Knowledge base intelligent question answering and management system based on retrieval enhancement generation technology

Abstract

The invention provides a knowledge base intelligent question answering and management system based on a retrieval enhancement generation technology, belongs to the technical field of information retrieval, adopts a mixed cloud service mode, combines the advantages of private cloud and public cloud, and provides flexible deployment options. The system architecture is divided into three layers: a model layer, a capability layer and an application layer. Through the functions, the system aims to improve the efficiency and accuracy of information retrieval, accelerate the document processing speed and improve the writing efficiency, and is suitable for the fields of government affairs, commerce, academic, content creation and the like.

Description

Technical Field

[0001] The present invention relates to the field of information retrieval technology, and in particular to a knowledge base intelligent question and answer and management system based on retrieval enhancement generation technology. Through a hybrid cloud service model and multi-model support, it realizes efficient data integration, intelligent processing, and safe and compliant knowledge question and answer and management, and is suitable for a variety of industry business scenarios. Background Art

[0002] Retrieval-Augmented Generation (RAG) is a natural language processing technology that combines retrieval and generation. It improves the output quality of large language models (LLMs) in a cost-effective way by retrieving relevant information from large document collections and using this information to assist the generative model in producing more accurate and richer text responses.

[0003] The working principle of RAG includes three key steps: indexing, retrieval and generation. In the indexing stage, documents are divided into blocks, encoded into vectors and stored in the database; in the retrieval stage, the system retrieves relevant documents based on the similarity between the user query and the document block; in the generation stage, the retrieved information is combined with the original question and input into the language model to generate an answer.

[0004] The technical development of RAG has gone through a process from naive RAG to advanced RAG and then to modular RAG. Naive RAG includes basic indexing, retrieval and generation steps. Advanced RAG adds additional processing to these steps, such as data cleaning, document structure design and metadata addition. Modular RAG further breaks the traditional framework and introduces more functional modules and technologies, such as query search engine, multi-answer fusion, etc.

[0005] RAG technology has a wide range of application scenarios, including but not limited to question-answering systems, document generation, intelligent assistants, information retrieval, and knowledge graph filling. It retrieves relevant information from large-scale document collections and then uses this information to guide text generation, thereby improving the quality and accuracy of predictions. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a knowledge base intelligent question and answer and management system based on retrieval enhancement generation technology, which improves knowledge management capabilities based on retrieval enhancement generation technology and provides services such as knowledge base question and answer, knowledge base document analysis, and AI-assisted writing. At the same time, the system provides a unified API interface, through which users can easily call AI understanding capabilities and AI generation capabilities, and quickly deploy AI services without having to deeply understand complex algorithms and models.

[0007] The technical solution of the present invention is:

[0008] A knowledge base intelligent question answering and management system based on retrieval enhancement generation technology, comprising:

[0009] Model layer: supports open source and closed source large models, and provides local private deployment and public cloud deployment;

[0010] Capability layer: Integrates AI understanding and analysis, AI retrieval, AI generation, and AI data analysis capabilities;

[0011] Application layer: Build intelligent search, intelligent analysis, and intelligent writing applications, and provide AI understanding and generation capabilities through API interfaces;

[0012] Furthermore,

[0013] The model layer adopts a hybrid cloud service model. Local private deployment supports the deployment of AI models in the local environment, and public cloud deployment provides SaaS services for large cloud models through the public cloud, which can be used on demand.

[0014] In local private deployment, it provides support for open source and closed source large models; in public cloud deployment, it integrates large cloud models.

[0015] It allows users to select the most appropriate model for a specific task, or run several models simultaneously to improve performance and accuracy; and provides a unified interface and framework to manage and convert AI models from different sources and with different characteristics.

[0016] Furthermore,

[0017] AI understanding and analysis capabilities include diverse document recognition, document graph analysis, key information extraction, text error correction, and entity analysis capabilities;

[0018] AI search capabilities include RAG search enhancement, search intent understanding, personalized recommendations, and search sorting and filtering capabilities;

[0019] AI generation capabilities include labeling, summary generation, article creation, and article classification capabilities;

[0020] AI data analysis capabilities include indicator intent recognition, NL2SQL, and BI tool calling capabilities.

[0021] in,

[0022] Intelligent search: provides natural language search, automatic answer summary and content tracing functions;

[0023] Intelligent analysis: parse document content, generate summaries and mind maps;

[0024] Intelligent writing: supports multi-document reference and automatic generation of full text.

[0025] Furthermore,

[0026] Intelligent search uses natural language to communicate with the system in a chatty manner. Users can enter complete questions or use everyday language to describe their needs. The system understands the user's query intent and provides relevant search results in natural language.

[0027] Smart search automatically extracts key information from the data and summarizes it into concise answers to help users get the information they need.

[0028] While answering user questions, the system will also return the document or documents through which the answer was obtained; users can view the source documents of the information to verify its reliability.

[0029] The beneficial effects of the present invention are

[0030] 1) The system is deployed in a hybrid cloud service model, which is a combination of local private deployment and public cloud deployment. This approach can provide maximum deployment flexibility, combining the security of private cloud and the scalability of public cloud. It allows users to select the most appropriate model for a specific task, or run multiple models simultaneously to improve performance and accuracy. It also provides a unified interface and framework, allowing users to easily and conveniently manage and convert AI model capabilities from different sources and with different characteristics, making management and calling easier.

[0031] 2) The system provides a unified API interface, through which users can easily call AI understanding capabilities and AI generation capabilities, and quickly deploy AI services without having to have an in-depth understanding of complex algorithms and models.

[0032] 3) Intelligent search applications provide users with a comprehensive, efficient and accurate information retrieval solution through natural language processing, automatic answer summarization and content tracing technologies. This not only improves the user's search experience, but also greatly improves the efficiency and accuracy of information retrieval.

[0033] 4) Intelligent analysis makes document processing more efficient and intuitive by quickly parsing text, generating summaries and mind maps, improving users' understanding of documents and helping them quickly grasp the core points of documents.

[0034] 5) The intelligent writing function helps users improve writing efficiency by integrating reference documents and automatically generating full texts, while ensuring the quality and personalization of the content, freeing users from tedious text work. This intelligent writing method is very suitable for business, academic and content creation fields that need to quickly produce a large number of documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic diagram of the system architecture of the present invention;

[0036] Figure 2 This is a schematic diagram of the AI ​​comprehension capability API;

[0037] Figure 3 This is a schematic diagram of the AI ​​generation capability API. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0039] The present invention provides a knowledge base intelligent question answering and management system based on retrieval enhancement generation technology. The system architecture is mainly divided into three layers, from top to bottom: industry application scenario layer, capability layer and model layer.

[0040] 1) Model layer: hybrid cloud service model + multi-model support

[0041] The model layer adopts a hybrid cloud service model, which is a combination of local private deployment and public cloud deployment. It is designed to provide maximum deployment flexibility. This model combines the security of private cloud and the scalability of public cloud. Specifically, local private deployment supports the deployment of AI models in the local environment to ensure the security and privacy of data. Public cloud deployment provides SaaS services for large models in the cloud through the public cloud, which can be used on demand without worrying about the maintenance of the underlying hardware and software.

[0042] At the same time, under local private deployment, the system provides support for a variety of open source and closed source large models. Under public cloud deployment, the system also integrates a variety of cloud large models. This design provides great flexibility and scalability, allowing users to select the most appropriate model for a specific task, or run multiple models at the same time to improve performance and accuracy. And it provides a unified interface and framework, so that users can easily and conveniently manage and convert AI model capabilities from different sources and with different characteristics, making it easier to manage and call.

[0043] 2) Capability layer: Building multiple AI capabilities based on the model layer

[0044] Based on the model layer's various open source, closed source, and public cloud models, the capability layer integrates and encapsulates AI understanding and analysis capabilities, AI retrieval capabilities, AI generation capabilities, and AI data analysis capabilities.

[0045] Among them, AI understanding and analysis capabilities include diversified document recognition, document graph analysis, key information extraction, text error correction and entity analysis capabilities. AI retrieval capabilities include RAG retrieval enhancement, retrieval intent understanding, personalized recommendation and retrieval sorting and filtering capabilities. AI generation capabilities include labeling, summary generation, article creation and article classification capabilities. AI data analysis capabilities include indicator intent recognition, NL2SQL and BI tool calling capabilities.

[0046] 3) Application layer: Build top-level applications based on the capability layer.

[0047] Based on the capability layer, the following applications are built: intelligent search, intelligent analysis, and intelligent writing, which are used to establish a full range of intelligent document services, which can build a knowledge base, parse complex document content, and generate high-quality automated documents. At the same time, by encapsulating the four AI capabilities of the capability layer, two major API interfaces are provided: AI understanding capability API and AI generation capability API, so that AI understanding capability and AI generation capability can be easily called through API without in-depth understanding of complex algorithms and models.

[0048] (1) Intelligent search

[0049] Smart search provides an efficient and convenient information retrieval platform. It has the following main features:

[0050] Natural language search, express yourself freely. Users can use natural language to talk to the system in a chat-like manner and ask what they want to know, just like talking to a person. Users can enter complete questions or describe their needs in everyday language. The system can understand the user's query intent and provide relevant search results in the form of natural language. This search method reduces the precise keyword knowledge required by users when constructing queries, making the search process more intuitive and user-friendly.

[0051] Automatic answer summary improves efficiency. Smart search applications can automatically extract key information from large amounts of data and summarize it into concise answers to help users quickly obtain the information they need. This automatic summary function greatly improves retrieval efficiency.

[0052] Support content tracing to ensure accuracy. In order to ensure the accuracy of the information provided, the smart search application also supports content tracing. While answering user questions, the application will also return the document or documents from which the answer was obtained. Users can view the source document of the information to verify its reliability. The tracing function enhances the user's trust in the search results and ensures the authority and accuracy of the information, which is particularly important for situations where decisions rely on accurate information.

[0053] In summary, intelligent search applications provide users with a comprehensive, efficient and accurate information retrieval solution through technologies such as natural language processing, automatic answer summarization and content tracing. This not only improves the user's search experience, but also greatly improves the efficiency and accuracy of information retrieval.

[0054] (2) Intelligent Analysis

[0055] Intelligent analysis is designed to help users understand and digest document content faster and more deeply.

[0056] Efficient parsing and faster understanding. Zhuoshu Think Tank uses the text understanding ability of the big model to quickly parse the document content, identify key information and main points, and generate summaries and display them to users. Intelligent analysis applications can quickly improve the user's understanding of documents, especially when dealing with large or complex documents, so that users can quickly grasp the core points of the document.

[0057] Article mind map, clear structure. Intelligent analysis applications can also convert document content into a mind map, intuitively displaying the structure and hierarchy of the document. Users can see the organizational structure of the document at a glance, understand the relationship between different parts, and deepen their understanding and memory of the document content.

[0058] Together, these features form an intelligent analysis application that makes document processing more efficient and intuitive. Whether you need to quickly browse a large amount of information or deeply analyze a specific document, this application can provide corresponding support to help users improve work efficiency and quality.

[0059] (3) Intelligent Writing

[0060] The intelligent writing application combines the text generation capabilities of large models to provide users with an efficient and convenient document creation platform.

[0061] Multiple document references, integrated information. Intelligent writing tools can process multiple documents at the same time. Whether it is TXT, PDF, Word documents or PPT slides, the system can quickly read and analyze the contents of these documents. Users can specify reference documents, and the system will automatically extract key information and data from the reference documents to help users integrate information from different sources and ensure the comprehensiveness and accuracy of the writing content.

[0062] Intelligently generate full text to improve efficiency. Using the text generation capabilities of the big model, the intelligent writing application can automatically write a complete article based on the reference documents provided by the user. This ability to intelligently generate full text greatly improves writing efficiency, and users can get a well-structured and coherent article in a very short time.

[0063] The intelligent writing function provides users with a new writing experience by integrating reference documents and automatically generating full text. It not only improves writing efficiency, but also ensures the quality and personalization of content, freeing users from tedious text work. This intelligent writing method is very suitable for business, academic and content creation fields that need to quickly produce a large number of documents.

[0064] (4) AI Understanding API

[0065] This API encapsulates the AI ​​understanding and analysis capabilities, and AI retrieval capabilities at the capability layer.

[0066] (5) AI generation capability API

[0067] This API encapsulates the AI ​​generation capability and AI data analysis capability of the capability layer.

[0068] The system provides a unified API interface, through which users can easily call AI understanding capabilities and AI generation capabilities.

[0069] The system provides an intelligent search application, offering users a comprehensive, efficient and accurate information retrieval solution.

[0070] The system provides an intelligent writing application that can integrate reference documents and automatically generate full text.

[0071] The above description is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, and is not used to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A knowledge base intelligent question answering and management system based on retrieval enhancement generation technology, characterized in that: include: Model layer: supports open source and closed source large models, and provides local private deployment and public cloud deployment; Capability layer: Integrates AI understanding and analysis, AI retrieval, AI generation, and AI data analysis capabilities; Application layer: Build intelligent search, intelligent analysis, and intelligent writing applications, and provide AI understanding and generation capabilities through API interfaces.

2. The system according to claim 1, characterized in that The model layer adopts a hybrid cloud service model. Local private deployment supports the deployment of AI models in the local environment, and public cloud deployment provides SaaS services for large cloud models through the public cloud, which can be used on demand.

3. The system according to claim 1, characterized in that In local private deployment, it provides support for open source and closed source large models; in public cloud deployment, it integrates large cloud models.

4. The system according to claim 1, characterized in that It allows users to select the most appropriate model for a specific task, or run several models simultaneously to improve performance and accuracy; and provides a unified interface and framework to manage and convert AI models from different sources and with different characteristics.

5. The system according to claim 1, characterized in that AI understanding and analysis capabilities include diverse document recognition, document graph analysis, key information extraction, text error correction, and entity analysis capabilities; AI search capabilities include RAG search enhancement, search intent understanding, personalized recommendations, and search sorting and filtering capabilities; AI generation capabilities include labeling, summary generation, article creation, and article classification capabilities; AI data analysis capabilities include indicator intent recognition, NL2SQL, and BI tool calling capabilities.

6. The system according to claim 5, characterized in that in, Intelligent search: provides natural language search, automatic answer summary and content tracing functions; Intelligent analysis: parse document content, generate summaries and mind maps; Intelligent writing: supports multi-document reference and automatic generation of full text.

7. The system according to claim 6, characterized in that Intelligent search uses natural language to communicate with the system in a chatty manner. Users can enter complete questions or describe their needs in everyday language. The system understands the user's query intent and provides relevant search results in natural language.

8. The system according to claim 7, characterized in that Smart search automatically extracts key information from the data and summarizes it into concise answers to help users get the information they need.

9. The system according to claim 6, 7 or 8, characterized in that: While answering user questions, the system will also return the document or documents through which the answer was obtained; users can view the source documents of the information to verify its reliability.

Images