Supervision task processing method and system based on unified data view

Abstract

The invention provides a supervision task processing method and system based on a unified data view, and relates to the technical field of data processing, and the method comprises the steps: obtaining original data related to supervision from a multi-source heterogeneous system; analyzing the original data to generate supervision task items; based on the supervision task items, gathering the corresponding attribute information and fusing the execution process data, and constructing a supervision data view; based on the supervision data view, the supervision task items are analyzed, and an analysis result is visually presented; monitoring and analyzing the supervision data view according to a preset risk prediction model, and identifying to obtain an abnormal state of the supervision task item; and based on the abnormal state, triggering graded early warning, and pushing early warning information to the corresponding person in charge. According to the invention, the accuracy and efficiency of supervision and handling are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a supervision task processing method and system based on unified data view. BACKGROUND

[0002] In large enterprise operation management, supervision is a key link to ensure the implementation of strategies, but there are many pain points at present. The supervision related information such as enterprise internal target tasks and weekly key work is scattered in various departments and multi-source heterogeneous systems, forming information islands, and the cross-system data interaction and integration capability is weak. The existing scheme lacks pertinence adaptation, and it is difficult to fit the complex organization structure and business process of large enterprises, and the tracking mechanism is single and the reminding is not timely, which easily leads to task delay. At the same time, the data analysis only stays at the basic report level, and cannot realize deep correlation analysis and strategic matching degree evaluation, and the operation process is complicated and the permission management is extensive, which not only reduces the work efficiency, but also exists information security hidden danger.

[0003] Therefore, it is urgent to provide a supervision task processing method and system based on unified data view to improve the accuracy and efficiency of supervision. SUMMARY

[0004] In order to solve the above technical problems, the present application provides a supervision task processing method and system based on unified data view.

[0005] The first aspect of the embodiment of the present application provides a supervision task processing method based on unified data view, comprising: obtaining original data related to supervision from multi-source heterogeneous systems; analyzing the original data to generate a supervision task item; based on the supervision task item, gathering its corresponding attribute information and fusing its execution process data to construct a supervision data view; based on the supervision data view, analyzing the supervision task item and visualizing the analysis result; According to a preset risk prediction model, the supervision data view is monitored and analyzed to identify the abnormal state of the supervision task item; based on the abnormal state, triggering a hierarchical early warning and pushing the early warning information to the corresponding person in charge.

[0006] The second aspect of the embodiment of the present application provides a supervision task processing system based on unified data view, comprising: a data acquisition module for obtaining original data related to supervision from multi-source heterogeneous systems; a task generation module for analyzing the original data to generate a supervision task item; The data view module is used to collect the corresponding attribute information and integrate the execution process data of the supervision task items to construct a supervision data view. The view analysis module is used to analyze the supervision task items based on the supervision data view and to visualize the analysis results. The risk analysis module is used to monitor and analyze the supervision data view according to the preset risk prediction model, and identify the abnormal status of the supervision task item. The anomaly push module is used to trigger tiered early warnings based on the anomaly status and push the early warning information to the corresponding responsible person.

[0007] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described supervision task processing method based on a unified data view.

[0008] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for processing supervisory tasks based on a unified data view.

[0009] The beneficial effects of the supervision task processing method and system based on a unified data view provided in this application are as follows: By constructing a unified data view, this application effectively integrates supervision-related information scattered across multiple heterogeneous systems, solving the problems of information silos, difficult data integration, and superficial analysis in traditional supervision and enforcement. Through in-depth analysis of supervision task items, attribute aggregation, and process data fusion, comprehensive and accurate management of the entire task lifecycle is achieved. Simultaneously, the introduction of a risk prediction model for intelligent monitoring and analysis, and the triggering of tiered early warnings, improves the efficiency of abnormal state identification and the timeliness of early warnings, thereby effectively avoiding task delays and ensuring the accuracy and efficiency of supervision and enforcement. Attached Figure Description

[0010] Figure 1 A flowchart illustrating a supervision task processing method based on a unified data view, provided as an embodiment of this application; Figure 2 This is a schematic diagram of the application architecture of a supervision task processing system provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the interaction architecture and data flow of the supervision task processing system, the meeting system, and the weekly key work system provided in an embodiment of this application; Figure 4A flowchart illustrating the closed-loop management logic based on meeting resolutions, key tasks, and supervision linkage provided in one embodiment of this application.

[0011] Figure 5 A structural block diagram of a supervision task processing system based on a unified data view provided in an embodiment of this application; Figure 6 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0013] To make the purpose, technical solution, and advantages of this application clearer, the following will be described in conjunction with the appendix. Figures 1-6 The following is an explanation using specific examples.

[0014] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a supervision task processing method based on a unified data view, provided in an embodiment of this application. The method includes: S101: Obtain raw data related to supervision from multi-source heterogeneous systems.

[0015] In this embodiment, this application is applied to a supervision task management platform, which aims to manage the entire lifecycle of various supervision tasks within an enterprise, ensuring effective task execution and risk control. Multi-source heterogeneous systems refer to multiple independent information systems existing within or outside the enterprise, differing in data format, storage method, and interface protocols, but all containing information related to supervision tasks. These systems include project management systems, OA systems, financial systems, human resource systems, etc., and particularly include meeting systems and weekly key work reporting systems. By constructing an Enterprise Service Bus (ESB) and a standardized data interaction interface system, this platform can achieve efficient and real-time data integration with meeting systems, weekly key work systems, etc., supporting closed-loop management of the entire process from meeting deployment to weekly key work execution, supervision and follow-up, and meeting debriefing and optimization.

[0016] Raw data exists in various forms, such as structured database records, semi-structured log files, unstructured documents, or email content. Data acquisition can be achieved in several ways, for example, by establishing direct database connections with various business systems and periodically performing data extraction operations; or by calling the application programming interfaces (APIs) provided by each business system to programmatically obtain the required data. For meeting systems, key information such as resolutions, responsible departments, and completion deadlines can be automatically captured from meeting minutes through real-time interfaces; for weekly task management systems, asynchronous data synchronization can be achieved through message queues to ensure real-time updates of task status.

[0017] S102: Parse the raw data to generate inspection task items.

[0018] In this embodiment, a supervision task item refers to a set of core elements that, after parsing and structuring, can describe a specific supervision task, such as the task title, responsible department, and deadline. It is the basic unit for constructing the supervision data view.

[0019] Due to their diverse sources, raw data often exhibits inconsistent formats and content, making it unsuitable for direct use in supervision and management. Therefore, the parsing process aims to transform heterogeneous data into standardized, system-understandable task items. Specifically, text analysis techniques can be used to identify and extract key information from the raw data. For example, keyword matching and regular expressions can be used to extract elements such as task titles, responsible persons, and deadlines from unstructured text. Through these parsing methods, raw, scattered data is transformed into clear and manageable supervision task items.

[0020] In a specific enterprise application scenario of this application, the generation of supervision tasks particularly emphasizes the linkage with the meeting system and the weekly key work system. The system connects to the meeting system, automatically extracts resolutions from meeting minutes, and breaks down complex resolutions into independent weekly key work tasks according to four dimensions: task name, responsible party, completion milestone, and acceptance criteria. These tasks are automatically synchronized to the weekly key work system, and a unique supervision number is generated, ensuring that supervision tasks are generated immediately after the meeting. Simultaneously, the system associates with the weekly key work system through personnel IDs and responsible party IDs to ensure real-time synchronization and feedback of progress data during task execution.

[0021] S103: Based on the inspection task items, gather their corresponding attribute information and integrate their execution process data to construct an inspection data view.

[0022] In this embodiment, the unified data view refers to a data set with a unified structure and semantics, formed by aggregating and associating static attribute information and dynamic execution process data of the inspection task item as the core. This view provides the foundation for comprehensive analysis, monitoring, and early warning of inspection tasks through data integration.

[0023] Attribute information includes basic descriptions of the supervision task items, such as static information like creation time, project affiliation, and priority. Execution process data represents the dynamic changes of the task from initiation to completion, such as progress updates, communication records with relevant personnel, and attachment upload status. Building a supervision data view involves associating supervision task items with relevant attribute information, for example, by establishing a master-slave table relationship in a database. Simultaneously, various event data generated during task execution, such as logs and feedback, are bound to the corresponding task items, forming a comprehensive data set that includes both static task characteristics and dynamic progress.

[0024] In one embodiment of this application, the system employs a star schema to construct a unified supervision data center. A fact table is centered around supervision tasks, linked to multi-dimensional tables such as meeting resolution tables, departmental information tables, and performance data tables. Real-time data cleaning, standardization, and integration are achieved through an ETL data processing module. The system assigns a globally unique task identifier to each supervision task item and a data identifier to each execution process data entry, binding them through associated fields. This establishes a bidirectional index between static attribute information and the process data chain, forming a traceable and highly correlated supervision data view.

[0025] S104: Based on the inspection data view, analyze the inspection task items and present the analysis results in a visual format.

[0026] In this embodiment, the analysis can include statistics on task progress, assessment of resource allocation, and preliminary identification of potential risks. For example, it can calculate the completion rate and overdue rate of all current tasks, or analyze the task load of different departments. The visualization of the analysis results aims to present data insights in an intuitive and easy-to-understand way, such as by generating various chart formats like bar charts, pie charts, and line charts, or by using dashboards and Gantt charts to clearly display task status, trends, and distribution information to users. This helps managers quickly grasp the overall situation and key details of the supervision tasks.

[0027] In one embodiment of this application, the system provides an interactive, visual decision support platform. It uses visual chart components to display core indicators (such as completion rate, top 5 overdue items, and strategic alignment trends), supporting drill-down viewing of details. Through a drag-and-drop report generation tool, users can customize dimensions and filter conditions, export reports in Excel or PDF format, and set up scheduled push notifications. The system also integrates multi-dimensional analysis models, including basic statistical models, correlation analysis models (such as a model linking task completion rate and performance), and a strategic alignment model (built based on the Analytic Hierarchy Process), providing management with in-depth data insights and decision support.

[0028] S105: Based on the preset risk prediction model, monitor and analyze the supervision data view to identify abnormal states of supervision task items.

[0029] In this embodiment, the risk prediction model refers to a mathematical model that assesses the future execution status of inspection tasks based on historical data and preset algorithms, and predicts the risk level or abnormal state they may face. This model enables the early identification of potential problems.

[0030] In one embodiment of this application, the system achieves anomaly monitoring through an intelligent tracking and dynamic reminder mechanism. The system presets time thresholds (no progress submitted one day before the deadline, or incomplete upon deadline) and automatically marks delayed / abnormal events. A status recognition system is built based on a rule engine, supporting visual rule configuration (no submission one day before the deadline - warning of 24 hours overdue - severe overdue), and the rules can be flexibly adjusted. The system periodically calculates the risk index distribution of all supervised tasks and dynamically updates risk thresholds at each level based on statistical characteristics and preset warning coverage targets, ensuring the accuracy and adaptability of risk identification.

[0031] S106: Based on the abnormal state, trigger a graded early warning and push the warning information to the corresponding responsible person.

[0032] In this embodiment, the tiered early warning system refers to a system that classifies abnormal states of supervisory tasks into different levels (e.g., warning, alarm) based on their severity, and implements notification strategies and response mechanisms for each level. This mechanism aims to ensure that critical information reaches the relevant responsible parties in a timely and accurate manner.

[0033] The tiered early warning mechanism categorizes warnings into different levels based on the severity of the abnormal situation, such as general warnings and emergency warnings. Different levels of warnings correspond to different notification methods and response requirements. For example, general warnings can be communicated via internal system messages or emails, while emergency warnings will be communicated via SMS, telephone, or other more immediate methods. Warning information will be precisely pushed to the responsible personnel directly involved in the abnormal task, ensuring timely delivery and prompting them to take swift intervention measures to prevent further escalation of the risk.

[0034] In one specific implementation of this application, the system implements tiered automatic reminders and early warning pushes. When a task is delayed or abnormal, the system automatically sends reminders to the responsible department head and the supervisor in charge, supporting multiple channels such as OA in-site messages, SMS, and email. The intelligent priority algorithm is based on a weighted scoring model (40% urgency + 30% overdue time + 20% importance + 10% historical completion rate) to divide tasks into high, medium, and low priorities, and implements dynamic frequency control (high priority once per hour, medium priority once every 6 hours, and low priority once per day). At the same time, the core information of abnormal matters (responsible department, reason for delay, current progress, and rectification plan) is automatically synchronized to the meeting system and linked to the agenda of the next corresponding meeting, providing data support for key discussions in the meeting.

[0035] In addition, the system supports root cause analysis. Responding to a root cause analysis command for an inspection task in an abnormal state, the system, based on the corresponding static attribute information and process data chain in the inspection data view, identifies one or more potential root cause factors leading to the abnormal state (such as delays in preceding tasks, shortages of key resources, delayed responses from collaborating departments, defects in deliverable quality, or external shocks) through causal inference analysis. Based on the identification results, a structured root cause analysis report is generated and visualized, providing a basis for decision-making in problem rectification and process optimization.

[0036] In the enterprise-level application of this application, the system particularly emphasizes the closed-loop linkage between the supervision and implementation system and the meeting system and weekly key task system. The supervision and implementation system, as the central hub, integrates the decision-making information of the meeting system and the execution data of the weekly key task system to form a closed-loop management of the entire process of deployment, execution, tracking and optimization.

[0037] Precise conversion of meeting resolutions into oversight tasks: The system automatically retrieves resolutions from the meeting system and converts them into oversight tasks, which are then synchronized to the weekly key work system to ensure rapid task implementation.

[0038] Execution tracking and anomaly feedback: Information such as the completion status of key weekly tasks and existing problems is automatically summarized to the supervision and follow-up system and simultaneously fed back to the meeting system as a reference for meeting discussions.

[0039] Meeting debriefing and decision support: The system automatically records the entire lifecycle data of each task, forming a traceable supervision file. Before the next meeting, the system automatically generates a summary table of the implementation status of past meeting resolutions, highlighting incomplete and abnormal items, providing quantitative evidence for meeting debriefing, helping to optimize subsequent work deployment, and promoting the upgrading of management cycles.

[0040] Through the above mechanisms, this application achieves a seamless connection between meeting deployment, weekly key work execution, supervision and follow-up, and meeting debriefing and optimization, completely solving the management pain points of no follow-up after meetings, no feedback on task progress, and no closed loop for problem rectification, and improving the accuracy, efficiency, and scientific nature of supervision and follow-up in large enterprises.

[0041] As can be seen from the above, this application effectively integrates supervision-related information scattered across multiple heterogeneous systems by constructing a unified data view, solving the problems of information silos, difficulties in data integration, and superficial analysis in traditional supervision and enforcement. Through in-depth analysis of supervision task items, attribute aggregation, and process data fusion, comprehensive and precise management of the entire task lifecycle is achieved. Simultaneously, the introduction of a risk prediction model for intelligent monitoring and analysis, and the triggering of tiered early warnings, improves the efficiency of abnormal state identification and the timeliness of early warnings, thereby effectively avoiding task delays and ensuring the accuracy and efficiency of supervision and enforcement.

[0042] In one embodiment of this application, the raw data is parsed to generate inspection task items, including: The raw data is preprocessed to obtain data in a standard format; Based on a pre-defined ontology of supervisory elements, semantic alignment and mapping are performed on standard format data to obtain semantically unified data. By performing rule mapping and association queries on the structured data in the semantically unified data, the basic elements required to constitute the supervision task items are extracted; the basic elements include the task title, responsible department, responsible person, and specified deadline. Based on a pre-defined natural language processing model, deep semantic analysis is performed on unstructured data in semantically unified data to extract extended elements required to constitute inspection task items; extended elements include task background, core requirements, key deliverables, and collaborating departments. The basic elements and extended elements are linked and integrated to generate inspection task items, and a confidence score is assigned to each task item element in the inspection task item; the task item elements include: basic elements and extended elements; Validate and correct task item elements whose confidence scores are lower than the preset confidence threshold; Based on the verification and correction results, the verified supervision task items are obtained; and the verified supervision task items are used as the final supervision task items.

[0043] In this embodiment, the raw data is preprocessed to obtain standard format data, aiming to unify the data format, eliminate noise, ensure data quality and consistency, and lay the foundation for subsequent semantic processing. For example, ETL (Extract, Transform, Load) tools or custom scripts can be used to extract data from different sources according to predefined rules and perform structured transformations, such as converting unstructured or semi-structured data into relational database table structures or a unified JSON format.

[0044] Based on a pre-defined ontology library of supervisory elements, semantic alignment and mapping are performed on standard format data to obtain semantically unified data. This aims to solve the problem of using different terms or expressions to represent the same concepts in different systems, ensuring data consistency at the semantic level. For example, an ontology matching algorithm can be constructed to match fields or values ​​in standard format data with concepts in the ontology library, uniformly mapping fields such as "responsible person" and "handler" from different systems to the concept of "responsible person" in the ontology library.

[0045] By performing rule mapping and relational queries on structured data within semantically unified data, the basic elements required to constitute inspection tasks are extracted. The aim is to accurately identify and extract the core attributes of inspection tasks from this standardized data based on pre-defined business logic and extraction rules. For example, by configuring a series of data extraction rules, such as SQL queries, XPath expressions, or JSONPath expressions, the values ​​of fields such as task title, responsible department, responsible person, and stipulated deadline can be directly queried and extracted from the semantically unified structured data.

[0046] Based on a pre-defined natural language processing model, deep semantic analysis is performed on unstructured data in semantically unified datasets to extract extended elements necessary for the supervisory task items. The aim is to understand the deeper meaning of this textual content and extract detailed information that supplements the supervisory task. For example, rule-based and dictionary-based methods, combined with keyword matching, syntactic analysis, and semantic pattern recognition, can be used to extract extended elements from unstructured text. For instance, a keyword list and syntactic rules can be defined to identify task background paragraphs or core requirement statements.

[0047] This process links and merges basic and extended elements to generate inspection task items. Each task item element within these items is assigned a confidence score. The aim is to integrate these disparate elements into a complete inspection task item, forming a comprehensive task description, and to evaluate the reliability of each extracted task item element, providing a basis for subsequent verification and correction. For example, a unique task identifier can be defined as the association key to match and merge basic elements extracted from structured data and extended elements extracted from unstructured data. The confidence score can be calculated comprehensively based on factors such as the reliability of the extraction method, the quality of the data source, and the degree of matching.

[0048] Validating and correcting task elements with confidence scores below a preset confidence threshold aims to identify and correct potential errors, thereby improving the overall quality and accuracy of the supervised tasks. For example, preliminary corrections can be made using validation rules, adjusting incorrect date formats, or standardizing the names of responsible persons. Those that cannot be corrected can be marked for manual review. Alternatively, a manual intervention process can be initiated, submitting low-confidence task elements and their contextual information to human reviewers for confirmation, modification, or supplementation.

[0049] Based on the verification and correction results, verified supervisory task items are obtained. These verified supervisory task items are then used as the final supervisory task items. This ensures that all supervisory task items undergo rigorous quality control and verification before entering subsequent processing flows, thereby guaranteeing the accuracy and reliability of the data view. For example, all verification and correction operations can be recorded, and the original task item elements can be updated based on these operations. Once all low-confidence elements have been processed, whether through correction or manual confirmation, the task item is marked as verified. Furthermore, a quality threshold can be set in the data processing pipeline; only task items that pass all verification and correction stages can proceed to the next stage.

[0050] Through the above technical solution, this application first preprocesses the multi-source heterogeneous raw data to unify the data format, solving the problem of inconsistent formats caused by the diversity of data sources. Next, it uses a pre-defined ontology library of supervisory elements for semantic alignment and mapping, eliminating semantic ambiguity in terminology and expression between different systems and ensuring the consistency of data meaning. Based on this, for structured data, rule mapping and relational queries are used to efficiently and accurately extract basic elements such as task titles and responsible departments; simultaneously, for unstructured data, deep semantic analysis is performed using a natural language processing model to comprehensively extract extended elements such as task background and core requirements, greatly improving the completeness of task item elements. Furthermore, basic elements and extended elements are correlated and integrated to form complete supervisory task items, and a confidence score is assigned to each element, quantifying the reliability of the extraction results. For elements with confidence scores below a preset threshold, the system triggers a verification and correction mechanism to effectively identify and correct potential extraction errors or inaccuracies, thereby ensuring the accuracy and reliability of supervisory task items. Ultimately, the verified inspection task items provided a high-quality and reliable data foundation for the subsequent construction of a unified data view. This enabled more accurate and effective processing of inspection tasks, such as analysis, visualization, monitoring, and risk prediction, based on this data view, thereby improving the accuracy and efficiency of inspection and supervision.

[0051] In one embodiment of this application, assigning a confidence score to each element of the supervision task item includes: For basic elements, the first confidence sub-score is calculated based on the confidence weight of the data source, the completeness of the data fields, and the matching degree of the mapping rules. For extended elements, the second confidence sub-score is calculated based on the entity recognition probability output by the natural language processing model, the confidence of relation extraction, and the consistency of contextual semantics. Based on the type of inspection task, different weights are assigned to basic elements and extended elements respectively; For each task item element, the final confidence score is calculated based on its corresponding confidence sub-score and its weight coefficient.

[0052] In this embodiment, when calculating the first confidence sub-score, the credibility weight of the data source is used to measure the reliability of the data source. For example, data sources from the enterprise's internal core business system are given a higher weight, while data sources from unofficial channels or manually entered data are given a lower weight. This weight can be pre-configured in the system or dynamically adjusted based on historical data quality assessment results. The completeness of data fields reflects the degree to which the required fields of basic elements are filled. For example, for a supervision task item, if key basic element fields such as its responsible person or stipulated deadline are empty or missing, its completeness is low. This completeness can be obtained by calculating the ratio of the number of filled fields to the total number of fields. The matching degree of mapping rules is used to measure the degree of conformity between the original data and the mapping rules in the preset supervision element ontology library. For example, when a field value in the original data completely matches a standard term in the ontology library, the matching degree is high; if there is fuzzy matching or complex transformation is required, the matching degree will decrease accordingly. This matching degree can be calculated using string similarity algorithms, regular expression matching, or semantic matching algorithms. The first confidence sub-score is based on the confidence weight of the aforementioned data source, the completeness of the data fields, and the matching degree of the mapping rules. For example, it can be calculated using a weighted average method, a fuzzy comprehensive evaluation method, or a rule-based scoring mechanism.

[0053] When calculating the second confidence sub-score, the entity recognition probability output by the natural language processing model measures the accuracy or confidence of the model in identifying specific entities within extended elements (such as task context, core requirements, key deliverables, and collaborating departments) from unstructured data. For example, the model identifies the project launch meeting as a key event entity in the text and assigns a recognition probability of 0.9. The relation extraction confidence measures the accuracy or confidence of the model in extracting relationships between entities from the text. For example, the model identifies a causal or correlated relationship between the task context and core requirements and assigns a confidence score of 0.8. Contextual semantic consistency measures the semantic coherence of the extracted extended elements with the surrounding textual environment, for example, by calculating the similarity between the word vectors of the extracted elements and the word vectors of the context. The second confidence sub-score is based on the entity recognition probability, relation extraction confidence, and contextual semantic consistency mentioned above, and can be calculated, for example, by training a specialized confidence prediction model.

[0054] Furthermore, different weights are assigned to basic and extended elements based on the type of inspection task. The types of inspection tasks can include, but are not limited to, project inspections, routine task inspections, and risk warning inspections. Different types of inspection tasks will have different levels of emphasis on basic and extended elements. For example, for project inspection tasks, extended elements such as core requirements and key deliverables are more important, so higher weights can be assigned to these extended elements; while for routine task inspection tasks, basic elements such as responsible departments, responsible persons, and stipulated deadlines are more critical, so higher weights can be assigned to these basic elements. This dynamic weighting mechanism makes confidence calculation more adaptable and flexible.

[0055] Finally, for each task item element, the final confidence score is calculated based on its corresponding confidence sub-score and its weight coefficient. For example, a weighted sum can be used, multiplying the first confidence sub-score by the weight of the basic element, multiplying the second confidence sub-score by the weight of the extended element, and then adding the two together to obtain the final confidence score.

[0056] Through the above technical solution, the comprehensive confidence calculation mechanism of this application can effectively improve the accuracy of the confidence scores of the elements of the supervision task items, thereby providing a more reliable basis for the subsequent verification and correction of the elements of the task items with confidence scores less than the preset confidence threshold, improving the efficiency of verification and correction, reducing the cost of manual intervention, and thus ensuring the quality and credibility of the finally generated supervision task items.

[0057] In one embodiment of this application, the verification and correction of task item elements with confidence scores lower than a preset confidence threshold includes: When the confidence score of a task item element is less than the preset confidence threshold, a manual confirmation request is generated and pushed. Convert the elements of task items with confidence scores lower than a preset confidence threshold and their contextual information into vector representations; In the historical task database, retrieve K historical inspection task items and their corresponding task item elements that are similar to the vector representation; Extract the values ​​of the corresponding task item elements from K historical task items and cluster them. Then, push the centroid of the largest cluster as a candidate correction suggestion to the manual confirmation request. Update the corresponding task item elements based on the results of manual operations.

[0058] In this embodiment, the specific steps include: when the confidence score of a task item element is less than a preset confidence threshold, the system will generate and push a manual confirmation request. This step aims to trigger a manual intervention process when the confidence score of a certain element in the supervision task item (such as task title, responsible department, task background, etc.) is lower than a preset reliability standard. For example, the system can generate a notification message within the supervision task management platform and add it to the to-do list of the relevant responsible person or reviewer, while also providing auxiliary reminders via email or SMS.

[0059] Subsequently, task item elements with confidence scores below a pre-set confidence threshold, along with their contextual information, are transformed into vector representations. The purpose of this step is to convert unstructured or semi-structured task item elements and their related background information (such as task titles, responsible departments, and related document fragments) into machine-understandable numerical vector forms for efficient similarity calculation. A pre-trained Transformer model can be used, inputting the task item elements and their context into the model to extract their corresponding word vectors or sentence vectors, and then performing average pooling to obtain a fixed-dimensional vector representation.

[0060] Based on this, the K historical inspection task items and their corresponding task item elements that are similar to the vector representations are retrieved from the historical task database. By calculating the cosine similarity between the vector representations of the current low-confidence task item elements and all vectorized task item elements in the historical task database, the K historical inspection task items most relevant to the current problem can be identified and extracted.

[0061] Furthermore, the values ​​of the corresponding task item elements from the K historical task items are extracted and clustered. The centroid of the largest cluster is then used as a candidate correction suggestion and pushed to the manual confirmation request. This step aims to extract the most representative or reliable correction suggestions from multiple retrieved historical similar cases. For example, for textual elements, the text of the corresponding elements in the K historical task items can be vectorized and then K-Means clustered. The text corresponding to the centroid vector of the largest cluster (or the text with the highest frequency in that cluster) is selected as the correction suggestion. For numerical elements, K-Means clustering can be performed directly on the numerical values, and the mean or median of the largest cluster is selected as the correction suggestion.

[0062] Finally, the corresponding task item elements are updated based on the results of manual operations. This means that after receiving candidate correction suggestions from the system, the personnel making the manual confirmation can choose to accept the suggestions or manually enter new correction values, depending on the actual situation. The final value after manual confirmation is directly written into the corresponding element field of the supervision task item, ensuring the final accuracy of the correction. To improve the system's traceability, when updating task item elements, the system can not only record the final correction value, but also retain the original value, candidate suggestions, and information such as the personnel and time of the manual operation, forming a version history for easy subsequent auditing and traceability.

[0063] Through the above technical solution, this application updates task item elements based on the results of manual operation, realizing human-machine collaborative decision-making. This not only ensures the final accuracy of the correction, but also makes the entire verification and correction process more efficient, accurate and intelligent. It effectively solves the problems of low efficiency of manual verification and lack of historical data support for correction suggestions, and improves the overall data quality of supervision task items.

[0064] In one embodiment of this application, a method for processing supervisory tasks based on a unified data view further includes: Based on the correction data accumulated during manual verification, a model optimization sample set is constructed; Based on the model optimization sample set, the parameters of the natural language processing model are optimized and adjusted through a preset optimization algorithm to obtain the target parameters; The natural language processing model is iteratively updated and redeployed based on the target parameters.

[0065] In this embodiment, the correction data accumulated during manual verification refers to the records generated by human operators intervening, modifying, or supplementing the model's extracted results during the verification and correction of task item elements with confidence scores lower than a preset confidence threshold. This data includes information about model identification errors or uncertainties, as well as correct or more accurate information provided manually. Constructing a model optimization sample set involves structuring this manually corrected data to form a dataset that can be used for model training. For example, the original task item elements that the model predicted incorrectly, along with their contextual information, can be paired with the manually corrected values ​​of the correct task item elements to serve as training samples.

[0066] Pre-defined optimization algorithms refer to those used to adjust the internal parameters of a natural language processing (NLP) model to improve its performance. These algorithms aim to minimize or maximize a certain objective function, such as using metaheuristic algorithms like genetic algorithms and particle swarm optimization, which search for the optimal parameter combination by simulating optimization processes in nature. Optimizing and tuning the parameters of a NLP model involves using a pre-built model optimization sample set and the aforementioned optimization algorithms to systematically adjust the internal parameters of the NLP model, such as weights and biases, as well as hyperparameters that affect model performance. This process aims to enable the model to extract task-related elements more accurately when processing similar data, thereby improving the model's generalization ability and accuracy. Ultimately, the target parameters are the set of parameters adjusted by the optimization algorithms to achieve optimal model performance or meet pre-defined requirements.

[0067] Iterative updates refer to loading the optimized and tuned target parameters into the natural language processing model, replacing the model's original parameters, thereby enhancing the model's feature extraction capabilities. This process can be periodic or triggered based on the accumulation of manually corrected data. Redeployment refers to redeploying the updated natural language processing model to the supervision task management platform, enabling it to process new supervision task data. For example, a rolling update approach can be used to gradually replace old model instances, ensuring service continuity. Through iterative updates and redeployment, the model can continuously learn and adapt to new data patterns, thereby continuously improving the accuracy and efficiency of feature extraction for supervision task items.

[0068] Through the above technical solution, this application effectively solves the problem that the lack of full utilization of manually corrected data prevents natural language processing models from self-improving. By transforming the correction data accumulated during manual verification into a model optimization sample set, a real and targeted data foundation is provided for the continuous learning of the natural language processing model. Based on this, a pre-set optimization algorithm is used to optimize and adjust the model parameters, enabling the model to systematically learn from manual correction experience, thereby improving its accuracy in extracting elements for subsequent supervision tasks. Finally, the natural language processing model is iteratively updated and redeployed based on the optimized target parameters, achieving dynamic evolution and continuous performance improvement. This not only reduces reliance on manual verification, lowers the frequency and cost of manual intervention, but also improves the level of automation in the generation of supervision tasks and the overall processing efficiency, enabling the supervision task management platform to operate more accurately and efficiently.

[0069] In one embodiment of this application, the preset optimization algorithm is the ant colony optimization algorithm; Based on the model optimization sample set, the parameters of the natural language processing model are optimized and tuned using a pre-defined optimization algorithm, including: The combination of hyperparameters to be optimized in a natural language processing model is defined as the solution space of the ant colony optimization algorithm. Hyperparameters include learning rate, batch size, and number of neural network layers. The boundary of the solution space is initialized based on the current parameter configuration and model complexity of the natural language processing model.

[0070] In this embodiment, the preset optimization algorithm is the ant colony optimization algorithm. The ant colony optimization algorithm is a heuristic optimization algorithm that simulates the foraging behavior of ants. Through the accumulation and volatilization mechanism of pheromones, it can effectively perform a global search in a complex solution space, avoiding getting trapped in local optima. In this application, the ant colony optimization algorithm is chosen as the preset optimization algorithm to leverage its powerful global search capability and adaptability to efficiently explore hyperparameter combinations of natural language processing models, thereby overcoming the problems of inefficiency and susceptibility to local optima in traditional manual parameter tuning, grid search, and random search methods.

[0071] The model optimization sample set, built using correction data accumulated through manual verification, is used to adjust the internal parameters of the natural language processing model through intelligent algorithms, thereby improving the model's accuracy and robustness in task-specific feature extraction. This optimization and tuning mechanism reduces the need for manual intervention, improves the efficiency and accuracy of parameter adjustments, and ensures the model can continuously adapt to new data patterns and business needs. For example, regular or triggered optimization of model parameters can be achieved through scripted settings or integration into continuous integration / continuous deployment (CI / CD) processes.

[0072] The solution space is the set of all hyperparameter combinations. Defining the hyperparameter combinations to be optimized in a natural language processing model as the solution space of the ant colony optimization algorithm means that each hyperparameter combination is considered a potential path or solution within the ant colony algorithm. The ant colony algorithm searches within this defined solution space to find the optimal hyperparameter combination.

[0073] Hyperparameters include the learning rate, batch size, and number of neural network layers. The learning rate controls the step size at which the model updates weights in each iteration; its magnitude directly affects the model's convergence speed and final performance. The batch size determines the number of samples processed in each training iteration, affecting training stability and computational efficiency. The number of neural network layers reflects the model's depth and complexity, significantly impacting its feature extraction and generalization abilities. Besides these hyperparameters, natural language processing models also include other hyperparameters, such as the choice of activation function, regularization strength, Dropout ratio, and word embedding dimension. These hyperparameters can also be optimized by incorporating them into the solution space in a similar manner.

[0074] Based on the current parameter configuration and model complexity of the natural language processing model, the boundary of the solution space is initialized. Initializing the boundary of the solution space means setting a reasonable range of values ​​for each hyperparameter. For example, the learning rate boundary might be set between 0.0001 and 0.1, the batch size boundary between 16 and 256, and the number of neural network layers between 2 and 10. This initialization is based on considerations of the current performance of the natural language processing model and computational resource constraints. For example, if the current model shows a tendency to overfit, the upper limit of regularization strength or Dropout ratio will be increased; if the model training speed is too slow, the batch size boundary will be adjusted. By reasonably setting the boundaries, the search range can be effectively narrowed, optimization efficiency improved, and the algorithm avoided exploring unrealistic or inefficient parameter combinations.

[0075] By employing the aforementioned technical solution, an ant colony optimization algorithm is introduced to optimize and tune the hyperparameters of a natural language processing model, effectively solving the problems of low efficiency and difficulty in adjusting hyperparameter combinations to maximize model performance in traditional parameter optimization methods. By defining hyperparameters (such as learning rate, batch size, and number of neural network layers) as the solution space of the ant colony algorithm and initializing the search boundary based on the model's current configuration and complexity, the optimization process is ensured to be targeted and efficient. This not only improves the accuracy and robustness of the natural language processing model in extracting task-related elements but also significantly reduces the workload and time cost of manual parameter tuning, enabling the model to adapt more quickly and intelligently to constantly changing supervisory task data, thereby improving the overall efficiency and accuracy of supervisory task processing.

[0076] In one embodiment of this application, optimizing and adjusting the parameters of a natural language processing model using a preset optimization algorithm further includes: Initialize the pheromone distribution and ant population in the solution space; Each ant is guided to traverse the solution space based on pheromone concentration and heuristic information to construct candidate hyperparameter combination solutions; Based on the model optimization sample set, the natural language processing model is quickly trained and validated using the hyperparameter combination solution of each candidate, and the task item element extraction accuracy of the model on the validation set is used as the objective function value of the candidate hyperparameter combination solution. Update the global pheromone distribution based on the objective function values ​​of all candidate hyperparameter combinations.

[0077] In this embodiment, pheromone distribution and ant colony are initialized in the solution space. Pheromone distribution refers to the pheromone concentration left on paths with different hyperparameter combinations in the hyperparameter solution space, reflecting the quality of these paths during the historical search process. During initialization, the pheromone concentration on all paths is set to a small constant to ensure exploratory activity in the initial stage. The ant colony refers to individual ants simulating ant colony optimization algorithms; each ant represents an independent search agent responsible for constructing a candidate hyperparameter combination solution in the solution space. Initializing the ant colony involves randomly placing a certain number of ants at different starting points in the solution space, or distributing them evenly throughout the solution space, to ensure initial coverage of the solution space.

[0078] Furthermore, each ant is guided to traverse the solution space based on pheromone concentration and heuristic information, constructing candidate hyperparameter combination solutions. During this process, pheromone concentration causes ants to tend to choose paths with higher pheromone concentrations when selecting the next hyperparameter value, reflecting a preference for historically high-quality solutions and accumulated experience. Heuristic information refers to prior knowledge or domain experience related to the hyperparameter combination solution; for example, certain hyperparameter combinations are theoretically or empirically considered to produce better results. Heuristic information can guide ants to move to more promising regions, accelerating convergence. Each ant starts from its current hyperparameter state and, based on pheromone concentration and heuristic information, selects the next hyperparameter value with a certain probability, gradually constructing a complete hyperparameter combination solution.

[0079] Building upon this foundation, the natural language processing (NLP) model is rapidly trained and validated using each candidate hyperparameter combination solution on the model optimization sample set. The task item extraction accuracy of the model on the validation set is used as the objective function value of the candidate hyperparameter combination solution. The model optimization sample set is a dataset specifically designed to evaluate and optimize the performance of NLP models. It includes cleaned and labeled supervisory task item data to simulate the model's performance in real-world applications. Rapid training and validation refers to employing lightweight or simplified training strategies when evaluating each candidate hyperparameter combination solution. This includes using fewer training epochs, smaller data subsets, or simpler model architectures to quickly obtain preliminary evaluation results of the model's performance. The validation process is conducted on an independent validation set to assess the model's generalization ability. Task item extraction accuracy is a key metric for measuring the performance of a NLP model, reflecting its ability to correctly identify and extract supervisory task item elements from text. Using it as the objective function value directly quantifies the impact of each hyperparameter combination solution on the model's actual performance. This step is the core of evaluating the merits of candidate hyperparameter combinations. By rapidly training and validating the model on the optimization sample set, and using the accuracy of task item feature extraction as the objective function value, the performance of each hyperparameter combination can be quantified efficiently and accurately, providing a reliable basis for subsequent pheromone updates.

[0080] Simultaneously, the global pheromone distribution is updated based on the objective function values ​​of all candidate hyperparameter combination solutions. After each iteration, the algorithm updates the pheromone distribution in the solution space based on the candidate hyperparameter combination solutions constructed by all ants and their corresponding objective function values. Higher-performing hyperparameter combination solutions leave more pheromones on their paths, or the pheromones evaporate more slowly. The global pheromone distribution refers to the set of pheromone concentrations on all paths in the entire solution space. Updating the global pheromone distribution means that the algorithm dynamically adjusts its preference for different hyperparameter combination paths based on the search results of the current iteration. Pheromone updating is a key mechanism for ant colony optimization to achieve positive feedback and adaptive learning. By strengthening pheromones on high-performing hyperparameter combination paths and weakening pheromones on poorly performing paths, the algorithm can gradually guide subsequent ants to concentrate on better solution regions, thereby accelerating the convergence speed of the algorithm and increasing the probability of finding the globally optimal or near-optimal hyperparameter combination. This application solves the efficiency and stability problems in the parameter optimization process by specifying the execution steps of the ant colony optimization algorithm; it also improves the reliability and effectiveness of parameter optimization.

[0081] In one embodiment of this application, the heuristic information is controlled by a pheromone heuristic factor and a desired heuristic factor; A method for handling supervisory tasks based on a unified data view also includes: The values ​​of the pheromone heuristic factor and the expected heuristic factor are adjusted based on the degree of dispersion of the objective function values ​​of all solutions during the iteration process. When the dispersion of the objective function value is greater than the first threshold, the pheromone heuristic factor is increased based on the first step length and the expected heuristic factor is decreased based on the first step length. When the dispersion of the objective function value is less than or equal to the first threshold, the pheromone heuristic factor is decreased based on the second step size, and the expected heuristic factor is increased based on the second step size.

[0082] In this embodiment, the pheromone heuristic factor (denoted as α) controls the degree of influence of pheromone trajectories on ant path selection. The larger the value, the more likely the ant is to choose the path with higher pheromone concentration, reflecting the utilization of historical experience. The expectation heuristic factor (denoted as β) controls the degree of influence of heuristic information (such as the reciprocal of the path length or the objective function value) on ant path selection. The larger the value, the more likely the ant is to choose the path that currently appears to be better, reflecting the exploration of the current environment.

[0083] The dispersion of the objective function values ​​of all solutions during iteration refers to the degree of dispersion of the objective function values ​​(e.g., the model's accuracy in extracting task features on the validation set) corresponding to all candidate solutions (i.e., hyperparameter combination solutions) found by the current ant colony in each iteration of the algorithm. This dispersion can be represented by variance. By monitoring this dispersion, it is possible to perceive in real time whether the current search state is in the extensive exploration stage (high dispersion) or the convergence stage (low dispersion). Besides adjustments based on the dispersion of the objective function values...

[0084] When the objective function value is highly discrete, it indicates that the current ant population has a wide search range and the solutions are scattered, suggesting that it is in the exploration stage or has not yet found the optimal region. In this case, increasing the pheromone heuristic factor can enhance the guiding effect of pheromones on ant path selection, prompting ants to utilize existing, proven, and superior path information more effectively, thereby accelerating the algorithm's convergence towards promising regions. Conversely, decreasing the expectation heuristic factor will relatively weaken the attractiveness of current heuristic information (such as local optima), preventing ants from blindly exploring regions that have not yet been fully validated. The first threshold is a preset boundary value used to judge the degree of dispersion; it can be set empirically or dynamically adjusted experimentally. The first step length is a fixed amount that is increased or decreased each time the factor is adjusted; its magnitude affects the sensitivity of the adjustment and can be a constant or a dynamically changing amount based on the number of iterations or the degree of dispersion.

[0085] When the dispersion of the objective function value is low, it indicates that the search results of the current ant population tend to be concentrated and have converged to a certain region, but there is a risk of getting trapped in a local optimum. In this case, reducing the pheromone heuristic factor can reduce the forced guidance of pheromones on ant path selection, encouraging ants to conduct more random exploration, thereby increasing the chance of escaping local optima and finding better solutions. At the same time, increasing the expectation heuristic factor will enhance the effect of the current heuristic information, prompting ants to conduct a more refined search within the local range in order to discover better local solutions. The second step size here is a fixed amount that is increased or decreased each time the factor is adjusted. Its value can be the same as or different from the first step size to adapt to different adjustment strategies. For example, it can be set to a small constant for fine-tuning, or adaptively adjusted according to the degree of convergence.

[0086] Through the above technical solution, this application can adaptively adjust the pheromone heuristic factor and the expectation heuristic factor according to the dynamic dispersion of the objective function values ​​of all solutions during the ant colony algorithm iteration process. This dynamic adjustment mechanism makes the ant colony algorithm more flexible and robust in the parameter optimization process, effectively solving the problems of low search efficiency and unstable convergence caused by fixed factors, and ensuring the efficiency and reliability of parameter optimization in natural language processing models.

[0087] In one embodiment of this application, a supervision data view is constructed by aggregating the corresponding attribute information and fusing the execution process data based on the supervision task item, including: The generated inspection task items are used as the core nodes, and their corresponding basic elements and extended elements are associated and mounted as static attribute information. Extract execution process data associated with the supervision task items from the multi-source business execution system; the execution process data includes progress feedback, supporting materials, communication records, and status change logs; Establish the temporal and logical relationships between the inspection task items and their associated execution process data to form a process data chain; Based on a unified identification system, static attribute information and process data chains are integrated and aggregated to construct a supervision data view.

[0088] In this embodiment, using the generated inspection task item as the core node means that the inspection task item is regarded as the core entity of inspection management, representing the specific work items that need to be tracked, executed, and evaluated. Setting it as the core node means that all information related to the task item will be organized and managed around it, ensuring the clarity of data ownership and logical centralization.

[0089] Simultaneously, the corresponding basic and extended elements are associated and attached as static attribute information. Static attribute information refers to descriptive information that is determined at the time of generation of the supervision task item or remains relatively stable during task execution. Basic elements refer to core information such as task title, responsible department, responsible person, and stipulated deadline, while extended elements include supplementary information such as task background, core requirements, key deliverables, and collaborating departments. Association and attachment refers to binding these elements to the core supervision task item in a structured manner, making them inherent attributes of the task item. This can be achieved by adding fields to the supervision task item table in the database, or by establishing one-to-one or one-to-many relationship tables. For example, the basic elements can be directly used as columns in the supervision task item table, and the extended elements can be stored in another relationship table and connected by the task item ID.

[0090] In addition, execution process data related to the supervision task items is extracted from multi-source business execution systems. Multi-source business execution systems refer to various information systems used internally or externally by an enterprise to support business operations, such as project management systems, OA systems, ERP systems, email systems, and instant messaging tools. These systems record dynamic data generated during the actual execution of supervision task items. Extraction refers to obtaining the required data from these heterogeneous systems through various technical means such as data interfaces, API calls, direct database connections, and log parsing. Association refers to identifying data directly or indirectly related to a specific supervision task item through preset rules, keyword matching, or unified identifiers. Execution process data includes progress feedback, supporting materials, communication records, and status change logs. These are key dynamic information generated during the execution lifecycle of the supervision task item. Progress feedback records the phased progress and completion status of task execution; supporting materials provide supporting evidence such as task-related documents, pictures, and tables; communication records include text, voice, or video information exchanged by all parties during task execution; and status change logs record in detail every change in task status (such as pending, in progress, completed, or delayed) and its timestamp. These data together depict the complete trajectory of the task execution.

[0091] Furthermore, establish temporal and logical relationships between the supervisory task items and their associated execution process data, forming a process data chain. Temporal relationships refer to clearly defining the chronological order in which execution process data occurs; for example, a progress feedback occurs after a communication record, or a status change occurs before an attachment is uploaded. Logical relationships refer to clearly defining the causal, dependency, or inclusion relationships between data; for example, an attachment material supports a progress feedback, or a communication record leads to a status change. These relationships can be established by adding timestamps, association IDs, parent-child relationship fields, etc., to each piece of process data, thereby enabling the tracing of the complete history and event chain of task execution. The process data chain refers to linking all execution process data related to a specific supervisory task item according to their temporal and logical relationships, forming a directed acyclic graph or chain structure. This data chain clearly shows the entire evolution of the task from initiation to completion (or anomaly), including all key events, decision points, and deliverables. Its purpose is to provide a complete and traceable view of task execution, facilitating subsequent analysis, auditing, and problem localization.

[0092] Finally, based on a unified identification system, static attribute information and process data chains are integrated and aggregated to construct a supervisory data view. A unified identification system is key to achieving heterogeneous data fusion, providing a common identity or anchor for data from different sources and of different types. This can be a globally unique task identifier or a set of cross-system shared coding rules or mapping mechanisms. For example, each supervisory task item can be assigned a globally unique task ID, requiring all related execution process data to include this task ID when recording, or these identifiers can be managed and distributed through a central registration service. Integration and aggregation refer to combining the core static attribute information (task title, responsible person) surrounding the supervisory task item with the dynamic execution process data chain (progress, communication, status changes) to form a unified and comprehensive data set. This can be achieved through data warehouse technology, data lake technology, or a real-time data integration platform, cleaning, transforming, and loading data from different sources into the same data model, and establishing the relationships between them. A supervisory data view is a logical collection of data that presents all relevant information (including static attributes and dynamic process data) for a supervisory task item in a structured, easily queryable, and analyzable format. This view can be a physical data table, a virtual view, a data cube, or a collection of nodes and edges in a graph database. Its purpose is to provide a unified, complete, and consistent data interface for upper-layer applications (such as analysis modules, visualization modules, and risk prediction models), avoiding the complexity of directly accessing underlying heterogeneous data sources.

[0093] Through the above technical solution, this application effectively solves the data silo problem, ensures the integrity, consistency and high relevance of the supervision data view, provides high-quality data support for subsequent supervision task analysis, risk prediction and visualization based on the view, and improves the accuracy and efficiency of supervision task processing.

[0094] In one embodiment of this application, based on a unified identification system, static attribute information and process data chains are integrated and aggregated to construct a supervision data view, including: Assign a globally unique task identifier to each inspection task item; Assign a data identifier to each piece of execution process data, and bind it to the task identifier of the corresponding supervision task item through the association field; Based on the binding relationship between task identifiers and data identifiers, a bidirectional index is established between static attribute information and process data chain to construct a supervision data view.

[0095] In this embodiment, a globally unique task identifier is assigned to each inspection task item to ensure that each inspection task item has a unique identity throughout the entire inspection task management platform. This task identifier can be a generated universally unique identifier (UUID) to guarantee uniqueness in a distributed environment.

[0096] Furthermore, a data identifier is assigned to each piece of execution process data, and this identifier is bound to the task identifier of the corresponding supervision task item through an association field. The data identifier is used to uniquely identify each specific piece of execution process data, such as progress feedback, attachments, communication records, or status change logs. This data identifier is an auto-incrementing sequence number.

[0097] Building upon this foundation, a bidirectional index is established between static attribute information and process data chains, based on the binding relationship between task identifiers and data identifiers, to construct a supervisory data view. The establishment of this bidirectional index means that the system can not only quickly locate all static attribute information of a supervisory task item and its associated execution process data through the task identifier, but also trace back to its corresponding supervisory task item and its static attributes through the data identifier of any execution process data. For example, this can be achieved using primary and foreign key associations and indexing mechanisms in a database, or by constructing a hash table or inverted index structure in memory. This bidirectional indexing mechanism greatly improves the efficiency of data querying and integration, laying the foundation for constructing a comprehensive and accurate supervisory data view.

[0098] By employing the aforementioned technical solutions, a globally unique task identifier is assigned to each inspection task item, fundamentally resolving the task identifier conflict problem and ensuring that each task item can be uniquely identified within the system, thereby eliminating the possibility of association errors. A data identifier is assigned to each execution process data entry, and this identifier is bound to the corresponding inspection task item's task identifier through an association field, establishing a direct link between process data and task items, effectively preventing data chain attachment deviations. Based on the binding relationship between task identifiers and data identifiers, a bidirectional index is established between static attribute information and the process data chain, enabling mutual lookup capabilities between attribute information and process data, and improving data query efficiency. Ultimately, through this unified identifier system, the constructed inspection data view structure is complete and the information is accurate, greatly improving the efficiency and accuracy of data fusion.

[0099] In one embodiment of this application, the monitoring and analysis of the supervision data view is performed according to a preset risk prediction model to identify abnormal states of supervision task items, including: Based on the inspection data view, obtain the current status information and process data flow of the inspection task items; Input the current status information and process data stream into the preset risk prediction model to obtain the risk index; The risk index is compared with preset risk thresholds at various levels to identify abnormal states that are at the warning or alarm level.

[0100] In this embodiment, when acquiring the current status information and process data flow of an inspection task item, the inspection task management platform can extract the latest status information related to a specific inspection task item in real time from the constructed inspection data view through API interfaces or database queries. This information includes task progress, completion status, and feedback from responsible persons, as well as historical process data flows, such as communication records, attachment uploads, and status change logs. Alternatively, the system can maintain a data subscription service. When relevant data in the inspection data view changes, the data capture module is triggered to push the updated current status information and newly added process data flows to the risk analysis module, thereby providing comprehensive and real-time input data for risk prediction.

[0101] When current state information and process data stream are input into a preset risk prediction model to obtain a risk index, the preset risk prediction model can be a machine learning model, such as a random forest model. This model, trained on historical monitoring data (including normal and abnormal tasks), can learn the correlation between different data features and task risk. After inputting the current state information and process data stream, the model outputs a continuous value between 0 and 1 as the risk index; a higher value indicates a greater risk.

[0102] When comparing the risk index with preset risk thresholds at various levels to identify abnormal states at the warning or alarm levels, the system can preset multiple risk thresholds, such as a low-risk threshold, a medium-risk threshold, and a high-risk threshold. When the risk index is less than the low-risk threshold, the task status is normal; when the risk index is between the low-risk and medium-risk thresholds, it is identified as a warning state; when the risk index is between the medium-risk and high-risk thresholds, it is identified as an alarm state; and when the risk index is higher than the high-risk threshold, it is identified as a severe alarm state. Alternatively, the risk thresholds can be dynamically adjusted based on business needs and historical data analysis. For example, a percentile method can be used to identify the top 10% of tasks by risk index as high-risk and the top 30% as medium-risk. The calculated risk index is compared with these dynamic thresholds to classify the risk level of the inspection task based on the quantified risk index and trigger the corresponding warning or alarm mechanism.

[0103] Through the aforementioned technical solution, this application effectively addresses the problem of fixed, unchanging preset risk thresholds that cannot be dynamically adjusted according to actual risk distribution, leading to insufficient early warning coverage or high false alarm rates. By using a unified supervisory data view, it comprehensively acquires the current status information and process data flow of supervisory tasks, ensuring the completeness and real-time nature of the data required for risk assessment. Subsequently, inputting this multi-dimensional data into a preset risk prediction model quantifies complex supervisory information into an objective risk index, avoiding the subjectivity and inefficiency of manual judgment. Finally, by comparing the risk index with preset risk thresholds at various levels, it can accurately identify abnormal states at the early warning or alarm levels, thereby achieving refined risk classification and timely early warning for supervisory tasks. This mechanism enables the supervisory task management platform to discover potential problems earlier, promptly notify responsible parties for intervention, improve the accuracy and timeliness of supervisory task processing, and effectively reduce the probability of task delays and risk occurrences.

[0104] In one embodiment of this application, a method for processing supervisory tasks based on a unified data view further includes: Periodically analyze the risk index distribution of all inspection tasks; Based on the statistical characteristics of the risk index distribution and the preset early warning coverage target, calculate and update the risk thresholds at each level.

[0105] In this embodiment, periodically calculating the risk index distribution of all inspection tasks aims to obtain real-time data on the overall risk situation of the current inspection tasks, providing a data basis for subsequent dynamic adjustment of the warning threshold. This calculation can be implemented in various ways. For example, the system can be set to perform a statistical task every fixed time interval (such as hourly, daily, or weekly) to summarize the risk index data of all active inspection tasks within that time period.

[0106] Based on this, risk thresholds at all levels are calculated and updated according to the statistical characteristics of the risk index distribution and the preset early warning coverage target. The statistical characteristics are indicators that quantify the risk index distribution, such as the mean, median, standard deviation, variance, skewness, kurtosis, and different percentiles (e.g., the 90th percentile, 95th percentile). These statistical characteristics comprehensively reflect the central tendency, dispersion, and distribution pattern of the current risk index. The preset early warning coverage target is a strategic indicator set by business or management. For example, it can be set to ensure that the top 5% of tasks with the highest risk index are identified as alarm-level, or that the top 15% of tasks with the highest risk index are identified as early warning-level. The methods for calculating and updating thresholds can be varied. For example, the percentile method can be used, directly determining the corresponding percentile from the risk index distribution as the threshold based on the early warning coverage target. If the target is to cover the top 15% of risky tasks, then the 85th percentile of the risk index distribution is set as the early warning threshold. In addition, machine learning or adaptive algorithms can be used to train a model that takes the statistical characteristics of the risk index distribution and the early warning coverage target as input, outputs the optimal risk thresholds at each level, and is optimized through continuous learning and feedback.

[0107] The aforementioned technical solution allows for the dynamic adjustment of warning and alarm thresholds based on the actual distribution of risk indices for each inspection task. This effectively addresses the problem of fixed thresholds failing to adapt to changing risks, preventing high-risk tasks from being missed due to excessively high thresholds or numerous false alarms due to excessively low thresholds. By combining this with preset warning coverage targets, it ensures that critical high-risk tasks can be identified and warned about in a timely and accurate manner, thereby improving the accuracy of identifying abnormal states in inspection tasks and the effectiveness of warnings, ensuring the timely processing of inspection tasks and effective risk management.

[0108] In one embodiment of this application, a method for processing supervisory tasks based on a unified data view further includes: In response to the root cause analysis command initiated on the supervisory task item in an abnormal state, based on the corresponding static attribute information and process data chain in the supervisory data view; Through causal inference analysis, one or more potential root causes leading to the abnormal state are identified; potential root causes include delays in preceding tasks, shortages of key resources, timeouts in response from collaborating departments, defects in the quality of deliverables, or external shocks. Based on the identified potential root causes and their associated data, a structured root cause analysis report is generated and presented visually.

[0109] In this embodiment, responding to a root cause analysis instruction for an inspection task item in an abnormal state means that when an inspection task item is identified as being in an abnormal state (e.g., identified as being at a warning or alarm level by the aforementioned risk prediction model), a clear signal to initiate root cause analysis is received. This instruction can be triggered manually by clicking the root cause analysis button for a specific abnormal inspection task item on the user interface of the inspection task management platform; or, when the duration of the abnormal state exceeds a preset threshold or the abnormal level reaches the highest level, the instruction is triggered according to preset rules to ensure timely response and in-depth analysis of critical abnormal tasks.

[0110] The data foundation for root cause analysis, based on the static attribute information and process data chain corresponding to the inspection data view, originates from a unified inspection data view. This view integrates all relevant information for inspection task items, including their static attribute information (such as basic elements like task title, responsible department, responsible person, and stipulated deadline, as well as extended elements like task background, core requirements, key deliverables, and collaborating departments) and dynamic process data chains (such as progress feedback, supporting materials, communication records, and status change logs). This unified data view allows for a comprehensive and accurate acquisition of the complete context of the task, avoiding analytical blind spots caused by scattered data. For example, root cause analysis can retrieve all static attribute information and process data chains related to the target inspection task item from the inspection data view through a unified data query interface.

[0111] Causal inference analysis refers to using analytical methods to determine the causal relationship between anomalies and various potential factors, rather than merely their correlation. This analysis aims to reveal why anomalies occur, thereby finding the root cause of the problem. For example, a directed acyclic graph can be constructed using Bayesian networks, where nodes represent potential root causes and anomalies, and edges represent causal relationships. By learning from historical data to determine the conditional probability distribution between nodes, the worst-case direct or indirect cause of the anomaly can be inferred.

[0112] Identifying one or more potential root causes of an abnormal state means that the result of causal inference analysis is to explicitly identify the specific factors leading to the abnormal state of the supervisory task item. These factors are verified through causal relationships, rather than superficial phenomena. For example, a causal inference model can output the probability or influence weight of each potential root cause factor leading to the abnormal state, rank these values, and select the top N factors with the highest probability or greatest influence as the identified root causes.

[0113] Generating and visualizing structured root cause analysis reports based on identified potential root causes and their associated data means organizing the analysis results into an easy-to-read and understand report format for easier comprehension and decision-making. Structured reports ensure the completeness and consistency of information, while visualization uses charts, graphs, and other methods to intuitively display complex causal relationships and data trends. For example, the system can preset multiple root cause analysis report templates. Based on the identified root cause types and associated data, users can fill in the corresponding fields in the templates to generate standardized report documents. A timeline view can be used to annotate the execution process of supervisory tasks, key events, anomaly occurrence times, and identified root cause elements on a timeline, helping users understand the chronological relationships of events.

[0114] Through the above technical solution, this application, based on identifying abnormal states of inspection tasks and triggering early warnings, further provides in-depth root cause analysis capabilities, greatly improving the efficiency and accuracy of problem localization, and providing strong support for subsequent inspection task processing and decision-making. This effectively solves the problem of being unable to accurately locate the root cause of the problem after an early warning, thus affecting the efficiency of subsequent processing.

[0115] Please see Figure 2 This paper demonstrates the system application architecture supporting this method from a technical implementation perspective. The architecture adopts a microservice design and a layered model, including: an access layer responsible for request routing, security authentication, and protocol adaptation; a business service layer, decomposed into independent microservices such as task generation service, view construction service, and risk analysis service, which collaborate to complete the core processing logic; a data layer employing a hybrid storage strategy, using a relational database to store core entities and relationships, and a distributed document database to store process logs and unstructured data, linked through a unified data identifier system; and a support layer providing monitoring, configuration management, and service governance capabilities. This architecture ensures the system's resilience, reliability, and maintainability when dealing with massive multi-source data aggregation and high-concurrency analysis requests, providing the technical foundation for the efficient implementation of the step of building the supervisory data view.

[0116] Please see Figure 3 This illustrates the interaction architecture between the supervision task processing system and external business systems (especially the meeting system and the weekly key work reporting system) in this embodiment. For example... Figure 3As shown, this system, acting as a data aggregation and processing hub, achieves decoupled integration with multi-source heterogeneous systems by constructing a standardized Enterprise Service Bus (ESB) and API gateway. Specifically, it synchronizes data with the conference system through a real-time RESTful interface, proactively capturing or receiving structured data packets pushed by the conference system, including resolutions, responsible parties, and completion deadlines. Asynchronous communication with the weekly key tasks system is achieved through message queues, ensuring reliable transmission and eventual consistency of process data such as task status changes and progress feedback. This interactive architecture ensures the real-time nature, reliability, and low coupling of raw data acquisition, laying a solid foundation for building a unified data view.

[0117] Please see Figure 4 The document clearly illustrates the meeting-supervision-key task closed-loop management model implemented in this application using a flowchart. The process begins with resolutions generated by the meeting system. After being obtained through an interface, this system automatically parses and standardizes the tasks, generating supervision tasks with unique identifiers, and synchronizes them to the weekly key task system as execution units. During task execution, progress data from the weekly key task system and process data monitored by this system converge in a unified data view. When the risk analysis module identifies an abnormal state, the system not only triggers tiered warnings but also synchronizes the abnormal information back to the meeting system, providing decision-making input for subsequent meeting agendas. Finally, based on the complete process data chain, the system generates traceable analysis reports to support meeting debriefing and continuous optimization, thus forming a complete, two-way data business closed loop from decision-making, execution, tracking to feedback optimization. Figure 4 This diagram visually demonstrates how this application breaks down system barriers to automate and intelligentize management processes. Specifically, each node in the diagram represents a different system processing module; solid arrows indicate the forward data flow of task issuance and execution, while dashed arrows indicate the reverse data flow of status feedback and optimization decisions. Specifically: after the meeting system generates a structured resolution, it is broken down by the supervision system and synchronized to the weekly key work system; data during execution is collected and analyzed in a unified data view; when a node identifies an anomaly, an alert is triggered and fed back to the meeting system; simultaneously, routine monitoring continues; based on the feedback data, the meeting debriefs to form a new resolution, initiating the next closed loop. This flowchart clearly reveals the cross-system, full-cycle, two-way feedback intelligent supervision and management mechanism implemented in this application.

[0118] Corresponding to the supervision task processing method based on a unified data view in the above embodiment, Figure 5 This is a structural block diagram of a supervision task processing system based on a unified data view, provided as an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 5The supervision task processing system 50 based on a unified data view includes: a data acquisition module 51, a task generation module 52, a data view module 53, a view analysis module 54, a risk analysis module 55, and an anomaly push module 56.

[0119] Among them, the data acquisition module 51 is used to acquire raw data related to supervision from multi-source heterogeneous systems; Task generation module 52 is used to parse the raw data and generate inspection task items; Data view module 53 is used to collect the corresponding attribute information and integrate the execution process data of the supervision task items to construct a supervision data view; The view analysis module 54 is used to analyze the supervision task items based on the supervision data view and to visualize the analysis results. The risk analysis module 55 is used to monitor and analyze the supervision data view according to the preset risk prediction model, and identify the abnormal status of the supervision task items. The anomaly push module 56 is used to trigger tiered early warnings based on anomaly states and push the early warning information to the corresponding responsible persons.

[0120] See Figure 6 , Figure 6 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 6 The electronic device in this embodiment may include one or more processors, one or more input devices, one or more output devices, and one or more memories. The processor, input devices, output devices, and memories communicate with each other via a communication bus. The memory stores computer programs, including program instructions. The processor executes the program instructions stored in the memory. The processor is configured to invoke the program instructions to perform the functions of the modules in the above-described device embodiments, for example... Figure 2 The functions of the data acquisition module 51, task generation module 52, data view module 53, view analysis module 54, risk analysis module 55, and anomaly push module 56 are shown.

[0121] It should be understood that, in the embodiments of this application, the processor may be a Central Processing Unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0122] Input devices may include touchpads, fingerprint sensors (for collecting the user's fingerprint information and fingerprint orientation information), microphones, etc., while output devices may include displays (LCDs, etc.), speakers, etc.

[0123] The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store information about the device type.

[0124] In specific implementations, the processor, input device, and output device described in the embodiments of this application can execute the implementation methods described in any embodiment of the supervision task processing method based on a unified data view provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0125] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0126] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0127] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0128] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0131] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0132] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for processing supervisory tasks based on a unified data view, characterized in that, The method, applied to a supervision task management platform, includes: Obtain raw data related to supervision from multi-source heterogeneous systems; The raw data is parsed to generate inspection task items; Based on the aforementioned inspection task items, their corresponding attribute information is collected and their execution process data is integrated to construct an inspection data view; Based on the aforementioned inspection data view, the inspection task items are analyzed, and the analysis results are presented visually. Based on the preset risk prediction model, the inspection data view is monitored and analyzed to identify the abnormal status of the inspection task items; Based on the abnormal state, a tiered early warning is triggered, and the warning information is pushed to the corresponding responsible person.

2. The method for processing supervisory tasks based on a unified data view according to claim 1, characterized in that, The step of parsing the raw data to generate inspection task items includes: The raw data is preprocessed to obtain standard format data; Based on a pre-defined ontology library of supervisory elements, the standard format data is semantically aligned and mapped to obtain semantically unified data. By performing rule mapping and association queries on the structured data in the semantically unified data, the basic elements required to constitute the supervision task items are extracted; Based on a pre-defined natural language processing model, deep semantic analysis is performed on the unstructured data in the semantically unified data to extract the extended elements required to constitute the inspection task items. The basic elements and the extended elements are associated and merged to generate the supervision task items, and a confidence score is assigned to each task item element in the supervision task items; the task item elements include: basic elements and extended elements; The task item elements whose confidence scores are less than the preset confidence threshold are verified and corrected; Based on the verification and correction results, the verified supervision task items are obtained, and the verified supervision task items are used as the final supervision task items.

3. The method for processing inspection tasks based on a unified data view according to claim 2, characterized in that, Assigning a confidence score to each element of the inspection task items includes: For the aforementioned basic elements, a first confidence sub-score is calculated based on the confidence weight of the data source, the completeness of the data fields, and the matching degree of the mapping rules. For the extended elements, a second confidence sub-score is calculated based on the entity recognition probability output by the natural language processing model, the confidence level of relation extraction, and the consistency of contextual semantics. Based on the type of the inspection task item, different weights are assigned to the basic elements and the extended elements respectively; For each task item element, the final confidence score is calculated based on its corresponding confidence sub-score and its weight coefficient.

4. The method for processing supervisory tasks based on a unified data view according to claim 2, characterized in that, The step of verifying and correcting task item elements whose confidence scores are less than a preset confidence threshold includes: When the confidence score of the task item element is less than the preset confidence threshold, a manual confirmation request is generated and pushed. The task item elements with confidence scores lower than the preset confidence threshold and their context information are converted into vector representations; In the historical task database, retrieve K historical inspection task items and their corresponding task item elements that are similar to the vector representation; Extract the values ​​of the corresponding task item elements from the K historical task items and perform clustering. Then, push the centroid of the largest cluster as a candidate correction suggestion to the manual confirmation request. Update the corresponding task item elements based on the results of manual operations.

5. The method for processing inspection tasks based on a unified data view according to claim 4, characterized in that, Also includes: Based on the correction data accumulated during the manual verification, a model optimization sample set is constructed; Based on the model optimization sample set, the parameters of the natural language processing model are optimized and adjusted using a preset optimization algorithm to obtain the target parameters; The natural language processing model is iteratively updated and redeployed based on the target parameters.

6. The method for processing supervisory tasks based on a unified data view according to claim 1, characterized in that, The process of constructing a supervision data view based on the supervision task item, by aggregating its corresponding attribute information and integrating its execution process data, includes: Using the generated inspection task item as the core node, its corresponding basic elements and extended elements are associated and mounted as static attribute information; Extract execution process data associated with the supervision task item from the multi-source business execution system; Establish the temporal and logical relationships between the inspection task items and their associated execution process data to form a process data chain; Based on a unified identification system, the static attribute information is integrated and aggregated with the process data chain to construct the supervision data view.

7. The method for processing inspection tasks based on a unified data view according to claim 6, characterized in that, The unified identifier system is used to integrate and aggregate the static attribute information with the process data chain to construct the supervision data view, including: Assign a globally unique task identifier to each of the aforementioned inspection tasks; Assign a data identifier to each piece of execution process data, and bind it to the task identifier of the corresponding supervision task item through an association field; Based on the binding relationship between the task identifier and the data identifier, a bidirectional index is established between the static attribute information and the process data chain to construct the supervision data view.

8. A supervision task processing system based on a unified data view, characterized in that, include: The data acquisition module is used to acquire raw data related to supervision from multi-source heterogeneous systems; The task generation module is used to parse the raw data and generate inspection task items; The data view module is used to collect the corresponding attribute information and integrate the execution process data of the supervision task items to construct a supervision data view. The view analysis module is used to analyze the supervision task items based on the supervision data view and to visualize the analysis results. The risk analysis module is used to monitor and analyze the supervision data view according to the preset risk prediction model, and identify the abnormal status of the supervision task item. The anomaly push module is used to trigger tiered early warnings based on the anomaly status and push the early warning information to the corresponding responsible person.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images