Large-model-based enabling security liability list management system, device and medium

Abstract

The invention provides a big-model-based enabling security liability list management system and device and a medium, and the system employs a big language model, is based on a knowledge service base, and is combined with project demands and personnel capabilities to allocate proper persons in charge for project tasks. Performing tracking, management and early warning on task progress based on responsibility competence process data and a security responsibility knowledge graph which are acquired in real time; performing multi-dimensional analysis according to the task allocation and responsibility competing data, and generating a visual report; notices are generated based on information of all the modules, and task cooperation is promoted through accurate pushing of an intelligent routing strategy; meanwhile, the user permission is dynamically managed based on the security responsibility knowledge graph. According to the method, data are analyzed and interpreted by means of a large language model and a knowledge service base, and key information and modes in a liability list can be accurately understood; the knowledge service base supports knowledge index construction, updating maintenance, sample dynamic assembly and retrieval enhancement generation, so that retrieval and pushing are more efficient, flexible and accurate.

Description

Technical Field

[0001] This invention relates to artificial intelligence, algorithms, and natural language processing, specifically to a safety responsibility list management system, device, and medium based on a large model. Background Technology

[0002] Safety management is a core aspect of modern enterprise operations, and the safety responsibility list is a core tool for clarifying and implementing the safety responsibilities of all levels of positions. It includes five stages: clarifying responsibilities, understanding responsibilities, fulfilling responsibilities, supervising responsibilities, and holding individuals accountable. These stages are interconnected and mutually reinforcing, forming a complete chain of the safety production responsibility system. Clarifying Responsibilities: Based on relevant national laws and regulations on safety production and company regulations, each unit establishes its own responsibility system, clarifying the responsibilities and obligations of managers and positions at all levels for safety production, ensuring that each responsibility is assigned to a specific person, and strengthening the safety awareness of all employees. Understanding Responsibilities: Clarifying the uniformity and standardization of the implementation of safety production responsibilities in each unit, deepening the safety responsibility awareness of all employees through list dissemination, list publicity, list training, list examinations, and external units, standardizing the interpretation and understanding of laws and regulations, and effectively implementing the safety responsibilities of their own positions through training and daily learning. Fulfilling Responsibilities: Ensuring the implementation of safety responsibilities corresponding to all positions by all employees, through safety work deployment and implementation, on-site attendance, supervision and inspection, on-site handling, communication and reporting, and performance evaluation, urging personnel at all levels to fulfill their respective safety responsibilities. The results of implementation will be uploaded and archived in conjunction with meeting minutes and work logs to ensure traceability. Supervision and accountability: Supervise the implementation of safety responsibilities at each position, utilizing methods such as inspections, safety production patrols, unannounced inspections, special inspections, and routine checks to urge units to establish and improve a supervision and evaluation mechanism for the implementation of safety production responsibilities, and to urge personnel at all levels to fulfill their safety production responsibilities. Accountability: All levels of the company will implement the system of holding higher-level units accountable for safety responsibilities of lower-level units, clarifying the principles, circumstances, targets, procedures, methods, and standards for accountability, and, in conjunction with the company-wide safety responsibility system, reasonably and legally determine responsibilities based on factors such as safety duties, performance of duties, and conditions for performance.

[0003] However, existing safety responsibility list management primarily relies on manual creation, static documents (such as Excel and Word), and offline communication processes. This approach has significant limitations. The creation process is inefficient and highly subjective: the list heavily depends on the personal experience of managers, lacks data-driven decision support, and struggles to ensure the comprehensiveness, rationality, and scientific nature of responsibility allocation. It also suffers from poor dynamic adaptability: project environments and risks are constantly changing, but traditional lists are slow to update and cannot be automatically adjusted and optimized based on real-time data (such as historical accident data and real-time monitoring information), leading to a disconnect between responsibility clauses and actual circumstances. Monitoring and traceability are difficult: monitoring the fulfillment of responsibilities largely relies on manual reporting, which is inefficient and prone to information delays or distortions. The entire chain of "clarifying responsibilities, knowing responsibilities, fulfilling responsibilities, supervising responsibilities, and holding people accountable" is difficult to record transparently and traceably. Finally, the level of intelligence is insufficient: it lacks the ability to deeply analyze list data, making it impossible to intelligently predict potential risks, provide optimization suggestions, or generate effective decision support reports.

[0004] While some information systems have been introduced in an attempt to address these issues, these systems often lack true intelligence and are unable to deeply understand the content of responsibilities, automatically verify the rationality of responsibilities, or provide forward-looking decision support. Therefore, a new solution is urgently needed to achieve intelligent, precise, and efficient management of the safety responsibility list across the entire chain of clarifying, understanding, fulfilling, supervising, and holding accountable responsibilities. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a security responsibility list management system based on a large model, comprising: The responsibility list module is used to assign tasks in a project to appropriate responsible persons by utilizing a large language model and based on a knowledge service foundation, combined with project requirements and personnel capabilities; the knowledge service foundation is used to manage the safety responsibility knowledge graph; The task management module is used to track and manage task progress and generate early warnings based on real-time data on the responsible persons' performance of their duties, combined with a safety responsibility knowledge graph. The data analysis and reporting module is used to perform multi-dimensional analysis and generate visual reports based on the task allocation information from the responsibility list module and the responsibility person's performance process data from the task management module. The notification collaboration module is used to generate notifications based on information from each module, and to push them accurately to relevant responsible persons based on intelligent routing strategies to achieve task collaboration. The user management module is used to manage user permissions based on the dynamic business relationships between various modules, using a security responsibility knowledge graph.

[0006] Preferably, the responsibility list module includes: a responsibility library submodule for constructing and storing a security responsibility knowledge graph; The construction of the security responsibility knowledge graph includes: pre-training a general model using a large language model based on large-scale security responsibility domain text, and then performing entity recognition and relation extraction through multi-task learning to simultaneously identify entities and relations. Preferably, the responsibility list module further includes: a responsibility matching submodule, the responsibility matching submodule including: The responsible person capability profile building unit is used to integrate employees' discrete data, unstructured text data, and numerical data to generate a unified high-dimensional capability vector; The task requirement vectorization unit is used to convert task description text or task description text generated from the security responsibility knowledge graph into task vectors. The intelligent matching algorithm unit is used to generate a list of recommended responsible persons for tasks by simultaneously optimizing capability matching and load balancing using a multi-objective optimization algorithm.

[0007] Preferably, the responsibility list module further includes a responsible person display submodule, which is configured as follows: Monitor changes in relationships related to liability assumption within the aforementioned safety responsibility knowledge graph; In response to the aforementioned change event, the graph data is dynamically queried, and the display interface for responsibility allocation and task progress is updated in real time through visualization components from both the manager's and employee's perspectives.

[0008] Preferably, the responsibility list module further includes a responsibility update submodule, which includes: The change identification unit is used to automatically identify responsibility change points from project documents and system instructions using text difference technology and information extraction technology; The impact analysis unit is used to perform path queries and logical rule reasoning based on the security responsibility knowledge graph to analyze the impact range of the change point. The update unit is used to generate a knowledge graph update scheme based on the impact analysis results, and update the security responsibility knowledge graph through an approvable workflow.

[0009] Preferably, the task management module includes a progress tracking submodule, which includes: The progress acquisition unit is used to collect data on the performance of duties from multiple channels, including text reports, spatiotemporal data, and visual data, and to perform analysis and feature extraction using NLP models, spatiotemporal verification rules, and computer vision models, respectively. The progress calculation unit is used to comprehensively judge the multi-source analysis results using an evidence fusion algorithm, determine the actual progress status of the task, and automatically calculate the progress percentage; the automatic early warning submodule is configured as follows: When the progress calculation unit determines that the task progress is lower than the preset threshold or exceeds the time limit, it automatically triggers an early warning message and pushes it to the relevant responsible persons and managers through the notification collaboration module.

[0010] Preferably, the data analysis and reporting module includes a performance evaluation sub-module, which is configured as follows: Based on the relevant data from each module, and using the assessment indicator system, multi-dimensional performance indicator values ​​are calculated to obtain a performance evaluation of the project and / or the responsible person; the multi-dimensional aspects include: project progress and / or the responsible person's task completion rate, completion quality, and behavior; Performance evaluation results can be presented using one or more visualization methods, such as radar charts, comparative bar charts, and trend line charts.

[0011] Preferably, the notification collaboration module includes: The message generation unit is used to select the corresponding template from the preset template library according to the system event type, and fill the event parameters through rules or lightweight large models to generate personalized notification messages. The message distribution unit is used to distribute messages based on their urgency and recipient preferences, using one or more channels such as in-system messaging, mobile app push notifications, SMS, and email.

[0012] Based on the same inventive concept, this application also provides an electronic device, including: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a security responsibility list management system based on a large model empowerment provided in this application is executed.

[0013] Based on the same inventive concept, this application also provides a readable storage medium having an executable program stored thereon, which, when executed, executes a security responsibility list management system based on a large model empowered by this application.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a safety responsibility list management system based on a large-scale model, comprising: a responsibility list module, used to allocate tasks in a project to appropriate responsible persons using a large language model and a knowledge service foundation, combined with project requirements and personnel capabilities; the knowledge service foundation is used to manage a safety responsibility knowledge graph; a task management module, used to track and manage task progress and generate early warnings based on real-time acquired data on the responsible persons' performance of their duties, combined with the safety responsibility knowledge graph; a data analysis and reporting module, used to perform multi-dimensional analysis and generate visualized reports based on task allocation information from the responsibility list module and data on the responsible persons' performance of their duties from the task management module; a notification and collaboration module, used to generate notifications based on information from each module and to accurately push them to relevant responsible persons based on intelligent routing strategies to achieve task collaboration; and a user management module, used to manage user permissions based on the dynamic business relationships of each module using the safety responsibility knowledge graph. This invention utilizes a large language model combined with a knowledge service foundation to support data analysis and interpretation, forming a basic data model. By training the large model, it can accurately understand the key information and patterns in the responsibility list data. This invention relies on the knowledge service foundation to complete knowledge index construction, knowledge updating and maintenance, dynamic sample assembly, and enhanced retrieval generation. These steps are key to achieving rapid knowledge retrieval, real-time updates, flexible configuration, and ensuring content accuracy, making retrieval and push more efficient, flexible, and precise. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of a safety responsibility list management system based on a large model, according to the present invention. Figure 2 This is a schematic diagram of an electronic device structure according to the present invention. Detailed Implementation

[0016] The responsibility list system, centered on clarifying, understanding, fulfilling, supervising, and holding accountable responsibilities, aims to enhance awareness of responsibility, improve work efficiency, and ensure the implementation of responsibilities. This invention provides a safety responsibility list management system based on a large-scale model. It primarily utilizes large-scale model technologies such as Natural Language Processing (NLP), Machine Learning (ML), and Deep Learning (DL), combined with a knowledge service foundation, to analyze and interpret data, forming a basic data model. By training this large-scale model, it can accurately understand the key information and patterns in the responsibility list data.

[0017] This invention relies on a knowledge service platform to complete the construction of knowledge indexes, knowledge updates and maintenance, dynamic assembly of samples, and retrieval enhancement generation. These steps are key to achieving rapid knowledge retrieval, real-time updates, flexible configuration, and ensuring content accuracy, making retrieval and push more efficient, flexible, and precise.

[0018] The knowledge service platform is the infrastructure and technology platform supporting the above functions. It integrates various technologies and tools to provide efficient, accurate, and personalized knowledge services. Its core workflow includes: Knowledge Integration and Storage: Information from various data sources (integrating text, images, and other formats) is collected through various channels and tools, and then integrated and cleaned to ensure data accuracy and integrity. Distributed storage, relational databases, and non-relational databases are employed to efficiently store and manage data within the knowledge service infrastructure, supporting rapid data access and updates while also ensuring the scalability and query efficiency of the knowledge index.

[0019] Knowledge Processing and Representation: Utilizing technologies such as natural language processing, data mining, and machine learning, the collected data is processed and analyzed to extract valuable information and knowledge, providing decision support for knowledge services. Simultaneously, through technologies such as ontology and semantic web, knowledge is represented in a structured manner, supporting various reasoning methods such as rule-based reasoning and statistical reasoning to achieve intelligent application of knowledge.

[0020] Knowledge Application and Generation: Based on the knowledge representation and reasoning capabilities provided by the knowledge service platform, Retrieval Enhanced Generation (RAG) can be realized. That is, based on the large model of the power industry, combined with integrated knowledge content, an accurate, professional and reliable knowledge base can be built, thereby providing users with more accurate and real-time professional knowledge services.

[0021] To better understand this invention, the following description, in conjunction with the accompanying drawings and examples, will further illustrate the invention.

[0022] Example 1: like Figure 1 As shown, this invention provides a security responsibility list management system based on a large model, employing a layered architecture design to support the flexibility and scalability for different business needs. It includes: User Management Module: Responsible for managing user permissions, ensuring that each user's access permissions are associated with their job title, and enabling secure information sharing.

[0023] Responsibility List Module: Provides functions for task assignment, updating, review and confirmation, supports dynamic adjustment and change of the responsibilities corresponding to the task, and ensures accurate recording of all tasks and responsibilities.

[0024] Task Management Module: Allows responsible personnel to create, assign, and track tasks, clearly display the task completion progress, and support functions such as task delay reminders.

[0025] Data Analysis and Reporting Module: Collects and analyzes data in real time, provides a variety of visual reports, and supports management decision-making and performance evaluation.

[0026] Notification and Collaboration Module: The system can push task notifications and change information in real time, and supports collaboration and communication among team members to improve work efficiency.

[0027] Furthermore, A key technical principle of the responsibility list system is data transparency. By establishing a unified responsibility database, the system records the relationship between each project, each task, and the specific responsible person, ensuring that all responsibilities are clearly defined. Therefore, the responsibility list module includes: a responsibility database submodule, a responsibility matching submodule, a responsible person display submodule, and a responsibility update submodule.

[0028] Specifically, the Responsibility Repository submodule manages and stores the safety responsibility knowledge graph, including entities, relationships, attributes, and logical rules / constraints. Entities are the "nodes" of the knowledge graph, representing specific objects within the safety responsibility system. These mainly include: projects, and the responsible parties, responsible objects, responsible actions, legal basis, and risk points associated with those projects. Responsible parties include one or more of the following attributes: position, department, skill level, and responsible person, such as: Project Manager (position), Welding Team (department), Zhang San (responsible person), Senior Fitter (skill level). Responsible objects include one or more attributes such as equipment model, location, and risk level. Responsible actions include one or more attributes such as execution frequency and time limit. Legal basis includes one or more attributes such as law name, clause content, and level of validity. Risk points are special entities in the safety field used to associate responsibility, such as falls from heights, electric shocks, and being struck by objects.

[0029] The security responsibility knowledge graph is constructed in the following way: A domain corpus is constructed based on a large collection of safety responsibility texts (including the Production Safety Law, industry standards, corporate regulations, and historical accident reports). A general-purpose model is pre-trained on this corpus using a large language model. Then, multi-task learning is employed to simultaneously identify entities and relationships for entity recognition and relationship extraction. Specifically, a BiLSTM is first used as a shared encoder to encode the pre-trained sentences, obtaining the contextual representation of each word. Then, for each entity type to be extracted (e.g., responsible entity, responsible object), a pointer network is set up. These pointer networks share the output of the same BiLSTM encoder but decode independently, predicting the start and end positions of the corresponding entity type. Simultaneously, to extract relationships between entities, a relationship classification module is added to the model. Based on the representations of two entities identified by the pointer networks, it determines whether a predefined relationship exists between them, thus obtaining triples. The extracted triples are stored in a graph database (e.g., Neo4j) to form a safety responsibility knowledge graph. This invention designs rich attributes and logical constraints for the graph: Attribute definition: Attributes are defined for each entity and relationship. For example, the "Responsible Behavior" node has attributes such as "Execution Frequency," "Time Limit Requirements," and "Risk Level." Logical Constraint Rules: Logical constraints are defined at the graph schema layer for subsequent omission detection. For example: Inclusion constraint: If "High-Altitude Operations" responsibility exists, then "Safety Belt Inspection" responsibility must also exist. Mutual Exclusion constraint: "Work Approval" responsibility and "Work Execution" responsibility cannot be borne by the same entity. Timing constraint: "Technical briefing" must be completed before "On-Site Operation." Furthermore, the safety responsibility knowledge graph also includes relationships such as affiliation, responsibility, and participation, used to describe the dynamic business relationships between users and projects, tasks, and equipment.

[0030] Specifically, the responsibility matching submodule is used in the early stages of a project to assign responsibilities to appropriate responsible persons based on project needs and personnel capabilities, ensuring the rationality and effectiveness of responsibility matching. It is implemented based on a responsibility matching method that combines multi-source data fusion profiling and multi-objective optimization, and includes: a responsible person capability profile construction unit, a task requirement vectorization unit, and an intelligent matching algorithm unit.

[0031] Furthermore, the responsible person capability profile building unit is used to construct a dynamically updated, quantifiable capability vector for each employee. This is achieved by fusing multi-source data, including: (1) Discretized data (such as skill certificates, department affiliation) are converted into numerical vectors using one-hot encoding technology; (2) For unstructured text data (such as historical performance records), semantic features are extracted using a fine-tuned BERT model to generate text representation vectors; (3) Numerical data (such as task completion rate) are normalized.

[0032] Finally, all vectors are concatenated and normalized to form a unified high-dimensional capability profile vector.

[0033] Furthermore, the task requirement vectorization unit is used to convert task description text or task description text generated from the security responsibility knowledge graph into task vectors. Specifically, this includes two cases: (1) For tasks that can be supported by a security responsibility knowledge graph, the system extracts relevant entities and relationships from the graph according to preset templates or rules, and automatically generates a natural language description text; (2) For new tasks not defined in the graph, the task description text input by the user is used directly. Then, the Sentence-BERT model is used to convert the above text into a task vector.

[0034] Furthermore, the intelligent matching algorithm unit employs a multi-objective optimization method to recommend a responsible person for the task. This method simultaneously optimizes two objectives: (1) Ability matching degree, which is obtained by calculating the cosine similarity between the task vector and the employee ability vector;

[0035] in, For capability matching degree; Represents the task vector. Represents the employee's capability vector. Represents cosine similarity.

[0036] (2) Load balancing, reflected by calculating the load factor, which is inversely proportional to the number of tasks currently assigned to an employee (e.g., , (This represents the number of tasks currently assigned to an employee).

[0037] The multi-objective comprehensive scoring function is transformed into a single-objective comprehensive scoring function using a linear weighting method:

[0038] in, For comprehensive scoring, These are configurable weighting coefficients, and the sum of the coefficients is 1.

[0039] The system ultimately sorts employees in descending order of their overall scores and generates a recommendation list.

[0040] Specifically, the "Responsible Person Display" submodule is used to showcase task progress and the responsibilities of the responsible persons from both managerial and employee perspectives. From a managerial perspective, visualization tools clearly display the responsible person for each task, enabling each team manager to clearly understand the situation within their jurisdiction. These visualization tools include: graph visualization components (such as force-directed diagrams, tree diagrams, etc.), and table / card components. The employee perspective display includes generating a dedicated responsibility workbench for each employee, displayed in Kanban format, including the responsible person's to-do list and scope of responsibility.

[0041] As projects progress, responsibilities and tasks may be adjusted. The system supports dynamic updates to the responsibility list to ensure information is always up-to-date and avoids delays caused by unclear responsibilities. Therefore, the responsibility update submodule is used to: automatically identify change points involving responsibility adjustments based on monitoring changes in the project environment, analyze their impact scope, and drive dynamic, retrospective updates to the responsibility list through a workflow to ensure that the responsibility list always remains consistent with the actual project situation. Specifically, it includes: a change identification unit, an impact analysis unit, and... The change identification unit automatically identifies and extracts valid change points that may trigger responsibility updates from various data sources. Specifically, the system continuously monitors the project document library, meeting minutes from the OA system, and user-submitted change requests. For unstructured text (such as new regulations): text difference technology is used to compare different versions of the document, combined with a finely tuned NLP model (such as BERT for sequence labeling) to identify added, deleted, and modified sentences. Subsequently, information extraction technology is used to identify entities (such as "add equipment X") and relationships (such as "the person in charge of equipment X is changed to department Y") in the changed content. For structured / semi-structured data (such as system instructions): instruction parameters are directly parsed, such as changes in task time caused by project plan adjustments. Then, the change actions are classified: the identified changes are classified as "adding responsibility", "deleting responsibility", "modifying responsibility attributes (such as responsible person, time limit)", etc., providing input for subsequent impact analysis.

[0042] The Impact Analysis unit is used to analyze the scope of impact of identified changes based on a safety responsibility knowledge graph. Specifically, it involves using the change point (e.g., "Adding Equipment X") as a query condition to perform path queries in the graph database. For example, it queries all responsibility behavior nodes such as "Inspection" and "Maintenance" associated with "Equipment" in the knowledge graph, along with their responsible entity nodes. Then, it uses predefined logical constraint rules from the Responsibility Library submodule for reasoning. This unit can also output a structured report, which includes: the change details, the set of directly affected responsible nodes, potential indirect impacts inferred from the rules (such as newly added responsibilities), and a list of affected stakeholders.

[0043] The update unit is used to execute update operations in a controlled and traceable manner based on the impact analysis results. Specifically, based on the change content and impact analysis results, a knowledge graph update plan is generated, including a list of entities and relationships to be added, deleted, or modified. This update plan can be executed immediately or sent to a predefined workflow for manual review before execution. For example, the system automatically creates a "Responsibility Update Approval Form" and pushes it sequentially to the change proposer, the responsible department head, and the security administrator for approval. A detailed impact analysis report can be viewed in the approval workflow. When the change is executed, the data in the knowledge graph is modified, and version management, notifications, rollbacks, and traceability are also performed.

[0044] This module utilizes NLP and text difference technology to automate the process of proactively sensing and parsing changes, significantly improving update timeliness. By leveraging knowledge graph-based relational queries and predefined logical rules, it performs systematic and intelligent impact chain analysis, avoiding omissions that might occur with manual analysis and ensuring the integrity and consistency of the accountability system. Furthermore, by placing update actions within the approval workflow, it ensures the compliance and seriousness of changes while achieving full traceability through version records, thus meeting the audit requirements of security management.

[0045] Furthermore, the responsibility list system is not only used for responsibility allocation, but more importantly, for real-time monitoring of responsibility fulfillment. Therefore, the task management module includes: Specifically, the progress tracking submodule: The system tracks the progress status of tasks in real time, accesses the start and completion times of tasks, and generates dynamic progress bars or Gantt charts, allowing managers to clearly understand the work progress at a glance. It includes: a progress acquisition unit and a progress calculation unit.

[0046] The progress acquisition unit is used to collect data on the performance of responsibilities by responsible persons from multiple channels, and to perform preliminary analysis and feature extraction.

[0047] The collection and parsing of text data includes: (1) Based on the unstructured text data such as work logs, acceptance reports, and shift handover records submitted by employees, the BERT model is used to perform named entity recognition and relation extraction to extract key progress information and form structured progress events. For example, from the log "Today, a comprehensive inspection of fire hydrants in area A has been completed, and 2 problems have been found", the following can be extracted: (Action: Completed, Object: Fire Hydrant Inspection, Result: 2 Problems Found).

[0048] (2) The collection and verification of spatiotemporal data includes: collecting GPS positioning data, clock-in timestamps, sensor data, etc. from mobile terminal APP, extracting location / time data from them, and comparing them with the scope / time limit required by the task to obtain the verification result. For example, by determining whether the employee's GPS coordinates are within the geofence of the "foundation pit monitoring point" and combining it with whether the timestamp is within the planned time, the "on-site monitoring" task is verified to be completed on time.

[0049] (3) The collection and intelligent analysis of visual data includes: collecting images and video streams uploaded from on-site photos / videos, and using computer vision models to analyze them to obtain the analysis results of visual evidence. For object detection models (such as YOLO), it identifies whether there are specific objects or safety measures in the photos. For example, it identifies whether the image contains a "fire extinguisher" and the pressure gauge pointer is in the green area. For scene classification models, it judges the overall scene of the image. For example, it distinguishes between scenes such as "preparation before construction", "construction in progress", and "cleanup after construction".

[0050] The progress calculation unit is used to: fuse and analyze the multi-source data processed and analyzed by the progress acquisition unit, and comprehensively judge the actual progress status of the task.

[0051] Specifically, rule-based evidence fusion strategies or algorithms such as DS evidence theory are employed to comprehensively evaluate evidence from multiple sources, including text, spatiotemporal, and visual evidence. For example, a rule might be: a task is considered "completed" if and only if the text report contains the keyword "completed," GPS verification confirms the task is within the task area, and visual analysis confirms key steps have been completed. Then, based on the fusion results, the task progress status is categorized into predefined classes, such as "not started," "in progress," "completed," "delayed," and "anomaly present." For tasks in progress, the system automatically calculates the progress percentage based on predefined task milestones or sub-steps, combined with currently verified completed steps. For example, in an inspection task with 5 standard steps, the progress automatically updates to 60% when visual and positioning data confirm 3 steps have been completed.

[0052] Once the task status is updated, the responsible person display sub-module can be used to display the updated task status and the responsible person's work content. If necessary, the knowledge graph can be updated using the update unit.

[0053] Specifically, the automatic early warning submodule: if the progress of a task is lower than expected or exceeds the set time threshold, the system will send a reminder notification to the person in charge and relevant management personnel to ensure that the task can be paid attention to and handled in a timely manner.

[0054] Specifically, the anomaly reporting submodule: In cases of frequent delays or poor performance of duties, the system can generate anomaly reports so that management can identify problems and take timely measures.

[0055] Furthermore, the responsibility list system also needs to have data analysis capabilities to help management make more informed decisions. Therefore, the data analysis and reporting module includes: Specifically, the performance evaluation submodule: The system aggregates various types of data generated by each module, such as task allocation information from the responsibility list module and data on the performance of responsibilities by responsible persons from the task management module. It utilizes a comprehensive performance quantification evaluation method with configurable multi-dimensional weights to conduct multi-dimensional analysis of project information, such as the task completion rate of responsible persons and project delay analysis, providing a basis for performance appraisal. The performance evaluation submodule includes: a performance indicator calculation unit and a visualization display unit.

[0056] The performance indicator calculation unit is used to acquire multi-dimensional data such as progress, quality, and behavior generated during the execution of each task in real time, based on a pre-set assessment indicator system. It dynamically analyzes the indicator calculation rules and performs automated calculations to obtain the values ​​of each performance indicator, enabling a quantitative assessment of each responsible person's performance indicators and providing an accurate and objective data foundation for performance evaluation. The assessment indicator system includes several performance indicators, each with assigned attributes such as indicator ID, indicator name, and calculation rules. The calculation rules include indicator type (quantitative / qualitative), assessment period, target value, and weight. The assessment results can also be linked to subsequent applications such as training recommendations, salary incentives, and job adjustments.

[0057] Furthermore, the visualization unit is used to display the assessment results in various graphics as needed, including but not limited to: (1) Performance radar chart: simultaneously displays the performance of the same person in different dimensions such as basic execution and process behavior.

[0058] (2) Comparison bar chart: Show the performance scores of different members in the project team.

[0059] (3) Trend line chart: Shows the performance trend of an individual or project team in different assessment cycles.

[0060] Furthermore, the notification and collaboration module is responsible for automatically generating personalized notification messages throughout the entire lifecycle of responsibility management (such as responsibility assignment, progress updates, alert triggering, and change notifications), and accurately pushing them to relevant responsible parties based on intelligent routing strategies. It also provides structured collaboration tools to ensure that communication records are strongly linked to specific tasks, forming a traceable collaboration loop. This includes: a message generation unit. The message generation unit is used to notify users of various events occurring within the system in the form of messages. This includes: (1) Event listening and capturing: Listen for event messages from other modules of the system (such as responsibility matching completion, progress update, early warning trigger, etc.) and capture key parameters (such as task ID, triggerer, timestamp, event type).

[0061] (2) Message Template Selection: Select the corresponding message template from the template library based on the event type. The template library has multiple preset message types, including but not limited to: task assignment notification templates, task reminder / early warning notification templates, responsibility change notification templates, and system announcement templates. For example, rule-based NLG or lightweight large models can be used to populate event parameters into the template to generate natural and fluent personalized messages. For example: Input: Event Type = Task Assignment, Task Name = "Monthly Fire Inspection", Responsible Person = "Zhang San", Deadline = "2024-06-30" Output: "Hello, Zhang San. You have been assigned the task of [Monthly Fire Inspection]. Please complete it before June 30, 2024. Click to view details." The message distribution unit is used to determine the final delivery path and method of a message based on the message type, urgency (achieved through push priority), and recipient preferences, using a multi-channel push strategy.

[0062] The push channels include: Internal system messages: All messages will be sent for unified archiving. (Default channel).

[0063] Mobile app push notifications: Used for timely reminders. (Triggered when push priority P > medium threshold).

[0064] SMS: Used for highly urgent messages. (Triggered when push priority P > high threshold, and the user has not read system messages for a long time).

[0065] Email: Used for non-urgent notifications, summary reports, and formal communications that require record keeping. (Defaults to use for summarizing information such as daily / weekly reports).

[0066] Furthermore, data security is paramount in a responsibility list system. In the user management system of this invention, in addition to setting operational permissions, users also have data permissions set based on a security responsibility knowledge graph and performance data from related modules. The system sets differentiated access permissions according to user roles, employing a hybrid model combining RBAC (Role-Based Access Control) and ABAC (Attribute-Based Access Control) to achieve dynamic allocation and real-time verification of permissions, ensuring that only authorized personnel can view or edit specific data, preventing information leakage or tampering. Therefore, the user management module includes: a user permission submodule, a permission calculation submodule, a data access filtering submodule, and a data encryption submodule.

[0067] Furthermore, the user permissions submodule includes the following components: Role library: Predefined system roles (such as system administrator, security director, project manager, security officer, and general employee), each role is associated with a set of basic permission templates.

[0068] Attribute set: Define user attributes (such as department, position, job level), data attributes (such as project ID, data sensitivity level, creation department), and environment attributes (such as access time, IP address, device type).

[0069] Based on the relationships of affiliation, responsibility, and participation in the security responsibility knowledge graph, it is used to describe the dynamic business relationships between users and projects, tasks, and devices.

[0070] Access control rules: Use the policy language to define access control rules.

[0071] Example rule (project manager): ALLOWUPDATE taskkWHERE user.role='Project Manager' AND task.project_idIN user.managed_projects ALLOWREAD__ALL__WHEREuser.department==data.owner_departmentORdata.sensitivity_level=='PUBLIC' Permission Calculation Submodule Access Request Interception: When a user performs any data operation (create, delete, modify, query), the engine intercepts the request and extracts the (user, operation, data object) triple.

[0072] Context information collection: Real-time acquisition of the current user's attributes, department, list of managed projects, and operating environment (time, IP).

[0073] Real-time permission determination: The engine performs the following logical calculations: Role permission check: Verify whether the user role has the basic permissions for this operation.

[0074] Attribute rule matching: User attributes, data attributes, and environment attributes are substituted into the policy rules for matching.

[0075] Relationship graph traversal: For requests requiring relationship determination, the knowledge graph is traversed in real time. For example, to determine whether "user A has permission for task B", the engine will query the graph to see if the path exists: User A - [responsible for / participating in] -> Project X <- [belongs to] - Task B.

[0076] Decision result return: The engine returns Allow, Deny, or Not Applicable based on the result calculated by the rules.

[0077] Specifically, the data access filtering submodule includes row-level filtering, field-level filtering, etc.

[0078] Row-level filtering: Dynamically appending WHERE conditions at the database query level. For example, when a safety officer queries the task list, the system automatically rewrites the SQL as: SELECT * FROM tasks WHERE project_id IN (SELECT project_id FROM user_managed_projects WHERE user_id=CURRENT_USER_ID), ensuring that they can only see tasks for the projects they are responsible for.

[0079] Field-level filtering: Before data is returned to the front end, fields are filtered on JSON objects or data rows. For example, when a regular employee views task details, the system automatically filters out sensitive fields such as cost_budget (cost budget).

[0080] The data encryption submodule includes an encryption system covering the entire data lifecycle (transmission, storage, and use). It employs national cryptographic standard algorithms to ensure that external attackers cannot obtain sensitive information and that even if sensitive data is stolen, it cannot be cracked. This module uses a layered encryption and centralized key management strategy.

[0081] This module uses TLS 1.3 protocol for encryption throughout the data transmission channel and uses a server certificate issued by a trusted CA to force the client to verify the certificate's validity and prevent man-in-the-middle attacks.

[0082] This module uses the CBC mode of the national cryptographic SM4 algorithm to encrypt extremely sensitive business fields (such as employee ID numbers and confidentiality assessment conclusions).

[0083] This invention utilizes the built-in TDE (Trusted Execution Environment) function of the database (such as MySQL's keyring_file or Oracle's TDE), with the encryption key managed by a hardware security module or cloud KMS. The entire database file or tablespace is encrypted to prevent data leakage due to the loss of physical storage media.

[0084] Log monitoring and auditing submodule: This module records the timestamp, user ID, IP address, operation type (create / delete / modify / query), operation object, operation content (previous mirror / backward mirror), and operation result for each operation.

[0085] At the application layer, Aspect-Oriented Programming (AOP) is used to uniformly embed entry points at the entry points of business methods; at the database layer, all SQL operations are captured by parsing the database's binlog or archive logs.

[0086] The system uses access logs to detect abnormal logins, high-frequency operations, unauthorized access attempts, and sensitive data operations. Real-time alerts are sent to the security administrator for any of these detected events. Abnormal login detection includes the same account logging in from different cities within a short period. High-frequency operation detection includes the same user performing large-scale data queries or exports outside of working hours. Unauthorized access attempts include users repeatedly attempting to access data outside their authorized scope. Sensitive data operations include batch queries or modifications of data defined as sensitive (such as employee salaries or core formulas).

[0087] By combining the above technical principles, the responsibility list system can effectively improve the management efficiency of various projects in an enterprise, clarify responsibilities, and enhance team collaboration capabilities, providing solid technical support for daily operations.

[0088] Example 2: like Figure 2 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0089] The processor may be a central processing unit (CPU), or it may 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. It is the computing core and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to execute one or more modules in the security responsibility list management system based on a large model empowered by this application.

[0090] Example 3 Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor loads and executes one or more instructions stored in the storage medium to execute one or more modules in the large-model-enabled security responsibility list management system provided in this application.

[0091] Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0092] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0094] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0096] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A safety responsibility list management system based on a large model, characterized in that, include: The Responsibility List module is used to assign tasks in a project to appropriate responsible persons by leveraging a large language model and a knowledge service foundation, combined with project requirements and personnel capabilities. The knowledge service platform is used to manage the security responsibility knowledge graph; The task management module is used to track and manage task progress and generate early warnings based on real-time data on the responsible persons' performance of their duties, using a safety responsibility knowledge graph. The data analysis and reporting module is used to perform multi-dimensional analysis and generate visual reports based on the task allocation information from the responsibility list module and the responsibility person's performance process data from the task management module. The notification collaboration module is used to generate notifications based on information from each module, and to push them accurately to relevant responsible persons based on intelligent routing strategies to achieve task collaboration. The user management module is used to manage user permissions based on the dynamic business relationships between various modules, using a security responsibility knowledge graph.

2. The system according to claim 1, characterized in that, The responsibility list module includes: a responsibility library submodule for building and storing a security responsibility knowledge graph; The construction of the security responsibility knowledge graph includes: pre-training a general model using a large language model based on large-scale security responsibility domain text, and then using multi-task learning to simultaneously identify entities and relationships to perform entity recognition and relationship extraction.

3. The system according to claim 1, characterized in that, The responsibility list module further includes: a responsibility matching submodule, which includes: The responsible person capability profile building unit is used to integrate employees' discrete data, unstructured text data, and numerical data to generate a unified high-dimensional capability vector; The task requirement vectorization unit is used to convert task description text or task description text generated from the security responsibility knowledge graph into task vectors. The intelligent matching algorithm unit is used to generate a list of recommended responsible persons for tasks by simultaneously optimizing capability matching and load balancing using a multi-objective optimization algorithm.

4. The system according to claim 1, characterized in that, The responsibility list module also includes a responsibility person display submodule, which is configured as follows: Monitor changes in relationships related to liability assumption within the aforementioned safety responsibility knowledge graph; In response to the aforementioned change event, the graph data is dynamically queried, and the display interface for responsibility allocation and task progress is updated in real time through visualization components from both the manager's and employee's perspectives.

5. The system according to claim 1, characterized in that, The responsibility list module also includes a responsibility update submodule, which includes: The change identification unit is used to automatically identify responsibility change points from project documents and system instructions using text difference technology and information extraction technology; The impact analysis unit is used to perform path queries and logical rule reasoning based on the security responsibility knowledge graph to analyze the impact range of the change point. The update unit is used to generate a knowledge graph update scheme based on the impact analysis results, and update the security responsibility knowledge graph through an approvable workflow.

6. The system according to claim 1, characterized in that, The task management module includes a progress tracking submodule, which includes: The progress acquisition unit is used to collect data on the performance of duties from multiple channels, including text reports, spatiotemporal data, and visual data, and to perform analysis and feature extraction using NLP models, spatiotemporal verification rules, and computer vision models, respectively. The progress calculation unit is used to comprehensively judge the multi-source analysis results using an evidence fusion algorithm, determine the actual progress status of the task, and automatically calculate the progress percentage; the automatic early warning submodule is configured as follows: When the progress calculation unit determines that the task progress is lower than the preset threshold or exceeds the time limit, it automatically triggers an early warning message and pushes it to the relevant responsible persons and managers through the notification collaboration module.

7. The system according to claim 1, characterized in that, The data analysis and reporting module includes a performance evaluation submodule, which is configured as follows: Based on the relevant data from each module, and using the assessment indicator system, multi-dimensional performance indicator values ​​are calculated to obtain a performance evaluation of the project and / or the responsible person; the multi-dimensional aspects include: project progress and / or the responsible person's task completion rate, completion quality, and behavior; Performance evaluation results can be presented using one or more visualization methods, such as radar charts, comparative bar charts, and trend line charts.

8. The system according to claim 1, characterized in that, The notification collaboration module includes: The message generation unit is used to select the corresponding template from the preset template library according to the system event type, and fill the event parameters through rules or lightweight large models to generate personalized notification messages. The message distribution unit is used to distribute messages based on their urgency and recipient preferences, using one or more channels such as in-system messaging, mobile app push notifications, SMS, and email.

9. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a security responsibility list management system based on a large model-enabled system as described in any one of claims 1 to 8 is executed.

10. A readable storage medium, characterized in that, It contains an executable program, which, when executed, implements a security responsibility list management system based on a large model as described in any one of claims 1 to 8.

Images