Government affair data processing system and method

Abstract

The invention provides a government affair data processing system and method. The government affair data processing system comprises a data access and preprocessing module, a multi-mode government affair large model and intelligent analysis module, a knowledge base module, a privacy calculation engine module, a data storage and management module, an application service and interface module and a system management and monitoring module. A cooperative work system integrating a multi-mode government affair large model, a data classification and grading sub-module, a privacy calculation engine module and a knowledge base is constructed, configuration, monitoring and scheduling are carried out through a unified interface and a management platform, and a closed-loop and sustainable optimization government affair data processing and application ecology is formed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of large models, in particular to a government data processing system and method. BACKGROUND

[0002] At present, the existing government data processing system has defects such as insufficient dynamic adaptability, imbalance between privacy computing efficiency and security, and lack of multi-modal collaboration capability. SUMMARY

[0003] Therefore, the present application aims to provide a government data processing system and method to configure, monitor and schedule through a unified interface and management platform, forming a closed loop and sustainable optimized government data processing and application ecology.

[0004] In a first aspect, the embodiments of the present application provide a government data processing system, which comprises a data access and preprocessing module, a multi-modal government large model and intelligent analysis module, a knowledge base module, a privacy computing engine module, a data storage and management module, an application service and interface module, and a system management and monitoring module. The data access and preprocessing module is used to access raw data from various government data sources and preprocess the raw data to obtain multi-modal data. The preprocessing operation includes cleaning, format conversion and metadata extraction. The multi-modal government large model and intelligent analysis module is used to intelligently analyze the multi-modal data based on a multi-modal government large model optimized based on government domain knowledge. The rules and models in the knowledge base module are called to classify and grade the multi-modal data. The intelligent analysis includes deep semantic understanding, feature extraction and entity recognition. The knowledge base module includes a knowledge base, which is a structured database storing professional knowledge in the government domain, classification and grading standards and historical cases. The knowledge base module provides a basis for classification and grading and privacy computing strategy formulation. The privacy computing engine module is used to encrypt, desensitize or process the classified and graded data using a privacy computing protocol according to the classification and grading results and the preset security strategy. The data storage and management module is used to store the data after classification, grading and security processing, and to perform data retrieval, access control, version management and life cycle management on the stored data. The application service and interface module is used to provide a unified data service interface and tool for upper government applications. The system management and monitoring module is used for user management, permission configuration, task scheduling, running state monitoring, log auditing and security alarm.

[0005] In an optional embodiment of the present application, the multi-modal government affair large model is trained by the following method: taking multi-source data in the government affair field as pre-training corpus; wherein the pre-training corpus includes: text data, image data, voice data and table data; performing independent pre-training and cross-modal semantic alignment based on the pre-training corpus to form a unified semantic space of the multi-modal government affair large model; performing task fine-tuning on the multi-modal government affair large model based on a government affair knowledge base and rules.

[0006] In an optional embodiment of the present application, the multi-modal government affair large model and the intelligent analysis module include: a data classification and grading sub-module; the data classification and grading sub-module is configured to receive pre-processed data, call the multi-modal government affair large model to perform content analysis on the data; wherein the content analysis includes: topic identification, entity extraction, relationship extraction, sentiment analysis of text data, and sensitive information detection in unstructured data; the data classification and grading sub-module is further configured to extract key features based on the results of the content analysis; query the classification and grading manner of the knowledge base module, classify the data set based on the key features and the classification and grading manner, and attach the classification and grading results to the data as labels.

[0007] In an optional embodiment of the present application, the privacy computing engine module integrates multiple privacy computing technologies, including: federated learning, multi-party secure computation, homomorphic encryption and trusted execution environment; the privacy computing engine module is configured to provide a unified interface for the multiple privacy computing technologies.

[0008] In an optional embodiment of the present application, the privacy computing engine module is configured to dynamically perform protocol matching on the multiple privacy computing technologies according to the classification and grading results of the data.

[0009] In an optional embodiment of the present application, the privacy computing engine module is further configured to record the execution, resource consumption and security events of the privacy computing task.

[0010] In an optional embodiment of the present application, the knowledge base is constructed by the following method: inputting classification and grading rules, laws and regulations, files and typical business scenario knowledge; giving natural language processing technology to perform named entity recognition, relationship extraction and text classification on the input content; extracting key knowledge elements from the input unstructured documents to construct a knowledge graph or a relational database as the knowledge base.

[0011] In an optional embodiment of the present application, the knowledge base is updated in the following ways: the content of the knowledge base is reviewed, revised and supplemented by manual review and maintenance; the decision results of data classification and grading and intelligent analysis of the government affair model, user feedback and newly generated compliance cases are used as new knowledge sources, and the content of the new knowledge sources is integrated into the knowledge base after screening and verification; the latest rules and knowledge are obtained from the network by the network crawler technology to update the knowledge base regularly.

[0012] In an optional embodiment of the present application, the government affair data processing system adopts a distributed micro-service architecture for deployment, and each module of the government affair data processing system is deployed as an independent micro-service on a containerized platform and is uniformly managed and scheduled through a gateway.

[0013] In a second aspect, the embodiments of the present application also provide a government affair data processing method applied to the above-mentioned government affair data processing system; the method comprises the following steps: a data access and preprocessing module accesses original data from each government affair data source and performs preprocessing operations on the original data to obtain multi-modal data; wherein the preprocessing operations include cleaning, format conversion and metadata extraction; a multi-modal government affair large model and intelligent analysis module performs intelligent analysis on the multi-modal data based on a multi-modal government affair large model optimized based on government affair field knowledge; rules and models in a knowledge base module are called to classify and grade the multi-modal data; wherein the intelligent analysis includes deep semantic understanding, feature extraction and entity recognition; the knowledge base module provides a basis for classification and grading and privacy calculation strategy making; the knowledge base module comprises a knowledge base, which is a structured database storing professional knowledge in the government affair field, classification and grading standards and historical cases; a privacy calculation engine module encrypts, desensitizes or processes the classified and graded data by using a privacy calculation protocol according to the classification and grading results and a preset security strategy; a data storage and management module stores the data processed by classification and security, and performs data retrieval, access control, version management and life cycle management on the stored data; an application service and interface module provides a unified data service interface and tool for an upper government affair application; and a system management and monitoring module performs user management, permission configuration, task scheduling, running state monitoring, log auditing and security alarm.

[0014] The embodiments of the present application have the following beneficial effects: The embodiments of the present application provide a government affair data processing system and method, and construct a collaborative working system integrating a multi-modal government affair large model, a data classification and grading sub-module, a privacy calculation engine module and a knowledge base, which are configured, monitored and scheduled through a unified interface and management platform to form a closed loop and sustainable optimized government affair data processing and application ecology.

[0015] Other features and advantages of the present disclosure will be set forth in the descriptions below, and in part will be apparent from the description, or can be learned by practice of the present disclosure as hereinafter more fully described, together with the advantages thereof. It will be appreciated that the features and advantages of the present disclosure can be realized and attained by means summarized fully in the written description and appended claims also hereinafter more fully described.

[0016] In order to make the above-mentioned purposes, features and advantages of the present disclosure more obvious and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or the prior art description. Obviously, the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0018] Figure 1 A structural schematic diagram of a government affair data processing system provided by an embodiment of the present application is provided. Figure 2 A general flow schematic diagram of a government affair data processing system provided by an embodiment of the present application is provided. Figure 3 A training flow schematic diagram of a multi-modal government affair large model provided by an embodiment of the present application is provided. Figure 4 A flow chart of a government affair data processing method provided by an embodiment of the present application is provided. DETAILED DESCRIPTION

[0019] In order to make the purposes, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the present application will be described clearly and completely in combination with the drawings below. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0020] The existing classification and grading technology includes: 1. Static rule dependence: preset classification and grading rule library, determine the data level through keyword matching (such as “ID number” “address”); 2. Large model assisted analysis: define the rules of data classification and grading, learn the rules of data classification and grading through large model, analyze and understand the input data through large model, judge the classification and grading rule characteristics that the input data conforms to, and combine the corresponding relationship between data category, data sensitivity level and classification and grading rule characteristics to obtain the classification and grading result of data; 3. Vector knowledge base matching: vectorize the rule base, match data and grading rules through similarity search (such as nearest neighbor algorithm).

[0021] Existing privacy computing and data sharing technologies include: 1. Fixed protocol application: full data uses a single privacy computing protocol (such as federated learning or homomorphic encryption), and is not differentiated according to data levels; 2. Blockchain permission management: data access is controlled through a smart contract, but the permission strategy is decoupled from the data sensitivity.

[0022] Existing knowledge base technologies include: 1. Single modality dominance: text data is analyzed using NLP (Natural Language Processing), image / table data is processed independently, and there is a lack of cross-modal association (such as separating certificate OCR (Optical Character Recognition) from anti-fake verification); 2. Static knowledge base update: rule base relies on manual input, has response delays, and cannot be updated dynamically.

[0023] Currently, the existing government data processing system has the following defects: 1. Lack of dynamic adaptability: classification rules are fixed and cannot respond to rule changes in real time (such as manually adjusting sensitive levels during a sudden public health event); semantic understanding is limited: keyword matching cannot identify synonyms (such as "medical record" vs. "health record"), with a high error rate.

[0024] 2. Imbalance between privacy computing efficiency and security: protocol selection is rigid: different security protection schemes cannot be matched according to different data levels, and there is no linkage with data classification results, which may result in over-encryption of low-risk data and waste of resources.

[0025] 3. Lack of multi-modal collaboration capability: heterogeneous data is processed separately, such as text analysis ignoring the coordinate precision in images (such as planning drawings), which increases the risk of missing confidential data.

[0026] Based on this, the government data processing system and method provided by the embodiments of the present application specifically provide an intelligent dynamic data classification and privacy protection processing system and method based on a multi-modal government big model.

[0027] To facilitate understanding of the present embodiment, first, a government data processing system disclosed by the present embodiment is described in detail.

[0028] Embodiment one: The present embodiment provides a government data processing system, referring to Figure 1 The structure schematic diagram of a government data processing system, the government data processing system comprises: a data access and preprocessing module, a multi-modal government big model and intelligent analysis module, a knowledge base module, a privacy computing engine module, a data storage and management module, an application service and interface module, and a system management and monitoring module.

[0029] See also Figure 2 Fig. 1 shows a general flowchart of a government data processing system, Figure 1 and Figure 2 Fig. 2 shows a government data processing system, which mainly includes the following core modules and works collaboratively through specific interaction processes: (1) Data access and preprocessing module, used for accessing raw data from various government data sources and performing preprocessing operations on the raw data to obtain multi-modal data; wherein the preprocessing operations include cleaning, format conversion and metadata extraction.

[0030] The data access and preprocessing module in this embodiment can be responsible for accessing raw data from various government data sources (such as databases, file systems, API interfaces (Application Programming Interface, API) and the like), and performing preliminary cleaning, format conversion, metadata extraction and other preprocessing operations, to prepare for subsequent classification, grading and analysis.

[0031] (2) Multi-modal government large model and intelligent analysis module, used for intelligent analysis of multi-modal data based on a multi-modal government large model optimized by government domain knowledge; calling rules and models in the knowledge base module to classify and grade multi-modal data; wherein intelligent analysis includes deep semantic understanding, feature extraction and entity recognition.

[0032] The multi-modal government large model and intelligent analysis module system in this embodiment is the core intelligent engine. The module has built-in pre-training large models optimized by government domain knowledge, which can perform intelligent analysis such as deep semantic understanding, feature extraction and entity recognition on preprocessed multi-modal data. At the same time, the module integrates data classification and grading logic, which can call rules and models in the knowledge base to automatically classify and grade data.

[0033] (3) Knowledge base module includes a knowledge base, which is a structured database storing professional knowledge in the government domain, classification and grading standards and historical cases; the knowledge base module is used to provide basis for classification and grading and privacy computing strategy formulation.

[0034] The knowledge base module in this embodiment includes a knowledge base, which is a structured database storing professional knowledge in the government domain, GB / T 43697-2024 national standard classification and grading standards, historical cases and the like. It provides knowledge support for the government large model and provides basis for classification and grading, privacy computing strategy formulation and the like.

[0035] (4) Privacy computing engine module, used for encrypting, desensitizing or processing classified and graded data using privacy computing protocols according to the results of classification and grading and pre-set security strategies.

[0036] The privacy computing engine module in this embodiment integrates various privacy computing technologies such as federated learning, multi-party secure computation, and trusted execution environment. According to the results of data classification and the preset security policy, this module is responsible for encrypting, desensitizing, or processing data using specific privacy computing protocols to ensure data security during sharing and computing.

[0037] (5) Data storage and management module, used for storing data after classification, grading and security processing, and performing data retrieval, access control, version management and life cycle management on stored data.

[0038] The data storage and management module in this embodiment can be responsible for storing data after classification, grading and security processing, and provide efficient data retrieval, access control, version management, life cycle management and other functions. This module ensures the integrity, availability and security of data.

[0039] (6) Application service and interface module, used for providing unified data service interface and tools for upper government affairs applications.

[0040] The application service and interface module in this embodiment can provide unified data service interface and tools for upper government affairs applications (such as decision support systems, public service platforms, risk early warning systems, etc.), supporting secure data query, analysis and visualization.

[0041] (7) System management and monitoring module, used for user management, permission configuration, task scheduling, running state monitoring, log auditing and security alarm.

[0042] The system management and monitoring module in this embodiment can provide user management, permission configuration, task scheduling, running state monitoring, log auditing, security alarm and other functions to ensure stable, secure and compliant operation of the entire system.

[0043] As shown in Figure 2 The typical interaction process of the government data processing system in this embodiment can include the following steps 1-5: Step 1, data access and preprocessing: raw government data enters the system through the data access module and is transmitted to the multi-modal government big model and intelligent analysis module after preprocessing.

[0044] Step 2, intelligent classification and grading: the multi-modal government big model combines rules and knowledge in the knowledge base to perform deep analysis and intelligent classification and grading of data, and apply corresponding labels.

[0045] Step 3: Privacy calculation strategy matching and execution: According to the classification and grading labels of the data and the security policy, the privacy calculation engine module selects the appropriate privacy calculation technology to safely process the data.

[0046] Step 4: Secure data storage and sharing: The processed secure data is stored in the data storage and management module, and authorized users or applications can securely access and use these data through the application service and interface module.

[0047] Step 5: Knowledge base update and model optimization: New classification cases, risk assessment results, and user feedback generated during system operation can be used to update the knowledge base and optimize the government affairs big model, forming a closed loop of continuous learning and improvement.

[0048] The embodiments of the present application provide a government data processing system, which builds a collaborative working system integrating a multi-modal government big model, a data classification and grading sub-module, a privacy calculation engine module, and a knowledge base. The system first uses the multi-modal government big model to perform deep analysis and feature extraction on the input raw government data, combines the built-in GB / T 43697-2024 "Data Security Technology Data Classification and Grading Rules" and other standards, and dynamically updates the government knowledge base to intelligently classify and grade the data and apply corresponding classification and grading labels. Subsequently, according to the classification and grading results of the data and the security policy, the system will use different privacy calculation technologies (such as federated learning, multi-party secure computation, trusted execution environment, etc.) to protect the data, ensuring the security of the data in subsequent sharing, transmission, storage, and calculation processes. The securely processed data can be securely accessed and used by authorized users or application systems to support various government application scenarios, such as intelligent decision-making, precise service, risk warning, etc. The entire system is configured, monitored, and scheduled through a unified interface and management platform, forming a closed loop of sustainable optimization of government data processing and application ecology.

[0049] Embodiment Two: The present embodiment provides another government data processing system, which is implemented based on the above-mentioned embodiments, and focuses on the specific implementation of the government data processing system.

[0050] I. Training of multi-modal government big model: In some embodiments, the above-mentioned multi-modal government big model is trained by the following method: using multi-source data in the government field as pre-training corpus; wherein the pre-training corpus includes: text data, image data, voice data and table data; based on the pre-training corpus, independent pre-training and cross-modal semantic alignment of each modality are performed to form a unified semantic space of the multi-modal government big model; based on the government knowledge base and rules, the multi-modal government big model is fine-tuned for tasks.

[0051] Referring to Figure 3 The training process of a multi-modal government big model is shown in FIG. 1. The training of the multi-modal government big model in this embodiment is the core step for the system to achieve intelligent data processing. First, a large amount of multi-source data in the government field is collected and prepared as pre-training corpus. These data can include files, laws and regulations, work reports, government information disclosure, historical records, professional literature, etc. The multi-modal government big model fuses heterogeneous data such as text, image, voice, table, and trajectory, independently pre-trains in each modality, aligns across modalities, and forms a unified semantic space. Then, task fine-tuning is performed relying on the government knowledge base and GB / T 43697-2024 rules, and human feedback reinforcement learning is introduced for continuous optimization.

[0052] II. Specific implementation of data classification and grading sub-module: In some embodiments, the multi-modal government big model and the intelligent analysis module include a data classification and grading sub-module. The data classification and grading sub-module is configured to receive pre-processed data, call the multi-modal government big model to analyze the content of the data, and extract key features based on the results of the content analysis. The data classification and grading sub-module is also configured to query the classification and grading method of the knowledge base module, classify the data set based on the key features and the classification and grading method, and attach the classification and grading results as labels to the data.

[0053] The specific implementation of the data classification and grading sub-module in this embodiment relies on the intelligent analysis capability of the multi-modal government big model and the rule support of the knowledge base. The module first receives pre-processed data, then calls the multi-modal government big model for deep content analysis, including topic identification, entity extraction, relationship extraction, sentiment analysis of text data, and sensitive information detection in unstructured data (such as images and videos). Based on the analysis results, the module extracts key features, including not only the attributes of the data itself, but also the sensitivity and potential risks reflected by the content. Next, the module queries the knowledge base to obtain the specific classification and grading standards, judgment conditions, and relevant legal requirements defined in GB / T 43697-2024 "Data Security Technology Data Classification and Grading Rules" and other standards. The government big model will combine the extracted features and the rules in the knowledge base, use its reasoning ability, and automatically determine the category and level of the data. The classification and grading results will be attached to the data in the form of labels, and the decision basis will be recorded for subsequent auditing and tracing. The module should also provide an interface for manual review and correction to ensure the accuracy of the classification results.

[0054] III. Integration and configuration of privacy computing engine module: In some embodiments, the privacy computing engine module integrates multiple privacy computing technologies, including federated learning, multi-party secure computation, homomorphic encryption, and trusted execution environment; and provides a unified interface for the multiple privacy computing technologies.

[0055] The integration and configuration of the privacy computing engine module in this embodiment is the key to achieving secure data sharing and computing. This module needs to integrate multiple mainstream privacy computing technologies such as federated learning (FL), multi-party secure computation (MPC), homomorphic encryption (HE), and trusted execution environment (TEE), and provide a unified API interface for system calls.

[0056] In some embodiments, the privacy computing engine module dynamically matches protocols for multiple privacy computing technologies according to the classification and grading results of the data.

[0057] The configuration of the privacy computing engine module in this embodiment needs to dynamically match protocols according to the classification and grading results of the data. In the specific execution process, the privacy computing engine module is responsible for coordinating the computing nodes of each participant, ensuring the correctness and security of the computing process.

[0058] In some embodiments, the privacy computing engine module also records the execution of privacy computing tasks, resource consumption, and security events.

[0059] In addition, the privacy computing engine module in this embodiment also needs to provide performance monitoring and auditing functions to record the execution of privacy computing tasks, resource consumption, and security events, facilitating subsequent optimization and tracing.

[0060] The classification and grading labels output by the multi-modal large model in this embodiment are directly mapped to the three-level capabilities of the privacy computing protocol according to the "core level, important level, and general level" in GB / T 43697-2024, forming a "one-key" automatic matching rule that can be directly embedded into the policy engine of the aforementioned system, as shown in Table 1.

[0061] Table 1

[0062] IV. Knowledge base construction and update mechanism: In some embodiments, the knowledge base is constructed by the following methods: entering classification and grading rules, laws and regulations, documents, and typical business scenario knowledge; giving natural language processing technology to perform named entity recognition, relationship extraction, and text classification on the entered content; extracting key knowledge elements from the entered unstructured documents, and constructing a knowledge graph or a relational database as a knowledge base.

[0063] The construction and updating mechanism of the knowledge base in this embodiment is the basis for ensuring the continuous intelligence and specialization of the system. In the initial stage of the construction of the knowledge base, domain experts are needed to systematically sort out and structureally input government data standards, GB / T 43697-2024 classification and grading rules, relevant laws and regulations, documents, and typical business scenario knowledge. Natural language processing techniques such as named entity recognition, relation extraction, and text classification can be used to automatically extract key knowledge elements from a large amount of unstructured documents to construct a preliminary knowledge graph or relational database.

[0064] In some embodiments, the above-mentioned knowledge base is updated by the following methods: the content of the knowledge base is reviewed, corrected and supplemented by artificial auditing and maintenance; the decision results of data classification and intelligent analysis of the government big model, user feedback, and newly generated compliance cases are used as new knowledge sources, and the content of the new knowledge sources is integrated into the knowledge base after screening and verification; the latest knowledge is obtained from the network by network crawler technology to update the knowledge base.

[0065] The updating mechanism of the knowledge base in this embodiment should be designed as a dynamic and multi-source driven process. First of all, artificial auditing and maintenance are essential, and expert teams need to regularly review, correct and supplement the content of the knowledge base. Secondly, the machine learning feedback loop, i.e. the decision results of data classification and intelligent analysis of the government big model, user feedback, and newly generated compliance cases, can be used as new knowledge sources, which are integrated into the knowledge base after screening and verification. Thirdly, automatic knowledge acquisition can be achieved by using network crawler technology to regularly obtain the latest knowledge from websites, legal databases, industry research reports, and other channels, and automatically or semi-automatically update the knowledge base. Finally, version control and change management mechanisms are also necessary to ensure the stability and traceability of the knowledge base.

[0066] Five, system deployment: In some embodiments, the above-mentioned government data processing system adopts a distributed micro-service architecture for deployment, and each module of the government data processing system is deployed as an independent micro-service on a containerized platform and is managed and scheduled through a gateway.

[0067] The government data processing system in this embodiment can be deployed using a distributed micro-service architecture to improve the scalability, reliability and maintainability of the system. Each core module (such as the data access module, the government big model service, the knowledge base service, the privacy computing engine, etc.) can be deployed as an independent micro-service on a containerized platform and managed and scheduled through an API gateway.

[0068] In summary, the government data processing system provided by the embodiments of the present application mainly provides the following contents: I. Joint Feature Extraction of Multi-modal Government Affairs Big Model: Break through the limitations of traditional single-modal processing by integrating heterogeneous data such as text, images, speech, and tables in a unified semantic space. Use cross-modal alignment technology to achieve joint analysis of multi-modal features (e.g., simultaneously analyzing image seals, text content, and speech keywords in documents). Dynamically enhance sensitive feature weights. Avoid single-modal missed detection, improve classification and grading accuracy (e.g., solve the problem of missed identification of seals in scanned documents), and reduce reliance on manual review.

[0069] II. Automatic Mapping Mechanism of Dynamic Grading Results and Privacy Computing Protocol: Establish dynamic matching rules between data grading labels (core, important, general) and privacy computing protocols. Based on GB / T43697-2024 standard, build three-level mapping rules (e.g., core level → multi-party secure computation; important level → federated learning; general level → desensitization). Achieve "one-key" protocol matching through a strategy engine without manual intervention. Solve the problem of resource waste (e.g., over-encryption of low-risk data) or security deficiency (e.g., insufficient protection of high-risk data) caused by static protocols, and achieve precise balance between security and efficiency.

[0070] III. Self-evolution and Dynamic Update Mechanism of Government Affairs Knowledge Base: Realize real-time dynamic update of classification and grading rules and knowledge base. Multi-source driven update: integrate artificial review, user feedback, compliance cases, web crawler (authoritative sources), etc. Combine NLP technology (entity recognition, relationship extraction) to automatically build knowledge graph and support model fine-tuning. Ensure the system responds to rule changes in real time (e.g., adjust sensitive level during public health emergencies), and eliminate compliance risks caused by rule lag.

[0071] IV. Grading-driven Privacy Computing Protocol Dynamic Selection Module: Dynamically adapt differential privacy protection technology according to data level. Protocol library integrates multiple technologies (federated learning / multi-party secure computation / homomorphic encryption / trusted execution environment). Based on grading labels, automatically trigger protocols (e.g., core data must enable multi-party secure computation). Avoid "one-size-fits-all" encryption causing wasted computing power (e.g., general data only needs desensitization), and improve security protection for high-sensitive data.

[0072] V. Loosely coupled microservice architecture and zero-downtime update capability: support independent update of system modules without affecting business continuity. Core modules (classification and grading, privacy computing, knowledge base) are containerized and deployed through API gateway. Dynamically load updated models / rules library to achieve seamless connection from "regulation / case update - rule taking effect". Solve the problem of service interruption caused by downtime update in government affairs systems, and improve system robustness.

[0073] The government affair data processing system provided by the embodiment of the application has the following advantages compared with the prior art. I. Multi-modal government affair large model and traditional single-modal / general model Many existing systems can only perform keyword matching on text, and when facing a ''file scan piece'', the failure to identify the words ''secret'' and ''confidential'' in the seal leads to level down, resulting in compliance risk.

[0074] The embodiment of the application simultaneously processes image seal, text content and voice keywords in a unified semantic space, and the cross-attention weight can automatically push the joint features of ''seal + sensitive words'' to the core level. Once reasoning, the multi-modal context can be comprehensively analyzed, the ''single-point loss'' can be avoided, the grading accuracy can be improved, and the artificial review workload can be reduced.

[0075] II. Dynamic grading protocol mapping and static rule library Once the static rule library of many existing systems is set, it cannot be adjusted with the changes of business scenarios.

[0076] The embodiment of the application can dynamically update and integrate the knowledge base by using user feedback and newly generated compliance cases as new knowledge sources. The ''regulation update-rule update-protocol update'' zero downtime closed loop can be realized to ensure business continuity.

[0077] III. Three-level privacy computing protocol and single encryption or plaintext sharing The traditional method uses AES encryption for all data, and although the core-level data is safe, the calculation cost is high; the general-level data also receives the same treatment, resulting in resource waste.

[0078] The embodiment of the application can automatically select different security protocols according to ''core-important-general'' three gears. The ''graded protection and precise investment'' can be realized.

[0079] In addition, the government affair data processing system provided by the embodiment of the application can also have an integrated data flow facility as an alternative solution, which can integrate data classification and grading, permission control and privacy computing, and has the following capabilities: environmental isolation (physical isolation of data between tenants); confidential storage (transmission encryption + secret storage); confidential computing (Trusted Execution Environment, TEE) (protecting runtime data); secure development (original data invisible to processing parties); permission control (multi-level active immunization protection). It can provide compliance data sources for AI (Artificial Intelligence) training and has been applied in the medical and energy fields.

[0080] Embodiment three Corresponding to the above-mentioned embodiment, the embodiment of the present application provides a government affair data processing method, which is applied to the government affair data processing system provided in the foregoing embodiment, and a flowchart of the government affair data processing method is shown in FIG. 4. Figure 4 The government affair data processing method includes the following steps: In step S402, the data access and preprocessing module accesses raw data from each government affair data source, and performs preprocessing operation on the raw data to obtain multi-modal data. The preprocessing operation includes cleaning, format conversion and metadata extraction. In step S404, the multi-modal government affair large model and intelligent analysis module perform intelligent analysis on the multi-modal data based on the multi-modal government affair large model optimized based on government affair field knowledge; and call rules and models in the knowledge base module to classify and grade the multi-modal data. The intelligent analysis includes deep semantic understanding, feature extraction and entity recognition. In step S406, the knowledge base module provides basis for classification and grading and privacy computing strategy formulation. The knowledge base module includes a knowledge base, which is a structured database storing professional knowledge in the government affair field, classification and grading standards and historical cases. In step S408, the privacy computing engine module encrypts, desensitizes or processes the data classified and graded according to the results of classification and grading and the preset security strategy by using a privacy computing protocol. In step S410, the data storage and management module stores the data classified and graded and processed in security, and performs data retrieval, access control, version management and life cycle management on the stored data. In step S412, the application service and interface module provides a unified data service interface and tool for upper government affair applications. In step S414, the system management and monitoring module performs user management, permission configuration, task scheduling, running state monitoring, log auditing and security alarm.

[0081] The embodiment of the present application provides a government affair data processing method, which constructs a collaborative working system integrating a multi-modal government affair large model, a data classification and grading submodule, a privacy computing engine module and a knowledge base, and configures, monitors and schedules through a unified interface and management platform to form a closed loop and sustainable optimized government affair data processing and application ecology.

[0082] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the government affair data processing method described above can refer to the corresponding process in the foregoing embodiment, which will not be described herein.

[0083] In addition, in the description of the embodiments of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connecting" should be understood in a broad sense, for example, can be fixedly connected, can be detachably connected, or integrally connected; can be mechanically connected, can be electrically connected; can be directly connected, can be indirectly connected through an intermediate medium, or can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0084] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the part of the present application that essentially contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0085] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0086] Finally, it should be noted that: the above embodiments are only specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, and are not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can make modifications or easily think of changes to the technical solutions recorded in the foregoing embodiments within the technical scope disclosed by the present application, or make equivalent replacements to some of the technical features; and these modifications, changes or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A government data processing system, characterized by, The government affair data processing system comprises a data access and preprocessing module, a multi-modal government affair large model and intelligent analysis module, a knowledge base module, a privacy computing engine module, a data storage and management module, an application service and interface module, and a system management and monitoring module. The data access and preprocessing module is configured to access raw data from various government affair data sources, and perform preprocessing operations on the raw data to obtain multi-modal data; wherein the preprocessing operations comprise cleaning, format conversion and metadata extraction. The multi-modal government affair large model and intelligent analysis module is configured to perform intelligent analysis on the multi-modal data based on a multi-modal government affair large model optimized based on government affair domain knowledge, and to call rules and models in the knowledge base module to classify and grade the multi-modal data; wherein the intelligent analysis comprises deep semantic understanding, feature extraction and entity recognition. The knowledge base module comprises a knowledge base, which is a structured database storing professional knowledge in the government affair domain, classification and grading standards, and historical cases; the knowledge base module is configured to provide a basis for classification and grading and privacy computing strategy formulation. The privacy computing engine module is configured to encrypt, desensitize, or process the classified and graded data using a privacy computing protocol according to the classification and grading results and a preset security strategy. The data storage and management module is configured to store the data after classification, grading and security processing, and to perform data retrieval, access control, version management and lifecycle management on the stored data. The application service and interface module is configured to provide a unified data service interface and tool for upper-layer government affair applications. The system management and monitoring module is configured to perform user management, permission configuration, task scheduling, running state monitoring, log auditing and security alarm.

2. The government data processing system of claim 1, wherein, The multi-modal government affair large model is trained in the following manner: Government affair domain multi-source data is used as pre-training corpus; wherein the pre-training corpus comprises text data, image data, voice data and table data; Based on the pre-training corpus, independent pre-training and cross-modal semantic alignment of each modality are performed to form a unified semantic space of the multi-modal government affair large model; Based on a government affair knowledge base and rules, task fine-tuning is performed on the multi-modal government affair large model.

3. The government data processing system of claim 1, wherein, The multi-modal government affair large model and intelligent analysis module comprises a data classification and grading submodule. The data classification and grading submodule is configured to receive preprocessed data, and to call the multi-modal government affair large model to perform content analysis on the data; wherein the content analysis comprises theme recognition, entity extraction, relationship extraction, sentiment analysis of text data, and sensitive information detection in unstructured data; The data classification and grading submodule is further configured to extract key features based on the results of content analysis, query a classification and grading manner of the knowledge base module, and classify and grade a data set based on the key features and the classification and grading manner, and attach the classification and grading results as labels to the data.

4. The government data processing system of claim 1, wherein, The privacy computing engine module integrates multiple privacy computing technologies, including federated learning, multi-party secure computation, homomorphic encryption and trusted execution environment. The privacy computing engine module is configured to provide a unified interface for multiple privacy computing technologies.

5. The government data processing system of claim 4, wherein, The privacy computing engine module is configured to dynamically perform protocol matching on the multiple privacy computing technologies according to the classification and grading result of the data.

6. The government data processing system of claim 4, wherein, The privacy computing engine module is further configured to record the execution, resource consumption and security events of the privacy computing task.

7. The government data processing system of claim 1, wherein, The knowledge base is constructed in the following manner: Enter classification and grading rules, laws and regulations, documents and typical business scenario knowledge; Give natural language processing technology to perform named entity recognition, relation extraction and text classification on the entered content; Extract key knowledge elements from the entered unstructured documents to construct a knowledge graph or a relational database as the knowledge base.

8. The government data processing system of claim 7, wherein, The knowledge base is updated in the following manner: Review, correct and supplement the content of the knowledge base through artificial review and maintenance; Integrate the content of the new knowledge source into the knowledge base after screening and verification; Update the knowledge base with the latest rules and knowledge by using network crawler technology.

9. The government data processing system of any of claims 1-8, wherein, The government data processing system adopts a distributed micro-service architecture for deployment, and each module of the government data processing system is deployed as an independent micro-service on a containerized platform and is uniformly managed and scheduled through a gateway.

10. A government data processing method characterized by comprising: The method is applied to the government data processing system of any one of claims 1-9; the method comprises: The data access and preprocessing module accesses raw data from various government data sources and performs preprocessing operations on the raw data to obtain multi-modal data; wherein the preprocessing operations include cleaning, format conversion and metadata extraction; The multi-modal government large model and intelligent analysis module performs intelligent analysis on the multi-modal data based on a multi-modal government large model optimized in the government field; calls rules and models in the knowledge base module to classify and grade the multi-modal data; wherein the intelligent analysis includes deep semantic understanding, feature extraction and entity recognition; The knowledge base module provides a basis for classification and grading and privacy computing strategy formulation; the knowledge base module includes a knowledge base, which is a structured database storing professional knowledge in the government field, classification and grading standards and historical cases; The privacy computing engine module encrypts, desensitizes or processes the classified and graded data using a privacy computing protocol according to the classification and grading result and the preset security strategy; The data storage and management module stores the data after classification and security processing, and performs data retrieval, access control, version management and life cycle management on the stored data; The application service and interface module provides a unified data service interface and tool for upper government applications; The system management and monitoring module performs user management, permission configuration, task scheduling, running state monitoring, log auditing and security alarm.

Images