business relationships. Based on the graph attention network, a pre-constructed domain knowledge graph and the semantic similarity between nodes of the business semantic graph are calculated. Based on the semantic similarity matching results, a mapping scheme from business entities to form controls is generated. The domain knowledge graph includes business domain type, form control type, and layout mode type. Based on the mapping scheme, a multi-objective optimization function for form layout is constructed, and the improved NSGA-II algorithm is used to solve the multi-objective optimization function to determine the optimal layout scheme. A logical table structure is generated based on the optimal layout scheme. The logical table structure is serialized into a standardized form description file. The low-code platform interface is called to complete the form creation. Multi-level access permissions are automatically configured based on the role information identified in the business requirement text.

2. The intelligent automatic form creation method according to claim 1, characterized in that, The intelligent automatic form creation method also includes: The system continuously collects user modification records of the generated forms, dynamically adjusts the weight parameters of the semantic parsing model and domain knowledge graph using an online learning mechanism, and introduces a reinforcement learning model with user feedback as a reward signal to iteratively optimize the form generation strategy.

3. The intelligent automatic form creation method according to claim 1, characterized in that, The information extraction from the requirement description text based on the semantic parsing model includes: The BERT pre-trained encoder based on the semantic parsing model obtains the context representation vector of each token in the requirement description text through a self-attention mechanism; The context representation vector is input into the BiLSTM sequence encoding layer based on the semantic parsing model. The forward LSTM propagates forward along the sequence to capture the preceding context dependency information, and the backward LSTM propagates backward along the sequence to capture the following context dependency information. The forward and backward hidden states are concatenated to output the feature representation of each position. The feature representation is input into the CRF conditional random field decoding layer based on the semantic parsing model. The legality of adjacent labels is constrained by the transition probability matrix between labels. The Viterbi algorithm is used to dynamically search for the globally optimal label sequence. Extract the set of business entities based on the token fragments marked as entity type in the globally optimal tag sequence, and extract the set of business relationships based on the dependency paths between token fragments or entity pairs marked as relationship type in the globally optimal tag sequence.

4. The intelligent automatic form creation method according to claim 1, characterized in that, The multi-objective optimization function is: F(X)= w1·U(X)+ w2·C(X)+ w3·S(X); Where F(X) is the multi-objective optimization function, U(X) is the user experience score, C(X) is the control compatibility score, S(X) is the structural standardization score, w1, w2, and w3 are the weight coefficients of each objective, and X is the form layout scheme to be evaluated.

5. The intelligent automatic form creation method according to claim 1, characterized in that, The improved NSGA-II algorithm includes: A hybrid encoding strategy is adopted, which combines integer encoding of control type, real number encoding of layout position, and tree-like encoding of nested relationships. A structure-preserving crossover operator based on subtree swapping while maintaining the integrity of the hierarchical structure; An adaptive mutation probability strategy is adopted to make the mutation probability decrease with the number of generations. Furthermore, an elite retention strategy is adopted to preserve all Pareto front non-dominated solutions, and the TOPSIS method is finally used to select the optimal layout scheme from the Pareto solution set through weighted Euclidean distance.

6. The intelligent automatic form creation method according to claim 1, characterized in that, The generation of the logical table structure based on the optimal layout scheme includes: The optimal layout scheme is parsed to determine the main table, sub-tables, and related sub-tables; Map the main form fields to the physical columns of the main form logical table, and automatically infer the field data type based on the control type; Map subforms to independent subform logical tables and automatically create parent fields to establish parent-child table relationships; Map associated subforms to associated subform logical tables, and implement many-to-many relationships through intermediate association tables; For multi-value control types, automatically create a multi-value association table containing three columns: record ID, field identifier, and value from the main table, and automatically create a B-Tree index for fields that need to support searching.

7. The intelligent automatic form creation method according to claim 2, characterized in that, The reward function of the reinforcement learning model is: R=α·R_accuracy+β·R_efficiency+γ·R_satisfaction; Where R_accuracy is the field matching accuracy reward, R_efficiency is the creation efficiency improvement reward, R is the reward signal, R_satisfaction is the user satisfaction reward, and α, β, and γ are reward weight coefficients.

8. A form intelligent automatic creation device, characterized in that, The intelligent automatic form creation device includes: The semantic parsing module is used to receive user input requirement description text, extract information from the requirement description text based on the semantic parsing model, and construct a business semantic graph based on the extracted set of business entities and set of business relationships. The knowledge graph reasoning module is used to calculate the semantic similarity between nodes of a pre-built domain knowledge graph and the business semantic graph based on a graph attention network, and to generate a mapping scheme from business entities to form controls based on the semantic similarity matching results; wherein, the domain knowledge graph includes business domain type, form control type and layout mode type; The layout optimization module is used to construct a multi-objective optimization function for the form layout based on the mapping scheme, and to solve the multi-objective optimization function using the improved NSGA-II algorithm to determine the optimal layout scheme. The form generation module is used to generate a logical table structure based on the optimal layout scheme, serialize the logical table structure into a standardized form description file, call the low-code platform interface to complete the form creation, and automatically configure multi-level access permissions based on the role information identified in the business requirement text.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the form intelligent automatic creation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the form intelligent automatic creation method as described in any one of claims 1 to 7.