Chat content processing method based on multilevel classification and relevance memory

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a chat content processing method based on multi-level classification and relevance memory, which realizes accurate management and efficient utilization of chat content through the steps of structured processing of dialogue content, multi-level classification, relevance memory extraction, sorting display and the like, and improves the chat content processing efficiency. The problem that the related chat content cannot be accurately extracted in the prior art is effectively solved, and the service quality and the user experience of the intelligent chat system are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a chat content processing method based on multi-level classification and associative memory. Background Art

[0002] With the development of artificial intelligence (AI) technology, intelligent chatbots have been widely used in various fields, such as online customer service, health consultation, and personalized recommendations. However, existing intelligent chatbot systems face technical bottlenecks when processing long-term conversations. In particular, during extended interactions between users and bots, traditional approaches typically rely on only the most recent chat logs to maintain conversation context. This results in a failure to fully utilize important information contained in past conversations, hindering the intelligent chatbot's ability to accurately understand and respond to user needs.

[0003] To address these issues in traditional intelligent chat systems, existing techniques currently process chat content by extracting conversation semantics to update users' long-term memories. However, this approach still has certain limitations, particularly its inability to accurately extract relevant chat content.

[0004] It can be seen that when the existing technology uses long-term memory to process the chat content of the intelligent chat system, there is a problem of being unable to accurately extract relevant chat content. Summary of the Invention

[0005] The purpose of this application is to provide a chat content processing method based on multi-level classification and associative memory, so as to solve the problem in the prior art that when using long-term memory to process the chat content of the intelligent chat system, relevant chat content cannot be accurately extracted.

[0006] In order to achieve the above objectives, the technical solutions adopted in this application are as follows: According to one aspect of an embodiment of the present application, a chat content processing method based on multi-level classification and associative memory is provided, including: receiving conversation content input by a user, and converting the conversation content into structured text data; using a large model to perform multi-level classification on the text data according to a preset classification standard to obtain a classification label, wherein the classification label is a secondary classification label including a main category label and a subcategory label, and the classification standard is used to define multiple main category labels and subcategory labels under each main category label; using the large model to extract associative memory data from a vector library according to the classification label and a preset association strength type, wherein the association strength type is used to indicate the semantic association strength between the conversation content and the classification label, and the vector library is dynamically updated based on the text vector corresponding to the text data using the large model; sorting the associative memory data according to the semantic association strength indicated by the association strength type, and displaying the sorted associative memory data to the user through a visual interface, wherein data with high association strength in the associative memory data is displayed first.

[0007] According to the above technical means, the processing format of unstructured conversation content is unified through structured text conversion, providing a standardized data foundation for subsequent analysis; the main-subcategory two-level classification system is adopted to realize the refined hierarchical organization of text data, which not only ensures the logical rigor of the classification framework, but also enhances the scenario coverage capability through subcategory expansion; based on the dynamically updated vector library and association strength model, the intelligent chat system can capture semantic association features in real time and quantify the relevance level of memory data; finally, through the sorting and display mechanism driven by association strength, it ensures that high-value information is presented first, thereby improving the efficiency of information retrieval while achieving the accuracy, timeliness and user orientation of conversation content processing through hierarchical classification and dynamic memory association technology.

[0008] Furthermore, the large model is used to perform multi-level classification on the text data according to the preset classification standards to obtain classification labels, including: using the large model to perform a first classification on the text data according to the main classification standards in the classification standards to obtain multiple main category labels, wherein the main classification standards are used to define the main category labels; using the large model to perform a second classification on the text data according to the sub-classification standards in the classification standards to obtain multiple sub-category labels, wherein the sub-classification standards are used to define sub-category labels under each main category label.

[0009] Based on the above technical means, the first classification using the main classification criteria can quickly classify text data into predefined main category labels, thereby establishing a clear information hierarchy framework and providing a structured foundation for subsequent analysis and processing. On this basis, the second classification using the sub-classification criteria further refines the information granularity within each main category, making the classification results more accurate and tailored to specific scenario needs. This two-stage, multi-level classification approach not only improves the logical rigor and scenario coverage of the classification, but also reduces the complexity of a single classification through layered processing, thereby improving overall processing efficiency and classification accuracy.

[0010] Furthermore, the main classification criteria include but are not limited to the following category tags: user basic information, health status, living habits, mental health, family member status, and others.

[0011] Based on the above technical means, by presetting the main classification standards covering core dimensions such as user basic information, health records, and living habits, a full-dimensional structured analysis of the conversation content is achieved. This classification framework not only improves the logical rigor of information organization, but also provides an expandable semantic navigation path for subsequent sub-category segmentation and cross-domain associated memory retrieval, thereby significantly enhancing the intelligent chat system's cognitive granularity and service professionalism for complex scenarios while ensuring data processing efficiency.

[0012] Furthermore, the association strength types include direct association, indirect association, and potential association. Direct association indicates that there is a direct association between the conversation content and the classification label, indirect association indicates that there is an indirect connection between the conversation content and the classification label, and potential association indicates that there is a potential connection between the conversation content and the classification label. The order of the three types of semantic association strength in the association strength type from high to low is direct association, indirect association, and potential association.

[0013] Based on the above technical means, by introducing three types of semantic association strength: direct association, indirect association and potential association, the intelligent chat system has built a multi-level association memory system. Direct association ensures the accurate matching of core relevant information, indirect association expands the boundaries of semantic understanding to capture implicit connections, and potential association explores the demand associations that are not clearly expressed through deep semantic analysis. This hierarchical association mechanism not only makes information retrieval present a gradient feature from precision to exploration, but also realizes the dynamic adaptation of user focus and the knowledge reserve of the intelligent chat system through association strength sorting. While ensuring the priority display of core information, it provides a progressive information discovery path for complex dialogue scenarios, thereby significantly improving the semantic understanding depth and personalized service capabilities of the intelligent chat system.

[0014] Furthermore, after obtaining the classification label, the method also includes: using the large model to dynamically update the vector library based on the text vector corresponding to the text data, wherein the step of dynamically updating the vector library includes: using the large model to vectorize the text data to obtain the corresponding text vector; storing the text vector and the classification label in the vector library.

[0015] Using these technical approaches, the intelligent chat system dynamically updates its vector library, capturing and storing the text vectors and classification labels for each conversation in real time. This ensures that the vector library always reflects the latest conversation context and user characteristics. This mechanism not only enhances the intelligent chat system's real-time responsiveness but also continuously accumulates personalized user information, providing a data foundation for deep semantic understanding and precise memory retrieval, thereby enhancing the personalization and intelligence of conversational services.

[0016] Furthermore, the text vector and the classification label are stored in the vector library, including: performing similarity matching between the text vector and all text vectors in the vector library to determine whether there is a target vector associated with the text vector in the vector library; if it is determined that there is no target vector associated with the text vector in the vector library, a summary of the text data is generated through a large model, and the summary, classification label, and text vector are stored as new records in the vector library; if it is determined that there is a target vector associated with the text vector in the vector library, the summary is updated based on the historical summary and text data corresponding to the target vector through the large model, and the classification label and text vector are stored as new records in the vector library.

[0017] According to the above technical means, dynamic deduplication and incremental updates of the vector library are achieved through a similarity matching mechanism, which not only avoids the accumulation of redundant data, but also extracts the core semantics of the conversation through summary generation technology. At the same time, a historical summary fusion update strategy is adopted when related vectors are detected to ensure that the vector library always remains streamlined and efficient and contains the latest conversation context. This dual-mode storage mechanism not only optimizes the storage efficiency of the knowledge base, but also provides structured and time-series data support for subsequent precise memory retrieval based on the strength of semantic association through the association and binding of classification labels and summaries. This improves the response speed of the intelligent chat system while enhancing the consistency of conversation history management and personalized service capabilities.

[0018] Furthermore, based on similarity matching between the text vector and all text vectors in the vector library, it is determined whether there is a target vector associated with the text vector in the vector library, including: calculating the similarity between the text vector and each text vector in the vector library using a preset similarity calculation method; if the similarity is greater than or equal to a preset similarity threshold, it is determined that there is a target vector associated with the text vector in the vector library, and the text vector with the highest similarity is selected as the target vector; otherwise, it is determined that there is no target vector associated with the text vector in the vector library.

[0019] Based on the above technical means, through similarity calculation and threshold comparison mechanism, the intelligent chat system can accurately identify the repetition or similarity of conversation content, thereby avoiding the accumulation of redundant data in the vector library and effectively optimizing storage resources; at the same time, this mechanism ensures that new record creation is triggered only when the conversation content is sufficiently novel, and when highly similar content is detected, the continuity update of historical semantics is achieved through target vector association. This dynamic balance strategy not only maintains the simplicity of the knowledge base, but also ensures the priority reuse of the most relevant historical context through similarity sorting, thereby improving the operating efficiency of the intelligent chat system while enhancing the coherence and semantic consistency of conversation memory management.

[0020] According to another aspect of an embodiment of the present application, a chat content processing device based on multi-level classification and associative memory is provided, comprising: The acquisition module is used to receive the conversation content input by the user and convert the conversation content into structured text data; The classification module is used to use the large model to perform multi-level classification on the text data according to the preset classification criteria to obtain classification labels, where the classification labels are secondary classification labels including main category labels and subcategory labels. The classification criteria are used to define multiple main category labels and subcategory labels under each main category label. A query module is used to extract associated memory data from the vector library using the large model based on the classification label and a preset association strength type, where the association strength type is used to indicate the semantic association strength between the conversation content and the classification label. The vector library is dynamically updated based on the text vector corresponding to the text data using the large model; The display module is used to sort the associated memory data according to the semantic association strength indicated by the association strength type, and display the sorted associated memory data to the user through a visual interface, wherein data with high association strength in the associated memory data are displayed first.

[0021] According to another aspect of the embodiments of the present application, an electronic device is also provided, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; wherein the memory is used to store computer programs; and the processor is used to execute the steps of the chat content processing method based on multi-level classification and associative memory in any of the above embodiments by running the computer program stored in the memory.

[0022] According to another aspect of the embodiments of the present application, a computer-readable storage medium is also provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of the chat content processing method based on multi-level classification and associative memory in any of the above embodiments when running.

[0023] Beneficial effects of this application: This application achieves accurate management and efficient utilization of chat content through steps such as structured processing of conversation content, multi-level classification, associative memory extraction and sorting display, effectively solving the problem of the inability to accurately extract relevant chat content in existing technologies, and significantly improving the service quality and user experience of the intelligent chat system. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0026] Figure 1 This is a schematic diagram of the hardware environment of an optional chat content processing method based on multi-level classification and associative memory provided in an embodiment of the present application; Figure 2 This is a flowchart of an optional chat content processing method based on multi-level classification and associative memory provided in an embodiment of the present application; Figure 3 This is a structural block diagram of an optional chat content processing device based on multi-level classification and associative memory provided in an embodiment of the present application; Figure 4 This is a structural block diagram of an optional electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] According to one aspect of an embodiment of the present application, a chat content processing method based on multi-level classification and associative memory is provided. Optionally, in this embodiment, the chat content processing method based on multi-level classification and associative memory can be applied to a hardware environment consisting of a terminal and a server. The server is connected to the terminal via a network and can be used to provide services to the terminal or a client installed on the terminal. A database can be set up on the server or independently of the server to provide data storage services for the server.

[0030] The aforementioned network may include, but is not limited to, at least one of the following: a wired network or a wireless network. The aforementioned wired network may include, but is not limited to, at least one of the following: a wide area network, a metropolitan area network, or a local area network. The aforementioned wireless network may include, but is not limited to, at least one of the following: Wi-Fi (Wireless Fidelity) or Bluetooth. The terminal may be, but is not limited to, a PC, a mobile phone, or a tablet computer.

[0031] The chat content processing method based on multi-level classification and associative memory of the embodiment of the present application can be executed by a server, a terminal, or a server and a terminal. The execution of the chat content processing method based on multi-level classification and associative memory of the embodiment of the present application by the terminal can also be executed by a client installed on the terminal.

[0032] Taking the chat content processing method based on multi-level classification and associative memory of this embodiment as an example, please refer to Figure 1 , Figure 1 This is a schematic diagram of the hardware environment of an optional chat content processing method based on multi-level classification and associative memory provided in an embodiment of the present application, such as Figure 1As shown, the hardware environment of the chat content processing method based on multi-level classification and associative memory includes: a terminal 102, and a server 104 connected to the terminal 102 via a network. The server 104 is used to deploy a large model, which executes the chat content processing method based on multi-level classification and associative memory according to an embodiment of the present application, performing multi-level classification and associative memory processing on the chat content; the terminal 102 is used to display the processed associative memory data, wherein the associative memory data can be obtained by classification, association, and sorting by the large model deployed by the server 104.

[0033] The chat content processing method based on multi-level classification and associative memory in this embodiment can be applied to scenarios such as intelligent customer service, intelligent assistants, and psychological counseling. For example, in the intelligent customer service scenario, multi-level classification and associative memory processing can more accurately understand user needs and provide personalized services; in the psychological counseling scenario, associative memory processing can tap into users' potential emotional needs and provide more intimate psychological support. In this embodiment of the application, the chat content processing method based on multi-level classification and associative memory is illustrated using the intelligent customer service scenario as an example.

[0034] With the development of artificial intelligence technology, intelligent chatbots have been widely used in the field of intelligent customer service. However, when processing long-term conversation content, existing intelligent chat systems update users' long-term memory by extracting conversation semantics. This method cannot accurately extract relevant chat content.

[0035] To solve the above problems, this embodiment provides a chat content processing method based on multi-level classification and associative memory running on the above server, please refer to Figure 2 , Figure 2 This is a flow chart of an optional method for processing chat content based on multi-level classification and associative memory provided by an embodiment of the present application. Figure 2 As shown, the chat content processing method based on multi-level classification and associative memory in the embodiment of the present application specifically includes the following steps: Step S201, receiving the conversation content input by the user and converting the conversation content into structured text data; Step S202: Using the large model, the text data is multi-classified according to a preset classification standard to obtain classification labels, wherein the classification labels are secondary classification labels including a main category label and a subcategory label. The classification standard is used to define multiple main category labels and subcategory labels under each main category label. Step S203: Using the large model, the associated memory data is extracted from the vector library based on the classification label and a preset association strength type, where the association strength type is used to indicate the semantic association strength between the conversation content and the classification label. The vector library is dynamically updated based on the text vector corresponding to the text data using the large model. Step S204 , sorting the associated memory data according to the semantic association strength indicated by the association strength type, and displaying the sorted associated memory data to the user through a visual interface, wherein data with high association strength in the associated memory data is displayed first.

[0036] Through the above steps S201 to S204, by means of structured processing of conversation content, multi-level classification, associative memory extraction and sorting display, etc., accurate management and efficient utilization of chat content are achieved, effectively solving the problem of the inability to accurately extract relevant chat content in the existing technology, and significantly improving the service quality and user experience of the intelligent chat system.

[0037] The following combination Figure 2 The chat content processing method based on multi-level classification and associative memory in the embodiment of the present application is explained.

[0038] In the technical solution of step S201, the conversation content input by the user is received and converted into structured text data.

[0039] During a conversation between an intelligent chatbot and a user, the user first asks a question or engages in conversation. The conversation content refers to the user's input, which can include any form of text, voice, or other multimodal information. The intelligent chat system converts the conversation content into structured text data for further analysis. To enhance the intelligent chat system's understanding capabilities, it not only performs basic lexical analysis of the text content but also uses large models (such as ChatGPT and DeepSeek) to analyze the user's underlying intent and emotional tone, laying a solid foundation for subsequent classification and memory storage.

[0040] In intelligent chat scenarios, receiving user input and converting it into structured text data is a critical first step in providing accurate services. This step is not simply text entry, but includes a series of intelligent processing steps aimed at converting natural language into structured information that can be understood and analyzed by machines.

[0041] Specifically, when a user enters conversation content through the intelligent chat system, the intelligent chat system will first receive this information. Next, the intelligent chat system will structure the received conversation content, converting the unstructured natural language text into a structured data format. It is understandable that structured data has a clear data type and format, which facilitates efficient computer processing and analysis. In this embodiment, structured processing is achieved through large models such as chatGPT and deepseek. Its structured processing generally includes the following aspects: Entity Recognition and Extraction: The intelligent chat system uses natural language processing technology to identify and extract key entities in conversations, such as names of people, places, times, and product names. For example, when a user says, "I want to check tomorrow's weather in City A," the intelligent chat system can recognize that "City A" is a place name and "tomorrow" is a time.

[0042] Intent Recognition: The intelligent chat system analyzes the semantics of the conversation to understand the user's true intent. For example, if a user says, "I've been suffering from insomnia lately," the intelligent chat system can recognize that the user's intention is to seek advice on solving insomnia.

[0043] Sentiment Analysis: The intelligent chat system determines the emotional tendency of the conversation content, such as positive, negative, or neutral. This helps the intelligent customer service better understand the user's emotional state and provide more attentive service.

[0044] In this embodiment, the structured data is stored in a JSON structure. It is understood that using JSON format to store structured data can significantly improve data processing efficiency, ensure data consistency, achieve efficient query retrieval, and thus improve classification accuracy. It also supports the application of complex classification models, bringing great convenience and advantages to data processing and analysis.

[0045] It's understandable that through structured processing, raw conversation content is transformed into structured text data containing rich information. This data is not only easier for computers to understand and process, but also provides a solid foundation for subsequent multi-level classification, associative memory processing, and personalized services. In intelligent customer service scenarios, structured text data is the cornerstone of the intelligent chat system, enabling it to more accurately understand user needs and provide more efficient and personalized services.

[0046] In the technical solution of step S202, the large model is used to perform multi-level classification on the text data according to the preset classification standard to obtain classification labels.

[0047] To achieve multi-level classification of text data and generate secondary classification labels consisting of primary and subcategory labels, we leveraged the powerful semantic understanding and intelligent classification capabilities of the large model and strictly adhered to pre-defined classification standards. Classification labels are secondary classification labels consisting of primary and subcategory labels. The classification standards define multiple primary and subcategory labels within each primary category label.

[0048] In this embodiment, the classification criteria are predefined, hierarchical classification criteria, which are divided into a primary classification criteria and a sub-classification criteria. The primary classification criteria are used to define a primary category label, which includes multiple primary category labels; the sub-classification criteria are used to define sub-category labels under each primary category label, which includes multiple sub-category labels. The primary category label is the highest level of classification, which summarizes the most core category attributes of the text data. The sub-category label is a further subdivision of the primary category label, used to more accurately describe the specific content of the text data.

[0049] As an optional embodiment, the main classification criteria include but are not limited to the following category tags: user basic information, health status, living habits, mental health, family member status, and others.

[0050] In practice, each main classification standard is further divided into multiple sub-category labels, and all sub-category labels constitute sub-classification standards, for example: User basic information is used to describe the basic attributes of the user. Its subcategory tags may include: name, age, gender, etc.

[0051] Health status is used to describe the user's health-related information. Its subcategory tags may include: physical examination status, disease history, medication history, etc.

[0052] Lifestyle habits are used to describe the user's lifestyle. Its subcategory tags may include: exercise habits, eating habits, sleeping habits, entertainment habits, etc.

[0053] Mental health is used to describe the user's mental state. Its subcategory tags may include: emotional state, stress level, etc.

[0054] Family member status is used to describe the user's family member information. Its subcategory tags may include: family member relationship, family member health status, etc.

[0055] In actual applications, each subcategory can be further subdivided according to needs. For example, medical history can be subdivided into chronic disease history, acute disease history, etc.

[0056] It is understandable that by presetting the main classification standards covering core dimensions such as user basic information, health records, and living habits, a full-dimensional structured analysis of the conversation content is achieved. This classification framework not only improves the logical rigor of information organization, but also provides an expandable semantic navigation path for subsequent sub-category segmentation and cross-domain associated memory retrieval, thereby significantly enhancing the intelligent chat system's cognitive granularity and service professionalism for complex scenarios while ensuring data processing efficiency.

[0057] As an optional embodiment, a large model is used to perform multi-level classification on text data according to preset classification standards to obtain classification labels, including: using the large model to perform a first classification on the text data according to the main classification standard in the classification standard to obtain multiple main category labels, wherein the main classification standard is used to define the main category label; using the large model to perform a second classification on the text data according to the sub-classification standard in the classification standard to obtain multiple sub-category labels, wherein the sub-classification standard is used to define the sub-category label under each main category label.

[0058] After receiving text data, the large model will deeply understand the semantics and context of the text, and intelligently classify the text data into the most appropriate main category label and subcategory label. The large model classification process is as follows: The large model first conducts an in-depth semantic understanding of the text data, grasping the core meaning and context of the text. For example, for the text "I feel more like playing badminton," the large model can understand that "playing badminton" reflects the user's lifestyle habits.

[0059] The large model then categorizes the text data into corresponding main categories based on its understanding of the text's semantics and pre-set classification criteria. For example, it might classify "playing badminton" into the "lifestyle habits" main category.

[0060] Furthermore, the large model will classify text data into more detailed subcategories based on the complex information in the recognized context, achieving multi-level fine-grained classification. For example, "playing badminton" can be classified under the "exercise habits" subcategory.

[0061] The above classified content is displayed in JSON format as follows: [{'id': '271fe04c-f262-4ac9-9e20-d488302e3a56', 'memory': 'I feel like playing badminton more', 'hash': '6facbeb0b6a40fc6e822aca174807f9c', 'metadata': {'business_line': 'Lifestyle', 'main_category': 'Exercise Habits', 'sub_category': 'Exercise Methods'}, 'created_at': '2025-01-15T19:52:15.328509-08:00', 'updated_at':None, 'user_id': '53454', 'agent_id': 'MasterAgent'}] Among them, memory is used to record text data, and metadata is used to record classification labels.

[0062] In this embodiment, the large model has a certain degree of adaptive ability and can flexibly adjust the classification method according to the dynamic changes in the conversation content to ensure the real-time and accuracy of the classification.

[0063] It is understandable that the first classification through the main classification criteria can quickly classify the text data into predefined main category labels, thereby establishing a clear information hierarchy framework and providing a structured basis for subsequent analysis and processing; on this basis, the second classification is carried out through the sub-classification criteria to further refine the information granularity under each main category, making the classification results more accurate and in line with the needs of specific scenarios. This two-stage multi-level classification method not only improves the logical rigor and scene coverage of the classification, but also reduces the complexity of a single classification through hierarchical processing, thereby improving the overall processing efficiency and classification accuracy. In the technical solution of step S203, a large model is used to extract associated memory data from the vector library according to the classification label and the preset association strength type, wherein the association strength type is used to indicate the semantic association strength between the conversation content and the classification label, and the vector library is dynamically updated based on the text vector corresponding to the text data using the large model.

[0064] The vector library is a database that stores text vectors and classification labels. It uses vectorized storage to enable computers to efficiently calculate similarity and quickly retrieve data.

[0065] To accurately extract the information users need from vast amounts of long-term memory, we leverage a large model, combined with classification labels and pre-defined association strength types, to efficiently retrieve associated memory data from a dynamically updated vector library. To more accurately understand user needs and extract the most relevant memory data from the vector library, we pre-defined association strength types. These types indicate the strength of the semantic connection between the conversation content and the classification labels, measuring the relevance of the memory data to the user's needs.

[0066] As an optional embodiment, the association strength types include: direct association, indirect association and potential association. Among them, direct association indicates that there is a direct association between the conversation content and the classification label, indirect association indicates that there is an indirect connection between the conversation content and the classification label, and potential association indicates that there is a potential connection between the conversation content and the classification label. The order of the semantic association strength of the three types in the association strength type from high to low is: direct association, indirect association, potential association.

[0067] A direct association indicates the most direct and obvious semantic connection between the memory data and the user's query intent. This type of memory data typically directly answers the user's question or is highly relevant to the user's question. For example, when a user asks about "treatments for high blood pressure," the directly associated memory data is "drug treatment options for high blood pressure," which the user has previously inquired about.

[0068] Indirect associations indicate a semantic connection between the memory data and the user's query intent, but this connection isn't immediately obvious and may require logical reasoning or background knowledge to establish. This type of memory data may provide users with background information, related cases, or expanded knowledge, helping them understand the question more comprehensively. For example, when a user asks about "treatments for high blood pressure," the indirectly associated memory data may be the user's previously learned information about "dietary restrictions for high blood pressure."

[0069] Latent associations indicate potential, yet not immediately apparent, semantic connections between memory data and the user's query intent. This type of memory data may appear unrelated to the user's question, but through in-depth analysis and exploration, a potential connection may be discovered, providing the user with unexpected insights or assistance. For example, when a user inquires about "treatments for high blood pressure," potentially related memory data may include information on "stress management and mental health" that the user has previously consulted, as stress management can also assist in controlling high blood pressure.

[0070] The current text data's corresponding association strength type is to match the corresponding association relationship according to the classification. This association relationship can be manually determined or automatically matched by a large model.

[0071] When a user asks a query, the large model combines the classification label and the preset association strength type to accurately extract the associated memory data from the dynamically updated vector library. The process of extracting the associated memory data by the large model is as follows: The large model first conducts a deep semantic understanding of the user's query, accurately grasping the user's true intention and needs. For example, when a user asks, "I've been suffering from insomnia lately. What should I do?" the large model will understand that the user's core need is to find a solution to their insomnia problem.

[0072] Based on its understanding of the user's query intent, the large model determines the long-term memory category to be queried, also known as the classification label. For example, for the question "insomnia," the large model might match classification labels such as "mental health" or "lifestyle > sleep habits."

[0073] The large model determines the strength of the association between the memory data stored in the vector library and the user's query intent based on a preset association strength type. For example, for the question "insomnia," the large model will comprehensively consider factors such as semantic similarity, logical relationships, and background knowledge to extract content such as "mental health" or "lifestyle habits > sleep habits."

[0074] It is understandable that by introducing three types of semantic association strength: direct association, indirect association and potential association, the intelligent chat system has built a multi-level association memory system. Direct association ensures the accurate matching of core relevant information, indirect association expands the boundaries of semantic understanding to capture implicit connections, and potential association uses deep semantic analysis to explore demand associations that are not clearly expressed. This hierarchical association mechanism not only makes information retrieval present a gradient feature from precision to exploration, but also realizes the dynamic adaptation of user focus and the knowledge reserve of the intelligent chat system through association strength sorting. While ensuring the priority display of core information, it provides a progressive information discovery path for complex dialogue scenarios, thereby significantly improving the semantic understanding depth and personalized service capabilities of the intelligent chat system.

[0075] In the technical solution of step S204, the associated memory data is sorted according to the semantic association strength indicated by the association strength type, and the sorted associated memory data is displayed to the user through a visual interface, wherein data with high association strength in the associated memory data is displayed first.

[0076] After extracting the associated memory data related to the user's query intent, in order to further improve the user experience and ensure that users can efficiently obtain the required information, these associated memory data will be sorted and intuitively displayed to the user through a visual interface. The entire associated memory data sorting and visualization process is as follows: The associative memory data extracted from the vector library is prioritized according to the strength of the association, ensuring that data with high association strength can be displayed to users first.

[0077] Using a large model, the association strength of each extracted piece of associated memory data is evaluated. This evaluation is based on multiple dimensions, including semantic similarity, logical relevance, and contextual connection. For example, for a user querying "treatment methods for hypertension," a memory piece directly describing "hypertension drug treatment plans" will clearly have a higher association strength than a memory piece describing "healthy dietary recommendations."

[0078] Based on the results of the association strength assessment, the intelligent chat system pre-establishes a set of sorting rules. Generally, directly associated memory data receives the highest sorting priority, followed by indirectly associated memory data, and finally potentially associated memory data. This sorting rule ensures that users first see the information most relevant and direct to their query intent. It should be noted that the sorting rules are not static. The intelligent chat system will dynamically adjust the sorting rules based on specific query scenarios and user needs. For example, in certain complex query scenarios, indirectly or potentially associated memory data may be more valuable to the user, and the intelligent chat system will accordingly increase the sorting priority of this memory data.

[0079] The sorted associated memory data must ultimately be presented to the user through an intuitive, easy-to-use visual interface. This visual interface acts as an "information visualization window," transforming abstract textual information into an intuitive visual presentation, helping users more easily understand and digest the information. The specific implementation and design details of the visual interface are based on general technical implementation and user experience principles and are not specifically limited in this embodiment.

[0080] It is understandable that the processing format of unstructured conversation content is unified through structured text conversion, providing a standardized data basis for subsequent analysis; the main-subcategory two-level classification system is adopted to realize the refined hierarchical organization of text data, which not only ensures the logical rigor of the classification framework, but also enhances the scene coverage capability through subcategory expansion; based on the dynamically updated vector library and association strength model, the intelligent chat system can capture semantic association features in real time and quantify the relevance level of memory data; finally, through the sorting and display mechanism driven by association strength, it ensures that high-value information is presented first, thereby improving the efficiency of information retrieval while achieving the accuracy, timeliness and user orientation of conversation content processing through hierarchical classification and dynamic memory association technology.

[0081] The vector library is the "memory warehouse" of the entire long-term memory system. It is not static but in dynamic update. Whenever new conversation content is generated, the intelligent chat system uses the large model to vectorize this text data, converting it into a vector form that the computer can efficiently understand and process, and stores it in the vector library. This dynamic update mechanism ensures that the memory data stored in the vector library is always the latest and most relevant. In this embodiment, after obtaining the classification label, the vector library is dynamically updated based on the text vector corresponding to the text data using the large model. As an optional embodiment, after obtaining the classification label, the method further includes: dynamically updating the vector library based on the text vector corresponding to the text data using the large model, where the step of dynamically updating the vector library includes: vectorizing the text data using the large model to obtain the corresponding text vector; storing the text vector and the classification label in the vector library.

[0082] A text vector is a vector form obtained by transforming text data through mathematical transformation. Vectorization enables the computer to understand and process text data. It captures the semantic information of the text, making texts with similar semantics close to each other in the vector space. A classification label is a label used to identify the category to which a text vector belongs. It helps the intelligent chat system manage the text data in a structured manner, enabling quick location of relevant category memory data during retrieval.

[0083] Vectorization is the process of converting text data into numerical vectors that the computer can understand and process. This process is usually completed by the large model, which can capture the semantic information of the text and encode it into vector form. The general steps of vectorization are as follows: Text preprocessing, including operations such as cleaning, tokenizing, and removing stop words from the original text data. Cleaning can remove noise in the text, such as special symbols, emojis, etc.; tokenizing is to split continuous text into meaningful word units; removing stop words is to filter out words that contribute less to semantic understanding, such as "de", "le", "zai", etc. The preprocessed text is more standardized and clean, which is beneficial to improving the quality of vectorization.

[0084] The preprocessed text data is input into the large model. The large model converts the words, phrases, and sentences in the text into high-dimensional numerical vectors. These vectors capture the semantic information of the text, making texts with similar semantics close to each other in the vector space. The vectorized text data can then be efficiently processed and analyzed by the computer.

[0085] Finally, the obtained text vector and the predefined classification label are stored in the vector library together. In this way, a new memory data is dynamically updated in the vector library, preparing for subsequent memory retrieval and application.

[0086] It's understandable that by dynamically updating the vector library, the intelligent chat system can capture and store the text vectors and classification labels of each conversation in real time, ensuring that the vector library always reflects the latest conversation context and user characteristics. This mechanism not only enhances the intelligent chat system's real-time responsiveness, but also provides a data foundation for deep semantic understanding and precise memory retrieval by continuously accumulating personalized user information, thereby enhancing the personalization and intelligence of conversational services.

[0087] As an optional embodiment, the text vector and the classification label are stored in the vector library, including: performing similarity matching between the text vector and all text vectors in the vector library to determine whether there is a target vector associated with the text vector in the vector library; if it is determined that there is no target vector associated with the text vector in the vector library, generating a summary of the text data through a large model, and storing the summary, classification label and text vector as a new record in the vector library; if it is determined that there is a target vector associated with the text vector in the vector library, updating the summary based on the historical summary and text data corresponding to the target vector through the large model, and storing the classification label and text vector as new records in the vector library.

[0088] Storing text vectors and classification labels in the vector library isn't simply a matter of piling up data; it involves a sophisticated process that integrates intelligent matching, information condensation, and dynamic updates. This process ensures that the memory data stored in the vector library is both comprehensive and accurate, and can continuously evolve as information is updated.

[0089] Before storing a new text vector in the vector library, the intelligent chat system first performs a similarity match. The purpose of similarity matching is to determine whether a "target vector" with similar semantics to the current text vector already exists in the vector library.

[0090] Based on the results of similarity matching, the intelligent chat system will adopt different strategies to process text data, one of the core operations of which is summary generation.

[0091] If the target vector associated with the current text vector does not exist in the vector library, it indicates that the current text data contains new information. At this point, the intelligent chat system uses a large model to summarize and refine the original text data, generating a concise summary. This summary captures the core content of the text data while significantly reducing the data size. The intelligent chat system then stores the generated summary, classification label, and text vector as a new record in the vector library.

[0092] If it is determined that a target vector associated with the current text vector exists in the vector library, this indicates that the current text data is highly similar or duplicated with the information already in the vector library, and may contain updates or supplements to the existing information. In this case, the intelligent chat system will not directly store the new summary. Instead, it will use the large model to conduct a comprehensive analysis based on the historical summary corresponding to the target vector and the current text data. The large model will determine whether the current text data is a supplement, correction, or update to the existing information, and will generate an updated summary accordingly. This updated summary will integrate the information of the historical summary and the current text data to ensure the accuracy and timeliness of the summary. The intelligent chat system will then overwrite the historical summary corresponding to the target vector with the updated summary, and at the same time store the classification label and text vector as a new record in the vector library.

[0093] It is understandable that the dynamic deduplication and incremental update of the vector library are achieved through the similarity matching mechanism. This dynamic update mechanism not only avoids the accumulation of redundant data, but also extracts the core semantics of the conversation through summary generation technology. At the same time, when the associated vector is detected, a historical summary fusion update strategy is adopted to ensure that the vector library always remains streamlined and efficient and contains the latest conversation context. This dual-mode storage mechanism not only optimizes the storage efficiency of the knowledge base, but also provides structured and time-series data support for subsequent precise memory retrieval based on the strength of semantic association through the association and binding of classification labels and summaries. This improves the response speed of the intelligent chat system while enhancing the consistency and personalized service capabilities of conversation history management.

[0094] As an optional embodiment, similarity matching is performed between the text vector and all text vectors in the vector library to determine whether there is a target vector associated with the text vector in the vector library, including: calculating the similarity between the text vector and each text vector in the vector library using a preset similarity calculation method; if the similarity is greater than or equal to a preset similarity threshold, determining that there is a target vector associated with the text vector in the vector library, and selecting the text vector with the highest similarity as the target vector; otherwise, determining that there is no target vector associated with the text vector in the vector library.

[0095] The intelligent chat system calculates the similarity between the current text vector and all text vectors in the vector library. Similarity calculation can use common vector similarity measurement methods such as cosine similarity and Euclidean distance.

[0096] After calculating the similarity, the intelligent chat system compares it with a preset similarity threshold. If the similarity is greater than or equal to the threshold, it is considered that the target vector associated with the current text vector exists in the vector library, indicating that the information expressed by the current text vector is highly similar or duplicates the information already in the vector library. If the similarity is less than the threshold, it is considered that the target vector associated with the current text vector does not exist in the vector library, indicating that the current text vector contains new information.

[0097] It can be understood that through the similarity calculation and threshold comparison mechanism, the intelligent chat system can accurately identify the repetitiveness or similarity of the conversation content, thereby avoiding the accumulation of redundant data in the vector library and effectively optimizing storage resources; at the same time, this mechanism ensures that new record creation is triggered only when the conversation content is sufficiently novel, and when highly similar content is detected, the continuity update of historical semantics is achieved through target vector association. This dynamic balance strategy not only maintains the simplicity of the knowledge base, but also ensures the priority reuse of the most relevant historical context through similarity sorting, thereby improving the operating efficiency of the intelligent chat system while enhancing the coherence and semantic consistency of conversation memory management. It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0098] Through the description of the above implementation methods, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus the necessary general hardware platform, or of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM (Read-Only Memory, Read-Only Memory) / RAM (Random Access Memory, Random Access Memory), a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of the various embodiments of the present application.

[0099] According to another aspect of the embodiment of the present application, a chat content processing device based on multi-level classification and associative memory for implementing the above method is also provided. Figure 3 , Figure 3This is a structural block diagram of an optional chat content processing device based on multi-level classification and associative memory provided by an embodiment of the present application, such as Figure 3 As shown, the chat content processing device 300 based on multi-level classification and associative memory may include: The acquisition module 301 is used to receive the conversation content input by the user and convert the conversation content into structured text data; A classification module 302 is configured to use the large model to perform multi-level classification on the text data according to a preset classification standard to obtain a classification label, wherein the classification label is a secondary classification label including a main category label and a subcategory label. The classification standard is used to define multiple main category labels and subcategory labels under each main category label; Query module 303, configured to extract associated memory data from a vector library using the large model based on the classification label and a preset association strength type, wherein the association strength type indicates the strength of the semantic association between the conversation content and the classification label. The vector library is dynamically updated based on the text vector corresponding to the text data using the large model; The display module 304 is used to sort the associated memory data according to the semantic association strength indicated by the association strength type, and display the sorted associated memory data to the user through a visual interface, wherein data with high association strength in the associated memory data are displayed first.

[0100] It should be noted that the acquisition module 301 in this embodiment can be used to execute the above step S201, the classification module 302 in this embodiment can be used to execute the above step S202, the query module 303 in this embodiment can be used to execute the above step S203, and the display module 304 in this embodiment can be used to execute the above step S204.

[0101] Regarding the chat content processing device based on multi-level classification and associative memory in this embodiment, the specific manner in which its acquisition module 301, classification module 302, query module 303 and display module 304 execute the above-mentioned chat content processing method based on multi-level classification and associative memory has been described in detail in the embodiment of the chat content processing method based on multi-level classification and associative memory, and will not be elaborated on here.

[0102] It can be understood that the technical solution provided by this embodiment, the various modules in the chat content processing device based on multi-level classification and associative memory, through structured processing of conversation content, multi-level classification, associative memory extraction and sorting display and other steps, achieve accurate management and efficient utilization of chat content, effectively solve the problem of the inability to accurately extract relevant chat content in the existing technology, and significantly improve the service quality and user experience of the intelligent chat system.

[0103] In addition to the above modules, the device in this embodiment may also include a module for executing any method in any of the aforementioned embodiments of the chat content processing method based on multi-level classification and associative memory.

[0104] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. Figure 1 The hardware environment of the method shown can be implemented through software or hardware, wherein the hardware environment includes a network environment.

[0105] According to another aspect of the embodiment of the present application, an electronic device for implementing the above-mentioned chat content processing method based on multi-level classification and associative memory is also provided. The electronic device can be a server, a terminal, or a combination thereof.

[0106] According to another embodiment of the present application, an electronic device is also provided. Figure 4 , Figure 4 This is a structural block diagram of an optional electronic device provided in an embodiment of the present application, such as Figure 4 As shown, the electronic device may include: a processor 1501 , a communication interface 1502 , a memory 1503 and a communication bus 1504 , wherein the processor 1501 , the communication interface 1502 , and the memory 1503 communicate with each other via the communication bus 1504 .

[0107] Memory 1503, used for storing computer programs; The processor 1501 is configured to execute the program stored in the memory 1503 to implement the following steps: Step S201, receiving the conversation content input by the user and converting the conversation content into structured text data; Step S202: Using the large model, the text data is multi-classified according to a preset classification standard to obtain classification labels, wherein the classification labels are secondary classification labels including a main category label and a subcategory label. The classification standard is used to define multiple main category labels and subcategory labels under each main category label. Step S203: Using the large model, the associated memory data is extracted from the vector library based on the classification label and a preset association strength type, where the association strength type is used to indicate the semantic association strength between the conversation content and the classification label. The vector library is dynamically updated based on the text vector corresponding to the text data using the large model. Step S204 , sorting the associated memory data according to the semantic association strength indicated by the association strength type, and displaying the sorted associated memory data to the user through a visual interface, wherein data with high association strength in the associated memory data is displayed first.

[0108] It can be understood that the technical solution provided in this embodiment, the processor of the electronic device, through steps such as structured processing of conversation content, multi-level classification, associative memory extraction and sorting display, realizes the precise management and efficient utilization of chat content, effectively solves the problem of the inability to accurately extract relevant chat content in the existing technology, and significantly improves the service quality and user experience of the intelligent chat system.

[0109] Alternatively, in this embodiment, the communication bus may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Such a communication bus may be classified as an address bus, a data bus, a control bus, and the like. For ease of illustration, the figure shows only one thick line, but this does not imply that there is only one bus or only one type of bus. The communication interface is used for communication between the electronic device and other devices.

[0110] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. Alternatively, the memory may be at least one storage device located away from the processor.

[0111] The above-mentioned processor can be a general-purpose processor, which can include but is not limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0112] An embodiment of the present application further provides a computer-readable storage medium, the storage medium including a stored program, wherein the method steps of the above method embodiment are executed when the program is run.

[0113] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a ROM, a RAM, a mobile hard disk, a magnetic disk, or an optical disk.

[0114] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0115] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above-mentioned computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods of each embodiment of the present application.

[0116] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is merely a logical functional division. In actual implementation, there may be other division methods, such as combining or integrating multiple units or components into another system, or ignoring or not implementing some features. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interface, indirect coupling or communication connection of units or modules, and may be electrical or other forms.

[0118] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected based on actual needs to achieve the purpose of the solution provided in this embodiment.

[0119] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0120] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A chat content processing method based on multi-level classification and associative memory, applied to an intelligent chat system, characterized in that: include: Receive the conversation content input by the user and convert it into structured text data; Using the large model to perform multi-level classification on the text data according to a preset classification standard to obtain classification labels, wherein the classification labels are secondary classification labels including a main category label and a subcategory label, and the classification standard is used to define multiple main category labels and subcategory labels under each main category label; Extracting associated memory data from a vector library using the large model based on the classification label and a preset association strength type, wherein the association strength type is used to indicate the semantic association strength between the conversation content and the classification label, and the vector library is dynamically updated based on the text vector corresponding to the text data using the large model; The associated memory data are sorted according to the semantic association strength indicated by the association strength type, and the sorted associated memory data are displayed to the user through a visual interface, wherein data with high association strength in the associated memory data are displayed first.

2. The chat content processing method based on multi-level classification and associative memory according to claim 1 is characterized in that: The text data is classified into multiple levels using a large model according to a preset classification standard to obtain classification labels, including: Using the large model, the text data is first classified according to a main classification standard in the classification standard to obtain a plurality of main category labels, wherein the main classification standard is used to define the main category labels; The text data is classified a second time using the large model according to the sub-classification criteria in the classification criteria to obtain a plurality of sub-category labels, wherein the sub-classification criteria are used to define sub-category labels under each main category label.

3. The chat content processing method based on multi-level classification and associative memory according to claim 2 is characterized in that: The main classification criteria include but are not limited to the following category tags: user basic information, health status, living habits, mental health, family member status, and others.

4. The chat content processing method based on multi-level classification and associative memory according to claim 1 is characterized in that: The association strength types include: direct association, indirect association and potential association. Among them, direct association indicates that there is a direct association between the conversation content and the classification label, indirect association indicates that there is an indirect connection between the conversation content and the classification label, and potential association indicates that there is a potential connection between the conversation content and the classification label. The order of the semantic association strength of the three types in the association strength type from high to low is: direct association, indirect association, potential association.

5. The chat content processing method based on multi-level classification and associative memory according to claim 1 is characterized in that: After obtaining the classification label, the method further includes: dynamically updating the vector library based on the text vector corresponding to the text data using the large model, wherein the step of dynamically updating the vector library includes: Vectorizing the text data using the large model to obtain corresponding text vectors; The text vector and the classification label are stored in the vector library.

6. The chat content processing method based on multi-level classification and associative memory according to claim 5 is characterized in that: Storing the text vector and the classification label in the vector library includes: Perform similarity matching between the text vector and all text vectors in the vector library, and determine whether there is a target vector associated with the text vector in the vector library; If it is determined that the target vector associated with the text vector does not exist in the vector library, a summary of the text data is generated by the large model, and the summary, the classification label, and the text vector are stored as a new record in the vector library; If it is determined that there is a target vector associated with the text vector in the vector library, the summary is updated based on the historical summary and text data corresponding to the target vector through the large model, and the classification label and the text vector are stored as new records in the vector library.

7. The chat content processing method based on multi-level classification and associative memory according to claim 6 is characterized in that: Performing similarity matching between the text vector and all text vectors in a vector library to determine whether a target vector associated with the text vector exists in the vector library includes: Calculate the similarity between the text vector and each text vector in the vector library using a preset similarity calculation method; If the similarity is greater than or equal to a preset similarity threshold, it is determined that there is a target vector associated with the text vector in the vector library, and the text vector with the highest similarity is selected as the target vector; otherwise, it is determined that there is no target vector associated with the text vector in the vector library.

8. A chat content processing device based on multi-level classification and associative memory, applied to an intelligent chat system, characterized in that: include: The acquisition module is used to receive the conversation content input by the user and convert the conversation content into structured text data; A classification module is used to use the large model to perform multi-level classification on the text data according to a preset classification standard to obtain a classification label, wherein the classification label is a secondary classification label including a main category label and a subcategory label, and the classification standard is used to define multiple main category labels and subcategory labels under each main category label; a query module configured to extract associated memory data from a vector library using the large model based on the classification label and a preset association strength type, wherein the association strength type indicates the strength of semantic association between the conversation content and the classification label, and the vector library is dynamically updated based on text vectors corresponding to the text data using the large model; A display module is used to sort the associated memory data according to the semantic association strength indicated by the association strength type, and display the sorted associated memory data to the user through a visual interface, wherein data with high association strength in the associated memory data is displayed first.

9. An electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein: The processor, the communication interface and the memory communicate with each other via the communication bus, wherein: The memory is used to store computer programs; The processor is configured to execute the steps of the chat content processing method based on multi-level classification and associative memory according to any one of claims 1 to 7 by running the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the steps of the chat content processing method based on multi-level classification and associative memory described in any one of claims 1 to 7 when running.

Images