Response information generation method and device, intelligent operation and maintenance system and electronic equipment

Abstract

The embodiment of the invention provides a response information generation method and device, an intelligent operation and maintenance system and electronic equipment, and the method comprises the steps: obtaining multi-source heterogeneous data in an operation and maintenance scene, the multi-source heterogeneous data being data of at least two different data formats generated in a process of executing an operation and maintenance task through the intelligent operation and maintenance system; the multi-source heterogeneous data is processed through a large language model, a private domain knowledge database for the operation and maintenance scene is generated, structured data with a target structure is stored in the private domain knowledge database, and the structured data is used for recording operation and maintenance knowledge for processing operation and maintenance problems; and in response to an inquiry instruction, calling the private domain knowledge database through the large language model, and generating response information corresponding to the inquiry instruction, the inquiry instruction being used for inquiring the operation and maintenance problem, and the response information being used for representing a processing flow for the operation and maintenance problem. The range of operation and maintenance problems which can be covered by the intelligent operation and maintenance system is expanded, and the accuracy of response information is improved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of intelligent operation and maintenance technology, and in particular to a response information generation method, device, intelligent operation and maintenance system, and electronic equipment. Background Art

[0002] At present, with the rapid increase in the scale of various Internet businesses, the corresponding business operation and maintenance costs are also increasing rapidly. In order to reduce the operation and maintenance costs of various businesses and equipment and improve operation and maintenance efficiency, relevant technologies have realized intelligent operation and maintenance for different businesses by utilizing operation and maintenance data and manual experience, combining big data, artificial intelligence and other technologies. For example, according to the operation and maintenance questions raised by users, corresponding response information is automatically generated.

[0003] However, the specific execution effect of the intelligent operation and maintenance technology in the prior art depends on the quality of training and tuning the operation and maintenance model in the intelligent operation and maintenance system based on the operation and maintenance knowledge. In actual applications, due to various types of private domain operation and maintenance knowledge, experienced operation and maintenance users are still required to manually collect it, and then train and configure the operation and maintenance model based on the above private domain operation and maintenance knowledge so that it can output some response information about operation and maintenance problems. Therefore, in the application scenarios of intelligent operation and maintenance, the solutions in the prior art have the problems of inaccurate response information output by the intelligent operation and maintenance system and small problem coverage. Summary of the invention

[0004] The embodiments of the present disclosure provide a response information generation method, device, intelligent operation and maintenance system, and electronic device to overcome the problems that the response information output by the intelligent operation and maintenance system is inaccurate and the problem coverage is small.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for generating response information, including:

[0006] Acquire multi-source heterogeneous data in an operation and maintenance scenario, wherein the multi-source heterogeneous data is data in at least two different data formats generated in the process of executing an operation and maintenance task through the intelligent operation and maintenance system; process the multi-source heterogeneous data through a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data with a target structure, and the structured data is used to record operation and maintenance knowledge for handling operation and maintenance problems; in response to an inquiry instruction, call the private domain knowledge database through the large language model to generate response information corresponding to the inquiry instruction, wherein the inquiry instruction is used to inquire about the operation and maintenance problem, and the response information is used to characterize the processing flow for the operation and maintenance problem.

[0007] In a second aspect, an embodiment of the present disclosure provides a response information generating device, including:

[0008] An acquisition module, used to acquire multi-source heterogeneous data in an operation and maintenance scenario, wherein the multi-source heterogeneous data is data in at least two different data formats generated during the execution of an operation and maintenance task by the intelligent operation and maintenance system;

[0009] A processing module, used to process the multi-source heterogeneous data through a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data having a target structure, and the structured data is used to record the operation and maintenance knowledge for handling operation and maintenance problems;

[0010] A generation module is used to respond to an inquiry instruction, call the private domain knowledge database through the large language model, and generate response information corresponding to the inquiry instruction, wherein the inquiry instruction is used to inquire about operation and maintenance issues, and the response information is used to characterize the processing flow for the operation and maintenance issues.

[0011] In a third aspect, an embodiment of the present disclosure provides an intelligent operation and maintenance system for executing the response information generation method described in the first aspect and various possible designs of the first aspect.

[0012] In a fourth aspect, an embodiment of the present disclosure provides an electronic device, including: a processor and a memory;

[0013] The memory stores computer-executable instructions;

[0014] The processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the response information generation method as described in the first aspect and various possible designs of the first aspect.

[0015] In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the response information generation method described in the first aspect and various possible designs of the first aspect is implemented.

[0016] In a sixth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements the response information generation method as described in the first aspect and various possible designs of the first aspect.

[0017] The response information generation method, device, intelligent operation and maintenance system and electronic device provided in the present embodiment obtain multi-source heterogeneous data in an operation and maintenance scenario, wherein the multi-source heterogeneous data is data in at least two different data formats generated in the process of performing operation and maintenance tasks by the intelligent operation and maintenance system; the multi-source heterogeneous data is processed by a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data with a target structure, and the structured data is used to record operation and maintenance knowledge for handling operation and maintenance problems; in response to a query instruction, the private domain knowledge database is called by the large language model to generate response information corresponding to the query instruction, wherein the query instruction is used to inquire about the operation and maintenance problem, and the response information is used to characterize the processing flow for the operation and maintenance problem. By acquiring multi-source heterogeneous data in the operation and maintenance scenario and converting it into a private domain knowledge database for the operation and maintenance scenario using a large language model, after receiving a query instruction, the operation and maintenance knowledge in the private domain knowledge database is used by the large language model to generate response information representing the processing flow for the operation and maintenance problem, thereby realizing full-link intelligent operation and maintenance, enabling the large language model to autonomously learn various types of private domain knowledge without the need for manual configuration by the operation and maintenance personnel, thereby increasing the scope of operation and maintenance problems that can be covered by the intelligent operation and maintenance system and improving the accuracy of the response information. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 An application scenario diagram of the response information generation method provided in an embodiment of the present disclosure;

[0020] Figure 2 Schematic diagram of the process of generating a response message provided by the embodiment of the present disclosure Figure 1 ;

[0021] Figure 3 A schematic diagram of the structure of an operation and maintenance knowledge graph provided in an embodiment of the present disclosure;

[0022] Figure 4 for Figure 2 A flowchart of a specific implementation method of step S102 in the illustrated embodiment;

[0023] Figure 5 for Figure 2 A flowchart of a specific implementation method of step S103 in the illustrated embodiment;

[0024] Figure 6 for Figure 5 A flowchart of a specific implementation method of step S1033 in the illustrated embodiment;

[0025] Figure 7 Schematic diagram of the process of generating a response message provided by the embodiment of the present disclosure Figure 2 ;

[0026] Figure 8 for Figure 7 A flowchart of a specific implementation method of step S203 in the illustrated embodiment;

[0027] Fig. 9 for Figure 8 A flowchart of a specific implementation method of step S2032 in the illustrated embodiment;

[0028] Fig.10 A schematic diagram of a process for implementing operation and maintenance knowledge question and answering through an intelligent operation and maintenance system provided in an embodiment of the present disclosure;

[0029] Fig.11 A structural block diagram of a response information generating device provided in an embodiment of the present disclosure;

[0030] Fig.12 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure;

[0031] Fig.13 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0033] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0034] The application scenarios of the embodiments of the present disclosure are explained below:

[0035] The response information generation method provided in the embodiment of the present disclosure can be applied to an intelligent operation and maintenance system, and more specifically, can be applied to application scenarios such as intelligent operation and maintenance question and answer, intelligent operation and maintenance diagnosis, etc. The execution subject of this embodiment can be a server that deploys an intelligent operation and maintenance system, or other electronic devices with computing capabilities.

[0036] Among them, in some embodiments, the server or other electronic devices can implement the response information generation method provided by the embodiment of the present disclosure by running various computer executable instructions or computer programs. For example, computer executable instructions can be program-level commands, machine instructions or software instructions. The computer program can be a native program or software module in the operating system; it can be a local application, that is, a program that needs to be installed in the operating system to run. In summary, the above-mentioned computer executable instructions can be instructions in any form, and the above-mentioned computer program can be an application, module or plug-in in any form, and the specific implementation form can be configured as needed. Further, taking the server as an example, in the process of implementing the response information generation method provided by the embodiment of the present disclosure, the server can execute the method by running a computer executable instruction or computer program set locally, or by calling a computer executable instruction or computer program set in an external server. In some embodiments, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud storage, cloud communications, cloud databases, cloud computing, cloud functions, network services, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), as well as big data and artificial intelligence platforms. Among them, cloud services can be interactive processing services for terminal devices to call.

[0037] Figure 1 This is an application scenario diagram of the response information generation method provided in the embodiment of the present disclosure, referring to Figure 1As shown in , the intelligent operation and maintenance system (at least a part of it) is deployed in the server. The operation and maintenance user communicates with the server through the terminal device, and sends a query statement for the operation and maintenance scenario to the server through the intelligent operation and maintenance system. The content is, for example, "The user-side delay of business APP_1 is too high, check the cause". After receiving the query statement, the server identifies the query statement, generates a corresponding query instruction and responds, and processes the query instruction by calling a large language model (LargeLanguage Model, LLM) deployed locally or externally, and generates corresponding response information. The response information is the answer to the above query statement, which may include text, images, address links, etc. Afterwards, the response information is returned to the terminal device and displayed on the side of the terminal device, thereby completing a process of intelligent operation and maintenance question and answer.

[0038] In the prior art, the execution effect of generating response information for operation and maintenance problems through a large language model depends on the quality of training and tuning the large language model in the intelligent operation and maintenance system based on operation and maintenance knowledge. Due to various types of private domain operation and maintenance knowledge, experienced operation and maintenance users are still required to manually collect them, and then train and configure the large language model based on the above private domain operation and maintenance knowledge so that it can output some response information about operation and maintenance problems. Therefore, the configuration cost of the large language model is very high, which makes it difficult for the large language model to obtain sufficient private domain operation and maintenance knowledge to accurately answer the operation and maintenance questions raised by users. Based on the above reasons, in the application scenarios of intelligent operation and maintenance, the solutions in the prior art have the problems of inaccurate response information output by the intelligent operation and maintenance system and small problem coverage.

[0039] The embodiment of the present disclosure provides a response information generating method to solve the above-mentioned problem.

[0040] refer to Figure 2 , Figure 2 Schematic diagram of the process of generating a response message provided by the embodiment of the present disclosure Figure 1 The method of this embodiment can be applied in an intelligent operation and maintenance system. The response information generation method includes:

[0041] Step S101: Acquire multi-source heterogeneous data in an operation and maintenance scenario, where the multi-source heterogeneous data is data in at least two different data formats generated in the process of executing an operation and maintenance task through an intelligent operation and maintenance system.

[0042] Step S102: Process multi-source heterogeneous data through a large language model to generate a private domain knowledge database for operation and maintenance scenarios. The private domain knowledge database stores structured data with a target structure. The structured data is used to record operation and maintenance knowledge for handling operation and maintenance problems.

[0043] For example, this embodiment is described by taking the server running the intelligent operation and maintenance system as the execution subject of the method provided in this embodiment, and referring to Figure 1 In the schematic diagram of the application scenario shown, the server will first collect multi-source heterogeneous data from the operation and maintenance scenario. Multi-source heterogeneous data is data in at least two different data formats generated in the process of executing the operation and maintenance task through the intelligent operation and maintenance system, such as text data, image data, log data, work order data, etc. The data in the above different data formats can be obtained by calling different interfaces. For example, text data and image data can be obtained by calling the interface of the instant messaging software associated with the intelligent operation and maintenance system. The text data and image data are, for example, data generated during the chat process of the operation and maintenance personnel through the instant messaging software. For another example, log data can be obtained by the intelligent operation and maintenance system by calling the reading interface corresponding to the historical maintenance record. The historical maintenance record is the historical processing flow recorded by the system when the operation and maintenance problem was handled. The specific implementation method of the data in the above different data formats can be set according to the specific situation, and there is no restriction this time. In short, the multi-source heterogeneous data in the operation and maintenance scenario can be understood as all data related to the operation and maintenance scenario. The intelligent operation and maintenance system obtains the multi-source heterogeneous data in the above operation and maintenance scenario through various services and devices connected to it. At the same time, since the multi-source heterogeneous data has different data formats, it cannot be used directly after being obtained. To become specific and relevant operation and maintenance knowledge, it needs to be further processed based on subsequent steps before it can be used.

[0044] Furthermore, after obtaining multi-source heterogeneous data, the intelligent operation and maintenance system processes the multi-source heterogeneous data to generate a private domain knowledge database of structured data with a target structure, wherein the structured data is used to record the operation and maintenance knowledge for handling operation and maintenance problems. This step sorts out the operation and maintenance knowledge and aligns the data format by screening, classifying, and fusing the information in the multi-source heterogeneous data, so that the generated private domain knowledge database becomes a formatted database that can be used by the big model. This process does not require human participation, so it can greatly improve the efficiency of the big model in learning and acquiring operation and maintenance knowledge.

[0045] In a possible implementation, the private domain knowledge database includes an operation and maintenance knowledge graph and / or an operation and maintenance task template, wherein the operation and maintenance knowledge graph is used to represent multiple operation and maintenance entities in an operation and maintenance scenario, as well as the causal relationship and dependency relationship between the operation and maintenance entities; the operation and maintenance task template is used to generate at least one text of the processing flow for the operation and maintenance task. The operation and maintenance knowledge graph records the solutions corresponding to various operation and maintenance problems, and the operation and maintenance task template can form the corresponding explanatory text, i.e., the response information, to achieve efficient and clean display of the solutions to the operation and maintenance problems.

[0046] Figure 3A schematic diagram of the structure of an operation and maintenance knowledge graph provided in an embodiment of the present disclosure, exemplarily, as Figure 3 As shown, the operation and maintenance knowledge graph is composed of the entity part and the connection between entities, wherein the entity part is the operation and maintenance entity, and the operation and maintenance entity is, for example, the operation and maintenance problems and execution actions corresponding to the equipment, service, configuration item, etc. The connection between entities is the causal relationship and dependency relationship between the operation and maintenance entities, such as alarm association, operation dependency, etc. Through the above operation and maintenance knowledge graph, all possible processing methods of the recorded operation and maintenance problems can be obtained, such as the operation and maintenance entity P1 shown in the figure (shown as P1 in the figure), which represents "Pod is in PENDING state". The operation and maintenance entities connected to it include operation and maintenance entity P2, operation and maintenance entity P3 and operation and maintenance entity P4 (shown as P2, P3, P4 respectively in the figure). Among them, the operation and maintenance entity P2 represents "viewing the cloud service management interface", the operation and maintenance entity P3 represents "speculating the automatic expansion problem", and the operation and maintenance entity P4 represents "requesting operation and maintenance personnel A to perform detection". As can be seen from the above example, the operation and maintenance entity P1 describes an operation and maintenance problem, while its downstream operation and maintenance entity P2, operation and maintenance entity P3, and operation and maintenance entity P4 respectively describe an action for handling the operation and maintenance problem. Furthermore, the operation and maintenance entity P1 can also be connected to the upstream operation and maintenance entity. For example, the operation and maintenance entity P1 is connected to the upstream operation and maintenance entity P0 (shown as P0 in the figure), and the operation and maintenance entity P0 represents "the occurrence of a platform alarm", that is, the alarm association between the two. Furthermore, the operation and maintenance entity P2, operation and maintenance entity P3, and operation and maintenance entity P4 can also further include downstream connected operation and maintenance entities to describe the specific implementation methods of the actions represented by each operation and maintenance entity. Through the above-mentioned operation and maintenance knowledge graph, the description of the processing methods and processing links corresponding to different operation and maintenance problems can be realized, that is, the operation and maintenance knowledge for handling operation and maintenance problems is recorded.

[0047] Furthermore, in a possible implementation, Figure 4 As shown, the specific implementation of step S102 includes:

[0048] Step S1021: using a large language model to perform reasoning based on multi-source heterogeneous data, multiple operation and maintenance entities, as well as causal relationships and dependency relationships between the multiple operation and maintenance entities are obtained;

[0049] Step S1022: Processing the causal relationship and dependency relationship between multiple operation and maintenance entities based on the graph reasoning algorithm to obtain a supplementary association relationship between at least two operation and maintenance entities;

[0050] Step S1023: Generate an operation and maintenance knowledge graph based on the causal relationship, dependency relationship and supplementary association relationship between each operation and maintenance entity.

[0051] For example, first, through the semantic recognition and logical reasoning capabilities of the large language model, effective information is extracted from multi-source heterogeneous data and reasoning and induction are performed to determine multiple operation and maintenance entities, as well as the causal and dependency relationships between operation and maintenance entities. Then, based on the graph reasoning algorithm, more connections that represent the implicit relationships between operation and maintenance entities are obtained based on the causal and dependency relationships between operation and maintenance entities, that is, supplementary association relationships, as a supplement to the above causal and dependency relationships. Finally, based on the causal relationships, dependencies, and supplementary association relationships between each operation and maintenance entity, an operation and maintenance knowledge graph is constructed.

[0052] In this embodiment, an operation and maintenance knowledge graph is constructed through entity recognition and classification technology. This process does not sample deep learning models (such as BERT, RoBERTa, etc.) for named entity recognition (NER), but autonomously identifies entities that can be extracted through the reasoning ability of a large language model. Through relationship extraction and graph reasoning, based on a pre-trained language model, the causal relationship and dependency relationship between entities, such as alarm association and operation dependency, are automatically identified. Combined with the graph reasoning algorithm, the supplementary association relationship that represents the implicit relationship is further discovered, and new knowledge is derived to ensure the comprehensiveness and real-time nature of the operation and maintenance knowledge graph. Through, with dynamic update and reasoning capabilities, the operation and maintenance knowledge graph obtains new knowledge from new operation and maintenance data (i.e., multi-source heterogeneous data in the operation and maintenance scenario) through incremental learning, and updates based on this. Through the incremental learning algorithm, the reconstruction of the entire graph is avoided, and the real-time nature is improved. Through reinforcement learning and adaptive reasoning algorithms, the system can continuously self-adjust in a complex environment and optimize the graph structure.

[0053] Step S103: In response to the query instruction, the private domain knowledge database is called through the large language model to generate response information corresponding to the query instruction, wherein the query instruction is used to inquire about operation and maintenance issues, and the response information is used to characterize the processing flow for the operation and maintenance issues.

[0054] Exemplarily, after the private domain knowledge database is generated, when the intelligent operation and maintenance system receives the query statement sent by the user through the terminal device, it will perform semantic understanding of the query statement and generate a corresponding query instruction, which is used to inquire about a specific operation and maintenance problem, such as "an alarm occurred on platform A", "high latency in service response", etc. Afterwards, the intelligent operation and maintenance system processes the query instruction by calling the large language model. During this process, the large language model can perform tasks such as problem reasoning and tool calling, thereby generating response information that represents the processing flow for the operation and maintenance problem. After that, the intelligent operation and maintenance system returns the response information generated by the large language model to the terminal device, thereby completing the operation and maintenance intelligent question and answer process for the query statement entered by the user.

[0055] Furthermore, in a possible implementation, Figure 5 As shown, the specific implementation of step S103 includes:

[0056] Step S1031: Obtaining the target inquiry category corresponding to the inquiry instruction;

[0057] Step S1032: generating a target prompt word corresponding to the inquiry instruction according to the target inquiry category, the target prompt word being used to represent the step of processing the operation and maintenance problem based on the target inquiry category;

[0058] Step S1033: input the target prompt word and the inquiry instruction into the large language model to control the large language model to obtain the target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the inquiry instruction based on the target operation and maintenance knowledge.

[0059] Exemplarily, first, the server first performs semantic recognition on the query instruction through the large language model to obtain the query category to which the query instruction belongs, that is, the target query category corresponding to the query instruction. Among them, the query category includes, for example, specific information query, oncall replay, standard operating procedure (SOP) document generation, fault analysis and suggestions, etc. Afterwards, according to the target query category corresponding to the query instruction, the target prompt word corresponding to the query instruction is generated, and the target prompt word is used to characterize the steps of handling operation and maintenance problems based on the target query category. Among them, the prompt word (prompt) is a text used to guide the model to generate a specific output or perform a specific task. In this embodiment, the prompt word or prompt word template can be pre-configured, and the server determines the target prompt word that matches it through the target query category. After that, by inputting the target prompt word and the corresponding query instruction into the large language model, the large language model is processed. The control of the content output. For example, when the target inquiry category is the generation of a standard operating procedure document, corresponding to the first prompt word template, the inquiry instruction and the content of the first prompt word template are combined and input into the large language model, so that the response information output by the large language model is the processing flow for the generation of the standard operating procedure document.

[0060] In the steps of this embodiment, a target prompt word matching the target inquiry category corresponding to the inquiry instruction is obtained, and then the large language model is called in combination with the prompt word to control the large language model to obtain target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the inquiry instruction based on the target operation and maintenance knowledge, so that the generated response information can match the content inquired by the inquiry instruction, thereby improving the accuracy of the response information.

[0061] Furthermore, in the process where the large language model obtains target operation and maintenance knowledge from the private domain knowledge database and generates response information corresponding to the query instruction based on the target operation and maintenance knowledge, the large language model can execute the corresponding reasoning steps by calling the target operation and maintenance tool and generating an automated diagnosis script, thereby obtaining response information that matches the query instruction.

[0062] Specifically, in a possible implementation, the target operation and maintenance knowledge includes platform tool information, such as Figure 6 As shown, the specific implementation of step S1033 includes:

[0063] Step S1033A-1: Obtain platform tool information corresponding to the query instruction from the private domain knowledge database through the large language model, where the platform tool information represents the operation and maintenance tool used in the external platform to execute the processing flow for the operation and maintenance problem corresponding to the query instruction.

[0064] Step S1033A-2: Register and call the target operation and maintenance tool indicated by the platform tool information based on the platform tool information through the large language model to generate response information corresponding to the query instruction.

[0065] Exemplarily, after obtaining the target prompt word, the server can obtain the platform tool information corresponding to the query instruction from the private domain knowledge database based on the target prompt word. The platform tool information represents the operation and maintenance tools used in the external platform to execute the processing flow of the operation and maintenance problems corresponding to the query instruction. In short, the platform tool information is a part of the operation and maintenance knowledge recorded in the private domain knowledge database. The private domain knowledge database records the external operation and maintenance tools used to handle various recorded operation and maintenance problems. The large language model obtains the corresponding operation and maintenance knowledge from the private domain knowledge database based on the target prompt word and the query instruction, including the platform tool information representing the operation and maintenance tools used in the external platform to execute the processing flow of the operation and maintenance problems corresponding to the query instruction. Afterwards, based on the platform tool information, the target operation and maintenance tool indicated by the platform tool information is registered and called, and then the response information corresponding to the query instruction is generated by calling the target operation and maintenance tool. In the steps of this embodiment, platform tool information is extracted from the operation and maintenance knowledge contained in various multi-source heterogeneous data, and the function-call capability of the large language model is used to register the platform tool information as executable tools, such as data analysis tools, trend forecasting tools, etc., thereby improving the ability to solve operation and maintenance problems and increasing the coverage of operation and maintenance problems.

[0066] In another possible implementation, after obtaining the target operation and maintenance knowledge, a specific implementation of generating response information corresponding to the query instruction based on the target operation and maintenance knowledge includes:

[0067] Step S1033B-1: Obtain real-time alarm data of the intelligent operation and maintenance system.

[0068] Step S1033B-2: Generate an automated diagnosis script using a large language model, real-time alarm data and target operation and maintenance knowledge, and register the diagnostic items provided by the external platform by running the automated diagnosis script.

[0069] Step S1033B-3: Generate response information by executing the diagnosis item.

[0070] Exemplarily, after obtaining the target operation and maintenance knowledge, the server can generate an automated diagnosis script through a large language model, using the reasoning ability of the large language model, combined with real-time alarm data and target operation and maintenance knowledge. The automated diagnosis script records one or more diagnostic items, and the diagnostic items are the targets or diagnostic steps that need to be diagnosed. After that, by running the automated diagnosis script, the diagnostic items provided by the external platform are registered to obtain the ability to execute the diagnostic items. After that, the diagnostic results, i.e., response information, are generated by sequentially executing each diagnostic item indicated by the automated diagnosis script. In the steps of this embodiment, the system obtains real-time alarm data through the monitoring platform, and uses the intelligent operation and maintenance model in the intelligent operation and maintenance system to perform time series analysis and predict potential faults. Based on the generation of the diagnostic document, the system can directly and automatically execute the diagnostic script, and interact with external tools through the REST API to register diagnostic items to generate response information, thereby improving the ability to solve operation and maintenance problems and increasing the coverage of operation and maintenance problems.

[0071] refer to Figure 7 , Figure 7 Schematic diagram of the process of generating a response message provided by the embodiment of the present disclosure Figure 2 In this embodiment Figure 2 On the basis of the illustrated embodiment, step S102 is further refined and applied to the intelligent operation and maintenance system. The response information generation method includes:

[0072] Step S201: Acquire multi-source heterogeneous data in an operation and maintenance scenario, where the multi-source heterogeneous data is data in at least two different data formats generated in the process of executing an operation and maintenance task through an intelligent operation and maintenance system.

[0073] Step S202: Filter the multi-source heterogeneous data through a large language model to obtain effective multi-source heterogeneous data.

[0074] Step S203: reconstruct the effective multi-source heterogeneous data to generate structured data with a target structure.

[0075] For example, after the intelligent operation and maintenance system or the server deploying the intelligent operation and maintenance system obtains multi-source heterogeneous data from different sources by configuring subscription information, it first screens the multi-source heterogeneous data to select meaningful data and removes meaningless, incomplete, and unrecognizable data, thereby obtaining effective multi-source heterogeneous data. Afterwards, the effective multi-source heterogeneous data is reconstructed and converted into structured data with a unified target structure.

[0076] In a possible implementation, the effective multi-source heterogeneous data includes first effective multi-source heterogeneous data and second effective multi-source heterogeneous data, the first effective multi-source heterogeneous data is non-text data, the second effective multi-source heterogeneous data is text data, and a multi-modal large model is configured in the intelligent operation and maintenance system, such as Figure 8 As shown, the specific implementation of step S203 includes:

[0077] Step S2031: Generate text data by calling the multimodal large model to fuse the first effective multi-source heterogeneous data and the second effective multi-source heterogeneous data.

[0078] Step S2032: Process the text data based on the target structure to generate structured data.

[0079] Exemplarily, the effective multi-source heterogeneous data includes first effective multi-source heterogeneous data and second effective multi-source heterogeneous data, wherein the first effective multi-source heterogeneous data is non-text data, such as image data, voice data, etc., and the second effective multi-source heterogeneous data is text data. Specifically, the first effective multi-source heterogeneous data is, for example, pictures in an operation and maintenance chat group or an alarm processing group. After obtaining such first effective multi-source heterogeneous data, in order to extract effective information therein, the text content in the first effective multi-source heterogeneous data is first extracted to obtain text data. The specific implementation process can be implemented by ORC technology, which will not be described in detail. Afterwards, the text data in the collected heterogeneous data, i.e., the second effective multi-source heterogeneous data, is merged, and the semantically identical parts are merged and deduplicated to generate text data. Afterwards, the text data is split and summarized based on a predetermined target format to generate standard data constructed based on the target format, i.e., structured data. Afterwards, a private domain knowledge database is constructed based on structured data, so that the operation and maintenance knowledge in the private domain knowledge database has consistency in data format, which can improve the efficiency of the large language model in utilizing the operation and maintenance knowledge in the private domain knowledge database. The above process is the implementation step of converting effective multi-source heterogeneous data into corresponding text data.

[0080] Furthermore, in a possible implementation, Fig. 9 As shown, the target structure includes at least two orderly arranged information nodes, and the specific implementation of step S2032 includes:

[0081] Step S2032-1: Use the large language model to process the text data to obtain the semantic feature encoding corresponding to the text data.

[0082] Step S2032-2: Based on the information category corresponding to each information node, information is extracted and classified for the semantic feature coding to obtain the information text corresponding to each information node.

[0083] Step S2032-3: Generate structured data according to the information text corresponding to each information node.

[0084] Illustratively, in the steps of this embodiment, the target format is represented based on orderly arranged information nodes. For example, the content of the target format is: ([p1], [p2], [p3], [p4], [p5], [p6]). [p1], [p2], [p3], [p4], [p5], [p6] each represent an information node, and each information node stores corresponding information content.

[0085] For example, in information node [p1], query information (Question) is stored, that is, the question raised by the user, the manifestation of this fault, problem, and alarm. In information node [p2], action information (Action) is stored, that is, the specific measures taken, which may be troubleshooting measures or stop-loss measures, such as "check the specific monitoring dashboard, view xx indicators", "check xx information in xx log". In information node [p3], tool information (Tools) is stored, that is, the platform / tools / links used by the above measures, such as: automatic diagnosis robot, monitoring face link. In information node [p4], observation information (Observation) is stored, that is, the description of specific alarm problems, or the manifestation of key faults found in the monitoring system or log. In information node [p5], recommended information (Thought) is stored, which is the adjustment of ideas or further verification during the analysis and troubleshooting process, such as, it may be xx cause or eliminate xx problem. In information node [p6], summary information (Conclusion) is stored, that is, the root cause of the final problem.

[0086] That is, illustratively, the information category of the information node includes at least one of the following: question information, action information, tool information, observation information, recommendation information, and summary information. By extracting and classifying the semantic feature encoding based on the information category corresponding to each information node, the information text corresponding to each information node can be obtained, and the information text belonging to each node can be obtained, thereby generating structured data. Using, the conversation text (text data) in the chat group is identified, split, and summarized, and summarized into six aspects of information: question information, action information, tool information, observation information, recommendation information, and summary information, thereby generating structured data corresponding to the conversation text in the chat group.

[0087] In the steps of this embodiment, firstly, the text data is feature extracted through a large language model to obtain a code representing semantic features, i.e., a semantic feature code. Then, based on the information category corresponding to each information node (e.g., the above-mentioned 6 information categories), the semantic feature code is processed to generate information text corresponding to each information node, and then structured data is generated based on the information text corresponding to each information node, so that the text data is expressed based on the information text corresponding to multiple information nodes constituting the target structure, and the purpose of formatting the text data is achieved.

[0088] Step S204: construct a private domain knowledge database based on the structured data.

[0089] Exemplarily, after generating structured data, the structured data corresponding to different heterogeneous data are used to perform information fusion to complete the construction of a private domain knowledge database. Exemplarily, the private domain knowledge database includes an operation and maintenance knowledge graph, which is used to characterize multiple operation and maintenance entities in an operation and maintenance scenario, as well as the causal relationship and dependency relationship between operation and maintenance entities. Operation and maintenance entities can be divided according to information nodes in the structured data, for example, operation and maintenance entities can be divided based on query information and action information, and the causal relationship and dependency relationship between operation and maintenance entities can be determined based on the attribution of information nodes (that is, the information content corresponding to multiple different information nodes in a structured data). For specific implementation methods, please refer to Figure 2 The corresponding introduction in the illustrated embodiment will not be repeated here.

[0090] Furthermore, in this embodiment, the steps are also included:

[0091] Step S203A: performing content detection on the text data corresponding to at least two valid multi-source heterogeneous data to obtain at least two associated valid multi-source heterogeneous data with content matching; performing content fusion on at least two associated valid multi-source heterogeneous data with content matching to obtain fused data.

[0092] Accordingly, the specific implementation of step S203 includes:

[0093] Step S203B: reconstruct the fused data to generate structured data with a target structure.

[0094] Exemplarily, on the other hand, the large language model will perform content detection on the text data corresponding to at least two valid multi-source heterogeneous data, and associate the first valid multi-source heterogeneous data with consistent or related content. For example, when a discussion on the same operation and maintenance problem occurs in two different operation and maintenance chat groups, the chat data in each operation and maintenance chat group will be collected as different data (multi-source heterogeneous data), and after processing based on the previous validity detection step, the corresponding (at least two) first valid multi-source heterogeneous data will be generated. After that, the large language model fuses the first valid multi-source heterogeneous data generated from different operation and maintenance chat groups but for the same operation and maintenance problem to obtain fused data. The fused data can provide more operation and maintenance information and operation and maintenance knowledge. After that, based on the reconstruction of the fused data, structured data is generated, and then a private domain knowledge database is constructed, so that the constructed private domain knowledge database contains more strongly correlated operation and maintenance knowledge, and the performance of subsequent inquiry and response based on the private domain knowledge database is improved.

[0095] Step S205: In response to the query instruction, the private domain knowledge database is called through the large language model to generate response information corresponding to the query instruction, wherein the query instruction is used to inquire about the operation and maintenance problem, and the response information is used to characterize the processing flow for the operation and maintenance problem.

[0096] Optionally, this embodiment also includes:

[0097] Step S206: Obtain historical operation and maintenance records through the large language model, and update the private domain knowledge in the private domain knowledge database through the historical operation and maintenance records.

[0098] Exemplarily, on the other hand, after the intelligent operation and maintenance system generates the private domain knowledge database based on the above steps, it will also periodically or based on other acquisition logics to obtain historical operation and maintenance records, which can be operation and maintenance records generated after manual processing, or operation and maintenance records generated after the intelligent operation and maintenance system calls the large language model for processing. The above historical operation and maintenance records contain feedback information used to characterize the processing results of the operation and maintenance problems, such as whether the operation and maintenance problems are solved, the cost and efficiency of solving the operation and maintenance problems, and other information. Obtaining historical operation and maintenance records through a large language model and updating the private domain knowledge in the private domain knowledge database through historical operation and maintenance records can further improve the accuracy and real-time nature of the private domain knowledge in the private domain knowledge database, thereby improving the intelligent query response capability of the intelligent operation and maintenance system.

[0099] Fig.10A schematic diagram of a process for implementing operation and maintenance knowledge question and answering through an intelligent operation and maintenance system provided in an embodiment of the present disclosure is shown below. Fig.10 For further introduction to the above steps provided in this embodiment, refer to Fig.10 As shown, exemplarily, first, by determining the access scenario, the subscription information is configured, and the subscription information is used to determine the data source of multi-source heterogeneous data, such as historical duty, alarm processing group, accident meeting records, relevant technical documents / standard operating procedure documents, alarm configuration information, platform work order records, etc. After that, data processing is performed through a large language model, that is, the steps of performing AI processing on multi-source heterogeneous data, such as data cleaning, image data translation, document processing, dialogue information extraction / classification, configuration management database / alarm and other configuration information association, construction of knowledge graph, etc. After that, the large language model is used for reasoning / generation to generate an operation and maintenance knowledge base and a large language model output, wherein the operation and maintenance knowledge base is the private domain knowledge database in this embodiment, which includes a knowledge graph (i.e., an operation and maintenance knowledge graph) and troubleshooting experience, and then the operation and maintenance knowledge base is used to obtain a large language model output, including a draft of a standard operating procedure, expert experience, a summary of historical knowledge, extraction of tool indicators, and the like. On the other hand, in response to the query instruction, the steps of knowledge generation and knowledge use are performed through the large language model to generate response information to solve the operation and maintenance problem, and the response information is returned to the data source for recording. Knowledge generation includes, for example, generating standard operating procedure documents, work order handover summaries, duty weekly review, and review optimization suggestions; knowledge use includes, for example, registering tool diagnostic items, generating diagnostic processes, binding robots, and calling platform tools. At the same time. The obtained knowledge (including the relevant knowledge of large language model generation and knowledge generation, and knowledge use) can be calibrated, supplemented, and debugged through the knowledge learning logic within the artificial or large language model to improve the richness and accuracy of the operation and maintenance knowledge that can be obtained by the large language model in the intelligent operation and maintenance system, thereby improving the efficiency and quality of intelligent operation and maintenance.

[0100] In this embodiment, the implementation method of step S201 and step S205 is the same as that of the present disclosure. Figure 2 The implementation methods of step S101 and step S103 in the illustrated embodiment are the same and will not be described in detail here.

[0101] Corresponding to the response information generation method of the above embodiment, Fig.11 This is a structural block diagram of a response information generation device provided in an embodiment of the present disclosure. The method described in the above embodiment can be executed by the response information generation device, which can be implemented by software and / or hardware, and can be integrated in an electronic device with certain data processing functions. Among them, the electronic device can include but is not limited to a mobile terminal with big data processing capabilities, and a fixed terminal with big data processing capabilities such as a desktop computer and a supercomputer.

[0102] For the sake of convenience, only the parts related to the embodiments of the present disclosure are shown. Fig.11 , the response information generating device 3 comprises:

[0103] An acquisition module 31 is used to acquire multi-source heterogeneous data in an operation and maintenance scenario, where the multi-source heterogeneous data is data in at least two different data formats generated during the execution of an operation and maintenance task by an intelligent operation and maintenance system;

[0104] A processing module 32 is used to process multi-source heterogeneous data through a large language model to generate a private domain knowledge database for operation and maintenance scenarios, wherein the private domain knowledge database stores structured data having a target structure, and the structured data is used to record operation and maintenance knowledge for handling operation and maintenance problems;

[0105] The generation module 33 is used to respond to the query instruction, call the private domain knowledge database through the large language model, and generate response information corresponding to the query instruction, wherein the query instruction is used to inquire about the operation and maintenance problem, and the response information is used to represent the processing flow for the operation and maintenance problem.

[0106] According to one or more embodiments of the present disclosure, the processing module 32 is specifically used to: filter multi-source heterogeneous data through a large language model to obtain effective multi-source heterogeneous data; reconstruct the effective multi-source heterogeneous data to generate structured data with a target structure; and build a private domain knowledge database based on the structured data.

[0107] According to one or more embodiments of the present disclosure, effective multi-source heterogeneous data includes first effective multi-source heterogeneous data and second effective multi-source heterogeneous data, the first effective multi-source heterogeneous data is non-text data, the second effective multi-source heterogeneous data is text data, and a multimodal big model is configured in the intelligent operation and maintenance system. When the processing module 32 reconstructs the effective multi-source heterogeneous data to generate structured data with a target structure, it is specifically used to: generate text data by calling the multimodal big model, fuse the first effective multi-source heterogeneous data and the second effective multi-source heterogeneous data; and process the text data based on the target structure to generate structured data.

[0108] According to one or more embodiments of the present disclosure, the target structure includes at least two orderly arranged information nodes. When the processing module 32 reconstructs the effective multi-source heterogeneous data to generate structured data with the target structure, it is specifically used to: convert the effective multi-source heterogeneous data into corresponding text data; use a large language model to process the text data to obtain semantic feature encodings corresponding to the text data; based on the information category corresponding to each of the information nodes, extract and classify the semantic feature encodings to obtain information texts corresponding to each of the information nodes; and generate structured data with the target structure according to the information texts corresponding to each of the information nodes.

[0109] According to one or more embodiments of the present disclosure, the information category of the information node includes at least one of the following: query information, action information, tool information, observation information, recommendation information, and summary information.

[0110] According to one or more embodiments of the present disclosure, the processing module 32 is further used to: perform content detection on text data corresponding to at least two valid multi-source heterogeneous data to obtain associated valid multi-source heterogeneous data with at least two content matches; perform content fusion on at least two associated valid multi-source heterogeneous data with at least two content matches to obtain fused data; when constructing a private domain knowledge database based on structured data, the processing module 32 is specifically used to: construct a private domain knowledge database based on structured data and fused data.

[0111] According to one or more embodiments of the present disclosure, the private domain knowledge database includes an operation and maintenance knowledge graph, and / or an operation and maintenance task template, wherein the operation and maintenance knowledge graph is used to characterize multiple operation and maintenance entities in an operation and maintenance scenario, as well as the causal relationship and dependency relationship between the operation and maintenance entities; the operation and maintenance task template is used to generate at least one text of a processing flow for an operation and maintenance task.

[0112] According to one or more embodiments of the present disclosure, the generation module 33 is specifically used to: perform reasoning based on multi-source heterogeneous data through a large language model to obtain multiple operation and maintenance entities, as well as causal relationships and dependency relationships between multiple operation and maintenance entities; process the causal relationships and dependency relationships between multiple operation and maintenance entities based on a graph reasoning algorithm to obtain a supplementary association relationship between at least two operation and maintenance entities; and generate an operation and maintenance knowledge graph based on the causal relationships, dependency relationships, and supplementary association relationships between each operation and maintenance entity.

[0113] According to one or more embodiments of the present disclosure, the generation module 33 is specifically used to: obtain a target inquiry category corresponding to the inquiry instruction; generate a target prompt word corresponding to the inquiry instruction according to the target inquiry category, the target prompt word is used to characterize the steps of processing the operation and maintenance problem based on the target inquiry category; input the target prompt word and the inquiry instruction into the large language model to control the large language model to obtain the target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the inquiry instruction based on the target operation and maintenance knowledge.

[0114] According to one or more embodiments of the present disclosure, the target operation and maintenance knowledge includes platform tool information. When the generation module 33 inputs the target prompt words and query instructions into the large language model to control the large language model to obtain the target operation and maintenance knowledge from the private domain knowledge database, and generates response information corresponding to the query instruction based on the target operation and maintenance knowledge, it is specifically used to: obtain the platform tool information corresponding to the query instruction from the private domain knowledge database through the large language model, the platform tool information represents the operation and maintenance tool used in the external platform to execute the processing flow of the operation and maintenance problem corresponding to the query instruction; register and call the target operation and maintenance tool indicated by the platform tool information based on the large language model to generate response information corresponding to the query instruction.

[0115] According to one or more embodiments of the present disclosure, when the generation module 33 generates response information corresponding to the query instruction based on the target operation and maintenance knowledge, it is specifically used to: obtain real-time alarm data of the intelligent operation and maintenance system; generate an automated diagnosis script through a large language model, using real-time alarm data and target operation and maintenance knowledge, and register the diagnostic items provided by the external platform by running the automated diagnosis script; generate response information by executing the diagnostic items.

[0116] According to one or more embodiments of the present disclosure, the processing module 32 is further used to: obtain historical operation and maintenance records through a large language model; and update the private domain knowledge in the private domain knowledge database through the historical operation and maintenance records.

[0117] The acquisition module 31, the processing module 32 and the generation module 33 are connected in sequence. The response information generation device 3 provided in this embodiment can implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated in this embodiment.

[0118] Fig.12 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure is shown in FIG. Fig.12 As shown, the electronic device 4 includes:

[0119] A processor 41, and a memory 42 communicatively connected to the processor 41;

[0120] The memory 42 stores computer executable instructions;

[0121] The processor 41 executes the computer execution instructions stored in the memory 42 to implement the following Figure 2-Figure 10 The response information generating method in the illustrated embodiment.

[0122] Optionally, the processor 41 and the memory 42 are connected via a bus 43 .

[0123] For related instructions, please refer to Figure 2-Figure 10The relevant descriptions and effects corresponding to the steps in the corresponding embodiments can be understood, and no further elaboration is made here.

[0124] The present invention provides an intelligent operation and maintenance system for implementing the present invention. Figure 2-Figure 10 The response information generation method provided by any one of the corresponding embodiments.

[0125] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, which are used to implement the present invention when executed by a processor. Figure 2-Figure 10 The response information generating method provided by any one of the corresponding embodiments.

[0126] The present invention provides a computer program product, including a computer program. When the computer program is executed by a processor, the present invention is realized. Figure 2-Figure 10 The response information generating method provided by any one of the corresponding embodiments.

[0127] In order to implement the above embodiment, the embodiment of the present disclosure also provides an electronic device.

[0128] refer to Fig.13 , which shows a schematic diagram of the structure of an electronic device 900 suitable for implementing the embodiment of the present disclosure, and the electronic device 900 may be a terminal device or a server. The terminal device may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, personal digital assistants (PDAs), tablet computers, portable multimedia players (PMPs), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.13 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0129] like Fig.13As shown, the electronic device 900 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage device 908 to a random access memory (RAM) 903. Various programs and data required for the operation of the electronic device 900 are also stored in the RAM 903. The processing device 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0130] Typically, the following devices may be connected to the I / O interface 905: input devices 906 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 907 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 908 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 909. The communication device 909 may allow the electronic device 900 to communicate with other devices wirelessly or by wire to exchange data. Although Fig.13 The electronic device 900 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0131] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 909, or installed from a storage device 908, or installed from a ROM 902. When the computer program is executed by the processing device 901, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0132] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, 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 device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0133] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0134] The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0135] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0137] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the name of a unit or module does not, in some cases, constitute a limitation on the unit itself.

[0138] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0139] 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 system, device, or equipment. 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, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, 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 device, a magnetic storage device, or any suitable combination of the foregoing.

[0140] In a first aspect, according to one or more embodiments of the present disclosure, a method for generating response information is provided, including:

[0141] Acquire multi-source heterogeneous data in an operation and maintenance scenario, wherein the multi-source heterogeneous data is data in at least two different data formats generated in the process of executing an operation and maintenance task through the intelligent operation and maintenance system; process the multi-source heterogeneous data through a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data with a target structure, and the structured data is used to record operation and maintenance knowledge for handling operation and maintenance problems; in response to an inquiry instruction, call the private domain knowledge database through the large language model to generate response information corresponding to the inquiry instruction, wherein the inquiry instruction is used to inquire about the operation and maintenance problem, and the response information is used to characterize the processing flow for the operation and maintenance problem.

[0142] According to one or more embodiments of the present disclosure, the multi-source heterogeneous data is processed by a large language model to generate a private domain knowledge database for the operation and maintenance scenario, including: screening the multi-source heterogeneous data by the large language model to obtain valid multi-source heterogeneous data; reconstructing the valid multi-source heterogeneous data to generate structured data with a target structure; and constructing the private domain knowledge database based on the structured data.

[0143] According to one or more embodiments of the present disclosure, the effective multi-source heterogeneous data includes first effective multi-source heterogeneous data and second effective multi-source heterogeneous data, the first effective multi-source heterogeneous data is non-text data, the second effective multi-source heterogeneous data is text data, a multimodal big model is configured in the intelligent operation and maintenance system, and the reconstructing the effective multi-source heterogeneous data to generate structured data with a target structure includes: generating text data by calling the multimodal big model and fusing the first effective multi-source heterogeneous data and the second effective multi-source heterogeneous data; and generating the structured data by processing the text data based on the target structure.

[0144] According to one or more embodiments of the present disclosure, the target structure includes at least two orderly arranged information nodes, and the processing of the text data based on the target structure to generate the structured data includes: processing the text data using a large language model to obtain semantic feature codes corresponding to the text data; extracting and classifying information from the semantic feature codes based on information categories corresponding to each of the information nodes to obtain information texts corresponding to each of the information nodes; and generating the structured data based on the information texts corresponding to each of the information nodes.

[0145] According to one or more embodiments of the present disclosure, the information category of the information node includes at least one of the following: query information, action information, tool information, observation information, recommendation information, and summary information.

[0146] According to one or more embodiments of the present disclosure, the method also includes: performing content detection on text data corresponding to at least two valid multi-source heterogeneous data to obtain associated valid multi-source heterogeneous data with at least two content matches; performing content fusion on the at least two associated valid multi-source heterogeneous data with content matches to obtain fused data; constructing the private domain knowledge database based on the structured data includes: generating structured data with a target structure based on reconstruction of the fused data.

[0147] According to one or more embodiments of the present disclosure, the private domain knowledge database includes an operation and maintenance knowledge graph, and / or an operation and maintenance task template, wherein the operation and maintenance knowledge graph is used to characterize multiple operation and maintenance entities in an operation and maintenance scenario, as well as the causal relationship and dependency relationship between operation and maintenance entities; the operation and maintenance task template is used to generate at least one text of a processing flow for an operation and maintenance task.

[0148] According to one or more embodiments of the present disclosure, the multi-source heterogeneous data is processed by a large language model to generate a private domain knowledge database for the operation and maintenance scenario, including: reasoning based on the multi-source heterogeneous data by a large language model to obtain multiple operation and maintenance entities, and causal relationships and dependency relationships between the multiple operation and maintenance entities; processing the causal relationships and dependency relationships between the multiple operation and maintenance entities based on a graph reasoning algorithm to obtain a supplementary association relationship between at least two operation and maintenance entities; generating the operation and maintenance knowledge graph according to the causal relationships, dependency relationships and supplementary association relationships between the operation and maintenance entities.

[0149] According to one or more embodiments of the present disclosure, in response to a query instruction, the private domain knowledge database is called through the large language model to generate response information corresponding to the query instruction, including: obtaining a target query category corresponding to the query instruction; generating a target prompt word corresponding to the query instruction according to the target query category, the target prompt word being used to characterize the step of processing the operation and maintenance problem based on the target query category; inputting the target prompt word and the query instruction into the large language model to control the large language model to obtain target operation and maintenance knowledge from the private domain knowledge database, and generating response information corresponding to the query instruction based on the target operation and maintenance knowledge.

[0150] According to one or more embodiments of the present disclosure, the target operation and maintenance knowledge includes platform tool information, and the target prompt word and the query instruction are input into the large language model to control the large language model to obtain the target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the query instruction based on the target operation and maintenance knowledge, including: obtaining the platform tool information corresponding to the query instruction from the private domain knowledge database through the large language model, the platform tool information representing the operation and maintenance tool used in the external platform to execute the processing flow of the operation and maintenance problem corresponding to the query instruction; registering and calling the target operation and maintenance tool indicated by the platform tool information based on the large language model to generate response information corresponding to the query instruction.

[0151] According to one or more embodiments of the present disclosure, generating response information corresponding to the query instruction based on the target operation and maintenance knowledge includes: obtaining real-time alarm data of the intelligent operation and maintenance system; generating an automated diagnosis script through the large language model using the real-time alarm data and the target operation and maintenance knowledge, and registering diagnostic items provided by an external platform by running the automated diagnosis script; and generating the response information by executing the diagnostic items.

[0152] According to one or more embodiments of the present disclosure, the method further includes: acquiring historical operation and maintenance records through the large language model; and updating the private domain knowledge in the private domain knowledge database through the historical operation and maintenance records.

[0153] In a second aspect, according to one or more embodiments of the present disclosure, a response information generating device is provided, including:

[0154] An acquisition module, used to acquire multi-source heterogeneous data in an operation and maintenance scenario, wherein the multi-source heterogeneous data is data in at least two different data formats generated during the execution of an operation and maintenance task by the intelligent operation and maintenance system;

[0155] A processing module, used to process the multi-source heterogeneous data through a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data having a target structure, and the structured data is used to record the operation and maintenance knowledge for handling operation and maintenance problems;

[0156] A generation module is used to respond to an inquiry instruction, call the private domain knowledge database through the large language model, and generate response information corresponding to the inquiry instruction, wherein the inquiry instruction is used to inquire about operation and maintenance issues, and the response information is used to characterize the processing flow for the operation and maintenance issues.

[0157] According to one or more embodiments of the present disclosure, the processing module is specifically used to: filter the multi-source heterogeneous data through the large language model to obtain effective multi-source heterogeneous data; reconstruct the effective multi-source heterogeneous data to generate structured data with a target structure; and construct the private domain knowledge database based on the structured data.

[0158] According to one or more embodiments of the present disclosure, the effective multi-source heterogeneous data includes first effective multi-source heterogeneous data and second effective multi-source heterogeneous data, the first effective multi-source heterogeneous data is non-text data, the second effective multi-source heterogeneous data is text data, and a multimodal large model is configured in the intelligent operation and maintenance system. When the processing module reconstructs the effective multi-source heterogeneous data to generate structured data with a target structure, it is specifically used to: generate text data by calling the multimodal large model to fuse the first effective multi-source heterogeneous data and the second effective multi-source heterogeneous data; and generate the structured data by processing the text data based on the target structure.

[0159] According to one or more embodiments of the present disclosure, the target structure includes at least two orderly arranged information nodes. When the processing module processes the text data based on the target structure to generate the structured data, it is specifically used to: process the text data using a large language model to obtain semantic feature codes corresponding to the text data; based on the information category corresponding to each of the information nodes, extract and classify the semantic feature codes to obtain information texts corresponding to each of the information nodes; and generate the structured data according to the information texts corresponding to each of the information nodes.

[0160] According to one or more embodiments of the present disclosure, the information category of the information node includes at least one of the following: query information, action information, tool information, observation information, recommendation information, and summary information.

[0161] According to one or more embodiments of the present disclosure, the processing module is further used to: perform content detection on text data corresponding to at least two valid multi-source heterogeneous data to obtain associated valid multi-source heterogeneous data with at least two content matches; perform content fusion on the at least two associated valid multi-source heterogeneous data with content matches to obtain fused data; when the processing module constructs the private domain knowledge database based on the structured data, the processing module is specifically used to: generate structured data with a target structure based on reconstruction of the fused data.

[0162] According to one or more embodiments of the present disclosure, the private domain knowledge database includes an operation and maintenance knowledge graph, and / or an operation and maintenance task template, wherein the operation and maintenance knowledge graph is used to characterize multiple operation and maintenance entities in an operation and maintenance scenario, as well as the causal relationship and dependency relationship between operation and maintenance entities; the operation and maintenance task template is used to generate at least one text of a processing flow for an operation and maintenance task.

[0163] According to one or more embodiments of the present disclosure, the generation module is specifically used for: performing reasoning based on the multi-source heterogeneous data through a large language model to obtain multiple operation and maintenance entities, as well as causal relationships and dependency relationships between the multiple operation and maintenance entities; processing the causal relationships and dependency relationships between the multiple operation and maintenance entities based on a graph reasoning algorithm to obtain a supplementary association relationship between at least two operation and maintenance entities; and generating the operation and maintenance knowledge graph based on the causal relationships, dependency relationships and supplementary association relationships between the operation and maintenance entities.

[0164] According to one or more embodiments of the present disclosure, the generation module is specifically used to: obtain a target inquiry category corresponding to the inquiry instruction; generate a target prompt word corresponding to the inquiry instruction according to the target inquiry category, and the target prompt word is used to characterize the steps of processing the operation and maintenance problem based on the target inquiry category; input the target prompt word and the inquiry instruction into the large language model to control the large language model to obtain target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the inquiry instruction based on the target operation and maintenance knowledge.

[0165] According to one or more embodiments of the present disclosure, the target operation and maintenance knowledge includes platform tool information. When the generation module inputs the target prompt word and the query instruction into the large language model to control the large language model to obtain the target operation and maintenance knowledge from the private domain knowledge database, and generates response information corresponding to the query instruction based on the target operation and maintenance knowledge, it is specifically used to: obtain the platform tool information corresponding to the query instruction from the private domain knowledge database through the large language model, and the platform tool information represents the operation and maintenance tool used in the external platform to execute the processing flow of the operation and maintenance problem corresponding to the query instruction; register and call the target operation and maintenance tool indicated by the platform tool information based on the large language model to generate response information corresponding to the query instruction.

[0166] According to one or more embodiments of the present disclosure, when the generation module generates response information corresponding to the query instruction based on the target operation and maintenance knowledge, it is specifically used to: obtain real-time alarm data of the intelligent operation and maintenance system; generate an automated diagnosis script through the large language model, using the real-time alarm data and the target operation and maintenance knowledge, and register the diagnostic items provided by the external platform by running the automated diagnosis script; and generate the response information by executing the diagnostic items.

[0167] According to one or more embodiments of the present disclosure, the processing module is further used to: obtain historical operation and maintenance records through the large language model; and update the private domain knowledge in the private domain knowledge database through the historical operation and maintenance records.

[0168] In a third aspect, according to one or more embodiments of the present disclosure, there is provided an electronic device, comprising: at least one processor and a memory;

[0169] The memory stores computer-executable instructions;

[0170] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the response information generation method as described in the first aspect and various possible designs of the first aspect.

[0171] In a fourth aspect, according to one or more embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer execution instructions. When a processor executes the computer execution instructions, the response information generation method described in the first aspect and various possible designs of the first aspect is implemented.

[0172] In a fifth aspect, according to one or more embodiments of the present disclosure, a computer program product is provided, including a computer program, which, when executed by a processor, implements the response information generation method described in the first aspect and various possible designs of the first aspect.

[0173] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0174] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0175] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A method for generating response information, characterized in that: Applied to intelligent operation and maintenance systems, including: Acquire multi-source heterogeneous data in an operation and maintenance scenario, where the multi-source heterogeneous data is data in at least two different data formats generated during the execution of an operation and maintenance task by the intelligent operation and maintenance system; Processing the multi-source heterogeneous data through a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data having a target structure, and the structured data is used to record operation and maintenance knowledge for handling operation and maintenance problems; In response to an inquiry instruction, the private domain knowledge database is called through the large language model to generate response information corresponding to the inquiry instruction, wherein the inquiry instruction is used to inquire about an operation and maintenance problem, and the response information is used to characterize a processing flow for the operation and maintenance problem.

2. The method according to claim 1, characterized in that The processing of the multi-source heterogeneous data by a large language model to generate a private domain knowledge database for the operation and maintenance scenario includes: The multi-source heterogeneous data is screened by the large language model to obtain effective multi-source heterogeneous data; Reconstructing the effective multi-source heterogeneous data to generate structured data with a target structure; Based on the structured data, the private domain knowledge database is constructed.

3. The method according to claim 2, characterized in that The effective multi-source heterogeneous data includes first effective multi-source heterogeneous data and second effective multi-source heterogeneous data, the first effective multi-source heterogeneous data is non-text data, the second effective multi-source heterogeneous data is text data, a multimodal large model is configured in the intelligent operation and maintenance system, and the reconstructing the effective multi-source heterogeneous data to generate structured data with a target structure includes: By calling a multimodal large model, the first effective multi-source heterogeneous data and the second effective multi-source heterogeneous data are integrated to generate text data; The text data is processed based on the target structure to generate the structured data.

4. The method according to claim 2, characterized in that: The target structure includes at least two orderly arranged information nodes, and the reconstructing the effective multi-source heterogeneous data to generate structured data with the target structure includes: Converting the valid multi-source heterogeneous data into corresponding text data; Processing the text data using a large language model to obtain semantic feature encoding corresponding to the text data; Based on the information category corresponding to each of the information nodes, information extraction and classification are performed on the semantic feature codes to obtain information text corresponding to each of the information nodes; According to the information text corresponding to each of the information nodes, structured data with a target structure is generated.

5. The method according to claim 2, characterized in that: Also includes: Performing content detection on text data corresponding to at least two valid multi-source heterogeneous data to obtain at least two associated valid multi-source heterogeneous data having matching contents; Performing content fusion on the at least two content-matched associated valid multi-source heterogeneous data to obtain fused data; The reconstructing the effective multi-source heterogeneous data to generate structured data with a target structure includes: Based on the reconstruction performed on the fused data, structured data with a target structure is generated.

6. The method according to claim 1, characterized in that The private domain knowledge database includes an operation and maintenance knowledge graph and / or an operation and maintenance task template, wherein the operation and maintenance knowledge graph is used to represent multiple operation and maintenance entities in an operation and maintenance scenario, as well as the causal relationship and dependency relationship between operation and maintenance entities; the operation and maintenance task template is used to generate at least one text of a processing flow for an operation and maintenance task.

7. The method according to claim 6, characterized in that The processing of the multi-source heterogeneous data by a large language model to generate a private domain knowledge database for the operation and maintenance scenario includes: The large language model is used to perform reasoning based on the multi-source heterogeneous data to obtain multiple operation and maintenance entities, as well as causal relationships and dependency relationships between the multiple operation and maintenance entities; Processing the causal relationship and dependency relationship between the multiple operation and maintenance entities based on a graph reasoning algorithm to obtain a supplementary association relationship between at least two operation and maintenance entities; An operation and maintenance knowledge graph in the private domain knowledge database is generated according to the causal relationship, the dependency relationship and the supplementary association relationship between the operation and maintenance entities.

8. The method according to claim 1, characterized in that The step of calling the private domain knowledge database through the large language model in response to the query instruction to generate response information corresponding to the query instruction includes: Obtaining a target inquiry category corresponding to the inquiry instruction; According to the target inquiry category, generating a target prompt word corresponding to the inquiry instruction, the target prompt word is used to represent the step of processing the operation and maintenance problem based on the target inquiry category; The target prompt word and the query instruction are input into the large language model to control the large language model to obtain target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the query instruction based on the target operation and maintenance knowledge.

9. The method according to claim 8, characterized in that The target operation and maintenance knowledge includes platform tool information, and the target prompt word and the query instruction are input into the large language model to control the large language model to obtain the target operation and maintenance knowledge from the private domain knowledge database, and generate response information corresponding to the query instruction based on the target operation and maintenance knowledge, including: Acquire platform tool information corresponding to the query instruction from the private domain knowledge database through the large language model, wherein the platform tool information represents an operation and maintenance tool used in an external platform to execute a processing flow for an operation and maintenance problem corresponding to the query instruction; The target operation and maintenance tool indicated by the platform tool information is registered and called based on the platform tool information through the large language model to generate response information corresponding to the query instruction.

10. The method according to claim 8, characterized in that The generating response information corresponding to the query instruction based on the target operation and maintenance knowledge includes: Obtaining real-time alarm data of the intelligent operation and maintenance system; Generate an automated diagnosis script by using the large language model, the real-time alarm data and the target operation and maintenance knowledge, and register the diagnostic items provided by the external platform by running the automated diagnosis script; The response information is generated by executing the diagnosis item.

11. The method according to claim 1, characterized in that: Also includes: Acquire historical operation and maintenance records through the large language model; The private domain knowledge in the private domain knowledge database is updated through the historical operation and maintenance records.

12. A response information generating device, characterized in that: Applied to intelligent operation and maintenance systems, including: An acquisition module, used to acquire multi-source heterogeneous data in an operation and maintenance scenario, wherein the multi-source heterogeneous data is data in at least two different data formats generated during the execution of an operation and maintenance task by the intelligent operation and maintenance system; A processing module, used to process the multi-source heterogeneous data through a large language model to generate a private domain knowledge database for the operation and maintenance scenario, wherein the private domain knowledge database stores structured data having a target structure, and the structured data is used to record the operation and maintenance knowledge for handling operation and maintenance problems; A generation module is used to respond to an inquiry instruction, call the private domain knowledge database through the large language model, and generate response information corresponding to the inquiry instruction, wherein the inquiry instruction is used to inquire about operation and maintenance issues, and the response information is used to characterize the processing flow for the operation and maintenance issues.

13. An intelligent operation and maintenance system, characterized in that: Used to execute the response information generating method according to any one of claims 1 to 11.

14. An electronic device, characterized in that: include: Processor and memory; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor executes the response information generation method according to any one of claims 1 to 11.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When a processor executes the computer-executable instructions, the response information generation method according to any one of claims 1 to 11 is implemented.

16. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the response information generating method according to any one of claims 1 to 11 is implemented.

Images