A large model-based data filling automation system and method

Abstract

The application discloses a kind of data filling automation system and method based on large model, it is related to electric digital data processing field, specifically includes: real-time monitoring and gathering the input behavior of user in different platform form, constructs semantic state vector, aggregates and forms semantic global vector;Structural disturbance tensor model is constructed, the stable position of each field is modeled, and the stable state representation vector of each target field under disturbance is obtained by tensor projection mode;According to the matching relationship between semantic global vector and structure mapping result, construct path cost function, solve the optimal execution path of field filling, according to path order drive browser to carry out automatic filling, realize semantic reasoning and data filling automation operation under structural disturbance. Realize the automation of cross-platform data filling, intelligent, effectively improve the efficiency of filling, with platform independence, semantic adaptability and structure fault tolerance ability, ensure data consistency and stability of filling process.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electric digital data processing, and in particular to a data reporting automation system and method based on a large model. BACKGROUND

[0002] With the continuous advancement of informatization construction, government affairs, enterprises, medical treatment, education and other industries widely use various business management platforms for data reporting, process approval and result aggregation. Users need to frequently perform data input, field matching and form submission operations in different systems, especially in multi-department cooperation and cross-system linkage scenarios. Data reporting has become a core link in daily work. At the same time, with the improvement of data security and consistency requirements, the reporting process not only requires efficiency and accuracy, but also needs to have strict constraints on data content, structure and logical order.

[0003] Due to significant differences in platform form structure, field naming, loading method and interaction logic, how to realize cross-platform data recognition, semantic unification, structure mapping and automatic reporting control under the premise of ensuring safety and stability has become a key problem to be solved in the current data processing field. Under this background, it has important practical significance and application value to build an automatic reporting system with scalability, fault tolerance and universality by combining the semantic reasoning ability of a large model and electric digital data processing methods.

[0004] The prior art has at least the following technical problems: most of the prior art relies on the interface or template import function provided by the platform, and cannot adapt to platforms without interfaces or with complex page structures, so the scope of application is limited; the prior art lacks intelligent analysis and adaptive ability when dealing with complex page behaviors such as field dependency, form linkage and asynchronous loading, and it is difficult to realize dynamic planning and fault tolerance control of the automatic reporting path, so it cannot guarantee the accuracy and stability of the reporting process. SUMMARY

[0005] The present application provides a data reporting automation system and method based on a large model to address the deficiencies in the prior art.

[0006] The present application provides a data reporting automation system and method based on a large model to address the deficiencies in the prior art.

[0007] A data reporting automation method based on a large model includes the following steps:

[0008] S1, real-time monitoring and collecting user input behavior in different platform forms, constructing a semantic state vector integrating semantic dimension, structure environment perception dimension and time behavior dimension, and aggregating to form a semantic global vector;

[0009] S2. To address the dynamic changes in the target page structure during loading, a structural perturbation tensor model is constructed. The stable position of each field is modeled, and the stable state representation vector of each target field under perturbation is obtained through tensor projection, providing a structural basis for data entry path planning.

[0010] S3. Based on the matching relationship between the semantic global vector and the structural mapping result, construct the path cost function, solve the optimal execution path for field filling, and drive the browser to automatically fill in the data according to the path order, so as to realize the automated operation of data filling under semantic reasoning and structural perturbation.

[0011] Preferably, in step S1, a semantic state vector is constructed that integrates the semantic dimension, the structural environment perception dimension, and the temporal behavior dimension, and its semantic state vector is defined as follows:

[0012] ;

[0013] in, It is the semantic state vector of the i-th field; the first term represents the semantic fit between the field label and the user input value; It is the set of labels for the i-th field, containing all semantic auxiliary information. It represents the number of tags in the i-th field; It is a tag The edit distance between the input value and the input value is used to evaluate the quality of semantic matching. It is the j-th semantic tag of the i-th field, which is a text string. It is the actual value that the user entered in the i-th field, which is also a string; This represents a symbol vector conversion function that converts labels... It is transformed into a semantic vector; the second term represents the influence of the field on the context structure of the page, and a weighted average of adjacent fields is introduced; It is the set of adjacent fields of the i-th field, referring to the other fields that are immediately adjacent to the i-th field in the page structure. It is the number of adjacent fields of the i-th field; It is the semantic coupling coefficient between the i-th field and its k-th adjacent field, representing the degree of their correlation; It is the semantic vector of the adjacency field, which represents the adjacency field. It is transformed into a semantic vector; the third term is the temporal density integral, which represents the weight density distribution of the current field in the user input behavior; It is the time decay coefficient, which controls the weighting emphasis on recent input behaviors; It is the maximum duration of the behavior time window; Let be the input behavior density function of the i-th field at time t, representing the frequency distribution of user usage of the i-th field.

[0014] Preferably, in step S2, a structural perturbation tensor model is constructed:

[0015]

[0016] in, It is a third-order tensor, representing the third order tensor in the structure space. The structural disturbance impact value of the field node in the x-th row and y-th column under each disturbance channel; This represents the static initial position coordinates of the field node within the page's DOM structure. Let be the structural disturbance term, representing the th Each disturbance channel The strength of the structural influence of field nodes at a given location is simulated by the positional vibration of field nodes caused by delays, animations, or script operations during page loading. Indicates time; The perturbation frequency represents the periodicity of the position change of the field node during loading or dynamic changes; The initial phase offset represents the shift in the start time of the disturbance, simulating the lag in asynchronous loading or field display. For semantically driven terms, from the semantically global vector The second-order gradient is extracted, corresponding to the semantic change rate of the field in the two-dimensional page layout, indicating the direction and strength of the semantic weight's influence on the structural layout. The structure-semantic coupling coefficient represents the coupling strength between the semantic gradient and the structural perturbation.

[0017] Preferably, in step S3, the path cost function is constructed as follows:

[0018]

[0019] in, It is the reporting path status function at time t, representing the execution status of the planned reporting field sequence at time t; and These are the start and end times of the task; The time derivative of the reported path, i.e., the path change rate; This indicates the difference between the current reported path status and the target field status; This is the penalty coefficient for the path change rate, used to control the smoothness of the reported path; The deviation penalty coefficient controls the degree of similarity matching between the path status and the target field; For dependency penalty terms, where It is a dependency penalty coefficient, used to control the degree to which sequential dependencies between fields are satisfied; Let be the edge weight in the dependency graph from the i-th field to the q-th field, representing the degree of strong dependency; To maintain stability and prevent division by zero; This is an indicator function for violating the order. It is 1 if the i-th field should be executed after the q-th field (violating the dependency), indicating that the path should be penalized; otherwise, it is 0.

[0020] An automated data entry system based on a large model includes the following components:

[0021] The system includes a data acquisition module, a semantic aggregation module, a structural modeling module, a path deduction module, and a data entry and execution module.

[0022] The data acquisition module is embedded in the target platform page accessed by the user through a self-developed browser plugin. It achieves real-time data acquisition in the front end of the page by listening. During the acquisition process, it establishes a set of field tags, a set of adjacent fields, and the distribution of input time behavior, and then passes the structured data records to the semantic aggregation module.

[0023] The semantic aggregation module constructs a three-dimensional semantic state vector for each field, including semantic fit dimension, adjacency context dimension, and temporal behavior dimension. All field vectors are globally aggregated to form a semantic global vector, which is then passed to the structural modeling module and the path inference module.

[0024] The structural modeling module is used to construct the logical structural representation of the target platform page and introduces a dynamic perturbation modeling mechanism to solve the problem of unstable field positions. The page is regarded as a two-dimensional logical coordinate grid, and perturbation modeling is performed on each field node in the time-structure space. A third-order structural perturbation tensor is established, and combined with the two-dimensional gradient of the semantic global vector, a semantic-structure coupling term is introduced to generate a stable set of field structure representation vectors, which is then passed to the path inference module.

[0025] The path deduction module constructs a path cost function, solves for a minimum-cost filling path, generates a globally optimal path function, and passes it to the filling execution module.

[0026] The data entry execution module schedules the control engine in the browser plugin to perform field operations sequentially based on the optimal path.

[0027] The beneficial effects of the technical solution of the present invention are:

[0028] 1. This invention constructs a field representation vector that jointly models multiple semantics, structures, and temporal behaviors, enhancing the accuracy and stability of field semantic matching and effectively solving problems such as inconsistent field naming and inconsistent structure distribution across different platforms;

[0029] 2. A perturbation tensor structure mapping mechanism is designed to dynamically perceive changes in field positions, reordering, and asynchronous rendering behavior during page loading, and still has strong fault tolerance and positioning accuracy in complex page environments;

[0030] 3. By formally modeling the filling order, field dependencies, and execution costs through path cost functions, an optimal filling path with reasonable order and consistent behavior is generated, which can adapt to the logical requirements of field linkage and triggering in dynamic interactive pages, ensuring the correctness and effectiveness of the filling process. Attached Figure Description

[0031] Figure 1 This is a system structure diagram of the present invention;

[0032] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0033] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] The following description, in conjunction with the accompanying drawings, details the specific scheme of the automated data entry system and method based on a large model provided by this invention.

[0036] See attached document Figure 1 The diagram illustrates an automated data entry system architecture based on a large model, according to an embodiment of the present invention. The system comprises the following components:

[0037] The system includes a data acquisition module, a semantic aggregation module, a structural modeling module, a path deduction module, and a data entry and execution module.

[0038] The data acquisition module is embedded in the target platform page accessed by the user through a self-developed browser plugin. It achieves real-time data acquisition in the front end of the page by listening. During the acquisition process, it establishes a set of field tags, a set of adjacent fields, and the distribution of input time behavior, and then passes the structured data records to the semantic aggregation module.

[0039] The semantic aggregation module constructs a three-dimensional semantic state vector for each field, including semantic fit dimension, adjacency context dimension, and temporal behavior dimension. All field vectors are globally aggregated to form a semantic global vector, which is then passed to the structural modeling module and the path inference module.

[0040] The structural modeling module is used to construct the logical structural representation of the target platform page and introduces a dynamic perturbation modeling mechanism to solve the problem of unstable field positions. The page is regarded as a two-dimensional logical coordinate grid, and perturbation modeling is performed on each field node in the time-structure space. A third-order structural perturbation tensor is established, and combined with the two-dimensional gradient of the semantic global vector, a semantic-structure coupling term is introduced to generate a stable set of field structure representation vectors, which is then passed to the path inference module.

[0041] The path deduction module constructs a path cost function, solves for a minimum-cost filling path, generates a globally optimal path function, and passes it to the filling execution module.

[0042] The data entry execution module schedules the control engine in the browser plugin to perform field operations sequentially based on the optimal path.

[0043] See attached document Figure 2 The diagram illustrates a flowchart of an automated data entry method based on a large model, provided by an embodiment of the present invention. The method includes the following steps:

[0044] S1. Monitor and collect user input behavior in forms on different platforms in real time, construct a semantic state vector that integrates semantic dimension, structural environment awareness dimension and time behavior dimension, and aggregate it to form a semantic global vector;

[0045] Data collection is achieved through a self-developed browser plugin. This plugin is embedded into the target platform page accessed by the user, collecting user input data in real time from multiple platforms. It captures user input behavior in real time based on a DOM structure listening mechanism. <input> `、` <select>`、`< / select> <textarea>The label text, attribute name, relative position, input content and modification time of the label (such as the label of the "name" field of the "contact information" field of the "personal information" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form" field of the "application form"< / textarea>

[0046] To achieve semantic consistency and structural alignment among heterogeneous form fields across platforms, it is necessary to unify the structure, fuse semantics, and model the context of field data involved in user input behavior from multiple data sources. This results in a stable, mappable, and highly discriminative field representation vector. Each field representation should not only include the original value information of the current field in the source page, but also the semantic center vector of its semantic tag set, its relative position in adjacent fields, and the user's access density and behavioral preferences for that field over historical time.

[0047] To achieve the above objectives, a triple semantic state vector is constructed, encompassing multiple semantic dimensions, a structural environment awareness dimension, and a temporal behavior dimension, to capture the multiple features of a field as a semantic node within its respective page. For the i-th field, its semantic state vector is defined as:

[0048]

[0049] in, It is the semantic state vector of the i-th field; the first term represents the semantic fit between the field label and the user input value; It is the set of labels for the i-th field, containing all semantic auxiliary information. It represents the number of tags in the i-th field; It is a tag The edit distance between the input value and the input value is used to evaluate the quality of semantic matching. It is the j-th semantic tag of the i-th field, which is a text string. It is the actual value that the user entered in the i-th field, which is also a string; This represents a symbol vector conversion function that converts labels... It is transformed into a semantic vector; the second term represents the influence of the field on the context structure of the page, and a weighted average of adjacent fields is introduced; It is the set of adjacent fields of the i-th field, referring to the other fields that are immediately adjacent to the i-th field in the page structure. It is the number of adjacent fields of the i-th field; It is the semantic coupling coefficient between the i-th field and its k-th adjacent field, representing the degree of their correlation; It is the semantic vector of the adjacency field, which represents the adjacency field. It is transformed into a semantic vector; the third term is the temporal density integral, which represents the weight density distribution of the current field in the user input behavior; It is the time decay coefficient, which controls the weighting emphasis on recent input behaviors; It is the maximum duration of the behavior time window; Let be the input behavior density function of the i-th field at time t, representing the frequency distribution of user usage of the i-th field.

[0050] All field vectors are aggregated to form a semantic global vector. ,in The number of fields entered by the current user; the aggregated semantic global vector contains both the semantic distribution of local fields and global behavioral and structural information.

[0051] S2. To address the dynamic changes in the target page structure during loading, a structural perturbation tensor model is constructed. The stable position of each field is modeled, and the stable state representation vector of each target field under perturbation is obtained through tensor projection, providing a structural basis for data entry path planning.

[0052] Based on the structure of the page to be filled (target platform), a field structure perturbation tensor model is established. To determine stable field position mappings in multi-platform data entry tasks, it is necessary to consider the non-deterministic perturbation behavior of the target platform page structure during the loading process in the actual operating environment. This includes differences in asynchronous rendering order caused by network latency, and changes in the relative displacement of fields caused by factors such as the front-end engine execution order, page DOM reordering, and dynamic stylesheet injection. During dynamic page loading, the positional state of field elements no longer has global static determinism. Therefore, it is impossible to achieve a precise correspondence between fields and semantics through static coordinates or DOM indexes. A structural perturbation modeling mechanism must be introduced to describe the spatiotemporal behavior of page fields as a high-dimensional tensor structure with multiple perturbation factors. This tensor is used to dynamically represent the fluctuation trend, semantic response gradient, and mapping position stability of fields in the structural space.

[0053] Specifically, the target page is encoded using two-dimensional coordinates. Considering the coupling effect between field loading order, structural layout perturbation, and semantic gradient driving, a third-order perturbation tensor is constructed for each field node to represent the perturbation response characteristics of the field at logical coordinates, in the time dimension, or in the structural loading dimension. Constructing the third-order tensor:

[0054]

[0055] in, It is a third-order tensor, representing the third order tensor in the structure space. The structural disturbance impact value of the field node in the x-th row and y-th column under each disturbance channel; This represents the static initial position coordinates of the field node within the page's DOM structure. Let be the structural disturbance term, representing the th Each disturbance channel The strength of the structural influence of field nodes at a given location is simulated by the positional vibration of field nodes caused by delays, animations, or script operations during page loading. Indicates time; The perturbation frequency represents the periodicity of the position change of the field node during loading or dynamic changes; The initial phase offset represents the shift in the start time of the disturbance, simulating the lag in asynchronous loading or field display. For semantically driven terms, from the semantically global vector The second-order gradient is extracted, corresponding to the semantic change rate of the field in the two-dimensional page layout, indicating the direction and strength of the semantic weight's influence on the structural layout. The structure-semantic coupling coefficient represents the coupling strength between the semantic gradient and the structural perturbation.

[0056] Projecting the structural perturbation tensor onto the principal directions of the structure yields the steady-state representation vector of each target field under perturbation:

[0057]

[0058] in, It is the set of stable structural position vectors of all fillable fields on the target platform under a disturbance environment; For the first The basis vectors of the perturbation principal directions corresponding to the tensors of each perturbation channel represent the main structural deformation directions of the page, such as vertical scrolling and horizontal layout switching. It is the number of vertical divisions or arrangement units of the logical field structure in the target platform page; It is the number of identifiable side-by-side field units in each logical row of the target platform page; This represents the number of disturbance channels.

[0059] S3. Based on the matching relationship between the semantic global vector and the structural mapping result, construct the path cost function, solve the optimal execution path for field filling, and drive the browser to automatically fill in the data according to the path order, so as to realize the automated operation of data filling under semantic reasoning and structural perturbation.

[0060] To achieve automated data entry at the field level, a formal path control model needs to be established based on the path differences between the current semantic global vector and the target set of mapped fields. Since different target platforms may exhibit interactive expansion, conditional rendering, or input fields triggering the visibility of other elements in their front-end page structure design, the execution order of the path must strictly adhere to the dependencies between fields; otherwise, automatic data entry will fail or form validation will fail. Therefore, to achieve a minimally perturbed, multi-target collaborative path from aggregated semantic state to structured mapped fields, a path cost function is constructed to formally describe the cost consumed by the execution path within a continuous time range, and its minimum value is obtained through variational analysis within the domain. The path cost function is constructed as follows:

[0061]

[0062] in, It is the reporting path status function at time t, representing the execution status of the planned reporting field sequence at time t; and These are the start and end times of the task; The time derivative of the reported path, i.e., the path change rate; This indicates the difference between the current reported path status and the target field status; This is the penalty coefficient for the path change rate, used to control the smoothness of the reported path; The deviation penalty coefficient controls the degree of similarity matching between the path status and the target field; For dependency penalty terms, where It is a dependency penalty coefficient, used to control the degree to which sequential dependencies between fields are satisfied; Let be the edge weight in the dependency graph from the i-th field to the q-th field, representing the degree of strong dependency; To maintain stability and prevent division by zero; The function is an indicator of order violation. It is 1 if the i-th field should be executed after the q-th field (violating dependency), indicating that the path needs to be penalized; otherwise, it is 0. The optimal path function is obtained by solving the objective function using existing gradient descent methods. .

[0063] Based on the field nodes and operation instructions corresponding to each moment in the optimal path, the browser control module (such as calling the DOM API through a self-developed plugin) sequentially executes field positioning, value filling, and event triggering operations. For each mapped field, its structural node in the page is accurately identified according to the position vector in the stable structural position vector set. The value matched in the semantic aggregation data is filled into the node by injecting scripts, and user interaction behaviors (such as clicking, selecting, focusing, defocusing, etc.) are simulated. Finally, the entire platform's automatic form filling process is completed according to the time sequence of the optimal path function, realizing an intelligent data filling automation method that is driven by the semantic aggregation reasoning results of a large model and under the fault-tolerant control of structural disturbances.

[0064] In summary, a data entry automation system and method based on a large model have been developed.

[0065] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0066] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0067] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An automated data entry method based on a large model, characterized in that, Including the following steps: S1. Monitor and collect user input behavior in forms on different platforms in real time, construct a semantic state vector that integrates semantic dimension, structural environment awareness dimension and time behavior dimension, and aggregate it to form a semantic global vector; In step S1, a semantic state vector is constructed that integrates semantic dimension, structural environment perception dimension, and temporal behavior dimension, and then aggregated to form a semantic global vector, including: Construct a semantic state vector that integrates semantic dimension, structural environment awareness dimension, and temporal behavior dimension, and define the semantic state vector of each field as follows: ； in, It is the semantic state vector of the i-th field; the first term represents the semantic fit between the field label and the user input value; It is the set of labels for the i-th field, containing all semantic auxiliary information. It represents the number of tags in the i-th field; It is a tag The edit distance between the input value and the input value is used to evaluate the quality of semantic matching. It is the j-th semantic tag of the i-th field, which is a text string. It is the actual value that the user entered in the i-th field, which is also a string; This represents a symbol vector conversion function that converts labels... It is transformed into a semantic vector; the second term represents the influence of the field on the context structure of the page, and a weighted average of adjacent fields is introduced; It is the set of adjacent fields of the i-th field, referring to the other fields that are immediately adjacent to the i-th field in the page structure. It is the number of adjacent fields of the i-th field; It is the semantic coupling coefficient between the i-th field and its k-th adjacent field, representing the degree of their correlation; It is the semantic vector of the adjacency field, which represents the adjacency field. It is transformed into a semantic vector; the third term is the temporal density integral, which represents the weight density distribution of the current field in the user input behavior; It is the time decay coefficient, which controls the weighting emphasis on recent input behaviors; It is the maximum duration of the behavior time window; Let be the input behavior density function for the i-th field at time t, representing the frequency distribution of user usage of the i-th field; All field vectors are aggregated to form a semantic global vector. ,in The number of fields entered by the current user; S2. To address the dynamic changes in the target page structure during loading, a structural perturbation tensor model is constructed. The stable position of each field is modeled, and the stable state representation vector of each target field under perturbation is obtained through tensor projection, providing a structural basis for data entry path planning. In S2, the structural perturbation tensor model is constructed as follows: ； in, It is a third-order tensor, representing the first order tensor in the structure space. The structural disturbance impact value of the field node in the x-th row and y-th column under each disturbance channel; This represents the static initial position coordinates of the field node within the page's DOM structure. Let be the structural disturbance term, representing the th Each disturbance channel The strength of the structural influence of field nodes at a given location is simulated by the positional vibration of field nodes caused by delays, animations, or script operations during page loading. Indicates time; The perturbation frequency represents the periodicity of the position change of the field node during loading or dynamic changes; The initial phase offset represents the shift in the start time of the disturbance, simulating the lag in asynchronous loading or field display. For semantically driven terms, from the semantically global vector The second-order gradient is extracted, corresponding to the semantic change rate of the field in the two-dimensional page layout, indicating the direction and strength of the semantic weight's influence on the structural layout. The structure-semantic coupling coefficient represents the coupling strength between the semantic gradient and the structural perturbation. S3. Based on the matching relationship between the semantic global vector and the structural mapping result, construct the path cost function, solve the optimal execution path for field filling, and drive the browser to automatically fill in the data according to the path order, so as to realize the automated operation of data filling under semantic reasoning and structural perturbation.

2. The automated data entry method based on a large model according to claim 1, characterized in that, In S3, the path cost function is constructed as follows: ； in, It is the reporting path status function at time t, representing the execution status of the planned reporting field sequence at time t; and These are the start and end times of the task; The time derivative of the reported path, i.e., the path change rate; This indicates the difference between the current reported path status and the target field status; This is the penalty coefficient for the path change rate, used to control the smoothness of the reported path; The deviation penalty coefficient controls the degree of similarity matching between the path status and the target field; For dependency penalty terms, where It is a dependency penalty coefficient, used to control the degree to which sequential dependencies between fields are satisfied; Let be the edge weight in the dependency graph from the i-th field to the q-th field, representing the degree of strong dependency; To maintain stability and prevent division by zero; This is an indicator function for violating the order. It is 1 if the i-th field should be executed after the q-th field, i.e., it violates the dependency, indicating that the path should be penalized; otherwise, it is 0.

3. An automated data entry system based on a large model, characterized in that, include: The system includes a data acquisition module, a semantic aggregation module, a structural modeling module, a path deduction module, and a data entry and execution module. The data acquisition module is embedded in the target platform page accessed by the user through a self-developed browser plugin. It achieves real-time data acquisition in the front end of the page by listening. During the acquisition process, it establishes a set of field tags, a set of adjacent fields, and the distribution of input time behavior, and then passes the structured data records to the semantic aggregation module. The semantic aggregation module constructs a three-dimensional semantic state vector for each field, including semantic fit dimension, adjacency context dimension, and temporal behavior dimension. All field vectors are globally aggregated to form a semantic global vector, which is then passed to the structural modeling module and the path inference module. The semantic aggregation module constructs a three-dimensional semantic state vector for each field, including semantic fit dimension, adjacency context dimension, and temporal behavior dimension. After global aggregation of all field vectors, a semantic global vector is formed and passed to the structural modeling module and the path inference module. The construction of a three-dimensional semantic state vector for each field, including semantic fit dimension, adjacency context dimension, and temporal behavior dimension, is as follows: ； in, It is the semantic state vector of the i-th field; the first term represents the semantic fit between the field label and the user input value; It is the set of labels for the i-th field, containing all semantic auxiliary information. It represents the number of tags in the i-th field; It is a tag The edit distance between the input value and the input value is used to evaluate the quality of semantic matching. It is the j-th semantic tag of the i-th field, which is a text string. It is the actual value that the user entered in the i-th field, which is also a string; This represents a symbol vector conversion function that converts labels... It is transformed into a semantic vector; the second term represents the influence of the field on the context structure of the page, and a weighted average of adjacent fields is introduced; It is the set of adjacent fields of the i-th field, referring to the other fields that are immediately adjacent to the i-th field in the page structure. It is the number of adjacent fields of the i-th field; It is the semantic coupling coefficient between the i-th field and its k-th adjacent field, representing the degree of their correlation; It is the semantic vector of the adjacency field, which represents the adjacency field. It is transformed into a semantic vector; the third term is the temporal density integral, which represents the weight density distribution of the current field in the user input behavior; It is the time decay coefficient, which controls the weighting emphasis on recent input behaviors; It is the maximum duration of the behavior time window; Let be the input behavior density function for the i-th field at time t, representing the frequency distribution of user usage of the i-th field; The semantic global vector formed by globally aggregating all the field vectors is as follows: All field vectors are aggregated to form a semantic global vector. ,in The number of fields entered by the current user; The structural modeling module is used to construct the logical structural representation of the target platform page and introduces a dynamic perturbation modeling mechanism to solve the problem of unstable field positions. The page is regarded as a two-dimensional logical coordinate grid, and perturbation modeling is performed on each field node in the time-structure space. A third-order structural perturbation tensor is established, and combined with the two-dimensional gradient of the semantic global vector, a semantic-structure coupling term is introduced to generate a stable set of field structure representation vectors, which is then passed to the path inference module. The structural modeling module is used to construct the logical structure representation of the target platform page and introduce a dynamic perturbation modeling mechanism to solve the problem of unstable field positions. The page is regarded as a two-dimensional logical coordinate grid, and perturbation modeling is performed on each field node in the time-structure space. A third-order structural perturbation tensor is established. Combined with the two-dimensional gradient of the semantic global vector, a semantic-structure coupling term is introduced to generate a stable set of field structure representation vectors, which is then passed to the path inference module. The third-order structure perturbation tensor is established as follows: ； in, It is a third-order tensor, representing the first order tensor in the structure space. The structural disturbance impact value of the field node in the x-th row and y-th column under each disturbance channel; This represents the static initial position coordinates of the field node within the page's DOM structure. Let be the structural disturbance term, representing the th Each disturbance channel The strength of the structural influence of field nodes at a given location is simulated by the positional vibration of field nodes caused by delays, animations, or script operations during page loading. Indicates time; The perturbation frequency represents the periodicity of the position change of the field node during loading or dynamic changes; The initial phase offset represents the shift in the start time of the disturbance, simulating the lag in asynchronous loading or field display. For semantically driven terms, from the semantically global vector The second-order gradient is extracted, corresponding to the semantic change rate of the field in the two-dimensional page layout, indicating the direction and strength of the semantic weight's influence on the structural layout. The structure-semantic coupling coefficient represents the coupling strength between the semantic gradient and the structural perturbation. The path deduction module constructs a path cost function, solves for a minimum-cost filling path, generates a globally optimal path function, and passes it to the filling execution module. The data entry execution module schedules the control engine in the browser plugin to perform field operations sequentially based on the optimal path.

Images