An AI-based multi-UI component framework code visualization page generation method and system

Abstract

The application provides an AI-based multi-UI component framework code visualization page generation method and system, relating to the technical field of data processing, which comprises the following steps: grouping the function description statements in the page configuration input data through an AI model, extracting component semantic fields and associated attribute fields, mapping the component attribute structures of different UI frameworks, converting the component semantic fields into unified description fields, dynamically adjusting the interaction weights between the component attributes, calculating the dynamic coupling strength between the component attributes, obtaining the structure coupling amount, calculating the deformability of the mapping relationship between the component attributes, obtaining the mapping deformation amount, updating the attribute mapping relationship in the standard semantic data, describing the unified conversion relationship between the attributes and events of different UI framework components, constructing a multi-UI component unified model, and generating a page configuration file. The application can realize the automatic construction and unified rendering of multi-framework visualization pages.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a code visualization page generation method and system based on an AI multi-UI component framework. BACKGROUND

[0002] At present, code or visual programming platforms are widely used in the field of front-end page generation. They mainly realize the rapid construction of pages by dragging components, configuring attributes, and event logic. However, most of these platforms are based on a single UI framework, and the component structure, attributes, and event interfaces of different frameworks may differ, resulting in the need for users to redesign interface logic when migrating platforms or replacing components, and possibly lacking cross-framework visual adaptation capabilities.

[0003] Some platforms attempt to introduce AI technology to assist in page generation, such as using natural language descriptions to generate page structures. However, the role of AI is mostly limited to the initial conversion of text to components, and may not be able to achieve intelligent modeling and dynamic mapping of cross-framework component attributes. Existing solutions may not be able to extract features and normalize learning of component semantics across different frameworks through AI, making it difficult to automatically generate cross-framework visual page configuration files based on user input or historical configuration data, and affecting the implementation of page generation and unified rendering. SUMMARY

[0004] The purpose of the present application is to provide a code visualization page generation method and system based on an AI multi-UI component framework, aiming to solve the problems mentioned in the background art.

[0005] To solve the above technical problems, the technical solution of the present application is as follows:

[0006] In a first aspect, a code visualization page generation method based on an AI multi-UI component framework is provided, the method comprising:

[0007] Obtaining page configuration input data;

[0008] Grouping the function description statements in the page configuration input data through an AI model, extracting component semantic fields and associated attribute fields, and obtaining semantic structure data;

[0009] According to the semantic structure data, mapping the component attribute structures of different UI frameworks, converting the component semantic fields into unified description fields, and obtaining standard semantic data;

[0010] According to the standard semantic data, the unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template to obtain semantic template data, the attribute feature vectors of each semantic template are combined into a multi-dimensional feature matrix, the co-occurrence relationship between the attribute feature vectors is counted to generate interaction weight data according to the multi-dimensional feature matrix, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data to obtain a structure matrix;

[0011] According to the structure matrix, the dynamic coupling strength between the component attributes in the structure matrix is calculated to obtain a structure coupling amount;

[0012] According to the structure matrix and the structure coupling amount, the deformability of the mapping relationship between the component attributes in the structure matrix is calculated to obtain a mapping deformation amount;

[0013] According to the mapping deformation amount, the attribute mapping relationship in the standard semantic data is updated to describe the unified conversion relationship between the attributes and events of different UI framework components, and a mapping rule set is obtained;

[0014] According to the mapping rule set, a multi-UI component unified model is constructed, and page configuration input data is input into the multi-UI component unified model to generate a page configuration file.

[0015] Further, the function description sentences in the page configuration input data are grouped by an AI model, and component semantic fields and associated attribute fields are extracted to obtain semantic structure data, including:

[0016] According to the page configuration input data, the function description sentences in the page configuration input data are arranged in order, and the sentences irrelevant to the page are removed to obtain sentence sequence data;

[0017] According to the AI model, the sentences in the sentence sequence data are analyzed, and each function description sentence is split into a component candidate word segment and an attribute candidate word segment to obtain semantic segment data;

[0018] According to the AI model, the association relationship between the component candidate word segments is inferred, and the segments belonging to the same component are merged into the same semantic group to obtain semantic group data;

[0019] According to the semantic group data, the fields representing the meaning of the component in each semantic group are determined and recorded as component semantic fields, and the attribute description fields corresponding to the component semantic fields are recorded as associated attribute fields to obtain semantic structure data.

[0020] Further, according to the standard semantic data, template division is performed, unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template, semantic template data is obtained, attribute feature vectors of each semantic template are combined into a multi-dimensional feature matrix, according to the multi-dimensional feature matrix, interaction weight data is generated by counting the co-occurrence relationship between the attribute feature vectors, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data, to obtain a structure matrix, including:

[0021] According to the standard semantic data, template division is performed, unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template, and semantic template data is obtained.

[0022] According to the semantic template data, unified description fields in each semantic template are arranged in order as attribute field sequences, and the attribute field sequences are converted into attribute feature vectors, to obtain feature vector data.

[0023] By sequentially filling each attribute feature vector in the feature vector data into a preset matrix structure according to the component and field position, a multi-dimensional feature matrix is obtained.

[0024] According to the multi-dimensional feature matrix, interaction weight data is generated by counting the co-occurrence relationship between the attribute feature vectors, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data, to obtain a structure matrix.

[0025] Further, according to the structure matrix, the dynamic coupling strength between the component attributes in the structure matrix is calculated, to obtain a structure coupling quantity, including:

[0026] According to the structure matrix and the interaction weight data, the joint action strength between the element values in the structure matrix and the weights in the interaction weight data is calculated, to obtain a local joint influence term.

[0027] According to the interaction weight data, the weight summary value of each attribute with the remaining attributes in the interaction weight data is calculated, to obtain an overall connection scale term.

[0028] According to the local joint influence term and the overall connection scale term, the concentration degree of the combination of the attributes inside the component under the constraint of the overall connection scale is calculated, to obtain a compactness comparison relationship term.

[0029] By nonlinearly transforming the compactness comparison relationship term, the compactness comparison relationship term is mapped to a comprehensive evaluation value in a limited numerical interval, to obtain the structure coupling quantity.

[0030] Further, according to the structure matrix and the structure coupling quantity, the deformability of the mapping relationship between the component attributes in the structure matrix is calculated, to obtain a mapping deformation quantity, including:

[0031] According to the structure matrix and the interaction weight data, the direction asymmetry degree of any two attributes in the same component under forward and reverse mapping is calculated, and a direction difference analysis item is obtained;

[0032] According to the structure matrix and the interaction weight data, the structure change amplitude of the structure value between any two attributes in the same component is calculated, and a local deformation strength item is obtained;

[0033] According to the local deformation strength item and the structure coupling quantity, the overall deformation sensitivity of the local deformation strength of each attribute under the constraint of the structure coupling quantity is calculated, and a normalized deformation sensitivity item is obtained;

[0034] The direction difference analysis item and the normalized deformation sensitivity item are nonlinearly transformed to calculate the asymmetry degree of the mapping relationship of the direction difference analysis item in the direction, and the overall variability degree of the normalized deformation sensitivity item under the constraint of the structure coupling quantity is calculated, and a mapping deformation quantity is obtained.

[0035] Further, according to the mapping deformation quantity, the attribute mapping relationship in the standard semantic data is updated to describe the unified conversion relationship between the attributes and the events of different UI framework components, and a mapping rule set is obtained, including:

[0036] According to the mapping deformation quantity, the attribute mapping relationship of each component in the standard semantic data is marked, and the attribute mapping relationship is associated with the mapping deformation quantity of the corresponding component, and a mapping update data is obtained;

[0037] According to the mapping update data, the attribute mapping relationships in the same numerical interval are merged into the same mapping group, and a mapping group data is obtained;

[0038] According to the mapping group data, the source attribute field and the target attribute field of the attribute mapping relationship in each mapping group are reserved, replaced and reorganized, and a mapping adjustment data is obtained;

[0039] According to the mapping adjustment data, a rule identifier is generated for each attribute mapping relationship, and the rule identifier and the attribute mapping relationship are archived and arranged according to the component and the event type, and a mapping rule set is obtained.

[0040] Further, according to the mapping rule set, a multi-UI component unified model is constructed, and page configuration input data is input into the multi-UI component unified model to generate a page configuration file, including:

[0041] According to the mapping rule set, the unified description field of each component is associated with the corresponding attribute mapping relationship in the mapping rule set, and a component mapping template data is obtained;

[0042] According to the component mapping template data, the source attribute field and the target attribute field corresponding to the same component semantic field are combined into a group of parameter descriptions, and a model parameter data is obtained.

[0043] According to the model parameter data, the parameter description of each group is hierarchically organized according to the component semantic field and the event type, and is assembled into a callable model structure to obtain a multi-UI component unified model;

[0044] By matching the component semantic field in the page configuration input data with the model structure in the multi-UI component unified model, page configuration analysis data is generated, and the model parameters of each component are filled and arranged according to the page configuration analysis data to obtain a page configuration file.

[0045] In a second aspect, a code visualization page generation system based on an AI-based multi-UI component framework, the system comprising:

[0046] A data module for obtaining page configuration input data;

[0047] A semantic module for grouping the function description statements in the page configuration input data by an AI model, extracting component semantic fields and associated attribute fields, and obtaining semantic structure data;

[0048] A mapping module for field mapping of component attribute structures of different UI frameworks according to the semantic structure data, converting the component semantic fields into unified description fields, and obtaining standard semantic data;

[0049] A matrix module for template division according to the standard semantic data, merging the unified description fields belonging to the same component in the standard semantic data into the same semantic template to obtain semantic template data, combining the attribute feature vectors of each semantic template into a multi-dimensional feature matrix, generating interaction weight data according to the co-occurrence relationship between the attribute feature vectors, and adjusting the element values in the multi-dimensional feature matrix according to the interaction weight data to obtain a structure matrix;

[0050] A coupling module for calculating the dynamic coupling strength between component attributes in the structure matrix according to the structure matrix to obtain a structure coupling amount;

[0051] A deformation module for calculating the deformability of the mapping relationship between component attributes in the structure matrix according to the structure matrix and the structure coupling amount to obtain a mapping deformation amount;

[0052] A rule module for updating the attribute mapping relationship in the standard semantic data according to the mapping deformation amount, describing the unified conversion relationship between attributes and events of components of different UI frameworks, and obtaining a mapping rule set;

[0053] A model module for constructing a multi-UI component unified model according to the mapping rule set, and inputting the page configuration input data into the multi-UI component unified model to generate a page configuration file.

[0054] The above scheme of the present application at least includes the following beneficial effects:

[0055] The present application converts the heterogeneous attribute structures of multiple UI frameworks into a single field structure by field mapping and converting the component attribute structure of different UI frameworks into a unified description field, so that the attribute naming, type and event parameters between frameworks are isomorphic at the semantic layer, the data set has a cross-framework compatible feature matrix input form, and the multi-template mapping failure caused by field naming differences in the traditional method is avoided. The data processing process no longer depends on the framework-specific attribute table, but indexes and calls through the unified description field, reduces the field conversion logic between heterogeneous structures, and realizes the structural consistency in the storage and calculation path.

[0056] The present application combines the semantic template attribute vector into a multi-dimensional feature matrix, dynamically adjusts the interaction weight between attributes to form a structure matrix, and maps the originally dispersed component attributes to a computable matrix space, so that the co-occurrence relationship and dependence degree between each field are represented in numerical form in the matrix unit, so that the relationship between component attributes is no longer only text logic or static rules, facilitating subsequent coupling degree and deformation degree calculation, making the relationship between attributes have storage sparsity and indexability, and any attribute associated value can be quickly retrieved according to the matrix position at runtime, realizing batch calculation for multiple components, providing a unified numerical expression basis for subsequent mapping and rule updating, and realizing the conversion of the calculation model from symbolic description to numerical matrix.

[0057] The present application describes the joint dependence degree between internal attributes of components at the numerical layer by calculating the structure coupling amount, uses the joint distribution between interaction weight and matrix value to obtain a measurable local influence amount and overall connection scale amount, and generates a comprehensive value in a limited interval through a nonlinear function mapping, objectively reflecting the concentration degree and correlation strength between component attributes, so that the system can distinguish between high and low related attributes at the data layer, so that the subsequent mapping calculation no longer depends on static rule screening, but realizes dynamic filtering through the range selection of the coupling value, realizes adaptive reduction of the attribute set according to the threshold of the coupling amount in the execution stage, completes the correlation compression and structure denoising at the data layer, provides a stable correlation basis for subsequent deformation amount calculation, and ensures the consistency and controllability of data input.

[0058] The application calculates the numerical offset degree of the attribute mapping relationship between different UI frameworks in the direction and amplitude by calculating the mapping deformation amount, forms a mapping matrix composed of direction difference analysis items and normalized deformation sensitive items, can identify the attribute mapping asymmetry of the same component under different frameworks at the data layer, records the difference as a continuous variable, directly aggregates and groups the data according to the deformation amount interval during subsequent mapping update, and enables the corresponding relationship of the component parameters under different frameworks to be stored, transmitted and updated in a calculable form, so that adaptive field replacement and parameter archiving can be realized during program generation or page compilation, and a continuously calculable semantic difference description basis is provided for cross-framework mapping.

[0059] The application forms a mapping rule set by updating the attribute mapping relationship in the standard semantic data, numerically marks and groups the attribute mapping relationship according to the deformation amount range, and recombines the source attribute field and the target attribute field into mapping adjustment data, so as to generate a reconfigurable mapping structure at the data processing layer, that is, the relationship between different attribute pairs is no longer fixed as one-way or static mapping, but is encapsulated as a rule entry set, which can be reindexed or combined at any time point, so that the system does not need to recalculate the mapping logic during subsequent page generation or code compilation, but only needs to call the corresponding rule entry to restore the attribute binding. The structured storage reduces the number of repeated executions of mapping calculation, provides data layer support for incremental update, and realizes a reusable and maintainable data organization mode of attribute mapping relationship. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 is a flow block diagram of a code visual page generation method based on an AI multi-UI component framework provided by an embodiment of the application. DETAILED DESCRIPTION

[0061] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0062] As Figure 1 shown, an embodiment of the application proposes a code visual page generation method based on an AI multi-UI component framework, which comprises:

[0063] obtaining page configuration input data;

[0064] grouping the function description statements in the page configuration input data by an AI model, extracting component semantic fields and associated attribute fields, and obtaining semantic structure data;

[0065] According to the semantic structure data, field mapping is performed on the component attribute structure of different UI frameworks, component semantic fields are converted into unified description fields, and standard semantic data is obtained;

[0066] According to the standard semantic data, template division is performed, unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template, semantic template data is obtained, attribute feature vectors of each semantic template are combined into a multi-dimensional feature matrix, according to the multi-dimensional feature matrix, interaction weight data is generated by counting the co-occurrence relationship between the attribute feature vectors, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data, and a structure matrix is obtained;

[0067] According to the structure matrix, the dynamic coupling strength between the component attributes in the structure matrix is calculated, and a structure coupling quantity is obtained;

[0068] According to the structure matrix and the structure coupling quantity, the deformability of the mapping relationship between the component attributes in the structure matrix is calculated, and a mapping deformation quantity is obtained;

[0069] According to the mapping deformation quantity, the attribute mapping relationship in the standard semantic data is updated, the unified conversion relationship between the attributes and events of different UI framework components is described, and a mapping rule set is obtained;

[0070] According to the mapping rule set, a multi-UI component unified model is constructed, and page configuration input data is input into the multi-UI component unified model, and a page configuration file is generated.

[0071] In the embodiment of the application, the page configuration input data is acquired, so that the AI model can directly perform semantic recognition and feature extraction on the statement content, and parsing ambiguity caused by natural language or configuration format difference is avoided; the AI model is used to group the function description statements in the page configuration input data, extract component semantic fields and associated attribute fields, obtain semantic structure data, and ensure that synonyms and morphological words can be mapped to unified semantic labels in the model, so as to eliminate the field confusion problem caused by semantic ambiguity; according to the semantic structure data, field mapping is performed on the component attribute structures of different UI frameworks, the component semantic fields are converted into unified description fields, standard semantic data is obtained, semantic isomorphism of component attributes and event interfaces between different UI frameworks is realized, and it is ensured that all components and attributes can be referenced, modified and combined in the same semantic space; according to the standard semantic data, templates are divided, the unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template, semantic template data is obtained, the attribute feature vectors of each semantic template are combined into a multi-dimensional feature matrix, according to the multi-dimensional feature matrix, interactive weight data is generated by counting the co-occurrence relationship between the attribute feature vectors, and the element values in the multi-dimensional feature matrix are adjusted according to the interactive weight data, a structure matrix is obtained, the limitation that the component attribute dependence can only be expressed at the text layer is eliminated, and data basis is provided for subsequent coupling and morphological analysis.

[0072] According to the structure matrix, the dynamic coupling strength between the component attributes in the structure matrix is calculated, a structure coupling quantity is obtained, a high-correlation attribute set and a weak-correlation attribute set are distinguished, and mathematical constraints are provided for subsequent mapping and morphological calculation and attribute screening; according to the structure matrix and the structure coupling quantity, the deformability of the mapping relationship between the component attributes in the structure matrix is calculated, a mapping deformation quantity is obtained, it is identified which attribute relationships are easy to distort when migrating and which relationships have stable correspondence, error mapping or incompatible attributes are avoided from being directly reused, and the semantic consistency and data continuity of the configuration generation process are ensured; according to the mapping deformation quantity, the attribute mapping relationship in the standard semantic data is updated, the unified conversion relationship between the attributes and events of the components between different UI frameworks is described, a mapping rule set is obtained, the component attribute relationship becomes rule data that can be called, and the system can realize automatic adaptation by calling existing rules when facing different frameworks or updated versions; a multi-UI component unified model is constructed according to the mapping rule set, and the page configuration input data is input into the multi-UI component unified model to generate a page configuration file, end-to-end traceable conversion from input semantics to rendering configuration file is realized, intermediate format conversion loss is avoided, and consistency and verifiability of the generated result are ensured.

[0073] The intermediate code is an intermediate page configuration representation form used in a page generation process, is located between page configuration input data and a specific UI framework page configuration file, and is used to carry semantic fields, attribute descriptions, event relationships of components in a page, and attribute mapping rules between different UI frameworks. The intermediate code exists in a structured data form, and its content does not directly correspond to a source code syntax of a specific UI framework, but through unified description fields, a multi-UI component unified model and a mapping rule set, abstractly expresses a page structure and component relationship, so that the intermediate code can be parsed, calculated and updated by an AI model, and is further converted to generate page configuration files adapted to different UI component frameworks. Therefore, the intermediate code in the application belongs to an intermediate representation layer facing page semantics and component structures, and is used to realize cross-UI framework page visualization generation, and does not constitute an identical relationship with intermediate code or executable program code in a compilation technology.

[0074] The AI model specifically includes:

[0075] The AI model in the application is a natural language processing model based on deep learning, and is used to automatically process page configuration input data, extract component semantic fields and attribute fields, and realize attribute mapping and adaptation between different UI frameworks. The core task of the AI model is to automatically identify components, attributes and their dependency relationships of a page from natural language descriptions provided by a user, so as to provide data support for subsequent cross-framework page configuration file generation. Main work of the AI model includes semantic extraction and labeling, semantic grouping and association inference, and attribute mapping and field generation. The principle of the AI model is based on current advanced natural language processing technology, especially deep learning and semantic representation models. The AI model uses a pre-trained language model such as BERT, RoBERTa or ERNIE based on deep learning to perform context-aware semantic understanding, and converts words in input text into semantic vectors. These word vectors reflect semantic relationships between words, so that the system can capture the internal relationship between components and attributes through semantic embedding. The AI model can automatically identify synonymous word fragments and classify them into the same category. Meanwhile, the AI model can analyze syntax structures, identify subject-predicate relationships and modification relationships, and determine which attributes belong to which components.

[0076] During the construction of the AI model, data collection and annotation are first required. The page configuration descriptions of different UI frameworks are collected, and each component and attribute is manually annotated to build a training data set. These annotated data will be used to train the AI model to recognize and extract components, attributes, and their dependencies in the page description. Model training usually relies on pre-trained language models such as BERT, which are pre-trained on a large amount of general text and then fine-tuned on specific domain data to ensure that the model can adapt to the specific task of page configuration description. During pre-training, the model learns general syntax and semantic knowledge, while in the fine-tuning stage, the model optimizes specific tasks according to the actual needs of page configuration, making it more suitable for extracting components and attributes from page configuration text. Through pre-training and fine-tuning, the AI model can accurately identify each component and its attributes in the page configuration, understand the dependencies between components, and automatically map these components and attributes to a unified semantic data structure. The model can also adapt to changes in different UI frameworks, automatically mapping and adjusting attributes between different frameworks to ensure consistent page rendering in different environments. The AI model also has functions such as anaphora resolution, semantic clustering, and field mapping, providing intelligent support for page generation and configuration. In different UI frameworks, the naming of component attributes, event handling mechanisms, and other differences often require manual adjustment and reconfiguration. The AI model can automatically handle these differences through intelligent inference and attribute mapping, making the generated page configuration file highly universal and portable.

[0077] The method comprises the following steps:

[0078] First, the front-end configuration interface receives the page configuration input data. This input data can come from various input forms, including natural language descriptions, configuration instructions generated by component dragging operations, or existing project exported page structure files. After receiving the input, a unified data access process is performed on the input content. The syntax of the input source is identified, the natural language description is segmented and dependency analyzed, and the structured input such as JSON, XML, or DSL description is parsed to extract meaningful sentence units.

[0079] In the natural language input scenario, the system uses a pre-trained semantic understanding model, such as a language parsing model based on the Transformer structure, to perform sentence segmentation and semantic labeling on the input sentence, identifying functional phrases in each sentence as component candidate units and identifying attribute or operation instruction phrases as attribute candidate units; for the visualization configuration scenario, the system directly reads the component configuration JSON structure generated by the interface dragging, and performs field verification and semantic label identification on the internal fields. Through syntax analysis and field identification of the input content, the system can convert the original input data into a data object containing sentence text, candidate component fragments, candidate attribute fragments, context position information, and other multi-dimensional features. Then the system enters the cleaning and normalization stage of the input data. The main task of this stage is to eliminate sentences unrelated to page generation from the input data, such as commentaries, logical hints, or non-functional descriptions. The cleaning algorithm automatically determines invalid fragments based on part-of-speech tagging and semantic category labels and filters them out, so that the remaining sentences all have structured meaning directly related to page construction. After cleaning, the system rearranges the data according to the order of sentence appearance to generate a sentence sequence dataset. This dataset provides a unified input interface for subsequent AI semantic models, and its format is a serialized text and semantic label pairing structure, where each data record contains sentence text, possible component candidates, attribute candidates, and position information index.

[0080] Among them, according to the semantic structure data, the component attribute structure of different UI frameworks is field mapped, the component semantic field is converted into a unified description field, and standard semantic data is obtained, specifically including:

[0081] The semantic structure data includes a plurality of semantic groups, each corresponding to a component semantic field and its associated attribute field. First, a preset multi-framework attribute structure reference library is read, which records the attribute names, type definitions and event binding methods of commonly used components under different UI frameworks, such as component attribute tables of frameworks such as React, Vue, Angular, etc. Each record in the reference library includes component name, attribute field name, field type, event interface and corresponding uniform semantic label. The system first matches the component semantic field in each semantic group, calculates the semantic similarity between the field and the framework component name through a word embedding model, and selects the highest similarity item as the candidate mapping target. If similar fields exist under multiple frameworks, the system further calculates the weighted average similarity according to the matching consistency of the context attribute fields to ensure the accuracy of semantic mapping. After determining the mapping target, the system reads the component attribute structure table under the corresponding framework, and compares the associated attribute fields extracted in the semantic group with the attribute names in the attribute structure table. A composite algorithm based on edit distance and semantic similarity is used in the comparison process to take into account both spelling differences and semantic approximations. For example, when the attribute field in the semantic group is a click event, the system can identify and match to the onClick field of the React framework or the v-on:click field of the Vue framework.

[0082] After completing the field matching, the system converts the matched attribute fields into a unified description field format according to the predefined uniform semantic field template. For example, onClick, v-on:click, and @click are uniformly mapped to the unified field clickEvent, and style.backgroundColor and v-bind:style.color are uniformly mapped to styleColorField, while recording the original source and target framework information of each field. For attribute fields that cannot be directly matched in the reference library, the system will call incremental learning to compare historical mapping records and context frequency to try to infer the possible uniform field corresponding to the attribute field from existing data; if it still cannot be matched, a temporary semantic field is generated and added to the uniform description template, which will be supplemented and improved by manual or model incremental training later. After mapping, all semantic structure data is converted into a standard semantic data set.

[0083] In a preferred embodiment of the present application, the AI model groups the function description statements in the page configuration input data, extracts the component semantic fields and associated attribute fields, and obtains semantic structure data, including:

[0084] According to the page configuration input data, the function description statements in the page configuration input data are arranged in order, and the statements unrelated to the page are removed to obtain statement sequence data.

[0085] According to the AI model, the sentences in the sentence sequence data are analyzed, each function description sentence is split into component candidate word segments and attribute candidate word segments, and semantic segment data is obtained;

[0086] According to the AI model, the association relationship between the component candidate word segments is inferred, the segments belonging to the same component are merged into the same semantic group, and semantic group data is obtained;

[0087] According to the semantic group data, the field representing the meaning of the component in each semantic group is determined and recorded as the component semantic field, and the attribute description field corresponding to the component semantic field is recorded as the associated attribute field, and semantic structure data is obtained.

[0088] In the embodiment of the application, according to the page configuration input data, the sentences are arranged in order, and the sentences irrelevant to the page are removed, to obtain the sentence sequence data, realizing the conversion from unstructured input text to ordered structured sentence set, ensuring the consistency and logical continuity of the input data, and laying a foundation for realizing page structuring generation; according to the AI model, the sentences in the sentence sequence data are analyzed, each function description sentence is split into component candidate word segments and attribute candidate word segments, and semantic segment data is obtained, realizing the decomposition from sentence-level description to semantic-level unit, so that the semantic information of each page element is expressed as an independent data segment in a fine-grained manner; according to the AI model, the association relationship between the component candidate word segments is inferred, the segments belonging to the same component are merged into the same semantic group, and semantic group data is obtained, realizing the automatic aggregation from semantic segments to component entities, providing consistent component boundaries for subsequent field extraction and semantic mapping; according to the semantic group data, the field representing the meaning of the component in each semantic group is determined and recorded as the component semantic field, and the attribute description field corresponding to the component semantic field is recorded as the associated attribute field, and semantic structure data is obtained, realizing the field storage of page description, so that the related attributes can be directly accessed by the field name without re-parsing the text.

[0089] According to the AI model, the sentences in the sentence sequence data are analyzed, each function description sentence is split into component candidate word segments and attribute candidate word segments, and semantic segment data is obtained, specifically including:

[0090] Firstly, the sentence is finely granulated by a semantic segmentation module based on deep learning. The module uses a context bidirectional encoding model such as BERT, ERNIE or RoBERTa to perform part-of-speech tagging and syntactic dependency analysis on each word in the sentence, enabling the system to identify subject-predicate structures, modification structures and functional phrases in the sentence. Then, a named entity recognition model is used to classify the segmented results by entity type, marking word segments representing functional entities, component entities or operation targets as component candidate word segments, and marking word segments containing attribute features, style descriptions, event trigger conditions, numerical parameters, etc. as attribute candidate word segments. In this process, the source sentence, position, and context dependency distance of the word segment are recorded, so that each segment is attached with traceable location information and semantic confidence. Then, through word vector similarity calculation, each word segment is matched with the vector space in the pre-constructed UI field semantic dictionary, enabling the system to identify the corresponding synonymous component words of buttons-btn, input boxes-input, and switch bars-tab in different natural language expressions, and aggregate them into a component segment candidate set. At the same time, attribute segments are also mapped to known attribute field spaces such as color-color, click event-clickEvent, and width-width through the same vector matching method. After completing entity recognition and semantic matching, the system performs dependency correction on the candidate segments, identifies the modification relationship and dependency direction between the segments through syntactic tree, for example, determines that blue in the blue button should be an attribute segment of the button, and finally all the segments that have completed semantic tagging, similarity matching and dependency binding are organized into semantic segment data.

[0091] Among them, the association relationship between the component candidate word segments is inferred according to the AI model, and the segments belonging to the same component are merged into the same semantic group to obtain semantic group data, which specifically includes:

[0092] The system startup component aggregation module performs correlation inference on all fragments marked as component candidate words. First, the semantic similarity between fragments is calculated based on the context information recorded in the semantic fragments. The system uses a vector comparison method based on cosine similarity to map each component fragment into a high-dimensional semantic space and calculate its similarity. At the same time, the model quantifies the distance between the source sentences of the fragments, constructs a sentence distance matrix, and judges whether the fragments belong to the same functional logic area. Then, by constructing a fragment co-occurrence matrix, the co-occurrence frequency of different component fragments in the same sentence or adjacent sentences is counted, and the parent-child nodes and dependency links in the syntax tree are weighted to make the component fragments with strong dependency relationship obtain higher correlation weight. After having the similarity matrix, distance matrix and co-occurrence matrix, the system uses hierarchical clustering algorithm to construct component grouping cluster. The clustering process will comprehensively calculate the similarity threshold, distance threshold and co-occurrence density between fragments, so that the system can automatically identify which fragments belong to the same component according to the semantic proximity. For example, fragments such as button, click button, main button, etc. may be highly similar in semantics, even if they appear in different sentences, they will be classified into the same component group because of the consistent context logic. The system also uses a coreference resolution model to handle cross-sentence references, such as pronouns like it and the control. This allows the model to backtrack to the original component fragment, ensuring that semantic clustering is not interrupted by changes in reference form. Finally, each clustering cluster is defined as an independent semantic group, and a unique identifier is assigned to the group, along with all related component fragments and their corresponding attribute fragment positions, forming semantic grouping data.

[0093] In a preferred embodiment of the present application, the template division is performed according to the standard semantic data, the unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template, and semantic template data is obtained. The attribute feature vectors of each semantic template are combined into a multi-dimensional feature matrix. According to the multi-dimensional feature matrix, the interaction weight data is generated by counting the co-occurrence relationship between the attribute feature vectors, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data to obtain a structure matrix, which includes:

[0094] The template division is performed according to the standard semantic data, the unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template, and semantic template data is obtained.

[0095] According to the semantic template data, the unified description fields in each semantic template are arranged in order as attribute field sequences, and the attribute field sequences are converted into attribute feature vectors to obtain feature vector data.

[0096] The multi-dimensional feature matrix is obtained by sequentially filling each attribute feature vector in the feature vector data into the preset matrix structure according to the component and field position.

[0097] According to the multi-dimensional feature matrix, interaction weight data is generated by counting the co-occurrence relationship between each attribute feature vector, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data to obtain a structure matrix.

[0098] In the embodiment of the application, the unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template according to the template division of the standard semantic data to obtain semantic template data, so as to realize the modularized organization mode of data, ensure that each component can be independently constructed for subsequent steps to build a feature vector, avoid the problems of field misplacement, vector fracture or repeated reference of data in the matrix filling stage; according to the semantic template data, the unified description fields in each semantic template are arranged in sequence as an attribute field sequence, and the attribute field sequence is converted into an attribute feature vector to obtain feature vector data, so as to ensure that different components and different fields can be uniformly calculated and processed, and avoid structural confusion and field calculation mismatch; by filling each attribute feature vector in the feature vector data into a preset matrix structure in sequence according to the component and field position, a multi-dimensional feature matrix is obtained, so as to avoid disordered splicing caused by inconsistent field numbers between components, make the entire data structure have a unified format that is scalable and computable, and lay a foundation for subsequent calculation; according to the multi-dimensional feature matrix, interaction weight data is generated by counting the co-occurrence relationship between each attribute feature vector, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data to obtain a structure matrix, so as to objectively reflect the co-dependence relationship between attributes, and enable the subsequent system to directly execute operation steps based on the matrix without repeatedly traversing the feature vector again.

[0099] According to the template division of the standard semantic data, the unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template to obtain semantic template data, which specifically includes:

[0100] First, each record in the standard semantic data is stored as a component attribute semantic entry, each record including at least component semantic field, uniform description field, page identification of the field, and appearance order identification of the field in the original configuration. A component index table is established according to the component semantic field, the component index table assigning a unique component identification to each different component semantic field and recording the uniform description field list corresponding to the component identification; then all records in the standard semantic data are traversed, for each record, the component semantic field is read, the corresponding component identification is found in the component index table, and the uniform description field in the record is appended to the field list corresponding to the component identification, when the uniform description field content is completely identical to the existing field in the list, it is judged as a repeated field through hash check or string comparison, and only one record is retained in the field list to eliminate redundant fields; in the field appending process, the appearance order of each uniform description field in the original page configuration input data, the function paragraph to which it belongs, and the semantic association label with other fields are recorded at the same time to form a field set with sequence information and association information. After the traversal is completed, the system performs a complete check on the field list corresponding to each component identification in the component index table, eliminates the empty set containing no uniform description field, and generates a semantic template identification for each non-empty field list, and encapsulates the component identification, semantic template identification, and corresponding uniform description field set as semantic template data.

[0101] Among them, according to the semantic template data, the uniform description fields in each semantic template are sorted into attribute field sequences according to the order, and the attribute field sequences are converted into attribute feature vectors to obtain feature vector data, specifically including:

[0102] Firstly, each semantic template data unit is processed respectively, the uniform description field set corresponding to the semantic template and the original appearance order, field category label and association marker of each field record in the set are read, a field sorting function is constructed according to a preset sorting rule, the sorting rule can determine the appearance order of the field in the original statement, the attribute type priority of the field in the component and the dependence order between the fields, and a strictly ordered attribute field sequence is output after the sorting function is executed. For each uniform description field in the attribute field sequence, the system parses the field name, field type, field value type, attribute category to which the field belongs such as style attribute, data binding attribute, event trigger attribute and reference marker of the field to other fields and other meta information from the field content, encodes the above meta information into numerical features, the encoding method can include assigning fixed dimension numbers to different field categories, generating binary features for Boolean attributes, generating sparse features for enumeration attributes, generating weight features for the association strength with other fields, and combining these features into an attribute feature vector according to a preset dimension order; when the feature dimension lengths generated by different fields are inconsistent, the system fills in zero values or default values for the missing dimensions, so that the dimension lengths of all attribute feature vectors remain consistent. Subsequently, all attribute feature vectors under the same semantic template are recorded as feature vector data in turn according to the field order.

[0103] Among them, according to the multi-dimensional feature matrix, the co-occurrence relationship between each attribute feature vector is counted to generate interaction weight data, and the element values in the multi-dimensional feature matrix are adjusted according to the interaction weight data to obtain a structure matrix, specifically including:

[0104] Firstly, each row or each column in the multi-dimensional feature matrix is regarded as a sequence of attribute feature vectors within a component, and by traversing any two attribute feature vectors in the sequence, the number of times they appear simultaneously in the same component and the number of times they appear together in different historical page configuration samples are counted to form basic co-occurrence count data; then, the basic co-occurrence count data is normalized to map the co-occurrence times to co-occurrence frequencies or conditional probabilities, and combined with the similarity index of the two attribute feature vectors in the feature space, such as the value obtained based on vector dot product or cosine similarity, to obtain interaction weight values reflecting the interaction closeness between attributes by combining the co-occurrence frequencies and the similarity values according to a preset function relationship, and the interaction weight values corresponding to each pair of attribute feature vectors are filled into the interaction weight matrix. The system adjusts the element values in the multi-dimensional feature matrix according to the interaction weight matrix, and when a certain attribute feature vector in the multi-dimensional feature matrix is taken as a target vector, other attribute vectors in the interaction weight matrix that have non-zero weights with the target vector are found, and these weights are taken as the adjacency coefficients of the target vector, and the co-occurrence relationship reflected by the interaction weights is superimposed into the matrix element where the target vector is located by weighting and superimposing or weighting and smoothing the feature values of the adjacent vectors, so that the target vector in the multi-dimensional feature matrix contains both its own attribute information and co-occurrence structure information with other attributes; the system performs the above weighting adjustment operation on all attribute feature vectors in the matrix in a loop, and after each adjustment, the result is subjected to numerical normalization processing or range truncation processing to ensure that the matrix element values are within a preset valid range, to obtain a structure matrix adjusted by the interaction weights.

[0105] In a preferred embodiment of the application, according to the structure matrix, the dynamic coupling strength between the component attributes in the structure matrix is calculated to obtain the structure coupling quantity, including:

[0106] According to the structure matrix and the interaction weight data, the joint action strength between the element values in the structure matrix and the weights in the interaction weight data of each attribute pair is calculated to obtain a local joint influence term;

[0107] According to the interaction weight data, the weight summary value of each attribute with the remaining attributes in the interaction weight data is calculated to obtain an overall connection scale term;

[0108] According to the local joint influence term and the overall connection scale term, the concentration degree of the attribute combination within the component under the constraint of the overall connection scale is calculated to obtain a compactness comparison relationship term;

[0109] The compactness comparison relationship term is nonlinearly transformed to map the compactness comparison relationship term to a comprehensive evaluation value within a limited numerical interval to obtain the structure coupling quantity.

[0110] In the embodiments of the present application, according to the structure matrix and the interaction weight data, the joint action strength of each attribute pair between the element value in the structure matrix and the weight in the interaction weight data is calculated to obtain a local joint influence item, which can accurately distinguish the relationship with close structure and low interaction weight or the relationship with frequent interaction and large structure value difference at the level of a single attribute pair, and lay a data foundation for subsequent construction of a more reasonable correlation evaluation in the overall component range; according to the interaction weight data, the weight summary value of each attribute in the interaction weight data and the remaining attributes is calculated to obtain an overall connection size item, which reflects the overall size of the connection formed by each attribute with other attributes in the component, and provides a necessary normalization reference for subsequent measurement of the concentration degree of attribute combination under a given constraint; according to the local joint influence item and the overall connection size item, the concentration degree of attribute combination in the component under the constraint of the overall connection size is calculated to obtain a compact comparison relationship item, which realizes the scale conversion from the influence of a single attribute pair to the aggregation degree of attribute combination, so that the system can subsequently uniformly evaluate the coupling degree of the attribute set at the component level; by performing nonlinear transformation on the compact comparison relationship item, the compact comparison relationship item is mapped to a comprehensive evaluation value in a limited numerical interval to obtain a structure coupling quantity, a constraint mapping deformation quantity calculation and attribute mapping relationship updating strategy, which realizes the reduction process from multiple source and multiple scale intermediate indicators to a single and bounded comprehensive indicator, so that the system has a clear component level coupling reference when performing cross-frame attribute mapping and rule updating.

[0111] According to the structure matrix and the interaction weight data, the joint action strength of each attribute pair between the element value in the structure matrix and the weight in the interaction weight data is calculated to obtain a local joint influence item, and the local joint influence item specifically includes:

[0112] Firstly, the joint action strength of attribute pairs is calculated based on the structure value and weight value recorded in the structure matrix and the interaction weight data. For each component in the page configuration model, the system traverses all attribute fields under the component and forms an attribute pair for each two different attributes. In the traversal process, the system first reads the corresponding matrix element value from the structure matrix according to the index position of the attribute pair. The matrix element value is used to represent the structural proximity or structural action strength of the attribute pair in the standard semantic vector space. Then the weight value corresponding to the attribute pair is read from the interaction weight data. The weight value reflects the frequency and importance of the co-occurrence or co-participation of the attribute pair in the semantic structure analysis and historical page configuration learning process. The system constructs a joint action record for each attribute pair, combines the element value obtained from the structure matrix and the weight value in the interaction weight data into a binary data item, and calculates the binary data item through a preset joint action function. The joint action function can use weighted summation, multiplication operation, weight-weighted nonlinear combination or other increasing functions to ensure that the structure value in the structure matrix and the weight value in the interaction weight data can jointly affect the calculation result. The system obtains the joint action strength of each attribute pair through the function and obtains the local joint influence item.

[0113] According to the interaction weight data, the weight summary value of each attribute with the remaining attributes in the interaction weight data is calculated to obtain the overall connection scale item, which specifically includes:

[0114] The overall connection scale description of the internal attributes of the component is established by counting the weight summary value between each attribute and other attributes in the component. First, for each component, the interaction weight data records belonging to the component are filtered, and the attribute fields in the records are used as the index basis to establish the mapping table of attributes and weights. For each attribute, the system traverses all attribute pairs involving the attribute in the interaction weight data and extracts all weight values related to the attribute. These weight values are processed according to the preset summary rule. The summary rule can be simple summation, which is used to represent the total connection scale between the attribute and other attributes; or it can be averaging, which is used to represent the overall level of connection between the attribute and other attributes; or it can use a weighted summary rule with suppression characteristics for abnormal values or discrete values, so that the overall connection scale item can stably reflect the connection degree of each attribute in the component. The overall connection scale value of each attribute is obtained through the summary rule and stored as the overall connection scale item in the component dimension.

[0115] According to the local joint influence item and the overall connection scale item, the concentration degree of the internal attribute combination of the component under the given overall connection scale constraint is calculated to obtain the compactness comparison relationship item, which specifically includes:

[0116] According to the attribute structure inside the component, the combined action strength values of all attribute pairs in the component are read from the local combined influence item data structure, and the overall connection scale values of the two attributes related to the attribute pair are read from the overall connection scale item. The system takes the combined action strength as the main variable, combines the two overall connection scale values as the normalization factor, so that the combined action value of each attribute pair is mapped to a relative scale that can reflect the concentration of the attribute combination set. The concentration calculation function can adopt the method of combined action strength divided by connection scale function value to avoid deviation of attribute pairs with high connection scale due to high combined action strength; a weighted normalization form can be adopted to combine the connection scale values of the two attributes into a normalization reference quantity, so that the combined action strength is reasonably compressed or enlarged under constraints. The concentration values of all attribute pairs in the component are counted, including the value distribution, concentration trend or smoothed comprehensive concentration value of the concentration, to obtain the compact contrast relationship item.

[0117] The compact contrast relationship item is mapped to a comprehensive evaluation value in a limited numerical interval by nonlinear transformation, and a structure coupling quantity is obtained, specifically including:

[0118] Firstly, the numerical range of the compact contrast relationship item between components is analyzed to determine whether there are extreme values, discrete peak values or dense distribution areas, so as to select an appropriate nonlinear transformation function. If the range of the compact contrast relationship item is large, the system can use a logarithmic compression function to compress the large value interval to a smaller range; if the compact contrast relationship item is concentrated in a small interval, the system can use an exponential stretching function to increase its resolution; if it is necessary to maintain the differentiation of the middle interval, the system can use a segmented function to obtain different mapping characteristics of the compact contrast relationship item in different intervals. The system takes the compact contrast relationship item as input and converts it into a structure coupling quantity in a preset interval such as 0 to 1 through a nonlinear function.

[0119] In a preferred embodiment of the present application, according to the structure matrix and the structure coupling quantity, the deformability of the mapping relationship between the component attributes in the structure matrix is calculated to obtain a mapping deformation quantity, including:

[0120] According to the structure matrix and the interaction weight data, the direction asymmetry degree of any two attributes in the same component under forward and reverse mapping is calculated to obtain a direction difference analysis item;

[0121] According to the structure matrix and the interaction weight data, the structure variation amplitude of the structure values between any two attributes in the same component is calculated to obtain a local deformation strength item;

[0122] According to the local deformation strength item and the structure coupling quantity, the overall deformation sensitivity of the local deformation strength of each attribute under the constraint of the structure coupling quantity is calculated to obtain a normalized deformation sensitivity item;

[0123] The direction difference analysis item and the normalized deformation sensitive item are nonlinearly transformed, the asymmetry degree of the mapping relationship of the direction difference analysis item in the direction is calculated, the overall variability of the normalized deformation sensitive item under the constraint of the structure coupling quantity is calculated, and the mapping deformation quantity is obtained.

[0124] In the embodiment of the application, according to the structure matrix and the interaction weight data, the direction asymmetry degree of any two attributes in the same component under forward and reverse mapping is calculated, and a direction difference analysis item is obtained, which describes the semantic shift, trigger mechanism difference or attribute dependency order change that may exist in the same attribute pair in forward and reverse mapping in different UI frameworks; according to the structure matrix and the interaction weight data, the structure change amplitude of the structure value between any two attributes in the same component is calculated, and a local deformation strength item is obtained, which quantifies the fluctuation degree of different attributes after being affected by external framework changes in the structure matrix, and objectively reflects the flexibility and stability of attribute mapping in the numerical structure; according to the local deformation strength item and the structure coupling quantity, the overall deformation sensitivity of the local deformation strength of each attribute under the constraint of the structure coupling quantity is calculated, and a normalized deformation sensitive item is obtained, which judges the real deformation sensitivity of the attribute under the constraint of the overall structure stability, and avoids false sensitivity judgment caused by local noise or accidental changes; the direction difference analysis item and the normalized deformation sensitive item are nonlinearly transformed, the asymmetry degree of the mapping relationship of the direction difference analysis item in the direction is calculated, and the overall variability of the normalized deformation sensitive item under the constraint of the structure coupling quantity is calculated, and a mapping deformation quantity is obtained, which identifies the shift degree of attribute mapping in different UI frameworks, and realizes batch generation and rapid update in a large-scale component attribute set, and provides a robust numerical basis for cross-framework page generation.

[0125] According to the structure matrix and the interaction weight data, the direction asymmetry degree of any two attributes in the same component under forward and reverse mapping is calculated, and a direction difference analysis item is obtained, which specifically includes:

[0126] First, the bidirectional mapping structure values between the attributes are read from the structure matrix. For the attributes A and B in the component, the system sequentially obtains the matrix element value representing A to B and the matrix element value representing B to A in the structure matrix, and further combines the interaction weight between A and B recorded in the interaction weight data as a reference. Based on the numerical difference, the direction difference calculation is performed to reflect the structure shift degree of the attributes in the forward structure mapping path and the reverse structure mapping path. In order to make the difference values of different attribute pairs comparable in the same order of magnitude, the system normalizes the direction difference value according to the weight value of the corresponding attribute pair in the interaction weight data, and converts the absolute difference value of the structure matrix into a direction asymmetry quantity constrained by the interaction intensity. Then the system aggregates and arranges the direction asymmetry quantities of each attribute pair in the same component to form a direction difference analysis item.

[0127] wherein, according to the structure matrix and the interaction weight data, a structure change range of the structure value between any two attributes within the same component is calculated, to obtain the local deformation strength item, specifically including:

[0128] First, for any attribute X in the component, all matrix units related to X in the structure matrix are comprehensively scanned, including the values of X to other attributes and the values of other attributes to X in the structure matrix. The system compares all the above related matrix values with the corresponding weight values in the interaction weight data, calculates the offset of each matrix value relative to the interaction weight expectation, and identifies the numerical change range of the structure relationship between the attributes. All offset values related to attribute X are accumulated and arranged by attribute dimension to form a local deformation strength value reflecting the overall change range between attribute X and other attributes. Then the system sequentially completes the above calculation according to the attribute list order within the component, so that all attributes in the component obtain their corresponding local deformation strength values, and these values are integrated into the local deformation strength item at the component level.

[0129] wherein, according to the local deformation strength item and the structure coupling amount, the overall deformation sensitivity of the local deformation strength of each attribute under the constraint of the structure coupling amount is calculated, to obtain the normalized deformation sensitivity item, specifically including:

[0130] First, read the local deformation strength item and combine the structure coupling amount to standardize the local strength value with the structure coupling amount as a global constraint. Take the component concentration corresponding to the structure coupling amount as the normalization factor, and calculate the local deformation strength value and the structure coupling amount in proportional relationship, so that the deformation sensitivity value can reflect both the local change characteristics and the stability of the overall structure of the component. For attributes with large local change range, if the structure coupling amount of the component to which they belong is strong, the system will suppress the excessive expansion of their sensitivity through the normalization factor; for attributes with small local change range, in components with weak overall coupling, their normalized sensitivity will present higher calculability, reflecting the deformation possibility brought by loose component structure. The system finally arranges the normalized sensitivity values in order of attributes to form the normalized deformation sensitivity item.

[0131] wherein, the direction difference analysis item and the normalized deformation sensitivity item are nonlinearly transformed, the asymmetric degree of the mapping relationship of the direction difference analysis item in the direction is calculated, and the overall variability of the normalized deformation sensitivity item under the constraint of the structure coupling amount is calculated, to obtain the mapping deformation amount, specifically including:

[0132] First, the direction difference analysis item and the normalized deformation sensitive item are respectively input as values representing the direction deviation degree and the structure sensitivity, and a preset nonlinear mapping function is selected to jointly calculate the two values. The nonlinear function can be an S-shaped function, an exponential function, a logarithmic stretching function or other function forms that can convey the structure gradient change. When the direction difference is small or the local sensitivity is low, the value is stable, and when the deviation is significant or the sensitivity is high, the output presents a continuous change that can express the mapping deformation trend across the framework. The system performs nonlinear transformation on the direction difference analysis item, so that the direction difference obtains a value weight corresponding to its deviation degree in the mapping deformation, and at the same time, the nonlinear interaction operation is performed on the normalized deformation sensitive item, and the normalized sensitive value is converted into the structure deformation expressing the overall variable structure characteristics. After the fusion calculation of the two types of data, the mapping deformation is output.

[0133] In a preferred embodiment of the present application, according to the mapping deformation, the attribute mapping relationship in the standard semantic data is updated, the unified conversion relationship between the attributes and events of different UI framework components is described, and a mapping rule set is obtained, including:

[0134] According to the mapping deformation, the attribute mapping relationship of each component in the standard semantic data is marked, and the attribute mapping relationship and the mapping deformation of the corresponding component are associated to obtain mapping update data.

[0135] According to the mapping update data, the attribute mapping relationships in the same numerical interval are merged into the same mapping group to obtain mapping group data.

[0136] According to the mapping group data, the source attribute field and the target attribute field of the attribute mapping relationship in each mapping group are reserved, replaced and reorganized to obtain mapping adjustment data.

[0137] According to the mapping adjustment data, a rule identifier is generated for each attribute mapping relationship, and the rule identifier and the attribute mapping relationship are archived and arranged according to the components and event types to obtain the mapping rule set.

[0138] In the embodiment of the present application, according to the mapping deformation amount, the attribute mapping relationship of each component in the standard semantic data is marked, the attribute mapping relationship is associated with the mapping deformation amount of the corresponding component, the mapping update data is obtained, the difference mapping relationship between different components is accurately identified, and a structured data basis is laid for subsequent mapping grouping and adjustment operation; according to the mapping update data, the attribute mapping relationship in the same numerical interval of the mapping deformation amount is merged into the same mapping group, the mapping grouping data is obtained, the conversion from continuous numerical value to discrete structure is realized, the system can be batch processed in a grouped manner, and the organization and consistency of mapping adjustment are improved; according to the mapping grouping data, the source attribute field and the target attribute field of the attribute mapping relationship in each mapping group are reserved, replaced and reorganized, the mapping adjustment data is obtained, the structure continuity and field consistency of the mapping relationship in the unified model are ensured, and field mismatch or event calling error caused by high deformation mapping in subsequent generation is avoided; according to the mapping adjustment data, a rule identifier is generated for each attribute mapping relationship, and the rule identifier and the attribute mapping relationship are archived and arranged according to the component and the event type, the mapping rule set is obtained, the system can accurately find the mapping relationship suitable for the semantic field of the current component in the unified model construction stage, and the mapping relationship between different components and different event types is ensured not to interfere with each other, and a deterministic rule entry is provided for subsequent unified model input.

[0139] In the embodiment of the present application, according to the mapping deformation amount, the attribute mapping relationship of each component in the standard semantic data is marked, the attribute mapping relationship is associated with the mapping deformation amount of the corresponding component, the mapping update data is obtained, the difference mapping relationship between different components is accurately identified, and a structured data basis is laid for subsequent mapping grouping and adjustment operation; according to the mapping update data, the attribute mapping relationship in the same numerical interval of the mapping deformation amount is merged into the same mapping group, the mapping grouping data is obtained, the conversion from continuous numerical value to discrete structure is realized, the system can be batch processed in a grouped manner, and the organization and consistency of mapping adjustment are improved; according to the mapping grouping data, the source attribute field and the target attribute field of the attribute mapping relationship in each mapping group are reserved, replaced and reorganized, the mapping adjustment data is obtained, the structure continuity and field consistency of the mapping relationship in the unified model are ensured, and field mismatch or event calling error caused by high deformation mapping in subsequent generation is avoided; according to the mapping adjustment data, a rule identifier is generated for each attribute mapping relationship, and the rule identifier and the attribute mapping relationship are archived and arranged according to the component and the event type, the mapping rule set is obtained, the system can accurately find the mapping relationship suitable for the semantic field of the current component in the unified model construction stage, and the mapping relationship between different components and different event types is ensured not to interfere with each other, and a deterministic rule entry is provided for subsequent unified model input.

[0140] First, the component-level attribute mapping entries are read from the standard semantic data one by one, and the source attribute field, the target attribute field and the event field of each mapping entry are packaged into an independent data record unit. The system then reads the deformation value corresponding to the component from the mapping deformation obtained in the structure matrix calculation stage, and writes the deformation value into the deformation identifier field of the data record unit with the field correspondence as the index. This writing action does not change the field structure of the original attribute mapping relationship, but extends the record unit in the form of a new field, so that the record unit is expanded from the source field to the target field to the event field structure to the source field to the target field to the event field to the deformation identifier field structure. In the marking process, a one-to-one field mapping method is adopted to ensure that each attribute mapping entry can be accurately associated with the deformation amount corresponding to the component. After the marking is completed, the system aggregates all the marked mapping record units into a unified data set, and writes them into the memory storage structure with the component as the index key, and finally forms the mapping update data.

[0141] In the embodiment of the present application, according to the mapping deformation amount, the attribute mapping relationship of each component in the standard semantic data is marked, the attribute mapping relationship is associated with the mapping deformation amount of the corresponding component, the mapping update data is obtained, the difference mapping relationship between different components is accurately identified, and a structured data basis is laid for subsequent mapping grouping and adjustment operation; according to the mapping update data, the attribute mapping relationship in the same numerical interval of the mapping deformation amount is merged into the same mapping group, the mapping grouping data is obtained, the conversion from continuous numerical value to discrete structure is realized, the system can be batch processed in a grouped manner, and the organization and consistency of mapping adjustment are improved; according to the mapping grouping data, the source attribute field and the target attribute field of the attribute mapping relationship in each mapping group are reserved, replaced and reorganized, the mapping adjustment data is obtained, the structure continuity and field consistency of the mapping relationship in the unified model are ensured, and field mismatch or event calling error caused by high deformation mapping in subsequent generation is avoided; according to the mapping adjustment data, a rule identifier is generated for each attribute mapping relationship, and the rule identifier and the attribute mapping relationship are archived and arranged according to the component and the event type, the mapping rule set is obtained, the system can accurately find the mapping relationship suitable for the semantic field of the current component in the unified model construction stage, and the mapping relationship between different components and different event types is ensured not to interfere with each other, and a deterministic rule entry is provided for subsequent unified model input.

[0142] First, an interval division table is established according to the numerical range of the deformation amount of the mapping, which can be divided into several continuous sections according to needs, for example, at fixed intervals, according to statistical distribution or at preset threshold sections. When the system traverses each mapping record unit, the value of the deformation identification field thereof is the retrieval condition, which is compared with each section in the interval division table, and it is classified into the corresponding group set according to the position of the section thereof. The field format of the records in the group is kept consistent during the classification process, that is, all the mapping record units classified into the same group must contain the source attribute field, the target attribute field, the event field and the deformation identification field. The order table is established in the group by sorting the field name and the field type, so that each group contains all the mapping records meeting the interval condition and has consistent access order and fixed field reading path. Finally, the group set corresponding to each numerical section is encapsulated as mapping group data.

[0143] Among them, according to the mapping group data, the source attribute field and the target attribute field of the attribute mapping relationship in each mapping group are reserved, replaced and recombined to obtain mapping adjustment data, which specifically includes:

[0144] First, each mapping group is read in sequence, and different processing methods are performed according to the deformation amount section characteristics of the group. For the mapping group with a low deformation amount, the system regards the source attribute field and the target attribute field as stable mapping relationship, does not change the structure, but directly copies the original mapping relationship to the mapping adjustment data as a reserved mapping entry. For the mapping group with a medium deformation amount, the system will find the target field matching the semantic of the source attribute field according to the unified description field in the standard semantic data, and use the judgment result of semantic consistency as the replacement basis to replace the original target attribute field with a new target field, and generate a new mapping record entry to write into the mapping adjustment data as a replacement mapping entry. For the mapping group with a high deformation amount, the system records the source attribute field and the target attribute field as a rearrangeable field set, respectively, restructures the new attribute corresponding relationship by comparing the field combination mode in the same group mapping entry, that is, recombines the high deformation field into a mapping mode with a structure more suitable for unified model calling, and writes it into the mapping adjustment data as a recombination mapping entry. The three-segment structure of the source field, the target field and the event field is always maintained during the operation process, and the mapping adjustment data is obtained.

[0145] Among them, according to the mapping adjustment data, a rule identifier is generated for each attribute mapping relationship, and the rule identifier and the attribute mapping relationship are archived and arranged according to components and event types to obtain a mapping rule set, which specifically includes:

[0146] First, the determined mapping entries are taken out one by one, and a rule identifier is automatically generated according to the component semantic field, the event type field and the position in the adjustment sequence in each mapping entry. The identifier is generated in the splicing manner of component identifier-event type-sequence number, so that each mapping rule can be retrieved by the system through the unique identifier. An archiving directory structure is established according to the component hierarchy, and a two-level directory is further established under each component directory according to the event type, and the mapping entries with rule identifiers are written into the corresponding directory nodes. During the writing process, the system keeps all the mapping rules under the event type arranged in sequence according to the sequence number, ensuring that all rule entries under the same event can be called in a determined sequence. Finally, all the sorted rule entries constitute a mapping rule set, which is logically stored in a tree structure with components as root nodes, event types as branch nodes and rule identifiers as leaf nodes.

[0147] In a preferred embodiment of the present application, a multi-UI component unified model is constructed according to the mapping rule set, and page configuration input data is input into the multi-UI component unified model to generate a page configuration file, including:

[0148] According to the mapping rule set, the unified description field of each component is associated with the corresponding attribute mapping relationship in the mapping rule set to obtain component mapping template data;

[0149] According to the component mapping template data, the source attribute field and the target attribute field corresponding to the same component semantic field are combined into a group of parameter descriptions to obtain model parameter data;

[0150] According to the model parameter data, each group of parameter descriptions is hierarchically organized according to the component semantic field and the event type, and is assembled into a model structure that can be called to obtain a multi-UI component unified model;

[0151] By matching the component semantic field in the page configuration input data with the model structure in the multi-UI component unified model, page configuration analysis data is generated, and the model parameters of each component are filled and arranged according to the page configuration analysis data to obtain a page configuration file.

[0152] In this embodiment of the invention, based on the mapping rule set, the unified description fields of each component are associated with the corresponding attribute mapping relationships in the mapping rule set to obtain component mapping template data. This avoids ambiguity caused by field name matching during dynamic parsing, and achieves the searchability, traceability, and reusability of the mapping logic, providing a stable data foundation for the parameter organization of model construction. Based on the component mapping template data, the source attribute fields and target attribute fields corresponding to the semantic fields of the same component are combined into a set of parameter descriptions to obtain model parameter data. This avoids the retrieval complexity caused by string matching, provides a stable input interface, and allows the subsequent model layer to directly reference this parameter structure when constructing the internal logic of the component, achieving the encapsulation and controllability of parameter data. Based on the model parameter data, each set of parameter descriptions is arranged according to the component. Semantic fields and event types are hierarchically organized and assembled into a callable model structure to obtain a unified model for multiple UI components. This avoids data access discrepancies caused by changes in field positions when calling across frames, and achieves unified storage and unified retrieval of cross-frame attribute conversion. This allows the page configuration generation process to complete parameter filling operations for different frames based on the same model. By matching the component semantic fields in the page configuration input data with the model structure in the unified model for multiple UI components, page configuration parsing data is generated. Based on this, the model parameters of each component are filled and arranged to obtain the page configuration file. This provides complete field placeholders and parameter mapping information for the serialization stage, avoiding problems such as missing fields or incomplete configurations, completing the conversion between cross-frame fields, and making the output file available across frames.

[0153] Specifically, based on the mapping rule set, the unified description field of each component is associated with the corresponding attribute mapping relationship in the mapping rule set to obtain component mapping template data, which includes:

[0154] First, the system needs to match and associate the unified description fields of each UI component with their corresponding attribute mapping relationships. This involves iterating through the description fields of each component and searching for matching attribute and event fields in the mapping rule set. For each successfully matched field, the system will associate the source field (the field in the current UI framework) with the target field (the field in the target UI framework) according to the mapping relationship defined by the mapping rule, forming a complete attribute mapping table. The system first extracts the description fields of each component from standard semantic data and groups them by component type. Each component field is compared one by one with the rules in the mapping rule set to find the most suitable mapping rule. These rules define the field mapping relationships between different UI frameworks, including attribute names, data types, event handling, etc. After completing the matching of all fields, the system will create a component mapping template. This template contains the field mapping relationships of each component, ensuring consistent mapping of components across different UI frameworks.

[0155] According to the component mapping template data, source attribute fields and target attribute fields corresponding to the same component semantic field are combined into a set of parameter descriptions to obtain model parameter data, specifically including:

[0156] First, according to the attribute fields and event fields of each component in the mapping template, it is identified which fields belong to the same component, and they are combined according to the logical relationship. Each parameter description is composed of a pair of source fields, i.e. fields in UI framework A and target fields, i.e. fields in UI framework B. These fields may include attribute fields such as the text content of a text box, the click event of a button or event fields such as user clicks, scroll events, etc. The system combines these fields into a parameter pair, ensuring that each parameter pair can be consistently processed and converted in different frameworks. For event fields, the system ensures that the correspondence between source events and target event types is correctly established. If multiple attributes of a component have different event handling requirements, an independent description is generated for each event field to ensure that it can be managed and called independently during subsequent conversion. Through this parameter combination, source attribute fields and target attribute fields are stored in a structured manner to generate model parameter data.

[0157] According to the model parameter data, each set of parameter descriptions is hierarchically organized according to the component semantic field and event type, and assembled into a callable model structure to obtain a unified model of multiple UI components, specifically including:

[0158] First, according to the semantic field of each component, all parameter descriptions are classified by component. Each component will form an independent hierarchical structure containing all attribute fields, event fields related to the component and their mapping relationships. In the hierarchical structure of each component, parameter fields are further classified according to event types. For each event type such as click event, hover event, etc., the system generates an independent sub-module to group all parameters related to the event. This hierarchical organization makes the relationship between the attributes and events of each component clear and visible, and lays the foundation for subsequent model structure assembly. Then, according to these hierarchically organized parameter descriptions, they are assembled into a callable unified model structure. This structure will contain all component semantic fields, event fields and their corresponding parameter mapping information. The attribute fields, event fields of each component and their mapping rules will be reflected in the model structure. Through this hierarchical organization and assembly process, it is ensured that the model structure can clearly represent the attributes, events of each component and their mapping relationships, while ensuring the scalability and reusability of the model, providing a basic data model for the final page configuration generation.

[0159] Wherein, by matching the component semantic field in the page configuration input data with the model structure in the multi-UI component unified model, page configuration analysis data is generated, and the model parameters of each component are filled and arranged according to the page configuration analysis data to obtain a page configuration file, which specifically includes:

[0160] First, according to the semantic field of each component provided in the page configuration input data, the corresponding component node in the model structure is found and matched. Under each component node, the system will fill the corresponding parameter field and event field to the specified position through the defined model parameter data. For each component, the system will fill the model parameters according to the field values provided in the page configuration input data, such as the default text value of the text box and the click event of the button. In this process, it is ensured that the field values in the input data are consistent with the parameter description in the model, avoiding data mismatch or omission. When the parameter values of all components are filled, the system will sort and arrange the fields of all components to ensure that the generated page configuration file meets the specifications of the target UI framework, and handle the mutual relationship between components to ensure that the generated configuration file can be correctly rendered in different frameworks. Finally, all filled component parameters are organized into a complete page configuration file, which contains the attributes, events and corresponding values of all components, and can be directly used for page generation or UI framework rendering. In this way, the user input data is deeply bound with the parameters in the model structure to generate a page configuration file that meets the requirements of the target framework.

[0161] Embodiments of the application also provide an AI-based multi-UI component framework code visualization page generation system, which comprises:

[0162] A data module for obtaining page configuration input data;

[0163] A semantic module for grouping the function description statements in the page configuration input data through an AI model, extracting component semantic fields and associated attribute fields, and obtaining semantic structure data;

[0164] A mapping module for field mapping of component attribute structures of different UI frameworks according to the semantic structure data, converting the component semantic fields into unified description fields, and obtaining standard semantic data;

[0165] A matrix module for template division according to the standard semantic data, merging the unified description fields belonging to the same component in the standard semantic data into the same semantic template, obtaining semantic template data, combining the attribute feature vectors of each semantic template into a multi-dimensional feature matrix, generating interaction weight data according to the co-occurrence relationship between the attribute feature vectors, and adjusting the element values in the multi-dimensional feature matrix according to the interaction weight data to obtain a structure matrix;

[0166] The coupling module is configured to calculate dynamic coupling strength between component attributes in the structure matrix according to the structure matrix, and obtain a structure coupling quantity;

[0167] The deformation module is configured to calculate a deformable degree of the mapping relationship between the component attributes in the structure matrix according to the structure matrix and the structure coupling quantity, and obtain a mapping deformation quantity;

[0168] The rule module is configured to update the attribute mapping relationship in the standard semantic data according to the mapping deformation quantity, describe a unified conversion relationship between attributes and events of different UI framework components, and obtain a mapping rule set.

[0169] The model module is configured to construct a multi-UI component unified model according to the mapping rule set, input the page configuration input data into the multi-UI component unified model, and generate a page configuration file.

[0170] It should be noted that the system is a system corresponding to the above method, and all implementation manners in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0171] Embodiments of the present application also provide a computing device, comprising a processor and a memory storing a computer program, wherein the computer program is executed by the processor to perform the method described above. All implementation manners in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0172] Embodiments of the present application also provide a computer readable storage medium storing instructions, which, when executed on a computer, cause the computer to perform the method described above. All implementation manners in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0173] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the scope of protection of the present application.

Claims

1. A method for generating visual pages using a multi-UI component framework based on AI, characterized in that, The method includes: Get the page configuration input data; The AI ​​model groups the functional description statements in the page configuration input data, extracts component semantic fields and related attribute fields, and obtains semantic structure data. Based on semantic structure data, field mapping is performed on the component attribute structure of different UI frameworks, and component semantic fields are transformed into unified description fields to obtain standard semantic data; Templates are divided based on standard semantic data. Unified description fields belonging to the same component in the standard semantic data are merged into the same semantic template to obtain semantic template data. The attribute feature vectors of each semantic template are combined into a multidimensional feature matrix. Based on the multidimensional feature matrix, the co-occurrence relationship between each attribute feature vector is statistically analyzed to generate interaction weight data. The element values ​​in the multidimensional feature matrix are adjusted based on the interaction weight data to obtain the structure matrix. Based on the structure matrix, calculate the dynamic coupling strength between component attributes in the structure matrix to obtain the structural coupling amount; Based on the structural matrix and structural coupling, the deformability of the mapping relationship between component attributes in the structural matrix is ​​calculated to obtain the mapping deformation amount; Based on the mapping deformation, the attribute mapping relationship in the standard semantic data is updated to describe the unified conversion relationship between attributes and events between different UI framework components, thus obtaining a mapping rule set; A unified model for multiple UI components is constructed based on the mapping rule set, and the page configuration input data is input into the unified model for multiple UI components to generate the page configuration file. Based on the structure matrix and structural coupling, the deformability of the mapping relationship between component attributes in the structure matrix is ​​calculated to obtain the mapping deformation amount, including: Based on the structural matrix and interaction weight data, the degree of directional asymmetry between any two attributes within the same component under forward and reverse mapping is calculated to obtain the directional difference analysis term. Based on the structural matrix and interaction weight data, the structural change amplitude of the structural values ​​between any two attributes within the same component is calculated to obtain the local deformation strength term. Based on the local deformation strength term and the structural coupling amount, the overall deformation sensitivity of the local deformation strength of each attribute under the constraint of the structural coupling amount is calculated, and the normalized deformation sensitivity term is obtained. Nonlinear transformations are performed on the directional difference analysis term and the normalized deformation sensitivity term to calculate the degree of asymmetry in the mapping relationship of the directional difference analysis term in the direction, and the overall variability of the normalized deformation sensitivity term under the constraint of structural coupling is calculated to obtain the mapped deformation amount. The code is located between the page configuration input data and the specific UI framework page configuration file. It is used to carry the semantic fields, attribute descriptions, event relationships, and attribute mapping rules between different UI frameworks of each component in the page.

2. The method for generating a code-visualized page based on an AI-driven multi-UI component framework according to claim 1, characterized in that, The AI ​​model groups the function description statements in the page configuration input data, extracts component semantic fields and related attribute fields, and obtains semantic structure data, including: The statements are organized according to the page configuration input data. The function description statements in the page configuration input data are arranged in order, and statements that are irrelevant to the page are removed to obtain statement sequence data. Based on the analysis of statements in the statement sequence data by the AI ​​model, the statements describing each function are split into component candidate word fragments and attribute candidate word fragments to obtain semantic fragment data; Based on the AI ​​model's inference of the relationships between candidate word fragments of components, fragments belonging to the same component are grouped into the same semantic group to obtain semantic grouping data; Based on the semantic grouping data, the fields representing the meaning of the components in each semantic group are determined and recorded as component semantic fields, and the attribute description fields corresponding to the component semantic fields are recorded as associated attribute fields, thus obtaining semantic structure data.

3. The method for generating a code-visualized page based on an AI-based multi-UI component framework according to claim 2, characterized in that, Template segmentation is performed based on standard semantic data. Unified description fields belonging to the same component in the standard semantic data are grouped into a single semantic template, resulting in semantic template data. The attribute feature vectors of each semantic template are combined into a multidimensional feature matrix. Based on the multidimensional feature matrix, the co-occurrence relationships between the attribute feature vectors are statistically analyzed to generate interaction weight data. The element values ​​in the multidimensional feature matrix are then adjusted based on the interaction weight data to obtain a structure matrix, including: Based on the semantic template data, the unified description fields in each semantic template are arranged in order into an attribute field sequence, and the attribute field sequence is converted into an attribute feature vector to obtain feature vector data; A multidimensional feature matrix is ​​obtained by filling the feature vectors of each attribute in the feature vector data into a preset matrix structure according to the component and field positions.

4. The method for generating a code-visualized page based on an AI-based multi-UI component framework according to claim 3, characterized in that, Based on the structure matrix, calculate the dynamic coupling strength between component attributes in the structure matrix to obtain the structural coupling amount, including: Based on the structure matrix and interaction weight data, calculate the joint effect strength between the element values ​​of each attribute pair in the structure matrix and the weights in the interaction weight data to obtain the local joint influence term; Based on the interaction weight data, calculate the sum of the weights of each attribute and the other attributes in the interaction weight data to obtain the overall connection scale item; Based on the local joint influence term and the overall connectivity scale term, the degree of concentration of the combination of internal attributes of the component is calculated under a given overall connectivity scale constraint, and the compact contrast relationship term is obtained. By performing a nonlinear transformation on the compact contrast relation term, the compact contrast relation term is mapped to a comprehensive evaluation value within a limited numerical range, thus obtaining the structural coupling quantity.

5. The method for generating a code-visualized page based on an AI-based multi-UI component framework according to claim 4, characterized in that, Based on the mapping deformation, the attribute mapping relationships in the standard semantic data are updated to describe the unified conversion relationship between attributes and events between different UI framework components, resulting in a mapping rule set, including: Based on the mapping deformation, the attribute mapping relationships of each component in the standard semantic data are marked, and each attribute mapping relationship is associated with the mapping deformation of the corresponding component to obtain the mapping update data; Based on the mapping update data, attribute mapping relationships with mapping deformation values ​​within the same numerical range are grouped into the same mapping group to obtain mapping group data; Based on the mapping grouping data, the source attribute fields and target attribute fields of the attribute mapping relationship in each mapping group are retained, replaced, and reorganized to obtain the mapping adjustment data; Based on the mapping adjustment data, rule identifiers are generated for each attribute mapping relationship, and the rule identifiers and attribute mapping relationships are archived and organized according to component and event type to obtain the mapping rule set.

6. The method for generating a code-visualized page based on an AI-based multi-UI component framework according to claim 5, characterized in that, A unified model for multiple UI components is constructed based on the mapping rule set, and page configuration input data is input into the unified model for multiple UI components to generate a page configuration file, including: Based on the mapping rule set, the unified description field of each component is associated with the corresponding attribute mapping relationship in the mapping rule set to obtain the component mapping template data; Based on the component mapping template data, the source attribute fields and target attribute fields corresponding to the semantic fields of the same component are combined into a set of parameter descriptions to obtain the model parameter data; Based on the model parameter data, the parameter descriptions of each group are organized hierarchically according to the component semantic fields and event types, and then assembled into a callable model structure to obtain a unified model for multiple UI components. By matching the component semantic fields in the page configuration input data with the model structure in the unified model of multiple UI components, page configuration parsing data is generated. Based on this data, the model parameters of each component are populated and arranged to obtain the page configuration file.

7. A code visualization page generation system based on an AI-driven multi-UI component framework, characterized in that, The system is used to perform the method as described in any one of claims 1 to 6, the system comprising: The data module is used to obtain page configuration input data; The semantic module is used to group the functional description statements in the page configuration input data using an AI model, extract component semantic fields and related attribute fields, and obtain semantic structure data. The mapping module is used to map the component attribute structure of different UI frameworks based on semantic structure data, and to convert the component semantic fields into unified description fields to obtain standard semantic data. The matrix module is used to divide templates based on standard semantic data. It merges the unified description fields of the same component in the standard semantic data into the same semantic template to obtain semantic template data. It combines the attribute feature vectors of each semantic template into a multi-dimensional feature matrix. Based on the multi-dimensional feature matrix, it calculates the co-occurrence relationship between each attribute feature vector to generate interaction weight data. Based on the interaction weight data, it adjusts the element values ​​in the multi-dimensional feature matrix to obtain the structure matrix. The coupling module is used to calculate the dynamic coupling strength between component attributes in the structure matrix based on the structure matrix, and obtain the structural coupling amount. The deformation module is used to calculate the degree of deformability of the mapping relationship between component attributes in the structural matrix based on the structural matrix and the structural coupling amount, and to obtain the mapping deformation amount. The rules module is used to update the attribute mapping relationship in the standard semantic data according to the mapping deformation, describe the unified conversion relationship between attributes and events between different UI framework components, and obtain the mapping rule set; The model module is used to build a unified model for multiple UI components based on the mapping rule set, and input page configuration data into the unified model for multiple UI components to generate page configuration files.

8. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 6.

Images