Form generation method, device and system based on AI enhancement and storage medium

Abstract

The embodiment of the invention provides a form generation method, device and system based on AI enhancement and a storage medium, and relates to the technical field of form generation. The method comprises the steps of generating form structure metadata according to a natural language processing technology, a reinforcement learning algorithm and / or a type security verification rule; according to the form structure metadata, the equipment sensor data and / or the user visual behavior data, the rendering priority of each field is determined; and generating a target form according to the rendering priority and the equipment parameters. By generating the form structure metadata which is reasonable in structure and conforms to the type safety specification, the automation degree and reliability of form design are improved. By fusing form structure metadata, device sensor data and user visual behavior data, a rendering strategy can reflect device characteristics and user attention distribution. The target form is generated by combining the rendering priority with the equipment parameters, so that the form layout is adaptively matched with the terminal equipment, and the loading performance and the interaction experience of the form are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of form generation, in particular, to an AI-enhanced form generation method and device, system and storage medium. BACKGROUND

[0002] At present, in the current digital office and business processing scene, as an important tool for data collection and interaction, forms are widely used in information collection, process approval and user feedback and other business scenarios. The traditional form design method usually relies on manual configuration, which not only consumes time and effort, but also is difficult to quickly respond and flexibly adjust in the face of complex and variable business needs, resulting in low form design efficiency and single style, which cannot meet the growing demand of users for personalized and intelligent forms.

[0003] To improve the efficiency and flexibility of form generation, related technologies realize automatic arrangement and verification of part of form fields by predefining form templates or based on rule engines, or use graphical interfaces to assist users in dragging components to generate form structures.

[0004] In the process of implementing the embodiments of the present application, it is found that at least the following problems exist in the related art:

[0005] The related technology reduces the workload of manual configuration to some extent and improves the speed of form construction. However, in actual application, the form structure generated by the related technology lacks dynamic optimization capability and cannot adaptively adjust the rendering order and display strategy of the fields according to the actual use scene, device characteristics or user behavior, resulting in low form rendering efficiency and long user operation path, which is more obvious in a multi-terminal device environment, seriously affecting the interactive experience and practical performance of the form. SUMMARY

[0006] The present application provides an AI-enhanced form generation method and device, system and storage medium.

[0007] The first aspect of the embodiments of the present application provides an AI-enhanced form generation method, comprising:

[0008] generating form structure metadata according to natural language processing technology, reinforcement learning algorithm and / or type safety verification rule;

[0009] determining the rendering priority of each field according to the form structure metadata, device sensor data and / or user visual behavior data;

[0010] generating a target form according to the rendering priority and device parameters.

[0011] In an optional embodiment of the present application, the form structure metadata is generated according to natural language processing technology, reinforcement learning algorithm and / or type safety checking rule, including:

[0012] The form generation requirement input by the user through natural language is received, and the form generation requirement is parsed and processed using natural language processing technology to extract form structure elements and business logic information therein;

[0013] Based on the parsed form structure elements and business logic information, a plurality of candidate form configurations are generated through artificial intelligence technology;

[0014] The candidate form configurations are iteratively optimized through a reinforcement learning algorithm, and an optimized initial form configuration is output;

[0015] The initial form configuration is combined and checked with a predefined type safety checking rule to generate standardized form structure metadata.

[0016] In an optional embodiment of the present application, based on the parsed form structure elements and business logic information, a plurality of candidate form configurations are generated through artificial intelligence technology, including:

[0017] An large language model is called to generate an initial form configuration scheme based on the form structure elements and business logic information;

[0018] A historical form database is accessed, and industry general form templates, enterprise customized form data and historical behavior data similar to the current requirement are obtained;

[0019] The initial form configuration scheme is optimized in combination with the industry general form templates, enterprise customized form data and historical behavior data to generate a plurality of candidate form configurations.

[0020] In an optional embodiment of the present application, the initial form configuration is combined and checked with a predefined type safety checking rule to generate standardized form structure metadata, including:

[0021] Each field definition in the initial form configuration is parsed;

[0022] Each field definition is matched and checked with a rule in a type safety checking rule library;

[0023] For a field that fails to pass the check, automatic correction or generation of prompt information is performed based on the type safety checking rule;

[0024] Standardized form structure metadata containing type safety guarantees of all fields is generated.

[0025] In an optional embodiment of the present application, the rendering priority of each field is determined according to the form structure metadata, device sensor data and / or user visual behavior data, including:

[0026] receiving multi-source input data; wherein the multi-source input data includes form structure metadata, device sensor data and / or user visual behavior data;

[0027] inputting the multi-source input data into a long short-term memory network model trained by historical interaction data, and predicting the use probability of each field defined by the form structure metadata through the long short-term memory network model;

[0028] outputting the rendering priority of each field according to the use probability of each field defined by the form structure metadata.

[0029] In an optional embodiment of the present application, the target form is generated according to the rendering priority and device parameters, including:

[0030] generating layout rules based on the rendering priority and device parameters;

[0031] generating the target form according to the layout rules.

[0032] In an optional embodiment of the present application, the AI-enhanced form generation method further includes:

[0033] collecting interaction data of the user with the target form;

[0034] adjusting the form structure metadata and the rendering priority based on the collected interaction data.

[0035] A second aspect of an embodiment of the present application provides an AI-enhanced form generation device, including a processor and a memory storing program instructions, the processor being configured to execute the AI-enhanced form generation method of the first aspect of the present application when running the program instructions.

[0036] A third aspect of an embodiment of the present application provides a system, including:

[0037] a system body; and

[0038] The AI-enhanced form generation device of the second aspect of the present application is installed in the system body.

[0039] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium storing program instructions, the program instructions being used to make a computer execute the AI-enhanced form generation method of the first aspect of the present application when running.

[0040] The AI-enhanced form generation method and device, system and storage medium provided in the embodiments of the present application have the following beneficial effects.

[0041] The embodiments of the present application replace the traditional method of relying on artificial design or static templates by comprehensively applying natural language processing technology, reinforcement learning algorithm and type safety checking rules, thereby automatically generating form structure metadata with reasonable structure and type safety specifications, and improving the automation and reliability of form design. By fusing the field information defined by the form structure metadata itself, the real-time collected device sensor data and the user visual behavior data, and determining the rendering priority of each field according to the above, the rendering strategy of the form can adaptively reflect the device characteristics and user attention distribution. Finally, by generating a target form according to the above dynamically determined rendering priority and combining device parameters, the adaptive matching of form layout and terminal device is realized, and the loading performance and interactive experience of the form are improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] The accompanying drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 is a schematic diagram of an AI-enhanced form generation method provided by an embodiment of the present application;

[0044] Figure 2 is a schematic diagram of an AI-enhanced form generation device provided by an embodiment of the present application.

[0045] Reference signs:

[0046] 800: AI-enhanced form generation device; 801: processor; 802: memory; 803: communication interface; 804: bus. DETAILED DESCRIPTION

[0047] In order to make the technical solutions and advantages in the embodiments of the present application clearer, the exemplary embodiments of the present application are further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not an exhaustive enumeration of all embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0048] This application proposes a system including a natural language requirement input module, an AI configuration optimization engine, a type-safe configuration generator, a predictive rendering control module, a layout rule generation module, a form generation module, and a form rendering module. The natural language requirement input module receives form generation requirements input by the user in natural language. A natural language parsing module, connected to the natural language requirement input module, parses the input content and extracts form structure elements and business logic information. The AI ​​configuration optimization engine, connected to the natural language parsing module, generates multiple candidate form configurations based on the parsed form structure elements and business logic information, and iteratively optimizes the candidate form configurations using a reinforcement learning algorithm, outputting an optimized initial form configuration. The type-safe configuration generator, connected to the AI ​​configuration optimization engine, receives the initial form configuration, combines and validates the initial form configuration using type-safe verification rules, and generates standardized form structure metadata. The predictive rendering control module, connected to the type-safe configuration generator, receives multi-source input data and processes this data using a long short-term memory network model trained on historical interaction data to predict the usage probability of each field defined in the form structure metadata and output rendering priorities. The layout rule generation module is connected to the predictive rendering control module and is used to generate layout rules based on rendering priority and device parameters. The form generation module is connected to the layout rule generation module and is used to generate the final interactive form, i.e., the target form, according to the layout rules. The form rendering module is connected to the form generation module and is used to render and display the interactive target form, enabling adaptive form interaction across terminal devices. The system also includes a processor, electrically connected to the above modules, used to control the above modules to perform the aforementioned actions.

[0049] Figure 1 This is a schematic diagram of the AI-enhanced form generation method provided in the embodiments of this application. Any of the following methods can be executed in the system, or in a server or terminal device that is connected to the system.

[0050] Based on the above system structure, such as Figure 1 As shown, this application provides an AI-enhanced form generation method, including:

[0051] S01, Generate form structure metadata based on natural language processing technology, reinforcement learning algorithms and / or type safety validation rules.

[0052] S02, determine the rendering priority of each field based on form structure metadata, device sensor data and / or user visual behavior data.

[0053] S03 generates the target form based on rendering priority and device parameters.

[0054] In this embodiment, natural language processing (NLP) technology is used to receive form generation requests input by users in natural language, and to parse and process these requests to extract form structure elements and business logic information. For example, a user might request the creation of an employee onboarding form containing specific fields. Next, based on artificial intelligence technology, particularly Large Language Models (LLM), multiple candidate form configuration schemes are generated according to the extracted information. The LLM can be a GPT series model, capable of understanding complex semantic information and generating corresponding form configurations. Then, reinforcement learning algorithms are used to iteratively optimize these candidate configurations to improve form completion efficiency, reduce error rates, and enhance user experience. Finally, the optimized initial form configuration is combined with predefined type safety validation rules to ensure that each field conforms to type safety specifications, such as data type matching and input format validation. This not only improves the automation level of form design but also ensures the reliability and compliance of the form structure.

[0055] In this embodiment, to determine the rendering priority of each field, multi-source input data is first collected, including but not limited to form structure metadata, device sensor data (such as screen size and device type), and user visual behavior data (such as gaze focus distribution and dwell time). Next, a Long Short-Term Memory (LSTM) network model is used to process the fused feature data to predict the usage probability of each field defined by the form structure metadata. The prediction takes into account the usage habits and preferences of different users, as well as device characteristics. Based on the prediction results, a score is calculated for each field; fields with higher scores have higher rendering priority. For example, on mobile devices, considering the smaller screen size, high-priority fields may be displayed prominently or use a collapsible layout to optimize display space utilization, thereby improving the user's filling experience and efficiency.

[0056] In this embodiment, the process of generating a target form based on rendering priority and device parameters first involves selecting an appropriate layout mode based on the device type. Different layout strategies are applied to different device types, such as smartphones and tablets. Next, the rendering order, loading strategy, and visual prominence of fields are determined based on the field scores in the rendering priority. For example, high-priority fields may be pre-loaded and displayed prominently, while low-priority fields may be lazy-loaded or hidden. Furthermore, the overall layout and style of the form are adjusted according to the device screen characteristics to ensure the best user experience across various devices. Finally, all fields and their corresponding interaction logic are integrated to generate an interactive target form. This not only enhances the form's adaptability but also improves user convenience and satisfaction across different terminal devices.

[0057] In the embodiments of the present application, after the target form is generated according to the rendering priority and the device parameter, the target form is rendered and displayed to realize adaptive form interaction across terminal devices.

[0058] By comprehensively applying natural language processing technology, reinforcement learning algorithm and type safety checking rule, the AI-enhanced form generation method provided in the embodiments of the present application replaces the traditional method relying on manual design or static template, thereby being capable of automatically generating form structure metadata with reasonable structure and conforming to type safety specification, and improving the automation degree and reliability of form design. By fusing field information defined in the form structure metadata itself, real-time collected device sensor data and user visual behavior data, and determining the rendering priority of each field according to the field information, the rendering strategy of the form can adaptively reflect the device characteristics and user attention distribution. Finally, the target form is generated according to the dynamically determined rendering priority and the device parameter, thereby realizing adaptive matching of the form layout and the terminal device, and improving the loading performance and interactive experience of the form.

[0059] Optionally, the form structure metadata is generated according to the natural language processing technology, the reinforcement learning algorithm and / or the type safety checking rule, including: receiving a form generation requirement input by a user through natural language, and performing analysis and processing on the form generation requirement by using the natural language processing technology to extract form structure elements and business logic information therein; based on the form structure elements and the business logic information obtained by analysis, generating a plurality of candidate form configurations by using artificial intelligence technology; iteratively optimizing the candidate form configurations by using the reinforcement learning algorithm to output an optimized initial form configuration; and combining and checking the initial form configuration with a predefined type safety checking rule to generate standardized form structure metadata.

[0060] In the embodiments of the present application, the natural language processing technology is used to analyze and process the form generation requirements, extract form structure elements and business logic information, including: extracting the field name, field type and / or field grouping information declared in the natural language input form generation requirements to generate form structure elements; analyzing the business logic constraints in the natural language input form generation requirements to generate business logic information. The business logic constraints include mandatory item rules, field interrelation and / or conditional display logic. The natural language processing technology includes but is not limited to semantic understanding, entity recognition and dependency syntax analysis methods, which are used to extract information units with clear meaning from unstructured text. For example, when the user inputs "create an employee leave application form, including name, department, leave type, start time, end time, and need to associate the approval process", the system can identify that the form name is "employee leave application form", the field name includes "name", "department", etc., the field type can be inferred as text type or enumeration type, and the business logic information such as mandatory constraints, field interdependence and whether to integrate a workflow engine to support the approval process is identified. In addition, the natural language processing technology can also combine the speech recognition module to convert the voice input into text and perform subsequent analysis, thereby supporting multi-modal user demand input methods.

[0061] In the embodiments of the present application, artificial intelligence technology covers various intelligent algorithms such as large language models, generative adversarial networks and rule-based reasoning systems. For example, a large language model (such as GPT series, BERT derivative model) is called to generate an initial form configuration scheme according to the extracted form structure elements and business logic information, which includes field layout structure, grouping method and basic interaction logic; at the same time, access to the historical form database to obtain industry general form templates similar to the current requirements (such as personnel management standard forms) and enterprise customized form data (such as the existing leave approval form configuration in the company), and fuse and optimize these reference templates with the initial scheme to generate multiple differentiated and practical application scenario-compliant candidate form configurations, thereby improving the diversity and practicality of configuration generation.

[0062] In the embodiments of the present application, the candidate form configuration is iteratively optimized by a reinforcement learning algorithm, and an optimized initial form configuration is output, including: establishing a multi-dimensional evaluation system; wherein the multi-dimensional evaluation system includes a filling efficiency index, an error rate index and / or a user satisfaction index; in a simulated user filling environment, the multi-dimensional evaluation system is used to test and evaluate each candidate form configuration; the reward function value is calculated according to the test evaluation result, and one or more candidate form configurations are iteratively optimized according to the reward function value. The reinforcement learning algorithm is constructed on the basis of a Markov decision process framework, and a policy gradient method or a Q-learning variant is used as the core optimization mechanism. The system establishes a multi-dimensional evaluation system, including a filling efficiency index (such as average completion time), an error rate index (such as the number of field verification failures) and a user satisfaction index (such as operation path rationality score), and tests and evaluates each candidate form configuration in a simulated user filling environment. The reward function value is calculated according to the evaluation result, and the signal is used to guide the agent to adjust the form field order, the grouping structure and the interaction prompt strategy. After multiple rounds of training, the optimal configuration strategy is converged, and the initial form configuration with high usability and good user experience is output.

[0063] In the embodiments of the present application, the type safety check rule includes a field data type matching rule, an input format check rule and a cross-field constraint rule based on business logic. For example, for the field data type matching rule, it is verified whether the “date of birth” field is correctly defined as a date type instead of a text type; for the input format check rule, it is checked whether the “mailbox” field meets the email format requirement in the RFC 5322 specification; for the cross-field constraint rule, it is ensured that the value of the “end time” field is not earlier than the value of the “start time” field. The system automatically parses each field definition in the initial form configuration, matches it with the rules in the type safety check rule library one by one, performs automatic correction operation or generates prompt information for the field that fails the check, and finally outputs the standardized form structure metadata containing all field type safety guarantees, which can be in the form of JSON Schema, XML Schema or other extensible markup language, for data-driven in the subsequent rendering stage.

[0064] In this way, by receiving the form generation requirements input by the user in natural language, the complex configuration by professional personnel when manually designing the form is replaced, the technical threshold of form design is reduced, and non-professional users can also conveniently submit requirements. Then, the natural language processing technology is used to analyze the requirements, which can accurately extract the form structure elements (such as field name, field type, and field grouping) and business logic information (such as mandatory item rules and inter-field association relationships), avoiding information omission or understanding deviation that may occur when manually analyzing the requirements, and providing accurate basic data for subsequent form configuration generation. Then, based on the structure elements and business logic information obtained by analysis, multiple candidate form configurations are generated through artificial intelligence technology (such as calling a large language model and combining industry general form templates and enterprise customized form data in the historical form database), breaking the limitation of form generation relying on a single static template, so that the candidate configurations can cover more diversified business scenario requirements, providing sufficient selection space for subsequent optimization. Then, through the reinforcement learning algorithm, the filling efficiency, error rate, and user satisfaction are used as a multi-dimensional evaluation system to iteratively optimize the candidate form configurations in the simulated user filling environment and output the initial form configuration, which can make the initial configuration more suitable for the user requirements in the actual use scenario, improve the rationality of the configuration, and be more scientific than manually screening the configuration. Finally, the initial form configuration is combined with the pre-defined type safety checking rules for verification, each field definition is analyzed and matched with the rule library, and the fields that do not pass the verification are automatically corrected or prompts are generated, and finally the standardized form structure metadata is generated, effectively avoiding problems such as mismatched field data types, incorrect input formats, and cross-field logic conflicts, ensuring the compliance and reliability of the form structure, and improving the automation level of form design, reducing the workload and error probability of manual verification.

[0065] Optionally, based on the form structure elements and business logic information obtained by analysis, multiple candidate form configurations are generated through artificial intelligence technology, including: calling a large language model, generating an initial form configuration scheme based on the form structure elements and business logic information; accessing the historical form database and obtaining industry general form templates, enterprise customized form data, and historical behavior data similar to the current requirements; combining the industry general form templates, enterprise customized form data, and historical behavior data to optimize the initial form configuration scheme and generate multiple candidate form configurations.

[0066] In the embodiments of the present application, the large language model refers to a language generation model with a large number of parameters and a deep neural network structure, which can understand and generate natural language and related structured content that conforms to grammatical and semantic rules. For example, a pre-trained language model based on the Transformer architecture such as the GPT series, PaLM or LLaMA is used to receive the table structure elements (such as field name, field type, grouping information) and business logic information (such as mandatory rules, inter-field dependencies, conditional display logic) extracted by natural language processing technology, and guide the model to output structured form configuration description through prompt engineering. The initial form configuration scheme includes the arrangement order of the fields, layout structure suggestions, interactive control type recommendations, and basic validation rule settings, forming an initial configuration template that can be used for subsequent optimization.

[0067] In the embodiments of the present application, the historical form database is a data collection that stores past created form configuration information, which supports indexing and retrieval by category, purpose, industry or enterprise attribute. For example, the system matches the current form generation requirement with the templates stored in the database through a semantic similarity calculation method (such as cosine similarity combined with sentence embedding vector), identifies industry general form templates (such as the onboarding registration form in the human resources field and the expense reimbursement application form in the finance field) and enterprise customized form data (such as a project approval process form specific to a company and an internal asset requisition form) that are similar in function or structure to the current requirement. The historical form database includes a relational database, a non-relational database, a distributed file storage system, etc. The industry general form templates include employee onboarding registration forms and attendance statistics forms in the human resources management field, expense reimbursement application forms and invoice registration detail forms in the finance field, etc. The enterprise customized form data includes project approval forms, customer information collection forms, student registration forms, course evaluation forms, etc. used by specific enterprises or institutions. The historical behavior data includes the field access order, field dwell time, field modification frequency, form submission time, etc. of past users filling out similar forms. The above templates and data serve as reference samples in actual applications to enhance the practicality and compliance of the generated results.

[0068] In the embodiments of the present application, the initial form configuration scheme is optimized in combination with industry general form templates and enterprise customized form data to generate multiple candidate form configurations. The combination process includes mode alignment, structure fusion and rule inheritance operation, and the initial form configuration scheme generated from the large language model is integrated with the similar templates retrieved from the historical form database through the configuration merging algorithm. For example, the field definition in the initial scheme is unified with the standard field naming specification in the industry general form template, or the specific approval logic in the enterprise customized form data is embedded into the business rules of the initial scheme. By adjusting the field grouping method, optimizing the layout level, and supplementing the missing verification rules, multiple candidate form configurations are generated, which are reasonable in structure, adaptable in industry, and personalized in enterprise, providing a diversified optimization starting point for the subsequent reinforcement learning algorithm.

[0069] In the embodiments of the present application, the initial form configuration scheme includes field layout structure and / or interaction logic.

[0070] In this way, the large language model has the ability to understand semantics and convert them into structured content, and can convert abstract elements and logic into an initial scheme containing basic layout structure and interaction logic, without the need for manual construction of the form framework from scratch, reducing the workload of human participation in form basic design, while ensuring the basic matching degree of the initial scheme and user demand. Access to the historical form database and obtain similar industry general form templates, enterprise customized form data and historical behavior data, wherein the industry general form templates carry the form design experience commonly applicable in the industry, the enterprise customized form data reflects the specific form design needs of the enterprise itself, and the historical behavior data records the actual use of the past forms. One or more of the above data provides a reference basis for the initial form configuration scheme that fits the actual business scenario, avoiding the problem that the initial scheme does not meet the industry convention or enterprise specific needs due to lack of actual scene support. Optimizing the initial form configuration scheme in combination with the above industry general form templates, enterprise customized form data and historical behavior data can integrate industry experience, enterprise needs and past use feedback into the initial scheme, making up for possible design flaws or insufficient scene adaptation of the initial scheme, while generating multiple candidate form configurations, providing sufficient selection space for subsequent iterative optimization of candidate configurations through reinforcement learning algorithm, which helps to select a form configuration that better meets the actual use needs.

[0071] Optionally, the initial form configuration is combined and checked with predefined type safety checking rules to generate standardized form structure metadata, including: parsing each field definition in the initial form configuration; matching and checking each field definition with rules in a type safety checking rule library; automatically correcting or generating prompt information for fields that fail to pass the check based on type safety checking rules; generating standardized form structure metadata containing type safety guarantees for all fields.

[0072] In the embodiments of the present application, each field definition in the initial form configuration is parsed. The parsing process includes structurally deconstructing the semantic properties, data types, constraint conditions, and association relationships of the fields to extract meta-information that can be used for checking. For example, for a field definition, the system identifies its name, control type (such as input box, selector, date selector), whether it is required, default value, allowed input format (such as email, mobile phone number, ID number), and logical dependency relationship with other fields (such as conditional display, linkage update). This parsing operation is implemented based on an extensible markup language or a lightweight data exchange format syntax analyzer, supports reading configuration files in formats such as JSON, XML, or YAML, and converts them into an internal unified data model representation, providing standardized input for subsequent matching with rules in the type safety checking rule library.

[0073] In the embodiments of the present application, type safety checking rules include field data type matching rules, input format checking rules, and / or cross-field constraint rules based on business logic. The type safety checking rule library is a predefined set of checking rules for guaranteeing the legality and consistency of form fields, covering basic data type verification, format compliance checking, and cross-field business logic constraints. For example, basic data type verification includes judging whether the "age" field is correctly defined as an integer type rather than a string type; format compliance checking includes verifying whether the "email" field meets the RFC 5322 specification according to a regular expression or standard protocol, or whether the "mobile phone number" field meets the national communication coding rules; cross-field business logic constraints include verifying that the "end time" cannot be earlier than the "start time", or that the "emergency contact phone" must be filled in when the "emergency contact name" is not empty. The system performs rule matching on each field definition one by one, records the check result status, and identifies the fields that do not meet the rules and their specific violation items.

[0074] In the embodiments of the present application, for the fields that do not pass the verification, automatic correction or generation of prompt information is performed based on the type safety verification rules. Automatic correction refers to correcting errors that can be automatically adjusted according to the repair strategy defined in the type safety verification rules, for example, automatically correcting the "date of birth" field mistakenly set as a text type to a date type, or supplementing the "name" field missing the required mark with the required attribute. For cases that cannot be automatically corrected, the system generates prompt information, including error description, rule number violated, and suggested modification scheme, for reference by the user or subsequent processing module. The prompt information can be output in the form of log recording, visual interface annotation or API response, ensuring that the problem can be traced and easily intervened. This mechanism not only improves the automated processing capability of form configuration, but also retains the space for manual review and adjustment, balancing efficiency and accuracy.

[0075] In the embodiments of the present application, the standardized form structure metadata is the final form configuration data that has passed complete verification and ensured type safety compliance, encapsulated in a general and interoperable data format, supporting cross-platform and multi-terminal parsing and rendering. For example, the metadata is output in the form of a structured object conforming to the JSON Schema specification, where each field has an explicit data type declaration, verification rule definition and meta attribute annotation; or in the form of XML Schema definition, supporting data exchange and integration in enterprise-level information systems. The metadata not only contains basic configuration information of the fields, but also embeds meta information such as verification pass status identifier, version number and generation timestamp, ensuring its integrity, consistency and auditability, serving as the basis for subsequent determination of field rendering priority and generation of target forms.

[0076] In this way, each field definition in the initial form configuration is parsed, the field types, verification requirements and other information contained in the initial form configuration are disassembled one by one, the specific definition details of each field are determined, clear basic data is provided for subsequent accurate verification, and the problem of incomplete verification due to ambiguous field definition is avoided. The definition of each field is matched and verified with the rules in the type safety verification rule library. Since the type safety verification rule library contains preset compliance rules such as field data type matching, input format verification, and cross-field constraints based on business logic, it can pinpoint the cases where the field definitions in the initial form configuration do not comply with the specifications, such as incorrect field data types and input formats that do not meet business requirements, thereby reducing the compliance risk of field definitions. For fields that do not pass the verification, automatic correction or generation of prompt information is performed based on the type safety verification rules. Automatic correction can directly adjust some obvious compliance issues, and the generation of prompt information can guide relevant personnel to handle more complex non-compliant situations, avoiding the retention of non-compliant field definitions to subsequent links, while reducing the workload of manually checking and correcting errors one by one. The standardized form structure metadata containing all field type safety guarantees can not only ensure that the form structure complies with the type safety specification, but also ensure the structural consistency of the form when rendered on multiple platforms in the future. This provides a reliable structural basis for determining the rendering priority of fields based on form structure metadata and generating target forms in combination with device parameters, thereby improving the reliability of form structure and the adaptability of subsequent applications.

[0077] Optionally, according to the form structure metadata, device sensor data and / or user visual behavior data, the rendering priority of each field is determined, including: receiving multi-source input data; wherein the multi-source input data includes form structure metadata, device sensor data and / or user visual behavior data; inputting the multi-source input data into a long short-term memory network model trained by historical interaction data, and predicting the use probability of each field defined by the form structure metadata through the long short-term memory network model; according to the use probability of each field defined by the form structure metadata, outputting the rendering priority of each field.

[0078] In the embodiments of the present application, the multi-source input data includes form structure metadata, device sensor data and / or user visual behavior data. The multi-source input data refers to a heterogeneous data set derived from different system modules or perception channels, which is used to comprehensively evaluate the importance and user attention of form fields in a specific use scenario. For example, the form structure metadata provides basic definition information of the field, including field name, type, grouping structure and business logic constraints; the device sensor data is derived from the hardware sensing unit of the terminal device, including screen size, device type (such as mobile device, tablet, desktop terminal), screen orientation (landscape or portrait) and network status (Wi-Fi, 4G / 5G); the user visual behavior data reflects the attention distribution characteristics of the user in the process of browsing and filling out the form, including gaze focus coordinates, gaze duration, page scrolling trajectory and heat map distribution. These data are collected and fused through a unified data interface to form a complete input feature set to support subsequent model inference.

[0079] In the embodiments of the present application, the long short-term memory network model is a deep learning model capable of processing sequence data and capturing time-dependent relationships, which is suitable for modeling the behavior timing characteristics of the user in the form filling process. In the training phase, the model learns parameters using historical interaction data, which includes the field access order, dwell time, modification frequency and skip record when the user fills out the same or similar form in the past. In the inference phase, the model receives multi-source input data of the current session, and after embedding layer, normalization processing and sequence encoding, outputs the probability value of each field defined by the form structure metadata being accessed or filled by the user in the future. For example, for high-frequency mandatory fields such as “name” and “contact information”, the model predicts that the probability of use is close to 0.95; and for low-frequency fields such as “remark” or “additional information”, the predicted probability may be less than 0.3.

[0080] In the embodiments of the present application, the rendering priority is a ranking index calculated based on the field usage probability, which is used to guide the layout arrangement and resource loading strategy in the form rendering phase. For example, the system assigns a priority score to each field, which is positively correlated with the usage probability and can be generated by linear mapping, piecewise function or weighted combination. The field with high usage probability corresponds to high rendering priority, indicating that it should be displayed first, loaded in advance or enhanced in visual prominence in the form; the field with low usage probability corresponds to low rendering priority, which can use strategies such as delayed loading, folding and hiding or gray display. The output rendering priority exists in the form of structured data, such as a key-value pair list containing field identifier and priority score, which is called by the subsequent layout rule generation module to realize dynamic adaptive form rendering.

[0081] In the embodiment of the present application, the multi-source input data is input into the long short-term memory network model trained by the historical interaction data, and the use probability of each field defined by the form structure metadata is predicted through the long short-term memory network model, including: performing fusion processing on the multi-source input data to extract the features of the input data; processing the fused feature data through the long short-term memory network model to predict the use probability of each field defined by the form structure metadata.

[0082] In the embodiment of the present application, according to the use probability of each field defined by the form structure metadata, the rendering priority of each field is output, including: calculating a score for each field according to the use probability of each field defined by the form structure metadata; wherein the rendering priority includes the score of each field, and the use probability is positively correlated with the score.

[0083] In this way, multi-source input data containing form structure metadata, device sensor data and / or user visual behavior data is received, wherein the form structure metadata provides basic definition information of the fields, the device sensor data (such as screen size, device type) reflects the characteristics of the terminal device, and the user visual behavior data (such as gaze focus distribution, dwell time) embodies the user's attention to the fields. The integration of multi-source data can avoid the limitations of single data dimension and provide comprehensive information support for subsequent accurate judgment of field importance. The multi-source input data is input into the long short-term memory network model trained by the historical interaction data. Since the long short-term memory network model is good at processing time series related data, and has learned the use rules and correlation characteristics of the fields when the past users use the form after being trained by the historical interaction data, the association between the field use and the device characteristics and the user visual behavior can be more accurately captured after combining the multi-source input data, and then the use probability of each field defined by the form structure metadata can be reliably predicted. According to the use probability of each field, the rendering priority is output. The field with high use probability means that the user is more likely to use or use frequently when filling out the form, and a higher rendering priority is given accordingly, so that the rendering strategy of the field can be reasonably arranged according to the priority when generating the form subsequently, avoiding the problem of time-consuming user search for common fields due to disordered rendering of the fields, thereby laying a foundation for improving the form interaction efficiency.

[0084] Optionally, according to the rendering priority and the device parameters, a target form is generated, including: generating a layout rule based on the rendering priority and the device parameters; and generating the target form according to the layout rule.

[0085] In the embodiments of the present application, based on the rendering priority and the device parameters, the layout rule is generated, including: according to the device type in the device parameters, selecting a corresponding layout mode; according to the field score in the rendering priority, determining the rendering order, loading strategy and / or visual prominence of the field. The layout rule is a configuration strategy set for guiding the form interface structure arrangement and field presentation mode, and its generation process comprehensively considers the rendering priority of the field and the device parameters of the terminal device. The device parameters include device type (such as mobile device, tablet computer, desktop computer), screen size, resolution, pixel density and operating system type. According to the device parameters, the adaptive layout mode is selected, for example, the single-column vertical scrolling layout is adopted on the mobile device, and the multi-column grid layout or tab grouping layout is adopted on the desktop device. At the same time, combined with the field score in the rendering priority, the rendering order, loading strategy and visual prominence of the field are determined, for example, the high-priority field is arranged in the top area of the form, the preloading mechanism is adopted and the font size or background color contrast is enhanced, while the low-priority field is set to the folded state or rendered on demand through the lazy loading mode, thereby forming the optimal layout rule which takes into account the user attention distribution and device characteristics.

[0086] In the embodiments of the present application, the target form is generated according to the layout rule, including: determining the structural framework of the form according to the layout mode in the layout rule; arranging the loading order and rendering mode of the fields according to the rendering order and loading strategy in the layout rule; setting the display style of different fields according to the visual prominence in the layout rule; adjusting the form layout and style in the structural framework according to the device screen characteristics; integrating all fields and corresponding interaction logic to generate an interactive target form. According to the rendering order and loading strategy in the layout rule, the loading order and rendering mode of the fields are arranged, including: for fields with high priority, the rendering strategy of preloading and front-end display is adopted; for fields with low rendering priority, the rendering strategy of delayed loading or folding and hiding is implemented. The target form is the final interactive form instance presented to the user, and its generation process is dynamically assembled according to the structural framework, rendering order, loading strategy and visual style defined in the layout rule. According to the layout mode in the layout rule, the overall structural framework of the form is determined, for example, card grouping, step guide or multi-region panel layout is adopted; according to the rendering order and loading strategy in the layout rule, the loading order and rendering mode of the fields are arranged, preloading and immediate rendering are performed on fields with high rendering priority, and delayed loading or condition-triggered rendering is implemented on fields with low rendering priority; according to the visual prominence in the layout rule, the display style of different fields is set, including font bold, border highlight, icon identification or animation guide; at the same time, the field spacing, container width and responsive breakpoints are dynamically adjusted in the structural framework in combination with the device screen characteristics, to ensure that the form has good readability and operability on different devices. Finally, all fields and their associated interaction logic are integrated to generate a target form that meets the current context environment and deliver it to the rendering module for display.

[0087] In this way, the layout rule is generated based on the rendering priority and the device parameters, wherein the rendering priority is determined according to the field use probability predicted by a long short-term memory network model trained by historical interaction data, according to the form structure metadata, the device sensor data and the user visual behavior data, and can reflect the actual use demand of the user on each field (a high-priority field corresponds to a field that the user is more likely to use preferentially or frequently), and the device parameters cover terminal device characteristics such as screen size and device type. When the layout rule is generated in combination of the two, the rule can not only be arranged around the user use demand to preferentially ensure the reasonable presentation of the high-priority field, but also adapt to the hardware conditions of different terminals to avoid the problem of mismatch between the layout and the device characteristics (for example, field congestion on a small-screen device and loose layout on a large-screen device), so that the layout rule is more suitable for the actual application scenario. The target form is generated according to the layout rule. Since the layout rule has pre-integrated the double considerations of user use demand adaptation and device characteristic adaptation, the target form generated can make the high-priority field more easily accessible to the user, reduce the time consumption of the user in searching for and operating the frequently-used field, and at the same time ensure that the form can be clearly and reasonably displayed on different terminal devices, avoiding poor interaction caused by layout problems, thereby improving the display adaptability of the form and optimizing the interaction experience of the user with the form, and providing support for the improvement of the practical performance of the form.

[0088] Optionally, the AI-enhanced form generation method further includes: collecting interaction data of the user with the target form; and adjusting the form structure metadata and the rendering priority based on the collected interaction data.

[0089] In the embodiments of the present application, the user interaction data is continuously collected, the user interaction data includes field modification frequency, filling completion rate and / or error correction record, the collected interaction data is analyzed to identify optimization points, and the analysis result is fed back to the form structure metadata generation node and / or the rendering priority determination node to optimize the form structure metadata and / or the rendering priority.

[0090] In the embodiments of the present application, the interaction data of the user with the target form refers to the behavior log information generated by the user in the process of filling, browsing or operating the target form, and is used to reflect the user operation mode and experience characteristics in the actual use scenario. For example, the interaction data includes field modification frequency (such as the number of times a field is repeatedly edited), field filling completion rate (such as the proportion of a field that is successfully filled in multiple form submissions), error correction record (such as correction operations performed by the user due to format errors or logical conflicts), field dwell time (average time of the user staying on a field) and form overall submission time consumption. These data are collected at the user end through front-end burying, event listener or log reporting mechanism, and are summarized to the background analysis system after anonymization processing, and are used as the basis for subsequent optimization of form structure and rendering strategy.

[0091] In the embodiments of the present application, the form structure metadata and the rendering priority are adjusted based on the collected interaction data. The adjustment process includes statistical analysis and pattern recognition of the interaction data, identifying potential optimization points in the form design, and feeding back to the relevant links of the form generation process. For example, for fields with high modification frequency, analyze whether the data type or validation rule is inconsistent with the user's expectation, and then adjust the field definition in the form structure metadata, such as changing text input to drop-down selection to reduce input errors; for fields with low completion rate, determine whether they are non-essential information, and can be adjusted from mandatory to optional fields, or adjust their position in the form to improve visibility. At the same time, based on the actual operation path and the duration of the user, the importance of each field is re-evaluated, the usage probability prediction model of the field is updated, and the rendering priority is dynamically adjusted, so that high-frequency operation fields have higher display priority and resource loading weight, and the continuous adaptation of form configuration and user behavior habits is realized.

[0092] In this way, the interaction data of the user with the target form is collected, the target form is generated according to the rendering priority and the device parameters, and the interaction data can truly record the behavior characteristics of the user in the actual use of the target form, such as field modification frequency, completion rate, error correction record, etc. These data provide a basis for subsequent adjustment work that fits the actual use scenario, avoiding the situation of blind adjustment that is divorced from the user's real needs. Based on the collected interaction data, the form structure metadata and the rendering priority are adjusted, wherein the form structure metadata is generated by natural language processing technology, reinforcement learning algorithm and pre-defined type safety validation rules, and the rendering priority is determined by a long short-term memory network model trained by historical interaction data according to the form structure metadata, device sensor data and user visual behavior data. Adjusting the two based on actual interaction data can make the form structure metadata more suitable for the user's filling habits (for example, optimizing the type or validation rule of fields with high modification frequency), and the rendering priority can better meet the user's actual use needs (for example, improving the priority of high-frequency operation fields of the user), so that the new form generated based on the adjusted form structure metadata and rendering priority in the subsequent process is closer to the actual use scenario of the user in terms of interaction logic and display strategy, which helps to improve the convenience of form use and the filling experience of the user.

[0093] In combination with Figure 2As shown, the embodiment of the present application provides an AI-enhanced form generation device 800, which comprises a processor 801 and a memory 802. Optionally, the device can further comprise a communication interface 803 and a bus 804. The processor 801, the communication interface 803, and the memory 802 can communicate with each other through the bus 804. The communication interface 803 can be used for information transmission. The processor 801 can invoke the logical instructions in the memory 802 to execute the AI-enhanced form generation method of the above-mentioned embodiment.

[0094] In addition, the logical instructions in the memory 802 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer-readable storage medium.

[0095] The memory 802, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present application. The processor 801 executes the function application and data processing by running the program instructions / modules stored in the memory 802, that is, implements the AI-enhanced form generation method in the above-mentioned embodiment.

[0096] The memory 802 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; the data storage area can store data created according to the use of the terminal device, etc. In addition, the memory 802 can include a high-speed random access memory, and can also include a non-volatile memory.

[0097] The embodiment of the present application provides a system, which comprises a system body and the AI-enhanced form generation device 800 described above. The AI-enhanced form generation device 800 of the system is installed in the system body. The installation relationship described herein is not limited to placing in the system, but also includes installation connection with other components of the system, including but not limited to physical connection, electrical connection or signal transmission connection, etc. Those skilled in the art can understand that the AI-enhanced form generation device 800 of the system can be adapted to a feasible system body, and thus other feasible embodiments can be realized.

[0098] The embodiment of the present application provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are set to execute the AI-enhanced form generation method described above.

[0099] The technical scheme of the embodiments of the present application can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes one or more instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium can be a non-transitory storage medium, including a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0100] The technical scheme of the embodiments of the present application can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes one or more instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium can be a non-transitory storage medium, including a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0101] The above description and drawings are illustrative of embodiments of the application and are not to be construed as limiting the application. Other embodiments can include structural, logical, electrical, process, and other changes. The embodiments are merely examples of the many possible embodiments of the application. Unless explicitly stated otherwise, individual components and functions are optional and the order of operations can vary. Parts and features of some embodiments can be included or replaced in or by parts and features of other embodiments. Also, the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. As used in the description of the embodiments and the claims, the article "a" is intended to include one or more articles, by use of "a" or "an," it is intended to include one or more embodiments. Also, as used in the description of the embodiments and the claims, the term "including" and variations thereof, means that existence of the features, integers, steps, operations, elements, and / or components noted are an option, and that not all of them are required, but that some embodiments can include, but not require, several of those specialties features, integers, steps, operations, elements, components, and / or combinations thereof. The phrase "consisting of" means an exclusive inclusion, that is, the listed components which are mandatory, not optional. The phrase "consisting essentially of" means the components which are mandatory, not optional, with the exception of other components of the application. The sentence "includes a..." means that other components can also be present in the process, method, or device. Each embodiment is focused on the differences from other embodiments, and the same or similar parts between embodiments can be mutually referred to. For the method, product, etc. of the embodiments, if it corresponds to the method part of the embodiments, the relevant part can be referred to the description of the method part.

[0102] Those skilled in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments applied herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods for each specific application to realize the described functions, but such implementation should not be considered beyond the scope of the embodiments of the application. The skilled person can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0103] In the embodiments disclosed herein, the disclosed methods, products (including but not limited to apparatuses, devices, etc.), can be implemented in other manners. For example, the described apparatus embodiments can be implemented only in a form of a logical function, and can be implemented by using a manner such as software (for example, a computer program) or a combination of hardware and software. In a component, a plurality of components, or a combination thereof, can be implemented. In some embodiments, a plurality of components or a combination thereof can be integrated in a machine to realize an apparatus according to the embodiments. In some embodiments, a plurality of components or a combination thereof can be integrated in one physical device or in a virtual machine. In some embodiments, a plurality of components or a combination thereof can be implemented in an addressable storage medium by using instructions.

[0104] The flowcharts and block diagrams in the drawings show the possible implementation architectures, functions and operations of the system, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment or a part of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions noted in the blocks can occur in an order different from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the drawings, the operations or steps corresponding to different blocks can also occur in an order different from that disclosed in the descriptions, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. Each block in the block diagrams and / or flowcharts, and the combination of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A form generation method based on AI enhancement, characterized in that, include: Generate form structure metadata based on natural language processing techniques, reinforcement learning algorithms, and / or type safety validation rules; Determine the rendering priority of each field based on form structure metadata, device sensor data, and / or user visual behavior data; Generate the target form based on rendering priority and device parameters.

2. The method according to claim 1, characterized in that, Generate form structure metadata based on natural language processing techniques, reinforcement learning algorithms, and / or type safety validation rules, including: It receives form generation requests from users via natural language input, and uses natural language processing technology to parse and process the form generation requests, extracting form structure elements and business logic information. Based on the parsed form structure elements and business logic information, multiple candidate form configurations are generated using artificial intelligence technology. The candidate form configuration is iteratively optimized using a reinforcement learning algorithm, and the optimized initial form configuration is output. The initial form configuration is combined and validated with predefined type safety rules to generate standardized form structure metadata.

3. The method according to claim 2, characterized in that, Based on the parsed form structure elements and business logic information, multiple candidate form configurations are generated using artificial intelligence technology, including: The large language model is invoked to generate an initial form configuration scheme based on form structure elements and business logic information; Access the historical form database and obtain industry-wide general form templates, enterprise-customized form data, and historical behavior data that are similar to the current needs; The initial form configuration scheme is optimized by combining industry-standard form templates, enterprise-customized form data, and historical behavior data, generating multiple candidate form configurations.

4. The method according to claim 2, characterized in that, The initial form configuration is combined and validated with predefined type safety rules to generate standardized form structure metadata, including: Parse each field definition in the initial form configuration; The definition of each field is matched and validated against the rules in the type security validation rule base. For fields that fail validation, automatic correction or prompt messages are generated based on type safety validation rules; Generate standardized form structure metadata that includes security guarantees for all field types.

5. The method according to any one of claims 1 to 4, characterized in that, Based on form structure metadata, device sensor data, and / or user visual behavior data, determine the rendering priority of each field, including: Receive multi-source input data; wherein, the multi-source input data includes form structure metadata, device sensor data and / or user visual behavior data; Multi-source input data is fed into a long short-term memory network model trained with historical interaction data, and the long short-term memory network model is used to predict the usage probability of each field defined by the form structure metadata. Based on the usage probability of each field as defined in the form structure metadata, output the rendering priority of each field.

6. The method according to any one of claims 1 to 4, characterized in that, Based on rendering priority and device parameters, generate the target form, including: Layout rules are generated based on rendering priority and device parameters; Generate the target form according to the layout rules.

7. The method according to any one of claims 1 to 4, characterized in that, Also includes: Collect user interaction data with the target form; The form structure metadata and rendering priority are adjusted based on the collected interaction data.

8. An AI-enhanced form generation device, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute the AI-enhanced form generation method as described in any one of claims 1 to 7 when running the program instructions.

9. A system, characterized in that, include: System body; as well as, The AI-enhanced form generation device as described in claim 8 is installed on the system body.

10. A computer-readable storage medium storing program instructions, characterized in that, When the program instructions are executed, they cause the computer to perform the AI-enhanced form generation method as described in any one of claims 1 to 7.

Images