Method and device for realizing operation and maintenance service integration based on natural language dialogue, processor and computer readable storage medium thereof

Abstract

The invention relates to a method for realizing operation and maintenance service integration based on natural language dialogue, which comprises the following steps of: constructing a hierarchical intention recognition model containing a plurality of hierarchical branches, covering high-level intentions and subdivision intentions, and understanding the requirements of a user by judging a natural language input by the user layer by layer by utilizing a pre-trained large language model; the advanced intention and the subdivision intention output by the large language model are combined with a keyword matching rule defined in advance, so that the understanding of the user intention, the response precision and the operation execution efficiency are improved; and summarizing and persistently storing each round of dialogues by using a multi-round dialogue mechanism. By the adoption of the method and device for achieving operation and maintenance service integration based on natural language dialogue, the processor and the computer readable storage medium of the processor, operation and maintenance task management is simplified and optimized, and operation and maintenance service integration is achieved by combining a large language model with multi-level intention recognition, keyword matching and a multi-round dialogue mechanism. And efficient and accurate operation and maintenance service integration and intelligent management are realized.

Description

Technical Field

[0001] The present invention relates to the field of operation and maintenance systems, and particularly to the field of operation and maintenance service management methods. Specifically, it refers to a method, device, processor, and its computer-readable storage medium for realizing operation and maintenance service integration based on natural language dialogue. Background Art

[0002] Traditional operation and maintenance service management methods usually face many challenges, including complex tools, information silos, low decision-making efficiency, etc. Existing operation and maintenance systems often rely on multiple independent tools and platforms, resulting in fragmented information. Operation and maintenance personnel need to frequently switch between different systems, increasing the complexity of operations and the risk of errors. In addition, traditional operation and maintenance service management relies on manual operations and manual interventions, making it difficult to meet the requirements of efficient and intelligent operation and maintenance. Therefore, how to improve the integration degree and automation level of operation and maintenance services through technical means has become the key to improving operation and maintenance efficiency and quality.

[0003] To solve the above problems, an operation and maintenance service integration method based on natural language dialogue has emerged. This method realizes the unified management of different operation and maintenance tools and services by integrating large language models and multi-level intent recognition technologies. Operation and maintenance personnel can interact with the system through natural language to complete various operation and maintenance tasks. This method not only simplifies the management process of operation and maintenance services, but also effectively improves operation and maintenance efficiency, reduces human operation errors, and meets the requirements of various operation and maintenance scenarios such as cloud computing platforms, data centers, and enterprise IT infrastructures.

[0004] Large language model technology, such as the GPT series of OpenAI, has made remarkable progress in the field of natural language processing, demonstrating powerful text understanding and generation capabilities. Using this technology, operation and maintenance systems can automatically handle common problems, troubleshoot faults, execute routine tasks, and provide real-time support. By deeply integrating with intelligent operation and maintenance systems, large language models can help operation and maintenance personnel quickly obtain accurate information and solutions, thus significantly improving the stability and operation and maintenance efficiency of the system. In addition, combined with the intelligent analysis capabilities of large language models, it can also be used for tasks such as in-depth mining of operation and maintenance data, fault prediction and warning, and resource allocation optimization, providing more comprehensive decision-making support for operation and maintenance teams. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned disadvantages of the prior art and provide a method, device, processor, and its computer-readable storage medium for realizing operation and maintenance service integration based on natural language dialogue that meets the requirements of high efficiency, accuracy, and intelligence.

[0006] To achieve the above purpose, the method, device, processor, and its computer-readable storage medium for realizing operation and maintenance service integration based on natural language dialogue of the present invention are as follows:

[0007] The method for realizing operation and maintenance service integration based on natural language dialogue is mainly characterized in that the method includes the following steps:

[0008] (1) Construct a hierarchical intent recognition model containing multiple hierarchical branches, covering high-level intents and detailed intents. Utilize a pre-trained large language model to understand the user's needs by judging the natural language input by the user layer by layer.

[0009] (2) Combine the high-level intents and detailed intents output by the large language model with the pre-defined keyword matching rules to improve the understanding of the user's intent, response accuracy, and operation execution efficiency.

[0010] (3) Utilize a multi-round dialogue mechanism to summarize and persistently save each round of dialogue, continuously track the user's needs, and make intelligent reasoning decisions.

[0011] Preferably, the step (1) specifically includes the following steps:

[0012] (1.1) Construct a tree structure for intent judgment;

[0013] (1.2) Detailed intent recognition, analyze and extract domain parameters related to specific operation and maintenance service types;

[0014] (1.3) Through a hierarchical structure, refine the user's needs layer by layer;

[0015] (1.4) After identifying the detailed intent and related parameters, call the corresponding operation and maintenance service interface to execute the operation requested by the user;

[0016] (1.5) Feedback the execution result or status information, judge whether the operation is successful. If so, return the result and confirm that the task is completed; otherwise, provide fault diagnosis and make necessary corrections.

[0017] Preferably, the step (1.1) specifically includes the following steps:

[0018] (1.1.1) Identify high-level intents;

[0019] (1.1.2) Conduct detailed intent recognition to determine the specific operation and maintenance service category requested by the user.

[0020] Preferably, the step (1.2) specifically includes the following steps:

[0021] (1.2.1) Extract detailed parameters related to specific service operations from the user input;

[0022] (1.2.2) If there are necessary domain parameters missing, prompt the user to supplement the missing information; if there are no parameter missing, continue with step (1.3).

[0023] Preferably, step (1.3) specifically includes the following steps:

[0024] (1.3.1) Gradually simplify complex intent recognition tasks to improve recognition accuracy and system processing efficiency;

[0025] (1.3.2) Decompose the task into multiple subtasks to handle complex or multi-step requests;

[0026] (1.3.3) Perform intelligent reasoning and judgment based on context information at different levels, gradually deepen the understanding of user needs, and continuously track user intent in multi-round conversations.

[0027] Preferably, step (1) specifically includes the following steps:

[0028] (2.1) Preprocess the natural language input by the user;

[0029] (2.2) Pass the user input to a large language model for preliminary intent recognition;

[0030] (2.3) Verify or supplement the intent output by the large language model through predefined keyword matching rules;

[0031] (2.4) According to the verified intent and the extracted domain parameters, call the corresponding operation and maintenance service module to perform specific operations;

[0032] (2.5) Provide feedback to the user based on the current execution result. If the user does not get the expected result or there are problems in task execution, provide fault diagnosis suggestions or prompt input parameter adjustments.

[0033] Preferably, step (2.1) specifically includes the following steps:

[0034] (2.1.1) Decompose the text input by the user into basic lexical units;

[0035] (2.1.2) Remove unnecessary noise information;

[0036] (2.1.3) Identify key entities in the text.

[0037] Preferably, step (2.2) specifically includes the following steps:

[0038] (2.2.1) Through the large language model, comprehensively understand the user input and judge the user's core needs;

[0039] (2.2.2) Further refine the recognition task and judge the specific service type and requirements.

[0040] Preferably, step (2.3) specifically includes the following steps:

[0041] (2.3.1) Compare the high-level intent and sub-intent identified by the large language model with the predefined keyword rules;

[0042] (2.3.2) Through the secondary verification of keyword matching, identify the user's needs;

[0043] (2.3.3) If there is missing key information in the intent recognition output by the large language model, automatically supplement relevant parameters through the keyword matching rule; if there is no missing key information, continue with step (2.4).

[0044] Preferably, step (2.4) specifically includes the following steps:

[0045] (2.4.1) Select an appropriate service interface for invocation according to the finally confirmed intent type;

[0046] (2.4.2) Pass all the key parameters extracted from the user input to the service interface;

[0047] (2.4.3) After the system completes the task, feedback the execution result to the user and provide relevant information.

[0048] Preferably, step (3) specifically includes the following steps:

[0049] (3.1) Utilize the multi-round dialogue mechanism. After each round of dialogue, the system summarizes and saves the user's needs and the corresponding execution results;

[0050] (3.2) After each round of dialogue ends, summarize the context information of this round of dialogue and save it in a structured manner;

[0051] (3.3) When the parameter information required by the user input or the current task is insufficient, issue a prompt and request the user to supplement relevant information.

[0052] Preferably, step (3.1) specifically includes the following steps:

[0053] (3.1.1) In each round of dialogue, the system extracts key information through the natural language processing module. Context information extraction and persistent storage;

[0054] (3.1.2) Based on the saved historical dialogue information and execution status, the intelligent operation and maintenance robot can understand the current context and the user's historical needs in real time and perform intelligent reasoning;

[0055] (3.1.3) Continuously modify and improve its needs during the dialogue process to simulate the decision-making process.

[0056] Preferably, step (3.2) specifically includes the following steps:

[0057] (3.2.1) Organize information based on the content, requirements, tasks to be executed, parameters, and results of the current conversation to form a structured record;

[0058] (3.2.2) Automatically update the current status and mark the completion status or current execution status of each operation and maintenance task;

[0059] (3.2.3) Through structured storage, store all relevant information of the current conversation in the background database; in the next round of conversation, preferentially obtain historical data from the persistent storage and supplement and improve the user requirements in combination with the current input information.

[0060] The device for realizing the integration of operation and maintenance services based on natural language conversation is mainly characterized in that the device includes:

[0061] A processor configured to execute computer-executable instructions;

[0062] A memory storing one or more computer-executable instructions, and when the computer-executable instructions are executed by the processor, each step of the method for realizing the integration of operation and maintenance services based on natural language conversation as described above is realized.

[0063] The processor for realizing the integration of operation and maintenance services based on natural language conversation is mainly characterized in that the processor is configured to execute computer-executable instructions, and when the computer-executable instructions are executed by the processor, each step of the method for realizing the integration of operation and maintenance services based on natural language conversation as described above is realized.

[0064] The computer-readable storage medium is mainly characterized in that a computer program is stored thereon, and the computer program can be executed by a processor to realize each step of the method for realizing the integration of operation and maintenance services based on natural language conversation as described above.

[0065] By adopting the method, device, processor, and computer-readable storage medium for realizing the integration of operation and maintenance services based on natural language conversation of the present invention, it aims to simplify and optimize the operation and maintenance task management through intelligent dialogue interaction, improve the operation and maintenance efficiency, and reduce human operation errors. By combining the large language model with multi-level intention recognition, keyword matching, and multi-round dialogue mechanism, the present invention realizes efficient, accurate operation and maintenance service integration and intelligent management. Description of the Drawings

[0066] Figure 1 It is a flowchart of the method for realizing the integration of operation and maintenance services based on natural language conversation of the present invention.

[0067] Figure 2 It is a flowchart of hierarchical intention recognition of the large language model for the method of realizing operation and maintenance service integration based on natural language dialogue of the present invention.

[0068] Figure 3 It is a schematic diagram of the multi-round dialogue mechanism for the method of realizing operation and maintenance service integration based on natural language dialogue of the present invention. Specific Embodiments

[0069] In order to more clearly describe the technical content of the present invention, the following will be further described in conjunction with specific embodiments.

[0070] The method of realizing operation and maintenance service integration based on natural language dialogue of the present invention includes the following steps:

[0071] (1) Construct a hierarchical intention recognition model including multiple hierarchical branches, covering high-level intentions and detailed intentions, and use a pre-trained large language model to understand the user's needs by judging the natural language input by the user layer by layer;

[0072] (2) Combine the high-level intentions and detailed intentions output by the large language model with the predefined keyword matching rules to improve the understanding of the user's intentions, response accuracy, and operation execution efficiency;

[0073] (3) Use the multi-round dialogue mechanism to summarize and persistently save each round of dialogue, continuously track the user's needs, and make intelligent reasoning decisions.

[0074] As a preferred embodiment of the present invention, the step (1) specifically includes the following steps:

[0075] (1.1) Construct a tree structure for intention judgment;

[0076] (1.2) Detailed intention recognition, analyze and extract domain parameters related to specific operation and maintenance service types;

[0077] (1.3) Through a hierarchical structure, refine the user's needs layer by layer;

[0078] (1.4) After identifying the detailed intention and related parameters, call the corresponding operation and maintenance service interface to execute the operation requested by the user;

[0079] (1.5) Feedback the execution result or status information, judge whether the operation is successful, if so, return the result and confirm that the task is completed; otherwise, provide fault diagnosis and make necessary corrections.

[0080] As a preferred embodiment of the present invention, the step (1.1) specifically includes the following steps:

[0081] (1.1.1) Identify high-level intentions;

[0082] (1.1.2) Conduct a breakdown intention recognition to determine the specific operation and maintenance service category requested by the user.

[0083] As a preferred embodiment of the present invention, the step (1.2) specifically includes the following steps:

[0084] (1.2.1) Extract detailed parameters related to specific service operations from the user input;

[0085] (1.2.2) If there is a lack of necessary domain parameters, prompt the user to supplement the missing information; if there is no parameter missing, continue to step (1.3).

[0086] As a preferred embodiment of the present invention, the step (1.3) specifically includes the following steps:

[0087] (1.3.1) Gradually simplify the complex intention recognition task to improve the recognition accuracy and system processing efficiency;

[0088] (1.3.2) Decompose the task into multiple subtasks to handle complex or multi-step requests;

[0089] (1.3.3) Perform intelligent reasoning and judgment depending on the context information at different levels, gradually deepen the understanding of the user's needs, and continuously track the user's intention in multiple rounds of conversations.

[0090] As a preferred embodiment of the present invention, the step (1) specifically includes the following steps:

[0091] (2.1) Preprocess the natural language input by the user;

[0092] (2.2) Pass the user's input to a large language model for preliminary intention recognition;

[0093] (2.3) Verify or supplement the intention output by the large language model through predefined keyword matching rules;

[0094] (2.4) Call the corresponding operation and maintenance service module according to the verified intention and the extracted domain parameters to perform specific operations;

[0095] (2.5) Provide feedback to the user according to the current execution result. If the user does not get the expected result or there is a problem with the task execution, provide fault diagnosis suggestions or prompt input parameter adjustments.

[0096] As a preferred embodiment of the present invention, the step (2.1) specifically includes the following steps:

[0097] (2.1.1) Decompose the text input by the user into basic lexical units;

[0098] (2.1.2) Remove unnecessary noise information;

[0099] (2.1.3) Identify key entities in the text.

[0100] As a preferred embodiment of the present invention, step (2.2) specifically includes the following steps:

[0101] (2.2.1) Use a large language model to comprehensively understand the user's input and judge the user's core needs;

[0102] (2.2.2) Further refine the recognition task and judge the specific service type and requirements.

[0103] As a preferred embodiment of the present invention, step (2.3) specifically includes the following steps:

[0104] (2.3.1) Compare the high-level intentions and sub-intentions identified by the large language model with the predefined keyword rules;

[0105] (2.3.2) Through the secondary verification of keyword matching, identify the user's needs;

[0106] (2.3.3) If there is missing key information in the intention recognition output by the large language model, automatically supplement relevant parameters through the keyword matching rule; if there is no missing key information, continue with step (2.4).

[0107] As a preferred embodiment of the present invention, step (2.4) specifically includes the following steps:

[0108] (2.4.1) Select a suitable service interface for invocation according to the finally confirmed intention type;

[0109] (2.4.2) Pass all the key parameters extracted from the user input to the service interface;

[0110] (2.4.3) After the system completes the task, feedback the execution result to the user and provide relevant information.

[0111] As a preferred embodiment of the present invention, step (3) specifically includes the following steps:

[0112] (3.1) Utilize a multi-round dialogue mechanism. After each round of dialogue, the system summarizes and saves the user's needs and the corresponding execution results;

[0113] (3.2) After each round of dialogue ends, summarize the context information of this round of dialogue and save it in a structured manner;

[0114] (3.3) When the parameter information required by the user input or the current task is insufficient, a prompt is issued and the user is required to supplement the relevant information.

[0115] As a preferred embodiment of the present invention, the step (3.1) specifically includes the following steps:

[0116] (3.1.1) In each round of conversation, the system extracts key information through the natural language processing module. Context information extraction and persistent storage;

[0117] (3.1.2) Based on the saved historical conversation information and execution status, the intelligent operation and maintenance robot can understand the current context and the user's historical needs in real time and perform intelligent reasoning;

[0118] (3.1.3) Continuously modify and improve its requirements during the conversation process to simulate the decision-making process.

[0119] As a preferred embodiment of the present invention, the step (3.2) specifically includes the following steps:

[0120] (3.2.1) Organize the information according to the content, requirements, tasks to be executed, parameters and results of the current conversation to form a structured record;

[0121] (3.2.2) Automatically update the current status and mark the completion status or the current execution status of each operation and maintenance task;

[0122] (3.2.3) Through structured storage, store all relevant information of the current conversation in the background database; in the next round of conversation, preferentially obtain historical data from the persistent storage and supplement and improve the user's requirements in combination with the current input information.

[0123] The device for realizing the integration of operation and maintenance services based on natural language conversation of the present invention, wherein the device includes:

[0124] A processor configured to execute computer-executable instructions;

[0125] A memory storing one or more computer-executable instructions, and when the computer-executable instructions are executed by the processor, each step of the method for realizing the integration of operation and maintenance services based on natural language conversation as described above is realized.

[0126] The processor of the present invention for realizing the integration of operation and maintenance services based on natural language conversation, wherein the processor is configured to execute computer-executable instructions, and when the computer-executable instructions are executed by the processor, each step of the method for realizing the integration of operation and maintenance services based on natural language conversation as described above is realized.

[0127] The computer-readable storage medium of the present invention stores a computer program thereon, and the computer program can be executed by a processor to implement each step of the above method for integrating operation and maintenance services based on natural language dialogue.

[0128] In a specific embodiment of the present invention, it relates to a method for integrating operation and maintenance services. Specifically, the present invention proposes a method for integrating operation and maintenance services based on natural language dialogue. This method integrates the services provided by multiple operation and maintenance tools through large language models and multi-level intent recognition technologies. Users interact with the system through natural language to complete the management of different operation and maintenance tasks. This method aims to simplify and optimize the management process of operation and maintenance tasks, improve operation and maintenance efficiency, reduce human operation errors, and is widely applicable to multiple operation and maintenance scenarios such as cloud computing platforms, data centers, and enterprise IT infrastructures.

[0129] The present invention aims to enhance operation and maintenance collaboration capabilities by integrating various operation and maintenance services in a chat dialogue in the form of a unified natural language dialogue interface, thereby simplifying and optimizing the management of operation and maintenance tasks. To achieve this goal, the present invention proposes a hierarchical intent recognition method based on large language models and combines a double-layer intent recognition mechanism of large language models and keyword matching. This method aims to improve the intelligent level and operation efficiency of the operation and maintenance service management system. The operation and maintenance service integration method of the present invention mainly includes the following steps:

[0130] (1) Hierarchical intent recognition of large language models:

[0131] Construct a hierarchical intent recognition model with multiple hierarchical branches, covering high-level intents and detailed intents. This model uses pre-trained large language models (such as GPT series models) to judge the natural language input by users layer by layer to more accurately understand user needs. The design of the hierarchical model helps to improve the accuracy of intent recognition and ensure that the system can effectively handle different levels of operation and maintenance tasks.

[0132] (2) Double-layer intent recognition of large language models and keyword matching:

[0133] Combine the high-level intents and detailed intents output by the large language model with the pre-defined keyword matching rules to form a double-layer intent recognition mechanism. First, the system performs preliminary intent recognition through the large language model, and then conducts secondary verification or supplementation through the keyword matching method to ensure an accurate understanding of user intents and improve the response accuracy and operation execution efficiency of the system.

[0134] This double-layer intent recognition mechanism can meet the strictness and accuracy requirements of operation and maintenance operations while retaining the flexibility of natural language expression, ensuring the efficiency and robustness of the system.

[0135] (3) Multi-round dialogue mechanism:

[0136] Adopt a multi-round dialogue mechanism based on context information to summarize and persistently save each round of dialogue. This mechanism extracts the key information of historical conversations and saves the status, results, and relevant parameters of various operation and maintenance services currently used by the user. In this way, the intelligent operation and maintenance robot can understand the user's current status in real time and perform intelligent reasoning and decision-making based on historical information in subsequent conversations.

[0137] The multi-round dialogue mechanism allows users to continuously modify and improve their requirements during the dialogue process, simulate the decision-making process, and summarize after each round of dialogue to ensure the system's continuous tracking and real-time feedback on user requirements. After the context information of each conversation is saved, the intelligent operation and maintenance robot can adjust its operations based on the saved status and results in subsequent conversations to ensure the best operation and maintenance support for users.

[0138] In step (1), an intention judgment is made by constructing a tree structure through a program. The natural language input by the user first enters the root node for high-level intention recognition to determine whether to execute operation and maintenance service calls, update the context, or confirm operations and other high-level intentions. Then, the system enters the detailed intention recognition stage to determine the specific operation and maintenance service category, such as log retrieval, metric query, call chain topology query, configuration information retrieval, etc.

[0139] In the detailed intention recognition stage, the system will further extract domain parameters related to specific service types, such as log retrieval conditions, automated job names, etc., to ensure a detailed understanding and accurate processing of user requests.

[0140] In step (2), a two-layer recognition method of large language model and keyword matching is adopted for intention recognition at each level. First, the large language model performs preliminary intention recognition, and then the keyword matching method is used for secondary verification or supplementation to ensure that the system can handle the diverse inputs of users and meet the requirements of the standardization and accuracy of operation and maintenance operations.

[0141] In step (3), the system summarizes the context information after each round of dialogue and saves the relevant information in a structured manner. The saved content includes the participants in the current conversation, the parameters of the operation and maintenance services used, the execution results, and the current status information. When the user inputs the next round of dialogue, the system will obtain the current context from the persistent record and enrich the user's input in combination with the saved status information, so as to ensure that the processing of the current round is more in line with the user's expectations.

[0142] When the operation and maintenance service parameter information required in steps (1) and (2) is insufficient, step (3) will issue a prompt to guide the user to supplement the required additional information to ensure the normal access and execution of the operation and maintenance services.

[0143] In a specific embodiment, the present invention proposes an operation and maintenance service integration method based on natural language dialogue. This method integrates the services provided by multiple operation and maintenance tools through a large language model and multi-level intent recognition technology, and specifically includes the following steps:

[0144] (1) Hierarchical intent recognition of the large language model: As Figure 1 shown, the hierarchical intent recognition method of the large language model proposed by the present invention covers different levels of intent recognition, including high-level intents and detailed intents, by constructing a multi-level branch structure. This model uses a pre-trained large language model (such as the GPT series models) to more accurately understand the user's needs by judging the natural language input by the user layer by layer. The design of the hierarchical model helps to improve the accuracy of intent recognition and ensure that the system can effectively handle different levels of operation and maintenance tasks. The working principle and technical implementation of this hierarchical intent recognition process are described in detail below.

[0145] (1.1) Constructing a tree structure for intent judgment: In the method of the present invention, a tree structure is first constructed through a program for intent judgment. The natural language input by the user first enters the root node for high-level intent recognition. At this stage, the system judges the category of the user's needs and decides whether to perform high-level tasks such as operation and maintenance service calls, updating context information, or confirming certain operations according to the high-level intent.

[0146] (1.1.1) High-level intent recognition: This process involves the preliminary understanding of the user's natural language input. The main task is to identify the overall direction of the user's request. For example, the user may input "query system status" or "restart the server", and the system needs to judge whether this belongs to the query operation or execution operation type and make a preliminary classification at this stage.

[0147] (1.1.2) Entering the detailed intent recognition stage: Once the high-level intent is determined, the system enters the next stage for detailed intent recognition. At this stage, the system further determines the specific operation and maintenance service category requested by the user. For example, if the high-level intent is a query operation, the system will judge the specific content to be queried, such as "log retrieval", "metric query", "call chain topology query", or "configuration information retrieval", etc. This step ensures that the system has a more detailed understanding of the user's needs and can accurately classify and execute.

[0148] (1.2) Detailed intent recognition and domain parameter extraction: In the detailed intent recognition stage, the system will further analyze and extract domain parameters related to the specific operation and maintenance service type to ensure a detailed understanding and accurate processing of the user's request.

[0149] (1.2.1) Domain parameter extraction: The goal of this stage is to extract detailed parameters related to specific service operations from the user input. For example, when performing log retrieval, the system needs to extract conditions such as the type of log, time range, and log level; when executing an automated job, the system needs to identify parameters such as the job name and job type. These domain parameters help the system accurately execute user requests and ensure that the operations meet user expectations.

[0150] (1.2.2) Confirmation and supplementation: If some necessary domain parameters are missing in the fine-grained intent recognition stage, the system will actively prompt the user to supplement the missing information to ensure that all operations have sufficient context data support. This mechanism helps to avoid operation errors caused by missing parameters.

[0151] (1.3) Hierarchical structure to improve intent recognition accuracy: Through a hierarchical structure, the hierarchical intent recognition model can refine the user's needs layer by layer. First, a relatively simple high-level intent classification is performed at the root node, and then the specific operations are further decomposed and confirmed in the fine-grained intent recognition stage.

[0152] (1.3.1) Improve recognition accuracy: The hierarchical structure can gradually simplify complex intent recognition tasks, avoiding processing all complex information at once, thereby improving recognition accuracy and system processing efficiency.

[0153] (1.3.2) Efficiently handle complex instructions: The natural language input by the user may contain multiple operation steps or multiple service requests. Hierarchical intent recognition can more effectively handle complex or multi-step requests by decomposing the task into multiple subtasks.

[0154] (1.3.3) Enhance context understanding ability: The system can rely on context information at different levels for intelligent reasoning and judgment, gradually deepen the understanding of user needs, and can continuously track the user's intent in multi-round conversations.

[0155] (1.4) Execute operation and maintenance operations: After identifying the fine-grained intent and related parameters, the system will call the corresponding operation and maintenance service interface to execute the operations requested by the user to ensure that the operations meet expectations.

[0156] (1.5) System response and feedback: After executing the corresponding operation and maintenance operations, the system will feedback the execution result or status information. If the operation is successful, the system will return the result and confirm the completion of the task. If there is a problem with the operation, the system will provide fault diagnosis and guide the user to make necessary corrections.

[0157] (2) Double-layer Intent Recognition of Large Language Model and Keyword Matching: The double-layer intent recognition mechanism of large language model and keyword matching proposed by the present invention further improves the system's accurate understanding of user intent, response accuracy, and operation execution efficiency by combining the high-level intent and sub-intent output by the large language model and combining them with the pre-defined keyword matching rules. While retaining the flexibility of natural language expression, this mechanism ensures the efficiency, precision, and robustness of the system when processing operation and maintenance operations.

[0158] (2.1) Input Preprocessing: The natural language input by the user first undergoes text preprocessing, and the system ensures that the information input by the user can be accurately transmitted to the large language model for parsing. The accuracy of preprocessing directly affects the subsequent intent recognition effect, so this step needs to be executed efficiently and precisely. The preprocessing includes the following contents:

[0159] (2.1.1) Word Segmentation: Decompose the text input by the user into basic lexical units so that the system can better understand the meaning of each word.

[0160] (2.1.2) Denoising Processing: Remove unnecessary noise information, such as irrelevant punctuation marks, redundant spaces, or repeated words, to ensure that the input text is clean and conforms to grammar.

[0161] (2.1.3) Entity Recognition: Identify key entities in the text, such as dates, device names, log types, etc. These information are crucial for subsequent intent recognition.

[0162] (2.2) Preliminary Intent Recognition: After text preprocessing, the system passes the user's input to the large language model for preliminary intent recognition. The large language model (such as GPT series) can understand the user's needs based on language context, grammar, semantics, and common expression patterns and give a preliminary intent classification.

[0163] (2.2.1) High-level Intent Recognition: First, the system uses the large language model to comprehensively understand the user's input and judge the user's core needs. For example, when the user inputs "query system load", the system will first identify that this request belongs to the "query" category, which is a high-level intent to query the system status.

[0164] (2.2.2) Sub-intent Recognition: After determining the user's high-level intent, the system will further refine the recognition task and judge the specific service type and requirements. For example, in the request of "query system load", the system will further judge whether the specific content that the user needs to query is "CPU load" or "memory load", and process it as a sub-intent.

[0165] (2.3) Keyword Matching Verification and Supplement: Although large language models have powerful semantic understanding capabilities, due to the diversity and complexity of natural language, relying solely on the language model may lead to some misunderstandings or missed identifications. Therefore, the present invention designs a keyword matching mechanism as the second-level verification step. In this step, the system verifies or supplements the intent output by the large language model through predefined keyword matching rules. The specific steps are as follows:

[0166] (2.3.1) Keyword Matching: The system will compare the high-level intent and detailed intent identified by the large language model with the predefined keyword rules. The keyword rules include task categories, service types, operation parameters, etc. For example, if the large language model identifies "query system load", the system will check whether the keyword "load" is included and further determine whether operations such as "query" or "monitor" are required.

[0167] (2.3.2) Secondary Verification: Through the secondary verification of keyword matching, the system can more accurately identify the user's needs. For example, the user may input "view server health status", and the system will match the keyword "health status" and confirm that the request involves querying the system health status rather than other types of operations.

[0168] (2.3.3) Parameter Supplement: If there is missing key information in the intent recognition output by the large language model, the keyword matching rules will automatically supplement the relevant parameters. For example, when querying logs, the system will check whether necessary conditions such as the time range and log type of the logs are provided. If the user does not provide this information, the system will prompt the user to input supplementary parameters.

[0169] (2.4) Confirm and Execute the Task: After double-layer intent recognition, the system will finally confirm the user's intent and execute the corresponding operation and maintenance tasks. The system calls the corresponding operation and maintenance service module according to the verified intent and the extracted domain parameters to perform specific operations. The specific implementation steps are as follows:

[0170] (2.4.1) Service Invocation: The system selects the appropriate service interface for invocation according to the finally confirmed intent type (such as query, execute task, etc.). For example, if it is identified as "query CPU load", the system will call the API interface of the monitoring system to obtain the real-time load information of the CPU.

[0171] (2.4.2) Parameter Transfer: Before the task is executed, the system transfers all the key parameters (such as time range, log type, etc.) extracted from the user input to the service interface to ensure that the task can be accurately executed according to the user's specific needs.

[0172] (2.4.3) Execution Feedback: After the system completes a task, it will feedback the execution result to the user and provide relevant information. For example, if the user queries "CPU load", the system will return the real-time data of this request and inform the user of the query result.

[0173] (2.5) Feedback and Adjustment: After the task is executed, the system will provide feedback to the user according to the current execution result. If the user does not get the expected result or there is a problem with the task execution, the system will provide fault diagnosis suggestions or let the user adjust the input parameters. For example, if the query returns an empty result, the system will prompt the user to check whether the parameter settings (such as time range or device selection) are correct. The system can also perform self-learning and adjustment based on historical conversations and task execution records to provide more accurate services in future conversations.

[0174] (3) Multi-round Conversation Mechanism: The multi-round conversation mechanism in the present invention is designed for the intelligent interaction process in operation and maintenance services, aiming to summarize and persistently save each round of conversation, continuously track user needs, and make intelligent reasoning decisions. Through efficient information storage and processing, this mechanism ensures that the system can make reasonable and accurate responses in subsequent conversations based on historical conversation information, so as to provide personalized and continuous operation and maintenance support.

[0175] As Figure 2 shown, the present invention adopts the Llama causal language model and uses an attention mask in the form of a diagonal matrix during the encoding process of the model. The design of this mask ensures that each token can only perceive the tokens before it during encoding and cannot obtain information about subsequent tokens. Specifically, in the encoding output of part 1 of the description, the model can only access the content of part 1 of the description and has no perception ability of the information in its subsequent part. While in the encoding output of part 2 of the description, the model can not only perceive the content of part 2 of the description, but also access the relevant information of part 1 and reply 1, thus providing effective context support for predicting reply 2.

[0176] (3.1) Multi-round Conversation Mechanism: Under the multi-round conversation mechanism, the user interacts with the intelligent operation and maintenance robot in multiple rounds. After each round of conversation, the system will summarize and save the user's needs and the corresponding execution results. The specific working principle is as follows:

[0177] (3.1.1) Context Information Extraction and Persistent Saving: In each round of conversation, the system first extracts key information through the natural language processing module. Context information extraction and persistent saving include but are not limited to the following contents:

[0178] a) User input requirements: That is, the operation and maintenance tasks, service requests, and any related parameters mentioned by the user in this round of conversation. b) Execution status of the operation and maintenance service: Including the execution status of the current operation and maintenance service, whether it has been successfully completed, error information during execution, etc.

[0179] Information, etc.

[0180] c) Execution results: The data, results, warning information, etc. returned by the system during the execution process.

[0181] After structuring and saving this information, the system will persistently store it in a database or information management system for use in the next round of conversation. This process ensures the system's real-time tracking and feedback on user requirements.

[0182] (3.1.2) Real-time status tracking and intelligent reasoning: Based on the saved historical conversation information and execution status, the intelligent operation and maintenance robot can understand the current context and the user's historical requirements in real time and perform intelligent reasoning. Through the context information, the system can achieve the following functions:

[0183] a) Track the execution progress of the current operation and maintenance task: Whether a certain service has been completed, whether the user needs to continue with the next operation, or whether adjustments need to be made based on the results of the current task.

[0184] b) Intelligent decision-making and feedback: If the system identifies a change in the user's requirements or tasks, or there are certain associations between the current task and historical tasks, the system will automatically adjust subsequent operations based on the saved status information and feedback the adjusted suggestions to the user. For example, if the user previously requested to query the status of a certain server, the system can intelligently recommend other relevant operations (such as restarting the server or checking the logs) based on the historical records, or directly execute subsequent tasks.

[0185] (3.1.3) Decision-making process simulation and user requirement refinement: The multi-round conversation mechanism supports the user to continuously modify and improve their requirements during the conversation, simulating a decision-making process. This function has the following characteristics:

[0186] a) Dynamic requirement adjustment: The user can modify the original input requirements according to the actual situation, such as changing the query time range, adjusting the execution conditions of the service, etc. The system can judge the user's modification intention based on the historical information and the current state to ensure that the new requirements are recognized and executed in a timely manner based on the existing information.

[0187] b) Decision-making process support: The system will simulate and guide the user to gradually improve the decision-making process. For example, in a situation where multiple steps of tasks need to be executed, the system will infer the user's next requirements based on the previous conversation and actively ask or provide suggestions to help the user complete the operation.

[0188] In this way, the system can not only continuously track the user's needs during the conversation, but also provide flexible and intelligent operation and maintenance support for the user.

[0189] (3.2) Summary and structured storage after each round of conversation: Whenever a round of conversation ends, the system summarizes the context information of that round of conversation and saves it in a structured manner. This process includes the following steps:

[0190] (3.2.1) Information extraction and organization: The system organizes the information based on the participants, requirements, tasks executed, parameters, and results of the current conversation to form a structured record. For example, it organizes the participants, task categories, execution results, log information, etc. and stores them in the system database or persistent storage system.

[0191] (3.2.2) Status update: The system automatically updates the current status, marking the completion status or current execution status (such as "processing", "completed", "failed", etc.) of each operation and maintenance task. These status information helps the system track the task progress in real time.

[0192] (3.2.3) Data saving and archiving: Through structured storage, the system stores all relevant information of the current conversation in the background database. During the next round of conversation, the system will first obtain the historical data from the persistent storage and supplement and improve the user's needs in combination with the current input information.

[0193] (3.3) Ensure the integrity of the information required for task execution: In steps (1) and (2), when the parameter information required by the user input or the current task is insufficient, the system will issue a prompt asking the user to supplement the relevant information. The system will automatically identify the missing key information and guide the user to supplement it through prompts:

[0194] a) Identification of missing information: The system checks whether there is incomplete or missing key information in the user input. For example, if the user requests "query logs" but does not specify the time range or log type, the system will automatically identify and remind the user to supplement these necessary parameters.

[0195] b) Automatic supplement suggestions: In some cases, the system can also automatically supplement the missing parameters through historical conversation information or context reasoning. For example, if the user has made a similar query before, the system can intelligently provide default parameters based on the historical records or recommend information based on the current progress of the task.

[0196] c) User guidance and prompts: When the system detects that the required parameter information is missing, it will actively issue a supplement request to the user through a prompt generated by natural language. For example, "Please provide the time range for the query" or "Please specify the log type".

[0197] This information supplementation mechanism can ensure that each operation and maintenance task can obtain complete parameter support, thus guaranteeing the normal execution and efficient completion of the task.

[0198] The technical solution of the present invention has a multi-level intention recognition model and an operation and maintenance service dynamic integration mechanism, constructing a tree-shaped hierarchical model including high-level intentions (such as "fault handling", "resource monitoring") and detailed intentions (such as "CPU load query", "log retrieval"). The system analyzes the user input layer by layer through a large language model and conducts secondary verification in combination with the preset keyword rules in the operation and maintenance field. For example, when the user inputs "Check the CPU usage rate of server A", the system first recognizes the high-level intention as "resource monitoring", further breaks it down into the detailed intention of "CPU metric query", and extracts the parameters "server A" and "time range".

[0199] By decomposing the intention recognition task into multiple levels through a tree structure, the semantic ambiguity risk of the large language model is significantly reduced. Combining keyword rules to supplement missing parameters (such as automatically prompting "Please supplement the time range") ensures the integrity and accuracy of operation and maintenance operations. This design solves the problem of misoperations caused by ambiguous intentions in traditional operation and maintenance systems and improves the success rate of task execution.

[0200] The technical solution of the present invention can realize the dynamic call of operation and maintenance service interfaces and the integration of multiple tools. The system automatically calls the corresponding operation and maintenance service interfaces (such as using the InfluxDB API to obtain monitoring data and the log platform interface to retrieve logs) according to the finally recognized detailed intentions and parameters, and provides real-time feedback on the execution status. For example, after recognizing the "log retrieval" intention, the system passes the parameters to the ELK interface and automatically generates a user-readable summary after the result is returned.

[0201] The technical solution of the present invention realizes the end-to-end automation from natural language instructions to operation and maintenance tool APIs for the first time, and eliminates the information silos between tools through a dynamic interface mapping table (such as mapping the intention of "service restart" to the automated job execution interface).

[0202] The technical solution of the present invention can realize context-aware multi-round conversations and persistent storage. The system structurally stores the context information after each round of conversation (such as recording "Current task: log retrieval; parameters: time range = the past 1 hour") and preferentially calls the historical state in subsequent conversations. For example, when the user subsequently inputs "Check the memory usage again", the system automatically inherits the server name and time range of the previous task without the need for repeated input.

[0203] The technical solution of the present invention realizes the inheritance and correction of task context across multiple rounds of interactions through a dialogue state machine and a structured storage mechanism. This mechanism can improve the processing efficiency of complex operation and maintenance tasks (such as multi-step fault troubleshooting) while reducing the number of user interactions.

[0204] For the specific implementation solutions of this embodiment, reference can be made to the relevant descriptions in the above embodiments, which will not be elaborated here.

[0205] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be referred to the same or similar content in other embodiments.

[0206] It should be noted that in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "a plurality of" refers to at least two.

[0207] Any process or method description shown in the flowchart or described in other ways herein can be understood to represent a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process. The scope of the preferred embodiments of the present invention includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in the reverse order according to the involved functions, rather than in the order shown or discussed, which should be understood by those skilled in the technical field to which the embodiments of the present invention belong.

[0208] It should be understood that each part of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution device. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following well-known technologies in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0209] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the methods in the above embodiments can be completed by instructing relevant hardware through a program, and the corresponding program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0210] In addition, in each embodiment of the present invention, each functional unit can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0211] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, etc.

[0212] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0213] By adopting the method, device, processor, and computer-readable storage medium of the present invention for realizing the integration of operation and maintenance services based on natural language dialogue, it aims to simplify and optimize operation and maintenance task management through intelligent dialogue interaction, improve operation and maintenance efficiency, and reduce human operation errors. By combining large language models with multi-level intent recognition, keyword matching, and multi-round dialogue mechanisms, the present invention realizes efficient and accurate integration of operation and maintenance services and intelligent management.

[0214] In this specification, the present invention has been described with reference to its specific embodiments. However, it is obvious that various modifications and transformations can still be made without departing from the spirit and scope of the present invention. Therefore, the specification and the drawings should be regarded as illustrative rather than restrictive.

Claims

1. A method for realizing the integration of operation and maintenance services based on natural language conversations, characterized in that, The method described above includes the following steps: (1) Build a hierarchical intent recognition model with multiple hierarchical branches, covering high-level intents and detailed intents. Utilize a pre-trained large language model to understand the user's needs by judging the natural language input by the user layer by layer. (2) Combine the high-level intents and detailed intents output by the large language model with the predefined keyword matching rules to improve the understanding of the user's intent, response accuracy, and operation execution efficiency. (3) Utilize a multi-round dialogue mechanism to summarize and persistently save each round of dialogue, continuously track the user's needs, and make intelligent reasoning decisions.

2. The method for realizing the integration of operation and maintenance services based on natural language dialogue according to claim 1, wherein The specific steps of step (1) include the following: (1.1) Build a tree structure for intent judgment. (1.2) Conduct detailed intent recognition, analyze and extract domain parameters related to specific operation and maintenance service types. (1.3) Refine the user's needs layer by layer through a hierarchical structure. (1.4) After identifying the detailed intent and related parameters, call the corresponding operation and maintenance service interface to execute the operation requested by the user. (1.5) Feedback the execution result or status information, judge whether the operation is successful. If it is, return the result and confirm that the task is completed; otherwise, provide fault diagnosis and make necessary corrections.

3. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 2, wherein, The specific steps of step (1.1) include the following: (1.1.1) Identify high-level intents. (1.1.2) Conduct detailed intent recognition to determine the specific operation and maintenance service category requested by the user.

4. The method for realizing operation and maintenance service integration based on natural language dialogue according to claim 2, wherein The specific steps of step (1.2) include the following: (1.2.1) Extract detailed parameters related to specific service operations from the user input. (1.2.2) If there are necessary domain parameters missing, prompt the user to supplement the missing information; if there are no missing parameters, continue with step (1.3).

5. The method for realizing operation and maintenance service integration based on natural language dialogue according to claim 2, characterized in that The specific steps of step (1.3) include the following: (1.3.1) Gradually simplify complex intent recognition tasks to improve recognition accuracy and system processing efficiency. (1.3.2) Decompose the task into multiple subtasks to handle complex or multi-step requests. (1.3.3) Perform intelligent reasoning and judgment depending on context information at different levels, gradually deepen the understanding of the user's needs, and continuously track the user's intent in multi-round dialogues.

6. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 1, characterized in that, The specific steps of step (1) include the following: (2.1) Preprocess the natural language input by the user. (2.2) Pass the user input to the large language model for preliminary intent recognition. (2.3) Verify or supplement the intent output by the large language model through predefined keyword matching rules. (2.4) According to the verified intent and the extracted domain parameters, call the corresponding operation and maintenance service module to execute specific operations. (2.5) Provide feedback to the user based on the current execution result. If the user does not get the expected result or there are problems in task execution, provide fault diagnosis suggestions or prompt input parameter adjustments.

7. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 6, wherein The specific steps of step (2.1) include the following: (2.1.1) Decompose the text input by the user into basic lexical units. (2.1.2) Remove unnecessary noise information. (2.1.3) Identify key entities in the text.

8. The method for realizing operation and maintenance service integration based on natural language dialogue according to claim 6, wherein The specific steps of step (2.2) are as follows: (2.2.1) Use a large language model to comprehensively understand the user's input and judge the user's core needs; (2.2.2) Further refine the recognition task and judge the specific service type and requirements.

9. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 6, wherein, The specific steps of step (2.3) are as follows: (2.3.1) Compare the high-level intentions and sub-intentions recognized by the large language model with the predefined keyword rules; (2.3.2) Through the secondary verification of keyword matching, identify the user's needs; (2.3.3) If there is missing key information in the intention recognition output by the large language model, automatically supplement relevant parameters through the keyword matching rule; if there is no missing key information, continue with step (2.4).

10. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 6, wherein The specific steps of step (2.4) are as follows: (2.4.1) Select a suitable service interface for invocation according to the finally confirmed intention type; (2.4.2) Pass all the key parameters extracted from the user input to the service interface; (2.4.3) After the system completes the task, feedback the execution result to the user and provide relevant information.

11. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 1, wherein The specific steps of step (3) are as follows: (3.1) Utilize the multi-round dialogue mechanism. After each round of dialogue, the system summarizes and saves the user's needs and the corresponding execution results; (3.2) After each round of dialogue ends, summarize the context information of this round of dialogue and save it in a structured manner; (3.3) When the parameter information required by the user input or the current task is insufficient, give a prompt and ask the user to supplement relevant information.

12. The method for realizing operation and maintenance service integration based on natural language conversation according to claim 11, characterized in that The specific steps of step (3.1) are as follows: (3.1.1) In each round of dialogue, the system extracts key information through the natural language processing module. Context information extraction and persistent storage; (3.1.2) Based on the saved historical dialogue information and execution status, the intelligent operation and maintenance robot can understand the current context and the user's historical needs in real time and perform intelligent reasoning; (3.1.3) Continuously modify and improve its needs during the dialogue process to simulate the decision-making process.

13. The method for realizing operation and maintenance service integration based on natural language dialogue according to claim 11, wherein The specific steps of step (3.2) are as follows: (3.2.1) Organize information according to the participation content, requirements, execution tasks, parameters, and results of the current dialogue to form a structured record; (3.2.2) Automatically update the current status and mark the completion status or the current execution status of each operation and maintenance task; (3.2.3) Through structured storage, store all relevant information of the current dialogue in the background database; in the next round of dialogue, preferentially obtain historical data from the persistent storage and supplement and improve the user's needs in combination with the current input information.

14. An apparatus for implementing operation and maintenance service integration based on natural language conversations, characterized in that, The device includes: A processor configured to execute computer-executable instructions; A memory storing one or more computer-executable instructions, which, when executed by the processor, implement the steps of the method for realizing integrated operation and maintenance services based on natural language dialogue according to any one of claims 1 to 13.

15. A processor for implementing the integration of operation and maintenance services based on natural language conversations, characterized in that, The described processor is configured to execute computer-executable instructions, and when the computer-executable instructions are executed by the processor, each step of the method for realizing operation and maintenance service integration based on natural language conversation according to any one of claims 1 to 13 is realized.

16. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program can be executed by a processor to realize each step of the method for realizing operation and maintenance service integration based on natural language conversation according to any one of claims 1 to 13.

Images