Intelligent operation and maintenance method and device and storage medium

Abstract

The invention provides an intelligent operation and maintenance method and device and a storage medium. The method comprises the following steps: constructing an operation and maintenance knowledge graph; obtaining question data input by a user, and determining a first intention classification result and a first scene classification result corresponding to the question data; based on the first intention classification result, obtaining target key information corresponding to the problem data through an operation and maintenance knowledge graph; determining recommendation information corresponding to the problem data based on the first scene classification result; and inputting the target key information and the recommendation information into the large language model to obtain answer data of the question data. Therefore, the target key information and the recommendation information of the question data are input into the large language model to obtain the answer data of the question data, so that the answer data better meets the actual requirements of the user, the pertinence and effectiveness of auxiliary decision making are enhanced, the operation and maintenance accuracy is improved, efficient transmission of operation and maintenance knowledge is promoted, and the user experience is improved. And team cooperation and problem solving processes are accelerated.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to an intelligent operation and maintenance method, device and storage medium. Background Art

[0002] At present, the operation and maintenance platform mainly relies on manual operations and semi-automated processes to process data. Among them, log analysis relies on operation and maintenance personnel to log in to the server and manually screen information using command line tools, and backup and recovery rely on predetermined scripts. Manual intervention is required during data recovery. As a result, the operation and maintenance process is highly dependent on manual experience, making it difficult to achieve real-time analysis and intelligent processing of data, and the operation and maintenance efficiency is low. At the same time, the knowledge documents used for operation and maintenance issues lack efficient management and utilization in different media such as Word and Wiki, resulting in data islands, which seriously hinders the flow of information and efficient decision-making, and reduces the accuracy of operation and maintenance. Summary of the invention

[0003] In order to overcome the problems existing in the related art, the present disclosure provides an intelligent operation and maintenance method, device and storage medium.

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided an intelligent operation and maintenance method, the method comprising:

[0005] Build an operation and maintenance knowledge graph;

[0006] Obtaining question data input by a user, and determining a first intent classification result and a first scene classification result corresponding to the question data;

[0007] Based on the first intent classification result, obtaining target key information corresponding to the problem data through the operation and maintenance knowledge graph;

[0008] Determining recommendation information corresponding to the problem data based on the first scenario classification result;

[0009] The target key information and the recommended information are input into a large language model to obtain answer data for the question data.

[0010] Optionally, in the embodiment of the present disclosure, obtaining target key information corresponding to the problem data through the operation and maintenance knowledge graph based on the first intention classification result includes:

[0011] Based on the first intent classification result, a corresponding target prompt word template is obtained;

[0012] Extract the target entity in the target prompt word template,

[0013] Based on the target entity, extracting initial key information in the target prompt word template from the question data;

[0014] Based on the target entity and the initial key information, the target key information corresponding to the problem data is obtained through the operation and maintenance knowledge graph.

[0015] Optionally, in an embodiment of the present disclosure, the recommendation information includes: context information, task information, information display strategy and solution information.

[0016] Optionally, in the embodiment of the present disclosure, determining the recommended information corresponding to the problem data based on the first scene classification result includes:

[0017] Inputting the question data into a large language model to obtain corresponding query data;

[0018] Determining context information of the question data through the query data and the vector index system;

[0019] Based on the first scene classification result, determining the task information of the problem data through a sub-data block vector library;

[0020] Determine the user role to which the user belongs, and determine the information display strategy of the problem data according to the user role and the task information;

[0021] The solution information of the problem data is determined through the query data and the fault case library.

[0022] Optionally, in the embodiment of the present disclosure, determining the context information of the question data through the query data and the vector index system includes:

[0023] Determine a second intent classification result and a second scene classification result corresponding to the query data;

[0024] Determining a target prompt word for the query data according to the second intent classification result and the second scene classification result;

[0025] The target prompt word is input into a vector index system to obtain context information of the question data.

[0026] Optionally, in the embodiment of the present disclosure, determining the task information of the problem data through a sub-data block vector library based on the first scene classification result includes:

[0027] Based on the first scenario classification result, acquiring scenario information of the problem data through the operation and maintenance knowledge graph;

[0028] The scene information is converted into a target vector, and similarity calculation is performed between the target vector and the sub-data block vector in the sub-data block vector library to obtain a similarity calculation result;

[0029] The similarity calculation results are sorted in descending order, and sub-data blocks corresponding to a first preset number of front blocks are determined as task information corresponding to the problem data.

[0030] Optionally, in the embodiment of the present disclosure, before performing similarity calculation between the target vector and the sub-data block vectors in the sub-data block vector library to obtain a similarity calculation result, the method further includes:

[0031] Determine the target text classification model;

[0032] Based on the target text classification model, a sub-data block vector library is established.

[0033] Optionally, in an embodiment of the present disclosure, the task information includes a target task; and the determining the role to which the user belongs and determining the information display strategy of the problem data according to the role and the task information includes:

[0034] Through the operation and maintenance knowledge graph, a mapping relationship between roles, tasks and display contents is established;

[0035] Determining the user role to which the user belongs;

[0036] Based on the user role and target task, the information display strategy of the problem data is determined through the mapping relationship.

[0037] Optionally, in the embodiment of the present disclosure, determining the solution information of the problem data through the query data and the fault case library includes:

[0038] Searching the target prompt word in the fault case library to determine a second preset number of top solutions;

[0039] The second preset number of solutions are integrated, and the integrated solutions are determined as the solution information of the problem data.

[0040] According to a second aspect of an embodiment of the present disclosure, there is provided an intelligent operation and maintenance device, the device comprising:

[0041] Construction module, used to build operation and maintenance knowledge graph;

[0042] A first determination module, used to obtain question data input by a user, and determine a first intent classification result and a first scene classification result corresponding to the question data;

[0043] A second determination module is used to obtain target key information corresponding to the problem data through the operation and maintenance knowledge graph based on the first intention classification result;

[0044] A third determination module, configured to determine the recommended information corresponding to the problem data based on the first scenario classification result;

[0045] A processing module is used to input the target key information and the recommended information into a large language model to obtain answer data for the question data.

[0046] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0047] processor;

[0048] a memory for storing processor-executable instructions;

[0049] The processor is configured to implement the method described in the first aspect above.

[0050] According to a fourth aspect of an embodiment of the present disclosure, a non-temporary computer-readable storage medium is provided, wherein the computer storage medium stores computer-executable instructions; after the computer-executable instructions are executed by a processor, the method described in the first aspect above can be implemented.

[0051] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0052] The present disclosure proposes an intelligent operation and maintenance method, device and storage medium, the method comprising constructing an operation and maintenance knowledge graph; obtaining the problem data input by the user, and determining the first intent classification result and the first scenario classification result corresponding to the problem data; based on the first intent classification result, obtaining the target key information corresponding to the problem data through the operation and maintenance knowledge graph; based on the first scenario classification result, determining the recommended information corresponding to the problem data; inputting the target key information and the recommended information into a large language model to obtain the answer data of the problem data. Thus, the present disclosure can input the target key information and the recommended information of the problem data into a large language model to obtain the answer data of the problem data, so that the answer data is more in line with the actual needs of the user, enhances the pertinence and effectiveness of auxiliary decision-making, improves the accuracy of operation and maintenance, and then promotes the efficient transmission of operation and maintenance knowledge, and accelerates team collaboration and problem-solving processes.

[0053] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0055] Figure 1 is a flowchart of an intelligent operation and maintenance method according to some embodiments of the present disclosure;

[0056] Figure 2 is a structural schematic diagram of an intelligent operation and maintenance device according to some embodiments of the present disclosure;

[0057] Figure 3 It is a block diagram of an electronic device suitable for executing an intelligent operation and maintenance method according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0058] Some embodiments of the present disclosure will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. Various changes, modifications and equivalents of the methods, devices and / or systems described herein will become apparent after understanding the present disclosure. For example, the order of operations described herein is merely an example and is not limited to those orders set forth herein, but can be changed as becomes apparent after understanding the present disclosure, except for operations that must be performed in a specific order. In addition, for clarity and brevity, descriptions of features known in the art may be omitted.

[0059] The embodiments described in some embodiments of the present disclosure below do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0060] Figure 1 is a flowchart of an intelligent operation and maintenance method according to some embodiments of the present disclosure, such as Figure 1 As shown, the method may include the following steps:

[0061] Step 101: Build an operation and maintenance knowledge graph.

[0062] It should be noted that the executor of the above-mentioned intelligent operation and maintenance method is an intelligent operation and maintenance device, which can be implemented by software and / or hardware. The intelligent operation and maintenance device in this embodiment can be configured in an electronic device.

[0063] In this example embodiment, the electronic device may include devices such as terminal devices and servers, and this embodiment does not limit the electronic device.

[0064] In the disclosed embodiment, before constructing the operation and maintenance knowledge graph, the required data can be extracted from different data sources according to the preset query parameters through a dedicated data extraction adapter. The different data sources may include but are not limited to relational databases (MySQL, Oracle, SQL Server), non-relational databases (MongoDB, Redis), data warehouses (Teradata, Snowflake) and various file storage systems (CSV, Excel, JSON files). In the disclosed embodiment, the preset query parameters may include: the name field of personnel query, the time range and system and indicator name of indicator query, the alarm title and alarm system of alarm query, and the device name and device IP of device query.

[0065] And, in the embodiment of the present disclosure, through the above steps, personnel-centric, associated query personnel-related indicators, logs, alarms, changes, equipment information, to build a complete knowledge graph of personnel, and the integrated data is built into a unified graph database for subsequent query. In the embodiment of the present disclosure, the above steps realize the rapid integration, efficient conversion and secure transmission of data, providing a solid data foundation and strong technical support for downstream data analysis, business decision support and innovative application development.

[0066] Step 102: Obtain question data input by the user, and determine a first intent classification result and a first scene classification result corresponding to the question data.

[0067] In the embodiment of the present disclosure, after obtaining the question data input by the user, the first intent classification result and the first scene classification result corresponding to the question data can be determined according to the large language model. Among them, in the embodiment of the present disclosure, in the context of the deep integration of natural language processing and operation and maintenance scenarios, the present disclosure uses the Qwen-14B cutting-edge high-performance open source large language model and adopts large-scale model lora fine-tuning technology to specifically strengthen the semantic understanding ability in the operation and maintenance field, ensuring accurate analysis of the diversified query requirements in actual operation and maintenance scenarios.

[0068] And, in the disclosed embodiment, the large model can recognize the intent of the question data input by the user through the preset prompt word, and classify the user's intent into one of personnel information retrieval, system indicator query, alarm information query and device information query. Among them, in the disclosed embodiment, the above-mentioned intent classification result can also be the corresponding scene classification result, that is, if the user's intent classification result is determined to be personnel information retrieval through the above steps, the corresponding scene is also personnel information retrieval. Table 1 is a table of correspondence between different scenes and query content.

[0069] Table 1

[0070]

[0071]

[0072] As shown in Table 1, different scenarios correspond to different query contents. For example, if the determined scenario is personnel information retrieval, the corresponding query contents can be determined from Table 1 to include name, mobile phone number, system account, email address, and company role.

[0073] Step 103: Based on the first intention classification result, obtain the target key information corresponding to the problem data through the operation and maintenance knowledge graph.

[0074] In the disclosed embodiment, after obtaining the first intent classification result through the above steps, the target key information corresponding to the problem data can be obtained through the operation and maintenance knowledge graph based on the first intent classification result.

[0075] Specifically, in the embodiment of the present disclosure, the method for obtaining the target key information corresponding to the problem data through the operation and maintenance knowledge graph based on the first intention classification result may include the following steps:

[0076] Step 1031: based on the first intention classification result, obtain the corresponding target prompt word template;

[0077] Step 1032: extract the target entity in the target prompt word template;

[0078] Step 1033: extracting initial key information in the target prompt word template from the question data based on the target entity;

[0079] Step 1034: Based on the target entity and the initial key information, the target key information corresponding to the problem data is obtained through the operation and maintenance knowledge graph.

[0080] In the disclosed embodiment, a corresponding target prompt word template can be obtained based on the first intention classification result, and the target entity in the target prompt word template can be extracted, and based on the target entity, the initial key information in the target prompt word template can be extracted from the question data.

[0081] Specifically, in the embodiment of the present disclosure, the above can be based on the correspondence table between the intent classification and the target prompt word template to obtain the target prompt word template corresponding to the first intent classification result. For example, Table 2 is a correspondence table between an intent classification result and a target prompt word template proposed in the embodiment of the present disclosure.

[0082] Table 2

[0083]

[0084]

[0085]

[0086] In the disclosed embodiment, after obtaining the target prompt word template through the above steps, the target entity in the target prompt word template can be extracted, and based on the target entity, the initial key information in the target prompt word template can be extracted from the question data. For example, in the disclosed embodiment, if the first intent classification result is indicator query information, the initial key information obtained based on the target prompt word template is indicator, system and time information.

[0087] And, in the embodiment of the present disclosure, the above-mentioned method for obtaining the target key information corresponding to the problem data through the operation and maintenance knowledge graph based on the target entity and the initial key information may include: converting the target entity and the initial key information into an input parameter form corresponding to the query interface, and retrieving the target key information corresponding to the problem data from the operation and maintenance knowledge graph based on the input parameter form. In the embodiment of the present disclosure, the input parameter form corresponding to the target entity and the initial key information can be obtained according to the correspondence table between the intent classification result and the input parameter form. For example, Table 3 is a correspondence table between the intent classification result and the input parameter form proposed in the embodiment of the present disclosure.

[0088] Table 3

[0089]

[0090]

[0091] In the disclosed embodiment, based on the above-determined input parameter form, target key information corresponding to the problem data can be retrieved from the operation and maintenance knowledge graph.

[0092] Step 104: Determine the recommended information corresponding to the problem data based on the first scenario classification result.

[0093] In the disclosed embodiment, after obtaining the first scene classification result through the above steps, the recommended information corresponding to the question data can be determined based on the first scene classification result to obtain other information corresponding to the question data, thereby making the answer data corresponding to the question data more accurate.

[0094] Among them, in the embodiment of the present disclosure, the above-mentioned recommendation information may include context information, task information, information display strategy and solution information.

[0095] Furthermore, in the embodiment of the present disclosure, the method for determining the recommended information corresponding to the problem data based on the first scenario classification result may include the following steps:

[0096] Step 1041: Input the question data into the large language model to obtain corresponding query data;

[0097] Step 1042, determining context information of the problem data by querying the data and the vector index system;

[0098] Step 1043: Based on the first scene classification result, determine the task information of the problem data through the sub-data block vector library;

[0099] Step 1044: determine the user role to which the user belongs, and determine the information display strategy of the problem data according to the user role and task information;

[0100] Step 1045: Determine solution information for the problem data by querying the data and the fault case library.

[0101] In the disclosed embodiments, in order to support users in asking follow-up questions, processing references in context, or any commands related to the previous conversation context, question data can be input into a large language model to obtain corresponding query data, so that when processing the user's question data, the context of the question and answer can be taken into account, historical interaction data can be analyzed, and user preferences, common query patterns, and operating habits can be identified.

[0102] Specifically, in the disclosed embodiment, the question data can be input into the large language model, and the corresponding query data can be obtained through the query compression method. In the disclosed embodiment, each time a user asks a new question, the question data can be input into the large language model, and the large language model can compress the chat history of the user and the intelligent assistant and the last user question into a new query to obtain the corresponding query data.

[0103] And, in the embodiment of the present disclosure, after the corresponding query data is obtained through the above steps, the context information of the question data can be determined through the query data and the vector index system. Specifically, in the embodiment of the present disclosure, the method for determining the context information of the question data through the query data and the vector index system may include: determining the second intent classification result and the second scene classification result corresponding to the query data, determining the target prompt word of the query data according to the second intent classification result and the second scene classification result, inputting the target prompt word into the vector index system, and obtaining the context information of the question data.

[0104] Among them, in the embodiment of the present disclosure, the second intent classification result and the second scene classification result corresponding to the query data can be determined through a large language model, and the target prompt word of the query data can be determined based on the second intent classification result and the second scene classification result. There is a corresponding relationship between the target prompt word and the intent classification result and / or the scene classification result, and the target prompt word of the query data can be determined based on the corresponding relationship.

[0105] And, in the embodiment of the present disclosure, after the target prompt word is determined through the above steps, the target prompt word can be sent to a vector indexing system composed of related document blocks stored in the FAISS (Facebook AI Similarity Search) vector search library, and the vector indexing system can extract user preferences, query patterns and operating habits from historical interaction data to obtain contextual information of the question data. Among them, the target prompt word can be converted into a corresponding vector, and the contextual information of the question data can be obtained through the similarity between the vector and the vector in the vector indexing system.

[0106] Further, in the embodiment of the present disclosure, the method for determining the task information of the problem data through the sub-data block vector library based on the first scene classification result may include the following steps:

[0107] Step 10431: Based on the first scenario classification result, obtain scenario information of the problem data through the operation and maintenance knowledge graph;

[0108] Step 10432: convert the scene information into a target vector, and perform similarity calculation between the target vector and the sub-data block vector in the sub-data block vector library to obtain a similarity calculation result;

[0109] Step 10433: sort the similarity calculation results in descending order, and determine the sub-data blocks corresponding to the first preset number as the task information corresponding to the problem data.

[0110] Among them, in the embodiment of the present disclosure, the scenario information of obtaining problem data through the operation and maintenance knowledge graph may include timestamp, system name, service name, fault description and impact range data.

[0111] Furthermore, in the embodiment of the present disclosure, before the similarity calculation is performed between the target vector and the sub-data block vectors in the sub-data block vector library to obtain the similarity calculation result, the method may also include: determining a target text classification model, and establishing a sub-data block vector library based on the target text classification model.

[0112] Among them, in an embodiment of the present disclosure, the above-mentioned method for determining the target text classification model may include: preprocessing the original text data to obtain a training data set and a test data set for training the initial text classification model; using the training data set to train the initial text classification model to obtain a trained text classification model; using the test data set to test and tune the trained text classification model to obtain a target text classification model.

[0113] In the embodiment of the present disclosure, the original text data including operation and maintenance logs, fault reports and event records are preprocessed to obtain a training data set and a test data set for training the initial text classification model, which include timestamps, system names, service names, fault descriptions, impact ranges, and corresponding label text data including fault causes and solutions, wherein the corresponding fault causes and solutions are used as potential risk points and key tasks respectively. For example, Table 4 is a corresponding relationship table of data and content classification proposed in the embodiment of the present disclosure, and Table 5 is a corresponding relationship table of fault causes and solutions proposed in the embodiment of the present disclosure.

[0114] Table 4

[0115] Raw text data Operation and maintenance logs, fault reports, and event records Training text data Timestamp, system name, service name, fault description, impact scope Label text data Causes of failure and solutions

[0116] Table 5

[0117]

[0118]

[0119] In the disclosed embodiment, after obtaining the training data set through the above steps, the pre-trained BERT word vector model can be used for vector conversion to convert it into 512-dimensional vector data. At the same time, zero padding or truncation methods are used to process sentences with different sequence lengths. Finally, the vector sequence of uniform length is converted into a [32, 512, 512] dimensional tensor format for training the initial text classification model.

[0120] Furthermore, in the embodiment of the present disclosure, the preprocessed data can be divided into a training data set and a test data set for training the initial text classification model according to a preset ratio. For example, assuming that the preset ratio is 8:2, that is, 80% of the data in the data set is randomly divided into a training data set and 20% of the data is divided into a test data set.

[0121] Among them, in the embodiment of the present disclosure, the above-mentioned initial text classification model can be a Transformer model, and the initial text classification model can include an encoder of a multi-layer self-attention mechanism and a feedforward neural network for processing input semantic information of time, location, and event text data of operation and maintenance scenarios, and a decoder of a multi-layer self-attention mechanism and a feedforward neural network for generating classification results of corresponding key tasks and potential risk points. The use of an attention mechanism connection between the encoder and the decoder can realize the association between texts. Specifically, in the embodiment of the present disclosure, the number of layers of the above-mentioned encoder and decoder can be set to 12, the number of attention heads can be set to 8, the learning rate in other hyperparameters can be set to 0.0001, the optimizer uses the Adam optimizer, and the dropout ratio is set to 0.5.

[0122] Furthermore, in the embodiment of the present disclosure, the training method for the initial text classification model is the same as the existing method, and the embodiment of the present disclosure will not be described in detail here. Also, in the embodiment of the present disclosure, after the target text classification model is obtained through the above steps, the target text classification model can be used to classify the task-related documents of the operation and maintenance scenario in the same data processing method, thereby dividing different operation and maintenance scenarios into various detailed key tasks and corresponding potential risk points.

[0123] Among them, in the embodiment of the present disclosure, the classified small information blocks can be obtained through the above-mentioned target text classification model, and the small information blocks are used as child data blocks, and the child data blocks are also associated with their parent document data blocks at the previous level. The child data blocks can be stored in a child data block vector library, and a child data block vector library is established for subsequent matching retrieval.

[0124] And, in the embodiment of the present disclosure, after the sub-data block vector library is established through the above steps, the scene information can be converted into a target vector, and the similarity calculation result is obtained by performing similarity calculation between the target vector and the sub-data block vector in the sub-data block vector library, and the similarity calculation result is sorted in descending order, and the sub-data blocks corresponding to the first preset number at the front are determined as the task information corresponding to the problem data. Among them, in the embodiment of the present disclosure, the above-mentioned first preset number can be set as needed. And, in the embodiment of the present disclosure, if the above-mentioned first preset number of sub-data blocks are connected to the same parent node (i.e., a larger data block), the fault cause and solution corresponding to the parent node can be used as the output result.

[0125] Further, in the embodiment of the present disclosure, after obtaining the task information through the above steps, the key tasks in the task information can be used as the target tasks. Among them, in the embodiment of the present disclosure, the method of determining the role to which the user belongs and determining the information display strategy of the problem data according to the role and task information can include the following steps:

[0126] Step 10441: Establish a mapping relationship between roles, tasks, and display contents through the operation and maintenance knowledge graph;

[0127] Step 10442: Determine the user role to which the user belongs;

[0128] Step 10443: Based on the user role and target task, determine the information display strategy of the problem data through the mapping relationship.

[0129] Among them, in the embodiments of the present disclosure, the above-mentioned user roles, tasks and display contents can be regarded as different entities.

[0130] Furthermore, in the embodiment of the present disclosure, after determining the user role to which the user belongs, the information display strategy of the problem data can be determined based on the user role and the target task, as well as the mapping relationship between the role, task and display content, so as to dynamically adjust the content and order of the information display. Table 6 is a mapping relationship table of roles, tasks and display content provided in the embodiment of the present disclosure.

[0131] Table 6

[0132]

[0133]

[0134] Furthermore, in the embodiment of the present disclosure, in order to realize intelligent recommendation of fault solutions, it is also necessary to propose personalized recommendation strategies through a large language model to help operation and maintenance personnel quickly obtain fault solutions. Among them, in the embodiment of the present disclosure, the solution information of the problem data can be determined by querying the data and the fault case library.

[0135] Specifically, in the embodiment of the present disclosure, the method for determining the solution information of the problem data by querying the data and the fault case library may include: searching the target prompt word in the fault case library, determining the top second preset number of solutions, integrating the second preset number of solutions, and determining the integrated solutions as the solution information of the problem data. In the embodiment of the present disclosure, the second preset number of rows can be set as needed.

[0136] Among them, in the embodiments of the present disclosure, the solution information is obtained by combining the following information with a large language model, which can achieve rapid analysis and accurate matching of problem data. At the same time, it has good scalability and flexibility, can adapt to the fault solution recommendation needs in different fields, and provide users with more personalized and accurate fault resolution support.

[0137] Step 105: Input the target key information and the recommended information into the large language model to obtain answer data for the question data.

[0138] In the disclosed embodiment, after obtaining the target key information and recommended information through the above steps, the target key information and recommended information can be input into the large language model to obtain answer data for the question data, thereby making the answer data more in line with the actual needs of users, enhancing the pertinence and effectiveness of auxiliary decision-making, improving the accuracy of operation and maintenance, and further promoting the efficient transfer of operation and maintenance knowledge, accelerating team collaboration and problem-solving processes.

[0139] The present disclosure proposes an intelligent operation and maintenance method, which includes constructing an operation and maintenance knowledge graph; obtaining the problem data input by the user, and determining the first intent classification result and the first scenario classification result corresponding to the problem data; based on the first intent classification result, obtaining the target key information corresponding to the problem data through the operation and maintenance knowledge graph; based on the first scenario classification result, determining the recommended information corresponding to the problem data; inputting the target key information and the recommended information into a large language model to obtain the answer data of the problem data. Therefore, the present disclosure can input the target key information and the recommended information of the problem data into the large language model to obtain the answer data of the problem data, so that the answer data is more in line with the actual needs of the user, enhances the pertinence and effectiveness of auxiliary decision-making, improves the accuracy of operation and maintenance, and then promotes the efficient transmission of operation and maintenance knowledge, and accelerates team collaboration and problem solving process.

[0140] Figure 2 is an intelligent operation and maintenance device according to some embodiments of the present disclosure, such as Figure 2 As shown, the device may include:

[0141] Construction module 201, used to construct an operation and maintenance knowledge graph;

[0142] A first determination module 202 is used to obtain question data input by the user and determine a first intent classification result and a first scene classification result corresponding to the question data;

[0143] The second determination module 203 is used to obtain target key information corresponding to the problem data through the operation and maintenance knowledge graph based on the first intention classification result;

[0144] A third determination module 204 is used to determine the recommended information corresponding to the problem data based on the first scenario classification result;

[0145] The processing module 205 is used to input the target key information and the recommended information into the large language model to obtain answer data of the question data.

[0146] In the embodiment of the present disclosure, the second determining module 203 is specifically used to:

[0147] Based on the first intention classification result, the corresponding target prompt word template is obtained;

[0148] Extract the target entity in the target prompt word template;

[0149] Based on the target entity, extract the initial key information in the target prompt word template from the question data;

[0150] Based on the target entity and initial key information, the target key information corresponding to the problem data is obtained through the operation and maintenance knowledge graph.

[0151] Optionally, in an embodiment of the present disclosure, the above-mentioned recommendation information includes: context information, task information, information display strategy and solution information.

[0152] Optionally, in the embodiment of the present disclosure, the third determining module 204 is specifically configured to:

[0153] Input the question data into the large language model to obtain the corresponding query data;

[0154] Determine the contextual information of the problem data by querying the data and vector indexing system;

[0155] Based on the first scene classification result, determining the task information of the problem data through the sub-data block vector library;

[0156] Determine the user role to which the user belongs, and determine the information display strategy for the problem data based on the user role and task information;

[0157] By querying the data and fault case library, the solution information for the problem data is determined.

[0158] Optionally, in the embodiment of the present disclosure, the third determining module 204 is specifically configured to:

[0159] Determine a second intent classification result and a second scene classification result corresponding to the query data;

[0160] Determine a target prompt word for the query data according to the second intent classification result and the second scene classification result;

[0161] The target prompt word is input into the vector index system to obtain the context information of the question data.

[0162] Optionally, in the embodiment of the present disclosure, the third determining module 204 is specifically configured to:

[0163] Based on the first scenario classification result, the scenario information of the problem data is obtained through the operation and maintenance knowledge graph;

[0164] The scene information is converted into a target vector, and similarity calculation is performed between the target vector and the sub-data block vector in the sub-data block vector library to obtain a similarity calculation result;

[0165] The similarity calculation results are sorted in descending order, and the sub-data blocks corresponding to the first preset number at the front are determined as the task information corresponding to the problem data.

[0166] Optionally, in the embodiment of the present disclosure, the third determining module 204 is further configured to:

[0167] Determine the target text classification model;

[0168] Based on the target text classification model, a sub-data block vector library is established.

[0169] Optionally, in the embodiment of the present disclosure, the task information includes a target task; and the third determination module 204 is further configured to:

[0170] Establish a mapping relationship between roles, tasks, and display content through the operation and maintenance knowledge graph;

[0171] Determine the user role to which the user belongs;

[0172] Based on user roles and target tasks, the information display strategy of the problem data is determined through mapping relationships.

[0173] Optionally, in the embodiment of the present disclosure, the third determining module 204 is further configured to:

[0174] Searching the target prompt word in the fault case library to determine the top second preset number of solutions;

[0175] The second preset number of solutions are integrated, and the integrated solutions are determined as solution information of the problem data.

[0176] In one or more embodiments of the present disclosure, an operation and maintenance knowledge graph is constructed; the problem data input by the user is obtained, and the first intent classification result and the first scenario classification result corresponding to the problem data are determined; based on the first intent classification result, the target key information corresponding to the problem data is obtained through the operation and maintenance knowledge graph; based on the first scenario classification result, the recommended information corresponding to the problem data is determined; the target key information and the recommended information are input into the large language model to obtain the answer data of the problem data. Therefore, the present disclosure can input the target key information and the recommended information of the problem data into the large language model to obtain the answer data of the problem data, so that the answer data is more in line with the actual needs of the user, enhance the pertinence and effectiveness of auxiliary decision-making, improve the accuracy of operation and maintenance, and then promote the efficient transmission of operation and maintenance knowledge, and accelerate team collaboration and problem-solving processes.

[0177] like Figure 3 As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 302 or a computer program loaded from a storage unit 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 can also be stored. The computing unit 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0178] A number of components in the electronic device 300 are connected to the I / O interface 305, including: an input unit 308, such as a keyboard, a mouse, etc.; an output unit 307, such as various types of displays, speakers, etc.; a storage unit 308, such as a disk, an optical disk, etc.; and a communication unit 309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 309 allows the electronic device 300 to exchange information / data with other electronic devices through a computer network such as the Internet and / or various telecommunication networks.

[0179] The computing unit 301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 301 performs the various methods and processes described above, such as intelligent operation and maintenance methods. For example, in some embodiments, the intelligent operation and maintenance method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the intelligent operation and maintenance method described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform the intelligent operation and maintenance method in any other appropriate manner (e.g., by means of firmware).

[0180] Various embodiments of the devices and techniques described above herein can be implemented in digital electronic circuit devices, integrated circuit devices, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), devices on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable device including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage device, at least one input device, and at least one output device, and transmit data and instructions to the storage device, the at least one input device, and the at least one output device.

[0181] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0182] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution device, device, or electronic device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor devices, devices, or electronic devices, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage electronic device, a magnetic storage electronic device, or any suitable combination of the foregoing.

[0183] To provide interaction with a user, the devices and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0184] The apparatus and techniques described herein may be implemented in a computing device including backend components (e.g., as a data server), or a computing device including middleware components (e.g., an application server), or a computing device including frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the apparatus and techniques described herein), or a computing device including any combination of such backend components, middleware components, or frontend components. The components of the apparatus may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0185] A computer device may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed device, or a server combined with a blockchain.

[0186] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0187] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An intelligent operation and maintenance method, characterized in that: The method comprises: Build an operation and maintenance knowledge graph; Obtaining question data input by a user, and determining a first intent classification result and a first scene classification result corresponding to the question data; Based on the first intent classification result, obtaining target key information corresponding to the problem data through the operation and maintenance knowledge graph; Determining recommendation information corresponding to the problem data based on the first scenario classification result; The target key information and the recommended information are input into a large language model to obtain answer data for the question data.

2. The method according to claim 1, characterized in that The obtaining, based on the first intention classification result, target key information corresponding to the problem data through the operation and maintenance knowledge graph includes: Based on the first intent classification result, a corresponding target prompt word template is obtained; Extracting the target entity in the target prompt word template; Based on the target entity, extracting initial key information in the target prompt word template from the question data; Based on the target entity and the initial key information, the target key information corresponding to the problem data is obtained through the operation and maintenance knowledge graph.

3. The method according to claim 1, characterized in that The recommendation information includes: context information, task information, information display strategy and solution information.

4. The method according to claim 3, characterized in that The determining, based on the first scenario classification result, recommendation information corresponding to the problem data includes: Inputting the question data into a large language model to obtain corresponding query data; Determining context information of the question data through the query data and the vector index system; Based on the first scene classification result, determining the task information of the problem data through a sub-data block vector library; Determine the user role to which the user belongs, and determine the information display strategy of the problem data according to the user role and the task information; The solution information of the problem data is determined through the query data and the fault case library.

5. The method according to claim 4, characterized in that Determining the context information of the question data through the query data and the vector index system includes: Determine a second intent classification result and a second scene classification result corresponding to the query data; Determining a target prompt word for the query data according to the second intent classification result and the second scene classification result; The target prompt word is input into a vector index system to obtain context information of the question data.

6. The method according to claim 4, characterized in that The determining the task information of the problem data through a sub-data block vector library based on the first scene classification result includes: Based on the first scenario classification result, acquiring scenario information of the problem data through the operation and maintenance knowledge graph; The scene information is converted into a target vector, and similarity calculation is performed between the target vector and the sub-data block vector in the sub-data block vector library to obtain a similarity calculation result; The similarity calculation results are sorted in descending order, and sub-data blocks corresponding to a first preset number of front blocks are determined as task information corresponding to the problem data.

7. The method according to claim 6, characterized in that Before performing similarity calculation between the target vector and the sub-data block vectors in the sub-data block vector library to obtain a similarity calculation result, the method further includes: Determine the target text classification model; Based on the target text classification model, a sub-data block vector library is established.

8. The method according to claim 4, characterized in that The task information includes a target task; The determining of the role to which the user belongs, and determining the information display strategy of the problem data according to the role and the task information, includes: Through the operation and maintenance knowledge graph, a mapping relationship between roles, tasks and display contents is established; Determining the user role to which the user belongs; Based on the user role and target task, the information display strategy of the problem data is determined through the mapping relationship.

9. The method according to claim 5, characterized in that Determining the solution information of the problem data through the query data and the fault case library includes: Searching the target prompt word in the fault case library to determine a second preset number of top solutions; The second preset number of solutions are integrated, and the integrated solutions are determined as the solution information of the problem data.

10. An intelligent operation and maintenance device, characterized in that: The device comprises: Construction module, used to build operation and maintenance knowledge graph; A first determination module, used to obtain question data input by a user, and determine a first intent classification result and a first scene classification result corresponding to the question data; A second determination module is used to obtain target key information corresponding to the problem data through the operation and maintenance knowledge graph based on the first intention classification result; A third determination module, configured to determine the recommended information corresponding to the problem data based on the first scenario classification result; A processing module is used to input the target key information and the recommended information into a large language model to obtain answer data for the question data.

Images