RAG-based intelligent customer support method, system and device, and storage medium

Abstract

The embodiment of the invention discloses an intelligent customer support method, system and equipment based on RAG and a storage medium, and particularly provides specialized document retrieval and problem solving functions aiming at operation and maintenance requirements of a software defined storage system, can improve the understanding precision of current query by utilizing historical query and interaction records of a user, and improves the user experience. More accurate and related answers are provided, the answers are generated in combination with contents in various forms such as texts, pictures and videos, the richness and expressive force of the answers are improved, an adaptive learning mechanism is introduced, learning strategies and parameters are automatically adjusted according to user feedback, and the performance is continuously optimized and improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of retrieval enhancement generation, and in particular to a RAG-based intelligent customer support method, system, device and storage medium. Background Art

[0002] With the rapid development of information technology, the digital transformation of enterprises has become an irreversible trend. In this context, efficient management and access to storage systems are crucial.

[0003] At present, the research and development and application of intelligent customer support systems are gradually increasing. Many companies have begun to explore the application of artificial intelligence and natural language processing technologies in the field of customer support. Intelligent customer support systems have been widely used in recent years, especially in the context of enterprise informatization and customer service automation. Most of the major intelligent customer support systems on the market are based on FAQ systems or simple automated question-and-answer systems, which usually rely on predefined question-and-answer pairs to respond to user queries. The advantages of these systems are simple implementation and low cost, but they also have obvious limitations, mainly reflected in the inability to flexibly respond to complex and diverse user problems.

[0004] The existing intelligent customer support system has improved the efficiency of customer service to a certain extent, but there are still many limitations. FAQ-based systems and rule-based systems have poor flexibility and cannot deal with complex and diverse user questions; and although the retrieval-based system has improved in breadth, it still has deficiencies in the accuracy of information and the naturalness of answers. They generally lack self-learning and optimization capabilities and cannot improve the quality of their own answers over time.

[0005] Retrieval-Augmented Generation (RAG) is a new technology that combines information retrieval and generation models. It can search for relevant information from a large number of documents and generate natural language answers, and has very broad application prospects. Summary of the invention

[0006] To this end, the embodiments of the present invention provide a RAG-based intelligent customer support method, system, device and storage medium to solve the technical problems that the prior art has poor flexibility, is unable to cope with complex and diverse user problems, lacks self-learning and optimization capabilities, is unable to improve its own answer quality over time, and has deficiencies in the accuracy of information and the naturalness of answers.

[0007] In order to achieve the above purpose, the embodiment of the present invention provides the following technical solutions:

[0008] According to a first aspect of an embodiment of the present invention, there is provided a RAG-based intelligent customer support method, the method comprising:

[0009] S1. Receive a query statement input by a user, and use a preset RAG model to perform semantic recognition on the query statement to obtain the user's query intention;

[0010] S2, analyzing the query statement, understanding the deep semantics of the query statement and converting the deep semantics into computer instructions;

[0011] S3, obtaining the user's historical query records and interaction records, and using the historical query records and the interaction records to process the query statement to generate a high-precision query intent;

[0012] S4, searching and acquiring relevant information in a preset database according to the user query intention and the high-precision query intention combined with computer instructions;

[0013] S5. Generate a first natural language answer based on the relevant information and optimize the first natural language answer to generate a second natural language answer.

[0014] Further, receiving a query statement input by a user, using a preset RAG model to perform semantic recognition on the query statement, and obtaining the user's query intention, including:

[0015] The semantic recognition is natural language processing analysis;

[0016] The query intent includes: adding, nodes, and storage clusters.

[0017] Further, analyzing the query statement, understanding the deep semantics of the query statement and converting the deep semantics into computer instructions includes:

[0018] The natural language processing analysis includes: word segmentation, part-of-speech tagging and intent recognition;

[0019] Among them, word segmentation and part-of-speech tagging are performed by using a pre-trained natural language processing model, and intent recognition is performed based on a deep learning model.

[0020] Further, according to the user query intention and the high-precision query intention combined with the computer instruction, the relevant information is retrieved from the preset database, including:

[0021] Use ElasticSearch to perform a full-text search in a preset database to obtain the first search result;

[0022] The first search result is analyzed by the BM25 algorithm to obtain high-precision relevant information.

[0023] Further, generating a first natural language answer based on the relevant information and optimizing the first natural language answer to generate a second natural language answer includes:

[0024] Generate a first natural language answer using the preset RAG model and the relevant information;

[0025] The first natural language answer is optimized by the optimization algorithm in the preset RAG model to generate a second natural language answer.

[0026] Preferably, a RAG-based intelligent customer support method further includes:

[0027] Obtaining and analyzing user feedback on the second natural language answer to identify the pros and cons of the answer and changes in user needs;

[0028] The feedback data is used to adjust the strategy of the preset RAG model and optimize the parameters of the preset RAG model.

[0029] According to a second aspect of an embodiment of the present invention, there is provided an intelligent customer support system based on RAG, the system comprising:

[0030] A query processing module is used to receive a query statement input by a user, perform semantic recognition on the query statement using a preset RAG model, and obtain the user's query intention;

[0031] The deep semantic analysis module is used to extract the core concepts and context of the query statement and generate a deep semantic representation of the query statement to support the precise retrieval and intent analysis of subsequent modules;

[0032] High-precision query intent analysis module, which is used to optimize the intent analysis of the current query statement based on historical interaction data, and generate more accurate query intent by combining context and historical information to improve the relevance of subsequent retrieval;

[0033] A retrieval module, used to retrieve relevant information in a preset database according to the user query intention and the high-precision query intention combined with computer instructions;

[0034] An answer optimization module is used to generate a first natural language answer based on the relevant information and optimize the first natural language answer to generate a second natural language answer.

[0035] Furthermore, the system further comprises: a learning optimization module, configured to perform the following steps:

[0036] Obtaining and analyzing user feedback on the second natural language answer to identify the pros and cons of the answer and changes in user needs;

[0037] The feedback data is used to adjust the strategy of the preset RAG model and optimize the parameters of the preset RAG model.

[0038] According to a third aspect of an embodiment of the present invention, there is provided a RAG-based intelligent customer support device, the device comprising: a processor and a memory;

[0039] The memory is used to store one or more program instructions;

[0040] The processor is used to run one or more program instructions to execute the steps of a RAG-based intelligent customer support method as described in any one of the above items.

[0041] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of a RAG-based intelligent customer support method as described in any one of the above items are implemented.

[0042] The embodiments of the present invention have the following advantages:

[0043] The embodiments of the present invention specifically target the operation and maintenance requirements of software-defined storage systems, provide professional document retrieval and problem-solving functions, utilize historical query and interaction records of users to improve the accuracy of understanding current queries, provide more accurate and relevant answers, generate answers by combining content in various forms such as text, pictures, and videos, improve the richness and expressiveness of answers, introduce an adaptive learning mechanism, automatically adjust learning strategies and parameters based on user feedback, and continuously optimize and improve performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0045] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0046] Figure 1 A schematic diagram of the logical structure of an intelligent customer support system based on RAG provided in an embodiment of the present invention;

[0047] Figure 2 A flowchart of a RAG-based intelligent customer support method provided by an embodiment of the present invention shows the complete process from user input of a query to generation of an answer, and each step in the figure shows the data processing path and the calling relationship of the model;

[0048] Figure 3 A schematic diagram of the architecture of an intelligent customer support system based on RAG provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] At present, the research and development and application of intelligent customer support systems are gradually increasing. Many companies have begun to explore the application of artificial intelligence and natural language processing technologies in the field of customer support. Intelligent customer support systems have been widely used in recent years, especially in the context of enterprise informatization and customer service automation. Most of the major intelligent customer support systems on the market are based on FAQ systems or simple automated question-and-answer systems, which usually rely on predefined question-and-answer pairs to respond to user queries. The advantages of these systems are simple implementation and low cost, but they also have obvious limitations, mainly reflected in the inability to flexibly respond to complex and diverse user problems.

[0051] In order to solve the technical problems that the above-mentioned existing technologies have poor flexibility, are unable to cope with complex and diverse user questions, lack self-learning and optimization capabilities, are unable to improve their own answer quality over time, and have insufficient accuracy of information and naturalness of answers.

[0052] Example 1

[0053] refer to Figure 1 The embodiment of the present invention discloses an intelligent customer support system based on RAG, which includes: a query processing module 1; a deep semantic analysis module 2; a high-precision query intention analysis module 3; a retrieval module 4; an answer optimization module 5; and a learning optimization module 6.

[0054] refer to Figure 3,The user interface module is the entrance for users to interact with the ,system, and provides a friendly interface for users to enter ,queries and receive answers generated by the system.,This module includes an input box, a result display area and a feedback ,mechanism. Users can enter questions through the interface, view ,the system

[0055] Technical framework: Use the React framework to build the user interface to ensure dynamic updates and responsiveness of the interface.

[0056] Voice input: Introduce Web SpeechAPI for voice recognition, supporting users to query through voice input.

[0057] The query processing module is responsible for receiving the query input by the user and performing preliminary natural language processing (NLP) analysis, including word segmentation, part-of-speech tagging, and intent recognition. The processed query will be passed to the retrieval engine module for further processing.

[0058] NLP processing: Use the SpaCy library for word segmentation and part-of-speech tagging, and use the pre-trained BERT model for intent recognition.

[0059] Custom intent recognition algorithm: Implement a custom intent recognition algorithm and combine the user's historical query records to improve recognition accuracy.

[0060] The document management module is based on software-defined storage technology to achieve efficient and flexible document management and storage. This module stores product documents and operation and maintenance documents, and supports dynamic update and version management of documents. The document management module works together with the permission control module to ensure the security and access control of documents.

[0061] Storage system: Toyou Storage is used as the underlying storage system to provide high availability and scalability.

[0062] Version control: Implement document version control function and use Git to manage document version changes.

[0063] RAG technology is used to search for information related to user queries in documents stored in the document management module. This module combines information retrieval technology and generation models to quickly find relevant documents and extract useful information from them.

[0064] Retrieval technology: Use ElasticSearch for full-text retrieval, combined with the BM25 algorithm to improve retrieval accuracy.

[0065] Among them, ElasticSearch is used for full-text retrieval, and the retrieval results are sorted by the BM 25 algorithm. The RAG technology is combined in the generation model to further filter and enhance the relevance of the optimal results.

[0066] RAG technology application: Implement RAG technology, combine retrieval results and generative models, and generate high-quality natural language answers.

[0067] The answer generation module generates natural language answers based on the information returned by the search engine module. This module uses the generation capabilities of large models (such as GPT-4) to ensure the fluency and accuracy of the answers. The generated answers will be returned to the user through the user interface module.

[0068] Generative models: Use large models (such as GPT-4) to generate natural language answers.

[0069] Optimization algorithm: Through a unique optimization algorithm, the generated answers are polished and restructured to ensure the accuracy and fluency of the answers.

[0070] The learning and optimization module continuously learns and optimizes system performance through user feedback and interaction data. This module uses machine learning technology to analyze user feedback on answers, adjust and optimize retrieval and generation models, and improve system accuracy and user satisfaction.

[0071] Feedback analysis: Use machine learning algorithms to analyze user feedback and identify the pros and cons of answers and changes in user needs.

[0072] Model optimization: Use feedback data to adjust and optimize model parameters and strategies.

[0073] The permission control module ensures the security of the system and the control of document access. This module is responsible for user authentication and permission allocation, ensuring that only authorized users can access specific sensitive information and documents.

[0074] Authentication: Implement multi-factor authentication technology to improve system security.

[0075] Permission management: Dynamically adjust information access permissions based on user roles and permission levels.

[0076] Corresponding to the above-disclosed intelligent customer support system based on RAG, the embodiment of the present invention further discloses an intelligent customer support method based on RAG. The following describes in detail an intelligent customer support method based on RAG disclosed in the embodiment of the present invention in combination with the above-described intelligent customer support system based on RAG.

[0077] Example 2

[0078] refer to Figure 2The present invention discloses an intelligent customer support method based on RAG, the method comprising: S1, receiving a query statement input by a user, using a preset RAG model to perform semantic recognition on the query statement, and obtaining the user's query intention; S2, analyzing the query statement, understanding the deep semantics of the query statement and converting the deep semantics into computer instructions; S3, obtaining the user's historical query records and interaction records, using the historical query records and the interaction records to process the query statement, and generating a high-precision query intention; S4, according to the user's query intention and the high-precision query intention combined with the computer instructions, retrieve and obtain relevant information in a preset database; S5, generating a first natural language answer based on the relevant information and optimizing the first natural language answer to generate a second natural language answer.

[0079] Furthermore, a query statement input by a user is received, and semantic recognition is performed on the query statement using a preset RAG model to obtain the user's query intention, including: the semantic recognition is a natural language processing analysis; the query intention includes: adding, nodes, and storage clusters.

[0080] Further, analyzing the query statement, understanding the deep semantics of the query statement and converting the deep semantics into computer instructions includes: the natural language processing analysis includes: word segmentation, part-of-speech tagging and intent recognition.

[0081] Among them, word segmentation and part-of-speech tagging are performed by using a pre-trained natural language processing model, and intent recognition is performed based on a deep learning model.

[0082] The pre-trained natural language processing model is preferably the SpaCy library, and the deep learning model is preferably the BERT model.

[0083] Furthermore, according to the combination of the user query intention and the high-precision query intention and computer instructions, relevant information is retrieved in a preset database, including: using ElasticSearch to perform a full-text search in the preset database to obtain a first search result; analyzing the first search result through the BM25 algorithm to obtain high-precision relevant information.

[0084] Preset database: Through database design, including the structure, fields and relationships of each table, to ensure the efficiency and integrity of data management.

[0085] Database selection: Use PostgreSQL as a relational database to store user queries, feedback, and document metadata.

[0086] Table structure design: describe the fields and relationships of each table in detail.

[0087] The tables include: user table, query record table, document metadata table, and user feedback table.

[0088] Furthermore, a first natural language answer is generated based on the relevant information and the first natural language answer is optimized to generate a second natural language answer, including: using the preset RAG model and the relevant information to generate the first natural language answer; optimizing the first natural language answer through the optimization algorithm in the preset RAG model to generate a second natural language answer.

[0089] Preferably, an intelligent customer support method based on RAG also includes: obtaining and analyzing user feedback on the second natural language answer to identify the pros and cons of the answer and changes in user needs; using feedback data to make strategic adjustments to the preset RAG model and optimize the preset RAG model parameters.

[0090] Feedback collection: Provide a feedback portal in the user interface, where users can comment on and leave messages on answers.

[0091] Feedback analysis: Use natural language processing techniques to analyze user feedback and identify common problems and areas for improvement.

[0092] System improvement: Adjust and optimize system functions and performance based on feedback analysis results.

[0093] System performance optimization measures are taken to ensure system stability and response speed under high concurrency and large data volumes.

[0094] Cache mechanism: Use Redis to cache common query results to reduce database query pressure.

[0095] Load balancing: Use Nginx for load balancing to disperse request pressure and improve system throughput.

[0096] Ensure the security of user data and system operations through system security measures.

[0097] Data encryption: Sensitive data is encrypted when stored in the database, using the AES-256 encryption algorithm.

[0098] Secure communication: Use HTTPS protocol for data transmission to ensure communication security.

[0099] Example 3

[0100] When the user needs a query function, the user enters a query: "How to install the latest software version?".

[0101] The user interface module receives the user query and passes it to the query processing module. The query processing module analyzes the query and identifies the keywords "install", "latest", and "software version". The search engine module searches the document management module for relevant installation guides and software version information. The answer generation module generates a natural language answer: "To install the latest software version, please visit the official website to download the latest installation package and follow the installation guide." The user interface module displays the answer and allows the user to provide feedback. The learning optimization module analyzes the user feedback, and if the user feedback is not satisfactory, the system will optimize the retrieval and generation process.

[0102] When users need to deal with permission control, user queries involve sensitive information: "How do I access the system's management console?".

[0103] The user interface module receives the user query and passes it to the query processing module. The query processing module analyzes the query and identifies the keywords "access", "system", and "management console". The permission control module checks the user's permissions and finds that the user does not have permission to access the information. The system returns a prompt indicating insufficient permissions: "You do not have permission to access this information. Please contact the administrator for more help." The user interface module displays the prompt and records the user's query and feedback.

[0104] Example 4

[0105] 1) User query input

[0106] The user enters a query through the user interface module. For example, the user enters: "How to add a new node in the storage cluster?".

[0107] The user interface design combines the NLU technology of the large model, which can process the spoken input of natural language and enhance the user experience. The adaptive interface technology is introduced, which can automatically adjust the interface layout and functions according to the user's historical query records and preferences to provide a personalized experience.

[0108] 2) Query processing

[0109] The query processing module receives the query input by the user and uses the big model to perform the following processing:

[0110] Intent recognition: The big model directly recognizes the intent of the user's query, such as "add", "node", "storage cluster". Semantic parsing: The big model understands the deep semantics of the query and converts it into machine-processable instructions. Context understanding: The cross-query context understanding function is introduced, which can use the user's historical query and interaction records to improve the accuracy of understanding the current query.

[0111] The large model uses pre-trained knowledge to accurately identify and parse complex queries, improving the system's understanding ability. The context understanding function enables the system to maintain efficient understanding capabilities in continuous conversations and enhance user experience.

[0112] 3) Search engine startup

[0113] The retrieval engine module searches for relevant information in the document management module based on the query processed by the big model: Query expansion: Expand user queries through the big model to obtain more relevant documents. Relevant document retrieval: Search for content matching the query in operation and maintenance documents and product documents. Information extraction: Extract the most relevant paragraphs and sentences from the retrieved documents. Dynamic knowledge base update: Automatically extract information from new data sources and update the knowledge base to ensure the latest information.

[0114] The large model combined with RAG technology expands the query and retrieves more relevant documents, improving the accuracy and coverage of the search. The dynamic knowledge base update function ensures that the system can obtain the latest information in real time and provide accurate and timely answers.

[0115] 4) Answer generation

[0116] The answer generation module uses a large model to generate natural language answers: Answer generation: Generate natural language answers that match user queries based on the retrieved relevant information. Answer optimization: Polish and structure the generated answers through a unique optimization algorithm to ensure the accuracy and fluency of the answers. Multimodal fusion: Generate answers by combining text, pictures, videos and other forms of content to improve the richness and expressiveness of the answers.

[0117] The large model generation is combined with the information retrieval results to ensure that the answers are accurate and natural. Multimodal fusion technology provides a rich variety of answer forms to enhance the user experience.

[0118] 5) Answer display and user feedback

[0119] The user interface module displays the generated answers to the user and provides a feedback mechanism: Answer display: The generated answers are displayed intuitively, and the user can view detailed information through the interface. Feedback collection: Users can evaluate and provide feedback on the answers, and the system records user feedback information. Interactive feedback guidance: Through intelligent prompts and guidance, help users provide more effective feedback.

[0120] Friendly user interface design, combined with user experience design principles, enhances user interactivity and satisfaction. Interactive feedback guidance function improves user feedback quality and system optimization efficiency.

[0121] 6) Learning and Optimization

[0122] The learning optimization module performs the following operations based on user feedback and interaction data: Feedback analysis: Analyze user feedback to identify the pros and cons of answers and changes in user needs. Model optimization: Use machine learning technology to adjust and optimize large models to improve the overall performance of the system. Adaptive learning mechanism: Automatically adjust learning strategies and parameters based on user feedback to improve the system's self-optimization capabilities.

[0123] The feedback closed-loop mechanism enables system self-learning and optimization to ensure continuous improvement of system performance. The adaptive learning mechanism dynamically adjusts model parameters and strategies to improve system adaptability and response speed.

[0124] 7) Permission Control

[0125] The permission control module ensures system security and document access control: Authentication: Authenticate users and confirm user access rights. Permission check: Check user permissions during retrieval and answer generation to ensure the security of sensitive information. Hierarchical permission management: Dynamically adjust information access rights based on user roles and permission levels to ensure information security and compliance.

[0126] Multi-factor authentication technology improves system security. The hierarchical permission management function dynamically adjusts information access rights according to different user permission levels to ensure system flexibility and security.

[0127] The present invention has the following beneficial effects:

[0128] 1) Improve operation and maintenance efficiency and reduce fault recovery time

[0129] Causal reasoning: Traditional document retrieval and problem-solving systems have shortcomings such as untimely information acquisition, insufficient accuracy, and lack of intelligent support, which causes operation and maintenance personnel to spend a lot of time and energy in the process of troubleshooting and problem solving.

[0130] By introducing big models and RAG technology, the intelligent operation and maintenance assistant system can retrieve relevant documents and solutions in real time and provide accurate and timely answers in natural language. Operation and maintenance personnel do not need to spend a lot of time manually searching and analyzing information, and can quickly obtain the required solutions, improve operation and maintenance efficiency, and shorten fault recovery time.

[0131] 2) Improve the accuracy and reliability of problem solving

[0132] Causal reasoning: The existing system lacks accuracy and cannot understand the user's query intent, resulting in inaccurate retrieval results and affecting the effectiveness of problem solving.

[0133] The intelligent operation and maintenance assistant system uses large models and RAG technology to accurately understand the user's query intent and provide personalized and context-related support. The documents and solutions retrieved by the system are more accurate and reliable, helping operation and maintenance personnel solve problems more quickly and improving the stability and reliability of the system.

[0134] 3) Improve user experience and satisfaction

[0135] Causal reasoning: Existing systems lack intelligent support, resulting in poor user experience and failure to provide personalized and context-related support, which affects user satisfaction.

[0136] The intelligent operation and maintenance assistant system provides a personalized and smooth user experience through a friendly user interface design and adaptive learning mechanism. The system can understand the user's needs based on the user's historical query and interaction records, provide users with more accurate and considerate services, and improve user satisfaction.

[0137] 4) Continuously optimize system performance and improve intelligence level

[0138] Causal reasoning: Existing systems lack adaptive learning and optimization mechanisms and are unable to continuously improve system performance based on user feedback, making it difficult for the system to adapt to changing environments and user needs.

[0139] The intelligent operation and maintenance assistant system can continuously optimize system performance based on user feedback through feedback learning modules and adaptive learning mechanisms. The system continuously learns user preferences and behaviors, automatically adjusts learning strategies and parameters, improves the intelligence level and adaptability of the system, and ensures that the system can be continuously optimized and improved.

[0140] The embodiments of the present invention can improve operation and maintenance efficiency, enhance the accuracy and reliability of problem solving, enhance user experience and satisfaction, and continuously optimize system performance, bringing significant beneficial effects, and provide a new solution for the operation and maintenance of software-defined storage systems.

[0141] Example 5

[0142] In addition, an embodiment of the present invention also provides a RAG-based intelligent customer support device, which includes: a processor and a memory; the memory is used to store one or more program instructions; the processor is used to run one or more program instructions to execute the steps of a RAG-based intelligent customer support method as described in any of the above items.

[0143] Example 6

[0144] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a RAG-based intelligent customer support method as described in any one of the above items are implemented.

[0145] In the embodiment of the present invention, the processor may be an integrated circuit chip having the signal processing capability. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0146] The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and completes the steps of the above method in combination with its hardware.

[0147] The storage medium may be a memory, which may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0148] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0149] The volatile memory may be a random access memory (RAM) which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced synchronous DRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus DRAM (DRRAM).

[0150] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0151] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by general or special-purpose computers.

[0152] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. An intelligent customer support method based on RAG, characterized in that: The method comprises: S1. Receive a query statement input by a user, and use a preset RAG model to perform semantic recognition on the query statement to obtain the user's query intention; S2, analyzing the query statement, understanding the deep semantics of the query statement and converting the deep semantics into computer instructions; S3, obtaining the user's historical query records and interaction records, and using the historical query records and the interaction records to process the query statement to generate a high-precision query intent; S4, searching and acquiring relevant information in a preset database according to the user query intention and the high-precision query intention combined with computer instructions; S5. Generate a first natural language answer based on the relevant information and optimize the first natural language answer to generate a second natural language answer.

2. The RAG-based intelligent customer support method according to claim 1, characterized in that: Receive a query statement input by a user, use a preset RAG model to perform semantic recognition on the query statement, and obtain the user's query intention, including: The semantic recognition is natural language processing analysis; The query intent includes: adding, nodes, and storage clusters.

3. The RAG-based intelligent customer support method according to claim 2, characterized in that: Analyzing the query statement, understanding the deep semantics of the query statement and converting the deep semantics into computer instructions, including: The natural language processing analysis includes: word segmentation, part-of-speech tagging and intent recognition; Among them, word segmentation and part-of-speech tagging are performed by using a pre-trained natural language processing model, and intent recognition is performed based on a deep learning model.

4. The RAG-based intelligent customer support method according to claim 3, characterized in that: According to the user query intention and the high-precision query intention combined with the computer instruction, the relevant information is retrieved from the preset database, including: Use ElasticSearch to perform a full-text search in a preset database to obtain the first search result; The first search result is analyzed by the BM25 algorithm to obtain high-precision relevant information.

5. The RAG-based intelligent customer support method according to claim 4, characterized in that: Generating a first natural language answer based on the relevant information and optimizing the first natural language answer to generate a second natural language answer includes: Generate a first natural language answer using the preset RAG model and the relevant information; The first natural language answer is optimized by the optimization algorithm in the preset RAG model to generate a second natural language answer.

6. The RAG-based intelligent customer support method according to claim 5, characterized in that: The method further comprises: Obtaining and analyzing user feedback on the second natural language answer to identify the pros and cons of the answer and changes in user needs; The feedback data is used to adjust the strategy of the preset RAG model and optimize the parameters of the preset RAG model.

7. An intelligent customer support system based on RAG, characterized in that: The system comprises: A query processing module is used to receive a query statement input by a user, perform semantic recognition on the query statement using a preset RAG model, and obtain the user's query intention; The deep semantic analysis module is used to extract the core concepts and context of the query statement and generate a deep semantic representation of the query statement to support the precise retrieval and intent analysis of subsequent modules; High-precision query intent analysis module, which is used to optimize the intent analysis of the current query statement based on historical interaction data, and generate more accurate query intent by combining context and historical information to improve the relevance of subsequent retrieval; A retrieval module, used to retrieve relevant information in a preset database according to the user query intention and the high-precision query intention combined with computer instructions; The answer optimization module is used to generate a first natural language answer based on the relevant information and optimize the first natural language answer to generate a second natural language answer.

8. The RAG-based intelligent customer support system as claimed in claim 7, characterized in that: The system further comprises: a learning optimization module, configured to perform the following steps: Obtaining and analyzing user feedback on the second natural language answer to identify the pros and cons of the answer and changes in user needs; The feedback data is used to adjust the strategy of the preset RAG model and optimize the parameters of the preset RAG model.

9. An intelligent customer support device based on RAG, characterized in that: The device comprises: a processor and a memory; The memory is used to store one or more program instructions; The processor is used to run one or more program instructions to execute the steps of the RAG-based intelligent customer support method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the RAG-based intelligent customer support method as claimed in any one of claims 1 to 6 are implemented.

Images