Government affair information intelligent retrieval and generation system applying RAG technology

Abstract

The invention provides a government affair information intelligent retrieval and generation system applying an RAG technology, and relates to the technical field of government affair information intelligent retrieval and generation systems. Comprising an information collection module, a data preprocessing module, a knowledge base construction module, an RAG core engine module, an intelligent retrieval module, an information generation module, a user interaction module, a safety and authority management module and a system monitoring and maintenance module, and all the modules work cooperatively to achieve efficient retrieval and intelligent generation of government affair information. Wherein the information acquisition module can improve the information retrieval accuracy from a plurality of government affair data sources, the query intention of a user can be accurately understood by combining an RAG technology with a knowledge base and a powerful retrieval algorithm, the most relevant content can be quickly positioned from massive government affair information, and compared with a traditional retrieval system, the accuracy of a retrieval result is greatly improved, and the retrieval efficiency is greatly improved. And the information screening time and energy of the user are effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent retrieval and generation systems for government information, and more specifically, particularly relates to intelligent retrieval and generation systems for government information using RAG technology. Background Art

[0002] In the field of government affairs, with the acceleration of the digitalization process, government information has exploded. Traditional government information retrieval systems mainly rely on keyword matching, which has many problems. On the one hand, it is impossible to accurately understand the semantic needs of users, resulting in a large number of redundant or irrelevant retrieval results, and users need to spend a lot of time screening useful information. On the other hand, it is difficult to provide comprehensive and in-depth answers to complex government affairs issues, and it is impossible to meet the growing information needs of the public and government officials. At the same time, the security of government information is of vital importance, but the existing system has deficiencies in user authentication, authority management and data encryption, which easily leads to the risk of information leakage. In addition, the data between various government departments are often independent of each other, lacking an effective integration and utilization mechanism, resulting in information islands, which hinders the improvement of government service efficiency. Therefore, there is an urgent need for an intelligent retrieval and generation system for government information using advanced technology to solve the above problems and improve the management and service level of government information. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides an intelligent retrieval and generation system for government information using RAG technology to solve the above problems.

[0004] The intelligent retrieval and generation system of government information using RAG technology includes information collection module, data preprocessing module, knowledge base construction module, RAG core engine module, intelligent retrieval module, information generation module, user interaction module, security and authority management module and system monitoring and maintenance module. Each module works together to achieve efficient retrieval and intelligent generation of government information.

[0005] Preferably, the information collection module can collect information from multiple government data sources, such as government department databases, government websites, electronic document libraries, etc., using distributed collection technology and data extraction tools. The collection process follows specific data collection protocols and specifications. The formula is: ,in Represents a list of data sources. Represents the extraction tool, It represents the protocol rules and has the functions of data deduplication and preliminary screening. The data preprocessing module uses a text cleaning algorithm, including removing noise characters, correcting spelling errors, unifying text formats, etc., to preprocess the collected data. The text cleaning formula is: ,in Represents the original text, The representation cleaning algorithm uses natural language processing technology to perform lexical, syntactic analysis and semantic annotation to improve the comprehensibility and usability of the data. The knowledge base construction module uses knowledge graph construction technology and semantic indexing methods to create a government knowledge base based on preprocessed data. It can automatically identify and extract key information and entity relationships. The algorithms used in the construction process include entity recognition algorithm, relationship extraction algorithm and semantic similarity calculation algorithm. The formula is: ,in represents the preprocessed data, represents the entity recognition algorithm, represents the relation extraction algorithm, Represents the semantic similarity calculation algorithm to ensure the accuracy and completeness of the knowledge base.

[0006] Preferably, the RAG core engine module combines the knowledge base and user query requirements, and uses a retrieval algorithm and a generative model based on an attention mechanism to achieve accurate information retrieval and related content generation. The core formula of the retrieval and generation process is: ,in Representation Knowledge Base , represents user query, represents the attention model, It represents a generative model, effectively integrating retrieval and generation capabilities to provide high-quality answers. The intelligent retrieval module uses the ranking algorithm and relevance evaluation technology based on the results of the RAG core engine module to further screen and sort the retrieved information, providing users with the most relevant and valuable information. The ranking formula is: , where table Shows the RAG engine results, represents the sorting algorithm, It represents relevance evaluation technology to improve the accuracy and availability of retrieval results. The information generation module uses natural language generation technology and pre-trained language models to generate detailed and coherent text answers based on user queries and knowledge base content. The language style and context information are considered in the generation process. The generation formula is: ,in Represents a user query, represents the knowledge base, represents the language model, Represents contextual information and satisfies users' in-depth information needs.

[0007] Preferably, the task scheduling agent reasonably allocates system resources and schedules the task execution order of each agent according to the preset task priority and resource status. Its task scheduling algorithm is based on the priority queue and resource load balancing strategy, and the formula is: ,in Represents a task list, Indicates the resource status. It represents the priority rules to ensure the efficient operation of the system. The security and permission management module is responsible for user identity authentication, authorization management and data encryption protection. It adopts multi-factor authentication, role-based access control model and advanced encryption standard algorithm. The formula is: ,in Indicates user information. Indicates role information. Indicates permission rules. It represents the encryption algorithm to ensure the security of the system and the confidentiality of the data. The user interaction module provides a friendly user interface to support operations such as user input of queries, viewing of retrieval and generation of results, and feedback. The interface design follows the principle of human-computer interaction and has good usability and operability. The interactive response formula is: ,in Indicates user request, Indicates system functionality and ensures that users can use the system efficiently.

[0008] Compared with the prior art, the present invention has the following beneficial effects:

[0009] Improve the accuracy of information retrieval: By combining RAG technology with a knowledge base and a powerful retrieval algorithm, it is possible to accurately understand the user's query intent and quickly locate the most relevant content from massive government information. Compared with traditional retrieval systems, the accuracy of retrieval results is greatly improved, effectively reducing the user's time and energy in screening information.

[0010] Realize intelligent information generation: Using natural language generation technology and a rich knowledge base, the system can not only retrieve information, but also generate detailed and logically coherent text answers based on user needs, providing users with more comprehensive and in-depth government information services to meet the needs of complex government consultation scenarios.

[0011] Enhance data utilization efficiency: The data preprocessing and knowledge base construction modules deeply process and integrate the original government data, extract key information and entity relationships, make data easier to retrieve and use, avoid data fragmentation and isolation, and enhance the overall value of government data.

[0012] Ensure system security and reliability: The multi-level protection measures of the security and permission management module ensure that only legally authorized users can access the system, and encrypt data to prevent information leakage and illegal operations, maintain the security and confidentiality of government information, and enhance the trust of government departments and users in the system.

[0013] Optimize user experience: The user-friendly design and efficient response mechanism of the user interaction module enable users of different technical levels to easily use the system, quickly obtain the required information and receive timely feedback, improve the satisfaction and efficiency of government services, and promote information exchange and interaction between the government and the public. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the intelligent retrieval and generation system of government information of the present invention;

[0015] Figure 2 It is a flow chart of the information collection and preprocessing system of the present invention;

[0016] Figure 3 is a flow chart of the knowledge base construction system of the present invention;

[0017] Figure 4 is a flow chart of the RAG core engine operation system of the present invention;

[0018] Figure 5 It is a flow chart of the retrieval and generation optimization system of the present invention;

[0019] Figure 6 It is a flow chart of the security and authority management system of the present invention;

[0020] Figure 7 is a flow chart of the user interaction design system of the present invention;

[0021] Figure 8 It is a flow chart of the knowledge base construction module system of the present invention;

[0022] Fig. 9 is a system flow chart of the RAG core engine module of the present invention;

[0023] Fig.10 It is a flow chart of the intelligent retrieval module system of the present invention. DETAILED DESCRIPTION

[0024] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0025] See also Figure 1-Figure 10 The present invention provides an intelligent retrieval and generation system for government information using RAG technology, including an information collection module, a data preprocessing module, a knowledge base construction module, a RAG core engine module, an intelligent retrieval module, an information generation module, a user interaction module, a security and authority management module, and a system monitoring and maintenance module. The modules work together to achieve efficient retrieval and intelligent generation of government information.

[0026] The information collection module can collect information from multiple government data sources, such as government department databases, government websites, electronic document libraries, etc., using distributed collection technology and data extraction tools. The collection process follows specific data collection protocols and specifications. The formula is: ,in Represents a list of data sources. Represents the extraction tool, It represents the protocol rules and has the functions of data deduplication and preliminary screening. The data preprocessing module uses text cleaning algorithms, including removing noise characters, correcting spelling errors, unifying text formats, etc. to preprocess the collected data. The text cleaning formula is: ,in Represents the original text, The representation cleaning algorithm uses natural language processing technology to perform lexical, syntactic analysis and semantic annotation to improve the comprehensibility and usability of the data. The knowledge base construction module uses knowledge graph construction technology and semantic indexing methods to create a government knowledge base based on preprocessed data. It can automatically identify and extract key information and entity relationships. The algorithms used in the construction process include entity recognition algorithm, relationship extraction algorithm and semantic similarity calculation algorithm. The formula is: ,in represents the preprocessed data, represents the entity recognition algorithm, represents the relation extraction algorithm, Represents the semantic similarity calculation algorithm to ensure the accuracy and completeness of the knowledge base.

[0027] The RAG core engine module combines the knowledge base and user query requirements, and uses the retrieval algorithm and generative model based on the attention mechanism to achieve accurate information retrieval and relevant content generation. The core formula of its retrieval and generation process is: ,in Representation Knowledge Base , represents user query, represents the attention model, It represents a generative model, effectively integrating retrieval and generation capabilities to provide high-quality answers. The intelligent retrieval module uses the ranking algorithm and relevance evaluation technology based on the results of the RAG core engine module to further screen and sort the retrieved information, providing users with the most relevant and valuable information. The ranking formula is: , where table Shows the RAG engine results, represents the sorting algorithm, It represents relevance evaluation technology to improve the accuracy and availability of retrieval results. The information generation module uses natural language generation technology and pre-trained language models to generate detailed and coherent text answers based on user queries and knowledge base content. The language style and context information are considered in the generation process. The generation formula is: ,in Represents a user query, represents the knowledge base, represents the language model, Represents contextual information and satisfies users' in-depth information needs.

[0028] The task scheduling agent reasonably allocates system resources and schedules the task execution order of each agent according to the preset task priority and resource status. Its task scheduling algorithm is based on the priority queue and resource load balancing strategy. The formula is: ,in Represents a task list, Indicates the resource status. It represents the priority rules to ensure the efficient operation of the system. The security and permission management module is responsible for user identity authentication, authorization management and data encryption protection. It adopts multi-factor authentication, role-based access control model and advanced encryption standard algorithm. The formula is: ,in Indicates user information. Indicates role information. Indicates permission rules. It represents the encryption algorithm to ensure the security of the system and the confidentiality of the data. The user interaction module provides a friendly user interface to support operations such as user input of queries, viewing of retrieval and generation of results, and feedback. The interface design follows the principle of human-computer interaction and has good usability and operability. The interactive response formula is: ,in Indicates user request, Indicates system functionality and ensures that users can use the system efficiently.

[0029] Embodiment 1:

[0030] Hardware configuration:

[0031] Server: Dell PowerEdge R740 server is selected, equipped with Intel Xeon Gold 6240 processor, 128GB memory, and 4TB hard disk to ensure stable system operation and data storage.

[0032] Network equipment: Huawei S5720-56C-PWR Ethernet switch is used to provide stable network connection and high-speed data transmission.

[0033] Software system:

[0034] Information collection module: Use the Scrapy framework to write a crawler program for government websites, follow the robots.txt protocol, collect information on a daily basis, and use a simple string matching algorithm to remove duplicate URLs and page content that has been collected.

[0035] Data preprocessing module: Use Python's re module to remove HTML tags, special symbols and other noise in the text, perform spelling check based on the open source Hunspell library, and use the NLTK tool to perform simple lexical analysis and part-of-speech tagging.

[0036] Knowledge base construction module: Store knowledge data in MySQL database, identify entities such as government agencies and policy names through customized regular expression rules, extract the relationship between entities based on co-occurrence frequency, and use the Word2Vec model to calculate semantic similarity to build a simple knowledge graph.

[0037] RAG core engine module: It uses a simple neural network retrieval algorithm based on the attention mechanism combined with the GPT-2 model for information retrieval and generation. It sets a smaller model parameter scale to adapt to hardware performance. The attention mechanism focuses on matching user query keywords with knowledge base text.

[0038] Intelligent search module: It performs simple result sorting based on the frequency of keyword occurrence, uses the BM25 algorithm to perform a preliminary relevance assessment, and selects the top 20 results to provide to users.

[0039] Information generation module: Based on the GPT-2 model, it guides the generation of text by setting specific prompt words in the government field, controls the length of the generated text within 200 words, and keeps the language concise and clear.

[0040] Security and authority management module: It uses username and password login, combined with role-based access control, and is divided into ordinary users, government affairs specialists, and administrator roles. Ordinary users can only perform routine queries, government affairs specialists can perform in-depth queries and simple data submissions, and administrators are responsible for system management and user authority allocation, using the AES-128 algorithm to encrypt stored data.

[0041] User interaction module: Develop a web application based on the Bootstrap framework, provide a simple search box, result display area and paging function, and the system will respond within 5 seconds after the user submits the query.

[0042] Embodiment 2:

[0043] Hardware Optimization:

[0044] The server was upgraded to Huawei FusionServerPro2288HV5 server, equipped with two Intel Xeon Silver 4210R processors, 256GB memory, and 8TB hard disk, improving the overall system performance and data storage capacity.

[0045] The network equipment was replaced with Cisco Catalyst9300-L switches, which support higher network bandwidth and more stable connections.

[0046] Software Upgrade:

[0047] Information collection module: Introduce the Apache Flume and Sqoop combined framework to achieve efficient data collection from multiple government databases (such as Oracle, SQL Server) and government data lakes, and use Bloom filters to perform efficient data deduplication to ensure the uniqueness of the collected data.

[0048] Data preprocessing module: The deep learning-based TextCNN model is used for noise removal and text cleaning, and the AllenNLP tool is used for comprehensive syntactic analysis and semantic role labeling to provide more accurate semantic information for knowledge base construction.

[0049] Knowledge base construction module: Build a complex knowledge graph based on the Neo4j graph database, use the BERT-NER model for high-precision entity recognition, extract entity relationships through a relationship extraction model based on the attention mechanism, and use the BERT-based semantic similarity model to optimize the knowledge graph construction.

[0050] RAG core engine module: adopts the improved multi-head attention mechanism retrieval algorithm and GPT-3 model, and uses LoRA technology to efficiently fine-tune the model, increase the context window length to 1024 tokens, and improve the quality of information retrieval and generation.

[0051] Intelligent retrieval module: sorts and evaluates the relevance of search results based on the semantic features of the BERT model, combines the analysis of users' historical query behavior to provide users with more personalized search results, and screens out the top 15 results in terms of relevance.

[0052] Information generation module: Based on the GPT-3 model, a large number of government professional documents and cases are introduced for secondary training. The structure and content of the generated text are dynamically adjusted according to the user's query intention. The length of the generated text is controlled at around 500 words, and the language style is more in line with the formal style of government affairs.

[0053] Security and permission management module: It uses fingerprint recognition and password dual authentication, combined with the attribute-based access control (ABAC) model, to assign fine-grained permissions based on the user's department, position and other attributes, encrypt sensitive data using the RSA-2048 algorithm, and use blockchain technology to record data access logs to ensure data security and traceability.

[0054] User interaction module: Develop responsive web applications that are compatible with a variety of terminal devices, provide intelligent search suggestions, visual result display (such as a bar chart showing policy quantity trends) and user feedback functions, and shorten user query response time to within 3 seconds.

[0055] Embodiment 3:

[0056] Hardware Innovation:

[0057] Raspberry Pi 4B is deployed as an edge computing device in various branches of government departments. It is equipped with 8GB of memory and a 128GB storage card. It communicates with the central server through the 5G network to achieve localized data collection and preprocessing.

[0058] The central server adopts a cloud computing architecture, rents elastic computing resources from Alibaba Cloud, dynamically adjusts CPU and memory resources according to system load, and expands storage capacity on demand.

[0059] Software expansion:

[0060] Information collection module: Integrate with government IoT devices (such as environmental monitoring sensors and smart office equipment), collect data in real time through the MQTT protocol, use lightweight data processing programs on edge computing devices to clean and preliminarily classify the collected data in real time, and only upload valid data to the central server.

[0061] Data preprocessing module: Using federated learning technology, collaborate with government departments in other regions to clean and standardize data, improve the overall data quality without exchanging original data, and use the Transformer-based unsupervised learning model for semantic enhancement processing.

[0062] Knowledge base construction module: Combining knowledge graph and semantic network technology, it conducts deep semantic understanding and knowledge extraction on unstructured texts in the government field (such as subtitles of policy interpretation videos and minutes of expert seminars), constructs a multimodal knowledge graph, and enriches the content and structure of the knowledge base.

[0063] RAG core engine module: uses reinforcement learning algorithms to train intelligent agents to optimize the retrieval and generation strategies of the RAG model. The intelligent agent dynamically adjusts parameters based on user feedback and system performance indicators, introduces multimodal information processing capabilities, and can retrieve and generate government content containing multimodal information such as pictures and charts.

[0064] Intelligent retrieval module: It uses a hybrid sorting algorithm to comprehensively consider factors such as text relevance, information timeliness, and user preferences to sort results. It uses a deep learning model to predict user satisfaction with search results, continuously optimizes the sorting results, and provides the top 10 best results.

[0065] Information generation module: Use generative adversarial network (GAN) technology to improve the text generation process. The generator and the discriminator compete with each other to improve the logic and credibility of the generated text. The generated content is customized according to the language habits and needs of different user groups (such as corporate users and residential users), and the length of the generated text can be flexibly adjusted.

[0066] Security and permission management module: Use homomorphic encryption technology to perform retrieval and analysis operations on encrypted data to ensure data privacy, use artificial intelligence-driven intrusion detection systems to monitor network attacks and abnormal behaviors in real time, and redesign the security system based on a zero-trust architecture to ensure system security without blind spots.

[0067] User interaction module: Develop intelligent interactive applications that integrate voice interaction and gesture control, support users to quickly query information and obtain results through voice commands and gesture operations, use natural language processing technology and computer vision technology to achieve precise interaction, and control the interactive response time within 2 seconds.

[0068] Embodiment 4:

[0069] Hardware configuration adjustment:

[0070] The server uses the Supermicro SuperServer 7049GP-TRT server, equipped with an AMD EPYC 7742 processor, 512GB of memory, and a 16TB hard drive, providing the system with powerful computing power and massive storage support, suitable for large-scale government data centers.

[0071] The storage device uses Dell EMC PowerMax storage array, which provides high-speed data reading and writing and high-reliability data storage, and is configured with data redundancy and snapshot functions to ensure data security and recoverability.

[0072] Software system customization:

[0073] Information collection module: Develop dedicated collection programs for specific government affairs (such as administrative approval systems), use database triggers and message queue technology to achieve real-time data collection, ensure the timeliness and integrity of data, and use SSL / TLS encryption protocols to ensure data transmission security.

[0074] Data preprocessing module: formulate professional data cleaning rules based on the characteristics of administrative approval data, such as verifying the data format of application forms, reviewing the completeness of application materials, and using domain ontology knowledge for semantic annotation and classification to improve the accuracy and professionalism of data processing.

[0075] Knowledge base construction module: Build a government knowledge base based on domain ontology, use the Semantic Web Rule Language (SWRL) to define reasoning rules, realize automatic reasoning and expansion of knowledge, use semantic indexing technology to improve knowledge retrieval efficiency, and build a domain-specific knowledge graph to display the relationship between business processes and regulations.

[0076] RAG core engine module: Develop the RAG model for administrative approval business, use a large number of real administrative approval cases and regulatory documents for training and optimization, so that it can accurately answer questions related to approval process, conditions, materials, etc., and combine the business rule engine in the retrieval and generation process to ensure that the results meet regulatory requirements.

[0077] Intelligent retrieval module: It uses a hybrid sorting algorithm based on business rules and semantic similarity, and sorts the search results based on the approval business priority and timeliness. For example, it gives priority to displaying information related to urgent approval matters, and selects the top 5 results that best meet user needs.

[0078] Information generation module: Based on the standardized text template and language style of administrative approval, natural language generation technology is used to generate detailed and accurate approval guidelines, feedback and other text content. The length of the generated text can be flexibly adjusted according to business needs to ensure that the content is complete and easy to understand.

[0079] Security and permission management module: Use hardware security module (HSM) for key management and data encryption, implement strict multi-factor identity authentication (password, SMS verification code, smart card) and role- and task-based access control, and use the audit system to monitor user operations and system security events in real time to ensure the security and confidentiality of administrative approval data.

[0080] User interaction module: Develop dedicated client software for government affairs, integrate functions such as electronic signature and file upload and download, provide intuitive business process navigation and intelligent assistant, help users quickly complete administrative approval-related operations, and respond to user operation feedback within 1 second.

[0081] Embodiment 5:

[0082] Hardware adaptation for special scenarios:

[0083] For mobile government office scenarios, high-performance laptops are used as temporary servers, such as Lenovo ThinkPad P15v, which is equipped with an Intel Core i7-11800H processor, 32GB of memory, and a 1TB solid-state drive, and network connection is achieved through a mobile hotspot or 5G network module.

[0084] Paired with portable storage devices, such as the Samsung T7 portable solid-state drive, for data backup and expanded storage.

[0085] Software adaptation for special scenarios:

[0086] Information collection module: Utilize the lightweight data collection function within the mobile application to collect information through manual input by users, photo uploading (such as collecting evidence of on-site government affairs issues), etc., perform simple format verification and compression processing on the collected data to reduce the amount of data transmission.

[0087] Data preprocessing module: Run a simplified text processing algorithm on the mobile device to remove obvious noise characters, correct common spelling errors, and use the local vocabulary to perform simple part-of-speech tagging to ensure the basic quality of the data.

[0088] Knowledge base construction module: Build a streamlined knowledge base on mobile devices based on the SQLite database to store common government knowledge and local policies and regulations, use local computing resources to perform simple knowledge association and reasoning, and improve knowledge utilization efficiency.

[0089] RAG core engine module: uses a lightweight version of the RAG model, optimizes the performance of mobile devices, reduces model parameters and computational complexity, and uses cache technology to store commonly used search and generation results to speed up response.

[0090] Intelligent search module: It uses a simple sorting algorithm based on keyword matching and local popularity, giving priority to displaying local popular government information and results related to users' recent queries, and quickly providing the top 10 search results.

[0091] Information generation module: Generate concise and clear short text answers based on the mobile device screen display characteristics and user operation habits, control the text length within 150 words, and avoid information overload.

[0092] Security and permission management module: Use biometrics (such as facial recognition) to log in, combined with device encryption technology (such as full disk encryption) to protect data security, and use the application sandbox mechanism to limit application permissions to prevent data leakage and malicious operations.

[0093] User interaction module: Develop a simple and easy-to-use mobile application interface, using large icons, simple menus and voice prompts to facilitate users to operate quickly in a mobile environment. Use push notifications to push important government information and system reminders to users in a timely manner to ensure that users do not miss key information. The interactive response time is controlled within 3 seconds.

[0094] Through these five implementation examples, the diversity and flexibility of the intelligent retrieval and generation system for government information using RAG technology are demonstrated from the aspects of different hardware configurations, software function upgrades and adaptation to special scenarios. It can meet various government application needs and provide strong support for the construction of government informatization.

[0095] The embodiments of the present invention are given for the purpose of illustration and description, and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. The intelligent retrieval and generation system of government information using RAG technology is characterized by: It includes information collection module, data preprocessing module, knowledge base construction module, RAG core engine module, intelligent retrieval module, information generation module, user interaction module, security and authority management module and system monitoring and maintenance module. Each module works together to achieve efficient retrieval and intelligent generation of government information.

2. According to the intelligent retrieval and generation system of government information in claim 1, the information collection module can collect information from multiple government data sources, such as government department databases, government websites, and electronic document libraries, using distributed collection technology and data extraction tools. The collection process follows specific data collection protocols and specifications, and the formula is: ,in Represents a list of data sources. Represents the extraction tool, It represents the protocol rules and has the functions of data deduplication and preliminary screening.

3. The intelligent retrieval and generation system for government information as claimed in claim 1, wherein the data preprocessing module uses a text cleaning algorithm, including removing noise characters, correcting spelling errors, and unifying text format operations, to preprocess the collected data, and the text cleaning formula is: ,in Represents the original text, The representation cleaning algorithm uses natural language processing technology to perform lexical, syntactic analysis and semantic annotation to improve the comprehensibility and usability of the data.

4. According to the intelligent retrieval and generation system for government information as claimed in claim 1, the knowledge base construction module creates a government knowledge base based on preprocessed data using knowledge graph construction technology and semantic indexing methods, and can automatically identify and extract key information and entity relationships. The algorithms used in the construction process include entity recognition algorithms, relationship extraction algorithms, and semantic similarity calculation algorithms. The formula is: ,in represents the preprocessed data, represents the entity recognition algorithm, represents the relation extraction algorithm, Represents the semantic similarity calculation algorithm to ensure the accuracy and completeness of the knowledge base.

5. In the intelligent retrieval and generation system for government information as claimed in claim 1, the RAG core engine module combines the knowledge base and user query requirements, and uses a retrieval algorithm and a generative model based on an attention mechanism to achieve accurate information retrieval and related content generation. The core formula of the retrieval and generation process is: ,in Representation Knowledge Base , represents user query, represents the attention model, Represents a generative model that effectively integrates retrieval and generation capabilities to provide high-quality answers.

6. According to the intelligent retrieval and generation system of government information in claim 1, the intelligent retrieval module further screens and sorts the retrieved information based on the results of the RAG core engine module using a sorting algorithm and relevance evaluation technology to provide users with the most relevant and valuable information. The sorting formula is: , where table Shows the RAG engine results, represents the sorting algorithm, Represents relevance evaluation technology to improve the accuracy and availability of retrieval results.

7. The intelligent retrieval and generation system for government information as claimed in claim 1, wherein the information generation module uses natural language generation technology and pre-trained language models to generate detailed and coherent text answers based on user queries and knowledge base content, and takes language style and context information into consideration during the generation process. The generation formula is: ,in Represents a user query, represents the knowledge base, represents the language model, Represents contextual information and satisfies users' in-depth information needs.

8. According to the intelligent retrieval and generation system of government information as claimed in claim 1, the task scheduling agent reasonably allocates system resources and schedules the task execution order of each agent according to the preset task priority and resource status. Its task scheduling algorithm is based on the priority queue and resource load balancing strategy, and the formula is: ,in Represents a task list, Indicates the resource status. Represents priority rules to ensure efficient operation of the system.

9. The intelligent retrieval and generation system for government information as claimed in claim 1, wherein the security and authority management module is responsible for identity authentication, authorization management and data encryption protection for users, and adopts a multi-factor authentication method, a role-based access control model and an advanced encryption standard algorithm, and the formula is: ,in Indicates user information. Indicates role information. Indicates permission rules. Represents the encryption algorithm, which ensures the security of the system and the confidentiality of the data.

10. According to the intelligent retrieval and generation system of government information in claim 1, the user interaction module provides a friendly user interface, supports users to input queries, view retrieval and generation results, and provide feedback. The interface design follows the principle of human-computer interaction and has good usability and operability. The interactive response formula is: ,in Indicates user request, Indicates system functionality and ensures that users can use the system efficiently.

Images