Form filling and report number system, method and device based on multi-modal intelligent fusion and medium

Abstract

The application relates to the technical field of form filling and number reporting systems, in particular to a form filling and number reporting system and method based on multi-modal intelligent fusion, equipment and a medium. The application obtains multi-source heterogeneous data through a data acquisition module, establishes a standardized field mapping rule by using a semantic analysis module, analyzes user filling behavior to generate a dynamic data correlation rule in combination with an intention recognition module, an automatic calculation is performed based on the correlation rule by a calculation engine module, hierarchical encryption storage and operation log recording are realized by a secure storage module, and a collaborative distribution module dynamically allocates tasks according to task progress and dependency relationships. The application realizes unified processing of heterogeneous data, improves system adaptability through dynamic data correlation and intelligent task allocation, and effectively improves the efficiency and reliability of form filling and number reporting.

Description

Technical Field

[0001] This application relates to the technical field of form filling and reporting systems, and in particular to form filling and reporting systems, methods, devices and media based on multimodal intelligent fusion. Background Technology

[0002] The data sources are diverse, including structured statistical reports and unstructured documents, requiring standardized management and intelligent processing. At the same time, the need for cross-departmental collaborative reporting and data sharing is increasingly prominent, placing higher demands on the intelligence level of the reporting system.

[0003] Existing technologies typically employ preset data templates and fixed data entry processes. The system collects and processes data according to pre-configured rules, which cannot adapt to diverse data entry scenarios and lacks flexibility. This situation needs further improvement. Summary of the Invention

[0004] To address the shortcomings of existing form-filling and reporting systems in adapting to diverse reporting scenarios and lacking flexibility, this application provides a form-filling and reporting system, method, device, and medium based on multimodal intelligent fusion, employing the following technical solution:

[0005] Firstly, this application provides a form-filling and data reporting system based on multimodal intelligent fusion, comprising:

[0006] The data acquisition module is used to access multiple data sources and extract structured and unstructured raw data;

[0007] The semantic parsing module is used to extract standardized field mapping rules based on the field features in the original data;

[0008] The intent recognition module is used to parse the user's input intent and, in conjunction with the standardized field mapping rules, generate dynamic data association rules.

[0009] The calculation engine module is used to generate derived field data based on the dynamic data association rules;

[0010] A secure storage module is used to store the derived field data using a hierarchical encryption algorithm and generate an operation log containing the data status.

[0011] The collaborative distribution module is used to dynamically allocate data entry tasks to nodes at all levels based on the task progress and data dependencies in the operation log.

[0012] By adopting the above technical solution, this application first accesses various data sources through a data acquisition module, and extracts raw data from structured reports and unstructured documents; the semantic parsing module analyzes the field characteristics of these raw data and establishes standardized field mapping rules; based on this, the intent recognition module combines user intent with field mapping rules by parsing user filling behavior, and automatically generates dynamic data association rules that reflect actual business logic; the calculation engine module then performs automatic calculation of derived fields according to these association rules; the secure storage module uses a hierarchical encryption algorithm to store the data, and records detailed operation logs; finally, the collaborative distribution module dynamically allocates filling tasks to each node by analyzing the task progress and data dependencies in the operation logs, realizing intelligent task scheduling; it achieves unified processing of multi-source heterogeneous data, and improves the system's adaptability and flexibility through dynamic data association and intelligent task allocation.

[0013] Optionally, the data acquisition module includes:

[0014] The intelligent polling unit is used to dynamically adjust the polling interval according to the data update frequency and periodically access the big data center, departmental business systems and mobile terminals.

[0015] The structured data parsing unit is used to extract title and non-title field data using keyword recognition technology to generate structured reports;

[0016] The unstructured data parsing unit is used to perform semantic analysis and feature extraction on text documents and voice input using natural language processing techniques;

[0017] The data reuse unit is used to reuse user-submitted data and department-submitted data, and to manage historical data versions.

[0018] By adopting the above technical solution, this application first uses an intelligent polling unit to adaptively adjust the access interval based on the historical update patterns of each data source, and performs differentiated polling of the big data center, business system, and mobile terminal. For the collected data, the structured data parsing unit uses keyword recognition technology to accurately extract the title and content fields and convert them into structured reports in a standard format. At the same time, the unstructured data parsing unit uses natural language processing technology to perform deep semantic analysis on text documents and voice input to extract key feature information. In addition, the data reuse unit establishes a historical version management mechanism to achieve intelligent reuse of already entered data and avoid duplicate entry. This not only reduces system load and improves data real-time performance, but also significantly reduces manual intervention through automated data extraction and conversion processes.

[0019] Optionally, the intent recognition module includes:

[0020] The behavior analysis unit is used to parse the user's input intent and generate dynamic data association rules in combination with the standardized field mapping rules.

[0021] The task parsing unit is used to analyze task requirements based on the dynamic data association rules and automatically generate task work orders containing data dependencies.

[0022] The template generation unit is used to automatically generate a standardized template library based on the data dependencies in the task work order. The template fields in the standardized template library are organized according to the dynamic data association rules.

[0023] By adopting the above technical solution, this application first uses a behavior analysis unit to conduct in-depth analysis of the user's data entry operations, and combines these behavioral characteristics with standardized field mapping rules in the system to automatically generate dynamic data association rules that reflect the actual business logic. Based on these dynamic association rules, the task parsing unit analyzes the specific requirements of the data entry task in depth and automatically constructs a task work order containing complete data dependencies. The template generation unit constructs a standardized template library according to the data dependencies in the task work order, and intelligently organizes the template fields according to the dynamic data association rules, so that the template structure can adapt to different data entry scenarios. This achieves dynamic optimization of data entry templates and precise matching of task allocation, which not only improves the adaptability of templates, but also significantly improves the efficiency of data entry and reporting and the user experience through the automatic identification and maintenance of data dependencies.

[0024] Optionally, the behavior analysis unit includes:

[0025] The intent feature extraction subunit is used to extract temporal features, operational features, and content features from user data entry behavior based on a deep learning model.

[0026] The context association subunit is used to semantically match the intent features with the standardized field mapping rules to identify the logical relationships between fields.

[0027] The rule generation subunit is used to construct a data dependency graph based on the logical associations and generate the dynamic data association rules based on a graph structure algorithm, wherein:

[0028] The nodes in the data dependency graph represent the fields to be filled in, and the edges represent the dependencies between the fields.

[0029] The dynamic data association rules include field filling order, data validation rules, and inter-field calculation rules.

[0030] By adopting the above technical solution, this application first uses a deep learning model to extract multi-dimensional features from user data entry behavior through an intent feature extraction subunit. These features include capturing the order and time distribution of data entry through time series analysis, identifying data reference and modification patterns through operation analysis, and discovering the correlation features between data items through content analysis. The context association subunit performs semantic matching analysis on these extracted intent features and the standardized field mapping rules in the system to deeply identify the potential logical relationships between fields. Based on the identified logical relationships, the rule generation subunit constructs a data dependency graph, where the data entry fields are used as graph nodes and the dependencies between fields are used as graph edges. The graph structure algorithm is used to analyze and generate dynamic data association rules that include the field entry order, data validation rules, and inter-field calculation rules. This not only improves the accuracy of intent recognition but also enables the system to adaptively optimize the data entry process, significantly improving the intelligence level of data entry.

[0031] Optionally, the generation mechanism of the dynamic data association rules includes:

[0032] The relationships in the data dependency graph are divided into strong dependencies and weak dependencies. Strong dependencies refer to field relationships that strictly follow the pre-filling rules, while weak dependencies refer to field relationships that allow parallel filling.

[0033] Based on the strong and weak dependencies, the filling priority of each field is calculated. The filling priority is determined according to the in-degree, out-degree and business importance of the target field. The in-degree represents the number of other fields that depend on the target field, and the out-degree represents the number of other fields that the target field depends on. The target field is any field among the fields.

[0034] Based on the analysis of historical data, the deviation between the user's actual filling order and the rule recommendation order is analyzed. Based on the deviation statistics, the dependency weights between fields are dynamically adjusted to iteratively optimize the dynamic data association rules.

[0035] By adopting the above technical solution, this application first classifies the relationships in the data dependency graph, marking field relationships that must strictly follow the pre-filling rules as strong dependencies and field relationships that allow parallel filling as weak dependencies. Then, based on these two types of dependencies, the system calculates the field filling priority by analyzing the in-degree, out-degree, and business importance of each target field. The system continuously collects and analyzes historical data, and dynamically adjusts the weights of field dependencies based on statistical results by comparing the deviation between the user's actual filling order and the system's recommended order, thereby continuously optimizing the dynamic data association rules. This not only improves the parallel efficiency of the filling process but also achieves continuous improvement of the filling rules by learning user habits, making the system more aligned with actual business needs.

[0036] Optionally, the computing engine module includes:

[0037] A computational dependency unit is used to construct a computational dependency tree based on the strong dependency relationship and the weak dependency relationship. The computational dependency tree includes original field nodes and derived field nodes.

[0038] The priority execution unit is used to determine the calculation order according to the filling priority, and to obtain the original filling field data and generate the derived field data according to the calculation order;

[0039] The real-time update unit is used to monitor data changes in the original data entry fields. When a data update is detected, the affected derived fields are determined according to the dependency weights, and a recalculation is triggered.

[0040] The exception handling unit is used to detect data chain breaks during the calculation process, record historical data of the exception fields, push exception reminders to relevant users, and automatically resume calculation after the exception is fixed.

[0041] The dynamic optimization unit is used to continuously adjust the calculation order and dependencies in the calculation dependency tree based on the deviation statistics.

[0042] By adopting the above technical solution, this application first constructs a computation dependency tree based on strong and weak dependencies using a computation dependency unit, organizing the original fields and derived fields as tree nodes. A priority execution unit determines the optimal computation order based on the field filling priority, sequentially acquiring the original field data and generating the derived field data. Simultaneously, a real-time update unit continuously monitors changes in the original field data; when an update is detected, it intelligently determines the affected range of derived fields based on dependency weights and triggers recalculation accordingly. Anomaly handling units are responsible for detecting anomalies such as data disconnection, ensuring the reliability of the computation process by recording historical data and pushing anomaly alerts. A dynamic optimization unit continuously adjusts the computation order and dependencies in the computation dependency tree based on deviation statistics, achieving dynamic optimization of the computation strategy. Real-time monitoring and anomaly handling mechanisms ensure the timeliness and reliability of the data, and continuous optimization improves computational efficiency.

[0043] Optionally, a degradation processing module may also be included, the degradation processing module comprising:

[0044] The monitoring unit is used to monitor data source connection status, semantic parsing response time, computing engine resource utilization, and task processing queue length in real time.

[0045] The exception handling unit is used to switch to local cached data when an abnormal data source connection is detected, switch to basic field mapping mode when the semantic parsing response times out, suspend non-critical field calculation when computing resources are insufficient, and process tasks according to the filling priority level when the task queue is backlogged.

[0046] The data compensation unit is used to compare the differences between the local cache and the source data and synchronize the data after the system returns to normal, re-execute the field calculations that were suspended during the degradation period, and verify the integrity of the compensation data according to the dynamic data association rules.

[0047] The status recovery unit is used to restore the downgraded functional modules in batches, control the execution frequency of data compensation operations, and record the operation log of the downgrade process.

[0048] By adopting the above technical solutions, the operation monitoring unit of this application continuously tracks key indicators, including data source connection status, semantic parsing response time, computing engine resource utilization, and task processing queue length, to monitor the system's operating status in real time. When the anomaly handling unit detects an anomaly, the system will take corresponding degradation strategies according to the anomaly type, including switching to local cached data to continue service when the data source is abnormal, downgrading to the basic field mapping mode when semantic parsing times out, prioritizing the calculation of key fields when computing resources are scarce, and handling tasks in the queue according to filling priority when there is a backlog. After the system recovers, the data compensation unit is responsible for comparing the differences between the local cache and the source data and synchronizing them, re-executing the computing tasks suspended during the degradation period, and verifying the integrity of the compensation data through dynamic data association rules. Finally, the state recovery unit restores the functional modules in batches and reasonably controls the frequency of compensation operations to ensure that the system smoothly transitions to the normal state. Through monitoring and tiered degradation strategies, the system achieves continuous service capability under abnormal conditions, and through data compensation and state recovery mechanisms, it ensures business continuity and data consistency.

[0049] Secondly, this application provides a form-filling and reporting method based on multimodal intelligent fusion, which, when applied to the aforementioned form-filling and reporting system, includes the following steps:

[0050] Access multiple data sources to extract structured and unstructured raw data;

[0051] Based on the field characteristics in the original data, extract standardized field mapping rules;

[0052] The user's intent to fill in the information is analyzed, and dynamic data association rules are generated based on the standardized field mapping rules.

[0053] Based on the aforementioned dynamic data association rules, derived field data is generated;

[0054] The derived field data is stored using a hierarchical encryption algorithm, and an operation log containing the data status is generated.

[0055] Based on the task progress and data dependencies in the operation log, the reporting tasks are dynamically allocated to nodes at all levels.

[0056] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the functional steps of the above-mentioned form filling and reporting system based on multimodal intelligent fusion.

[0057] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the functional steps of the above-mentioned form filling and reporting system based on multimodal intelligent fusion.

[0058] In summary, this application includes at least one of the following beneficial technical effects:

[0059] 1. This application provides a form filling and reporting system based on multimodal intelligent fusion; it acquires multi-source heterogeneous data through a data acquisition module, establishes standardized field mapping rules using a semantic parsing module, and generates dynamic data association rules by analyzing user filling behavior in conjunction with an intent recognition module; the calculation engine module performs automatic calculations based on the association rules, the secure storage module implements hierarchical encrypted storage and operation log recording, and the collaborative distribution module dynamically allocates tasks according to task progress and dependencies; it realizes unified processing of heterogeneous data, improves system adaptability through dynamic data association and intelligent task allocation, and effectively improves the efficiency and reliability of form filling and reporting.

[0060] 2. This application achieves differentiated collection of multi-source data by using an intelligent polling unit to adaptively adjust the access frequency based on historical update patterns; a structured data parsing unit extracts and standardizes field data using keyword recognition; an unstructured data parsing unit performs semantic analysis using natural language processing technology; and a data reuse unit establishes historical version management to achieve intelligent reuse. This reduces system load, improves data real-time performance, reduces manual intervention through automated processing, and improves data collection efficiency.

[0061] 3. This application utilizes a deep learning model to extract multi-dimensional features from user data entry behavior through an intent feature extraction subunit. These features include capturing the order and time distribution of data entry through time series analysis, identifying data referencing and modification patterns through operation analysis, and discovering data item association features through content analysis. The context association subunit performs semantic matching between intent features and standardized field mapping rules to identify logical relationships between fields. The rule generation subunit constructs a data dependency graph and generates dynamic data association rules containing entry order, validation rules, and calculation rules through graph structure algorithms. Through deep learning, accurate modeling of user data entry behavior is achieved, improving the intelligence level of form filling and data reporting. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of a form filling and reporting system based on multimodal intelligent fusion according to an embodiment of this application;

[0063] Figure 2 This is a flowchart illustrating the dynamic data association rule generation mechanism in an embodiment of this application;

[0064] Figure 3 This is a flowchart illustrating a form-filling and data reporting method based on multimodal intelligent fusion according to an embodiment of this application;

[0065] Figure 4 This is an internal flowchart of an electronic device according to an embodiment of this application. Detailed Implementation

[0066] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0067] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0068] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0069] Firstly, this application provides a form-filling and data reporting system based on multimodal intelligent fusion, referring to... Figure 1 ,include:

[0070] The data acquisition module is used to access multiple data sources and extract structured and unstructured raw data.

[0071] In this embodiment, the multi-source data sources include the enterprise big data center, departmental business systems, and mobile terminal devices. Structured data refers to standard-formatted database records and spreadsheets, while unstructured data refers to text documents and voice input content.

[0072] The semantic parsing module is used to extract standardized field mapping rules based on the field characteristics in the original data.

[0073] In this embodiment, field characteristics include field name, data type, value range, and business attributes. Standardized field mapping rules define the conversion relationship from raw fields to standard fields, including field correspondence and data conversion rules.

[0074] Specifically, the system establishes a field feature library to store standard field definitions for common business scenarios. For the input raw data, it first extracts the naming and data features of the fields and performs pattern matching against the standard definitions in the feature library. The system maintains a field mapping rule table, recording successfully matched field conversion rules, including field name conversion, data format conversion, and unit conversion relationships.

[0075] The intent recognition module is used to parse the user's input intent and generate dynamic data association rules by combining standardized field mapping rules.

[0076] In this embodiment, user filling intent refers to the characteristics of a user's operational behavior when filling out a form, including the filling order, data referencing, and field associations. Dynamic data association rules describe the dependencies and calculation rules between fields.

[0077] The calculation engine module is used to generate derived field data based on dynamic data association rules.

[0078] In this embodiment, derived field data refers to data items automatically generated through calculation rules, including summary data, statistical indicators, and analysis results.

[0079] The secure storage module is used to store derived field data using a hierarchical encryption algorithm and generate operation logs containing data status.

[0080] In this embodiment, the hierarchical encryption algorithm uses different encryption strengths for data with different sensitivity levels, and the operation log records the lifecycle status of the data, including creation, modification, review, and archiving.

[0081] Specifically, the system establishes a data security level table, uses a combination of encryption algorithms to encrypt and store data, and records a log of the entire data processing process. The system maintains a state transition table to manage the state changes of data at different lifecycle stages.

[0082] The collaborative distribution module is used to dynamically allocate data entry tasks to nodes at all levels based on the task progress and data dependencies in the operation log.

[0083] In this embodiment, task progress refers to the completion status of the reporting task, data dependency refers to the sequential relationship between reporting tasks, and nodes at each level refer to different levels of departments participating in the reporting.

[0084] Specifically, the system establishes a task scheduling table to record task allocation rules and execution status. By analyzing operation logs, it identifies dependencies between tasks and rationally arranges the task execution order. The system maintains a node resource table and dynamically allocates data entry tasks based on the node's processing capacity and current load.

[0085] In one embodiment, the data acquisition module includes:

[0086] The intelligent polling unit is used to dynamically adjust the polling interval based on the data update frequency, and periodically access the big data center, departmental business systems, and mobile terminals.

[0087] In this embodiment, data update frequency refers to the number of times the data source changes per unit of time. Big data center refers to the enterprise's data warehouse system; departmental business systems refer to the financial and human resources systems; and mobile terminals refer to mobile office applications and data collection terminals. Polling interval refers to the time interval between two data collection sessions.

[0088] The structured data parsing unit is used to extract title and non-title field data using keyword recognition technology to generate structured reports.

[0089] In this embodiment, keyword recognition technology refers to a rule-based text matching method. The title field refers to the table header information, while non-title fields refer to the actual data content. A structured report refers to a standard-formatted data table.

[0090] The unstructured data parsing unit is used to perform semantic analysis and feature extraction on text documents and voice input using natural language processing techniques.

[0091] In this embodiment, natural language processing technology refers to rule-based and statistical text analysis methods. Text documents refer to electronic documents and scanned files, while voice input refers to audio recordings. Semantic analysis refers to extracting key information from the text, and feature extraction refers to identifying the attribute features of data items.

[0092] Specifically, the system establishes a business terminology dictionary, including professional terms, units of measurement, and numerical expressions. It performs word segmentation and part-of-speech tagging on text content, recognizing numbers, amounts, and dates; and it transcribes speech content, extracting key data items. The system then converts the extracted information into a structured format for storage.

[0093] The data reuse unit is used to reuse user-submitted data and department-submitted data, and to manage historical data versions.

[0094] In this embodiment, "reported data" refers to historical reporting records, and "version management" refers to maintaining multiple historical versions of the data.

[0095] Specifically, the system establishes a data version table to record the version number, modification time, and modified content of the data. When a user starts a new data entry task, the system automatically loads relevant historical data and provides intelligent data entry suggestions. The system retains the history of the ten most recent versions, supporting data backtracking and comparative analysis.

[0096] In one embodiment, the intent recognition module includes:

[0097] The behavior analysis unit is used to parse the user's input intent and generate dynamic data association rules by combining standardized field mapping rules.

[0098] In this embodiment, user intent refers to the data processing preferences and operational habits exhibited by the user during the form completion process. Standardized field mapping rules refer to the correspondence and transformation rules between fields. Dynamic data association rules refer to the data item association relationships generated based on user behavioral characteristics.

[0099] The task parsing unit is used to analyze task requirements based on dynamic data association rules and automatically generate task work orders that include data dependencies.

[0100] In this embodiment, task requirements refer to the completeness and timeliness requirements of the task being filled in. Data dependencies refer to the logical associations and computational relationships between the fields being filled in. Task work order refers to a standardized task description document.

[0101] Specifically, the system establishes a task rule base, including task types, completion criteria, and processing procedures. By analyzing data association rules, the system identifies required and optional fields involved in the task and determines the dependency order between fields. The system generates task work orders, clearly defining task objectives, data requirements, and completion deadlines, and divides the task processing into stages based on dependencies.

[0102] The template generation unit is used to automatically generate a standardized template library based on the data dependencies in the task work order. The template fields in the standardized template library are organized according to dynamic data association rules.

[0103] In this embodiment, the standardized template library refers to a pre-defined set of form templates. Template fields refer to the data items in the template. Template organization refers to the layout and association of fields.

[0104] Specifically, the system establishes a template structure table, defining the template's framework and field attributes. Based on the dependencies in the task order, the system organizes related fields into logical groups, such as classifying asset data and liability data separately. The system creates dedicated templates for different business scenarios, such as balance sheet templates and profit and loss statement templates, supporting template reuse and customization. Each template includes field definitions, filling rules, and validation conditions.

[0105] In one embodiment, the behavior analysis unit includes:

[0106] The intent feature extraction subunit is used to extract temporal features, operational features, and content features from user data entry behavior based on a deep learning model.

[0107] In this embodiment, the temporal characteristics refer to the time intervals and sequential patterns of user-entered fields. Operational characteristics refer to data input methods and shortcut key usage habits. Content characteristics refer to data entry format preferences and referencing methods.

[0108] The context association subunit is used to semantically match intent features with standardized field mapping rules to identify logical relationships between fields.

[0109] In this embodiment, semantic matching refers to similarity calculation based on the meaning of fields. Logical association refers to the business relationship between fields.

[0110] Specifically, the system establishes a field semantic index table, which includes field names, business attributes, and association rules. For the extracted intent features, the system first calculates the semantic similarity with standard fields, and then identifies the logical relationships between fields based on their business attributes, such as the relationship between total fields and detail fields, and the relationship between income fields and expense fields.

[0111] The rule generation subunit is used to construct a data dependency graph based on logical associations and generate dynamic data association rules based on graph structure algorithms. In the data dependency graph, nodes represent fields to be filled in, and edges represent the dependencies between fields. The dynamic data association rules include field filling order, data validation rules, and inter-field calculation rules.

[0112] In this embodiment, the data dependency graph is a directed graph structure used to represent the dependencies between fields. Graph structure algorithms refer to a set of algorithms used to analyze and optimize dependencies.

[0113] Specifically, the system first constructs an adjacency matrix to represent field dependencies. The matrix dimension depends on the number of fields involved in the task. A topological sorting algorithm is used to determine the order in which fields are filled in, and a minimum spanning tree algorithm is used to optimize the dependency path.

[0114] In one embodiment, refer to Figure 2 The generation mechanism for dynamic data association rules includes:

[0115] S210. Divide the relationships in the data dependency graph into strong dependencies and weak dependencies.

[0116] Among them, strong dependency relationship represents a field relationship that strictly follows the previous filling rule, while weak dependency relationship represents a field relationship that allows parallel filling.

[0117] Specifically, the system establishes a dependency classification table and classifies relationships by setting dependency strength thresholds. Strong dependencies have weights set between 0.8 and 1.0, representing a mandatory order requirement; weak dependencies have weights set between 0.3 and 0.7, allowing for adjustments to the processing order based on actual circumstances. The system maintains a dependency matrix, recording the dependency types and weights between fields.

[0118] S220. Calculate the filling priority of each field based on strong and weak dependencies.

[0119] The priority of filling in the target field is determined based on the in-degree, out-degree, and business importance of the target field. The in-degree represents the number of other fields that depend on the target field, and the out-degree represents the number of other fields that the target field depends on. The target field is any field in each field.

[0120] In this embodiment, in-degree refers to the number of references to the target field by other fields, and out-degree refers to the number of dependencies of the target field on other fields. Business importance refers to the degree of criticality of a field in the business process.

[0121] Specifically, the system establishes a field weight calculation table, uses a weighted calculation method to determine field priority, and normalizes the calculated priority values ​​for sorting and task allocation. For fields with similar priorities, the system allows parallel processing.

[0122] Furthermore, the priority score is calculated as follows: α * (in-degree weight) + β * (out-degree weight) + γ * (business importance), where α = 0.3 + log(number of in-degrees / total number of fields), β = 0.2 + log(number of out-degrees / total number of fields), and γ = 0.4 * (1 + historical anomaly rate). The historical anomaly rate is calculated by dividing the number of historical anomalies of a field by the total number of submissions, and is used to dynamically adjust the impact of business importance. This mechanism can adaptively adjust the weight coefficients based on the actual usage of the fields, improving the accuracy of priority calculation. The in-degree weight reflects the importance of a field in cross-departmental data sharing: it is obtained by multiplying the level coefficient of the referencing department by the data sharing frequency of that department, and then averaging the results across all referencing departments. The department level coefficient indicates the management level of the data-using department in the organizational structure, and the data sharing frequency indicates the activity level of that department in accessing this field's data. The out-degree weight reflects the impact of a field on the reporting work of other departments: it is obtained by multiplying the timeliness requirement of each related department by the data dependency, and then averaging the results across all related departments. The timeliness requirement reflects the urgency of the deadline for the reporting tasks of related departments, and the data dependency indicates the degree of impact of this field on the reporting work of related departments. The business importance is dynamically adjusted based on departmental collaboration characteristics: it is based on the field's basic importance, and multiplied by the collaborative reporting coefficient and the data quality score adjustment factor to obtain the final business importance. The collaborative reporting coefficient increases with the number of departments reporting online simultaneously, and the data quality score is a comprehensive evaluation based on the accuracy and timeliness compliance rates of the field's historical reporting.

[0123] S230. Analyze the deviation between the user's actual filling order and the rule recommendation order based on historical data, and dynamically adjust the dependency weights between fields based on the deviation statistics to iteratively optimize the dynamic data association rules.

[0124] In this embodiment, the filling order deviation refers to the difference between the user's actual operation order and the system's recommended order. The deviation statistics are used to evaluate the rationality of the rules.

[0125] Specifically, the system establishes a form to analyze user behavior and records the actual order in which users fill in the forms. Through comparative analysis, the system calculates the order deviation value for each field. Within an observation period of 5-10 business days, rule adjustments are triggered when the frequency of deviation exceeds 3-5 times, ensuring the smooth evolution of the rules.

[0126] In one embodiment, the computing engine module includes:

[0127] The computation dependency unit is used to construct a computation dependency tree based on strong and weak dependencies. The computation dependency tree includes original field nodes and derived field nodes.

[0128] In this embodiment, the computation dependency tree is a multi-level tree structure used to describe the computational relationships between fields. Original field nodes refer to basic data items directly entered by the user, such as sales revenue and procurement costs; derived field nodes refer to data items generated through calculation formulas, such as gross profit margin and asset turnover rate. Strong dependencies represent field relationships that must be calculated sequentially, while weak dependencies represent field relationships that can be calculated in parallel.

[0129] Specifically, the system establishes a computation dependency mapping table, containing three basic attributes: field identifier, calculation formula, and dependency type. Dependencies are stored using a tree-structured adjacency list, organizing fields according to computational hierarchy and controlling the depth of the dependency tree to between 3 and 5 levels. For strong dependencies, the system strictly processes them according to hierarchical order; for weak dependencies, the system allows parallel computation within the same level.

[0130] The priority execution unit is used to determine the calculation order based on the filling priority, and to obtain the original filling field data and generate the derived field data according to the calculation order.

[0131] In this embodiment, the fill priority refers to the execution order of field calculations. Original fill field data refers to directly collected data values, while derived field data refers to data values ​​generated through calculation rules. The calculation order is determined based on the field priority.

[0132] The real-time update unit is used to monitor data changes in the original data entry fields. When a data update is detected, the affected derived fields are determined based on the dependency weights, and a recalculation is triggered.

[0133] In this embodiment, data change refers to modification of the original field values. Dependency weight refers to the quantified value of the strength of the association between fields. Affected derived fields refer to fields that need to be recalculated due to data changes.

[0134] Specifically, the system establishes a data monitoring table to record the update timestamps and changed values ​​of fields. A data monitoring window is set up to detect changes by comparing data snapshots. When a data update is detected, the system searches downwards along the dependency tree to identify related fields with weights greater than a threshold, triggering a recalculation.

[0135] The exception handling unit is used to detect data disconnection during the calculation process, record historical data of the exception fields, push exception reminders to relevant users, and automatically resume calculation after the exception is fixed.

[0136] In this embodiment, "data chain break" refers to an interruption during the calculation process due to missing input data. "Abnormal field" refers to a field where the calculation failed or the data is abnormal. "Historical data" refers to the historical calculation results of the field.

[0137] Specifically, the system establishes an exception handling log table, including the exception type, occurrence time, and scope of impact. Historical data from the last 3-5 versions is retained for backtracking analysis. An exception waiting time is set, during which the system continuously attempts to resume calculations while simultaneously sending exception alert messages to the user.

[0138] The dynamic optimization unit is used to continuously adjust the computation order and dependencies in the computation dependency tree based on the deviation statistics.

[0139] In this embodiment, the deviation statistics result refers to the statistical difference between the actual calculation order and the preset order. The calculation order refers to the processing order of the fields, and the dependency relationship refers to the calculation association between the fields.

[0140] Specifically, the system establishes an optimization and adjustment table to record execution statistics for the computation process. At the end of each business cycle, the execution efficiency of the computation path is analyzed to identify computational bottlenecks. Based on the statistical results, the dependency tree structure is dynamically adjusted, with the optimization depth controlled within 2-3 levels to ensure continuous improvement in computational efficiency.

[0141] In one embodiment, the system further includes a degradation processing module, which includes:

[0142] The monitoring unit is used to monitor data source connection status, semantic parsing response time, computing engine resource utilization, and task processing queue length in real time.

[0143] The exception handling unit is used to switch to local cached data when an abnormal data source connection is detected, switch to basic field mapping mode when the semantic parsing response times out, suspend non-critical field calculation when computing resources are insufficient, and process tasks according to filling priority when the task queue is backlogged.

[0144] The data compensation unit is used to compare the differences between the local cache and the source data and synchronize the data after the system recovers to normal, re-execute the field calculations that were suspended during the degradation period, and verify the integrity of the compensation data according to the dynamic data association rules.

[0145] The status recovery unit is used to restore the downgraded functional modules in batches, control the execution frequency of data compensation operations, and record the operation log of the downgrade process.

[0146] Secondly, this application provides a form filling and reporting method based on multimodal intelligent fusion. The form filling and reporting method based on multimodal intelligent fusion of this application will be described below in conjunction with the above-mentioned form filling and reporting system based on multimodal intelligent fusion.

[0147] Reference Figure 3 A form-filling and data reporting method based on multimodal intelligent fusion includes the following steps:

[0148] S310. Access multiple data sources to extract structured and unstructured raw data.

[0149] S320. Extract standardized field mapping rules based on the field characteristics in the original data.

[0150] S330. Parse the user's intent to fill in the information, and generate dynamic data association rules by combining standardized field mapping rules.

[0151] S340. Generate derived field data based on dynamic data association rules.

[0152] S350 uses a hierarchical encryption algorithm to store derived field data and generates an operation log containing data status.

[0153] S360: Based on the task progress and data dependencies in the operation log, dynamically allocate the reporting tasks to nodes at all levels.

[0154] In one embodiment, this application provides an electronic device, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, this electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the functional steps of a multimodal intelligent fusion-based form-filling and reporting system.

[0155] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0156] In one embodiment, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0157] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0158] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A form filling and counting system based on multi-modal intelligent fusion, characterized in that, The application relates to a dynamic data filling system and method. The application comprises: a data collection module for accessing multi-source data sources and extracting structured and unstructured original data; a semantic analysis module for extracting standardized field mapping rules according to field characteristics in the original data; an intention recognition module for analyzing user filling intentions and generating dynamic data correlation rules in combination with the standardized field mapping rules; a calculation engine module for generating derived field data based on the dynamic data correlation rules; a secure storage module for storing the derived field data by using a hierarchical encryption algorithm and generating an operation log containing data states; a collaborative distribution module for dynamically distributing filling tasks to nodes at different levels according to task progress and data dependency relationships in the operation log. The data collection module comprises: an intelligent polling unit for dynamically adjusting polling intervals according to data update frequencies, periodically accessing big data centers, department business systems and mobile terminals; a structured data analysis unit for extracting title and non-title field data by using keyword recognition technology and generating structured reports; an unstructured data analysis unit for performing semantic analysis and feature extraction on text documents and voice inputs by using natural language processing technology; 2. The multi-modal intelligent fusion based form filling and tallying system as claimed in claim 1, wherein, a data multiplexing unit for multiplexing user-filled data and department-filled data and performing historical data version management. The intention recognition module comprises: a behavior analysis unit for analyzing user filling intentions and generating dynamic data correlation rules in combination with the standardized field mapping rules; a task analysis unit for analyzing task requirements based on the dynamic data correlation rules and automatically generating task work orders containing data dependency relationships; 3. The multi-modal intelligent fusion based form filling and tallying system as claimed in claim 2, wherein, a template generation unit for automatically generating a standardized template library according to data dependency relationships in the task work orders, and organizing template fields in the standardized template library according to the dynamic data correlation rules. The behavior analysis unit comprises: an intention feature extraction subunit for extracting time sequence features, operation features and content features in user filling behaviors based on a deep learning model; a context association subunit for performing semantic matching of the intention features and the standardized field mapping rules and identifying logical associations between fields; a rule generation subunit for constructing a data dependency graph according to the logical associations and generating the dynamic data correlation rules based on a graph structure algorithm, wherein: nodes of the data dependency graph represent filling fields, and edges represent dependency relationships between fields; 4. The multi-modal intelligent fusion based form filling and tallying system according to claim 3, wherein, the dynamic data correlation rules include field filling sequences, data checking rules and field inter-computation rules. The generation mechanism of the dynamic data correlation rules comprises: dividing association relationships in the data dependency graph into strong dependency relationships and weak dependency relationships, wherein the strong dependency relationships represent field relationships that strictly follow presequence filling rules, and the weak dependency relationships represent field relationships that allow parallel filling. Based on the strong dependency relationship and the weak dependency relationship, a filling priority of each field is calculated, wherein the filling priority is determined according to in-degree, out-degree and business importance of a target field, the in-degree represents a number of other fields depending on the target field, the out-degree represents a number of other fields on which the target field depends, and the target field is any one of the each field; According to the deviation of the actual filling sequence of the user from the rule recommended sequence based on historical filling data, the dependency relationship weight between fields is dynamically adjusted based on the deviation statistical result, so as to iteratively optimize the dynamic data association rule.

5. The multi-modal intelligent fusion based form filling and tallying system according to claim 4, wherein, The computing engine module comprises: A calculation dependency unit configured to construct a calculation dependency tree based on the strong dependency relationship and the weak dependency relationship, wherein the calculation dependency tree comprises an original filling field node and a derived field node; A priority execution unit configured to determine a calculation order according to the filling priority, and acquire original filling field data according to the calculation order to generate derived field data; A real-time updating unit configured to monitor data changes of the original filling field, and when detecting data updates, determine affected derived fields according to the dependency relationship weight and trigger re-computation; An exception handling unit configured to detect data link breakage in the calculation process, record historical data of the abnormal field, push an exception reminder to a related user, and automatically restore the calculation after the exception is repaired; A dynamic optimization unit configured to continuously adjust the calculation order and the dependency relationship in the calculation dependency tree according to the deviation statistical result.

6. The multi-modal intelligent fusion based form filling and tallying system as claimed in claim 5, wherein, Further comprising a degradation processing module, wherein the degradation processing module comprises: A running monitoring unit configured to monitor a data source connection state, a semantic parsing response time, a computing engine resource occupancy rate and a task processing queue length in real time; An exception handling unit configured to switch to local cache data when detecting a data source connection exception, switch to a basic field mapping mode when a semantic parsing response is timed out, pause non-critical field calculation when computing resources are insufficient, and process tasks according to the filling priority in stages when a task queue is accumulated; A data compensation unit configured to compare differences between local cache and source data after the system returns to normal, synchronize data, re-execute field calculation paused during the degradation, and verify completeness of compensation data according to the dynamic data association rule; A state recovery unit configured to recover degraded functional modules in batches, control execution frequency of data compensation operations, and record operation logs of the degradation processing process.

7. A form filling and counting method based on multi-modal intelligent fusion, characterized in that, Applied to the form filling and reporting system of any one of claims 1-6, comprising the following steps: Accessing multiple data sources to extract structured and unstructured original data; Extracting standardized field mapping rules according to field characteristics in the original data; Analyzing user filling intentions, combining the standardized field mapping rules, and generating dynamic data association rules; Generating derived field data based on the dynamic data association rules; Storing the derived field data by using a hierarchical encryption algorithm, and generating operation logs containing data states; Dynamically distributing filling tasks to nodes at all levels according to task progress and data dependency relationship in the operation logs.

8. An electronic device, comprising: Computer program product, comprising a memory, a processor, and a computer program stored on the memory and loadable into the processor, which, when executed by the processor, carries out the functional steps of the multi-modal intelligent fusion based form filling and counting system according to any one of claims 1-6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program product, when executed by the processor, carries out the functional steps of the multi-modal intelligent fusion based form filling and counting system according to any one of claims 1-6.

Images