Generation method and device of operation and maintenance answer of target system, and storage medium

Abstract

The embodiment of the invention provides a method and device for generating an operation and maintenance answer of a target system and a storage medium, and the method comprises the steps: executing a question conversion operation on an obtained initial question, and obtaining a target question; searching N first text vectors meeting a first preset threshold in an operation and maintenance vector database of the target system by utilizing the target problem, and determining N first texts corresponding to the N first text vectors; performing screening operation on the N texts to obtain M target texts; and generating a target operation and maintenance answer based on the M target texts and the target question. Through the method and the device, the problem that the generated answers of the problems in the operation and maintenance field are not accurate enough in the related technology is solved, and the effect of improving the accuracy of generating the answers of the problems in the operation and maintenance field is achieved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computers, and specifically, to a method and device for generating operation and maintenance answers for a target system, and a storage medium. Background Art

[0002] In the field of operations and maintenance, traditional knowledge retrieval typically relies on precise keyword matching or rule-based search engines. These methods struggle to provide accurate and comprehensive answers to complex and ambiguous operations and maintenance questions, especially when explicit keywords are lacking. Furthermore, with the increasing complexity of information technology infrastructure, traditional retrieval methods overly rely on keyword matching and ignore the semantic structure of questions, resulting in poor retrieval results for non-standard questions and an inability to fully understand the context of the question, limiting the accuracy of problem solving.

[0003] In recent years, with the rapid development of large language models, they have been applied to multiple fields, including operation and maintenance knowledge question answering. However, the application of large language models in the operation and maintenance field still faces challenges, especially when the model lacks in-depth training in a specific field. The accuracy, professionalism and reliability of its answers may be limited. In addition, the model may generate seemingly reasonable but actually wrong answers based on existing knowledge and patterns, which is particularly dangerous for operation and maintenance decision-making. Summary of the Invention

[0004] The embodiments of the present application provide a method and device for generating operation and maintenance answers for a target system, and a storage medium, so as to at least solve the problem in the related art of inaccurate answers to questions in the field of operation and maintenance.

[0005] According to one embodiment of the present application, a method for generating an operation and maintenance answer for a target system is provided, comprising: performing a question conversion operation on an acquired initial question to obtain a target question, wherein the target question and the initial question are both operation and maintenance questions for the target system; using the target question to search for N first text vectors that meet a first preset threshold in an operation and maintenance vector database of the target system, and determining N first texts corresponding to the N first text vectors, wherein the first text vectors include a first question vector and a first summary vector, one text vector corresponds to one first text, and N is a natural number greater than or equal to 1; performing a screening operation on the N texts to obtain M target texts, wherein M is a natural number less than or equal to N; and generating a target operation and maintenance answer based on the M target texts and the target question.

[0006] In an exemplary embodiment, a question conversion operation is performed on the obtained initial question to obtain a target question, including: performing a first text cleaning operation on the above initial question to obtain a first question, wherein the above first text cleaning operation includes an operation of removing abnormal symbols and stop words in the above initial question, and an operation of converting the above initial question into a question in a target format; inputting the above first question into a pre-trained first target model to obtain target information output by the above first target model, wherein the above target information includes the application field, keywords and semantic structure of the above first question; based on the above target information and the selected preset prompt words, performing a rewriting operation on the above first question to obtain the above target question.

[0007] In an exemplary embodiment, before using the target problem to search for N first text vectors that meet a first preset threshold in the operation and maintenance vector database of the target system, and determining the N texts corresponding to the N first text vectors, the method further includes: obtaining original text information from multiple data sources, and performing a preprocessing operation on the original text information to obtain target text information, wherein the preprocessing operation includes a second text cleaning operation, a text deduplication operation, and a text format conversion operation; performing a block operation on the target text information to obtain P text blocks, wherein the P text blocks include N first texts, and the P is a positive integer greater than or equal to the N; determining the second problem and the second summary corresponding to the P text blocks; and storing the text information in the P text blocks, the second problem corresponding to the P text blocks, and the second summary in the operation and maintenance vector database.

[0008] In an exemplary embodiment, the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries are all stored in an operation and maintenance vector database, including: performing a vector conversion operation on the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries to obtain P second text vectors, P second question vectors, and P second summary vectors; the text information in the P text blocks, the P second text vectors, the P second question vectors, and the P second summary vectors are all stored in the operation and maintenance vector database, wherein the P second question vectors include the first question vector in the N first text vectors, and the P second summary vectors include the first summary vector in the N first text vectors.

[0009] In an exemplary embodiment, the target problem is used to search for N first text vectors that meet a first preset threshold in the operation and maintenance vector database of the target system, and N texts corresponding to the N first text vectors are determined, including: performing a vector conversion operation on the target problem to obtain a target problem vector; using the target problem vector to search for a text vector that meets a first preset threshold in the operation and maintenance vector database to obtain N first text vectors; querying the text vectors corresponding to the N first problem vectors and the N first summary vectors included in the N first text vectors from a mapping relationship table to obtain N third text vectors, wherein the mapping relationship table is used to store the mapping relationship between the first problem vector and the third text vector, and the mapping relationship between the first summary vector and the third text vector; and determining the N texts corresponding to the N first text vectors based on the N third text vectors.

[0010] In an exemplary embodiment, before querying the text vectors corresponding to the N first question vectors and the N first summary vectors included in the N first text vectors from the mapping relationship table to obtain N third text vectors, the method further includes: establishing a first mapping relationship between the first text information in the N first texts and the N first text vectors; establishing a second mapping relationship between the N first text vectors and the N first question vectors; establishing a third mapping relationship between the N first text vectors and the N first summary vectors; and storing the first mapping relationship, the second mapping relationship, and the third mapping relationship to obtain the mapping relationship table.

[0011] In an exemplary embodiment, a target operation and maintenance answer is generated based on the M target texts and the target questions, including: determining the matching degrees between the M target texts and the target questions respectively to obtain M matching degrees; sorting the M target texts according to the M matching degrees; splicing the sorted M target texts to obtain target context information of the target questions; inputting the target context information and the target questions into a pre-trained second target model to obtain the target operation and maintenance answer output by the second target model.

[0012] According to another embodiment of the present application, a device for generating an operation and maintenance answer for a target system is provided, including: a conversion module, used to perform a question conversion operation on an acquired initial question to obtain a target question, wherein the above-mentioned target question and the above-mentioned initial question are both operation and maintenance questions for the target system; a search module, used to use the above-mentioned target question to search for N first text vectors that meet a first preset threshold in the operation and maintenance vector database of the above-mentioned target system, and determine N first texts corresponding to the N above-mentioned first text vectors, wherein the above-mentioned first text vectors include a first question vector and a first summary vector, one above-mentioned text vector corresponds to one above-mentioned first text, and the above-mentioned N is a natural number greater than or equal to 1; a screening module, used to perform a screening operation on the N above-mentioned texts to obtain M target texts, wherein the above-mentioned M is a natural number less than or equal to the above-mentioned N; a generation module, used to generate a target operation and maintenance answer based on the M above-mentioned target texts and the above-mentioned target question.

[0013] According to another embodiment of the present application, a computer program product is provided, including a computer program, which implements the steps of any of the above method embodiments when executed by a processor.

[0014] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is configured to execute the steps of any one of the above method embodiments when run.

[0015] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0016] Through this application, the obtained initial question is converted into a more standardized target question through a question conversion operation. The converted question is then used to search the operation and maintenance vector database for N first text vectors whose relevance meets a preset threshold, determine the corresponding N first texts, and perform a screening operation on these N texts to select the most relevant M target texts. Finally, based on the M target texts and the target question, the final operation and maintenance answer is generated. This solves the problem of inaccurate answers to questions in the operation and maintenance field in related technologies, thereby achieving the effect of improving the accuracy of answers to questions in the operation and maintenance field. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of a hardware environment for a method for generating an operation and maintenance answer for a target system according to an embodiment of the present application;

[0018] Figure 2is a flowchart of a method for generating an operation and maintenance answer for a target system according to an embodiment of the present application;

[0019] Figure 3 is a schematic diagram of a method for processing original text information according to an embodiment of the present application;

[0020] Figure 4 This is a process for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 1 ;

[0021] Figure 5 This is a process for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 2 ;

[0022] Figure 6 This is a structural block diagram of a device for generating operation and maintenance answers for a target system according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] It should be noted that the terms "first", "second", etc. in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0025] The method embodiments provided in the embodiments of the present application can be executed in a server device or a similar computing device. Taking running on a server device as an example, Figure 1 This is a hardware environment diagram of a method for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 1 As shown, the server device may include one or more ( Figure 1 Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data, wherein the above-mentioned server device may also include a transmission device 106 for communication functions and an input and output device 108. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above server device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0026] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to a method for generating an operation and maintenance answer for a target system in an embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implementing the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories may be connected to a server device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0027] The transmission device 106 is used to receive or send data via a network. A specific example of the aforementioned network may include a wireless network provided by a communication provider of the server device. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0028] In this embodiment, a method for generating an operation and maintenance answer of a target system is provided. Figure 2 This is a flow chart of a method for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 2 As shown, the process includes the following steps:

[0029] Step S202: performing a problem conversion operation on the obtained initial problem to obtain a target problem, wherein the target problem and the initial problem are both operation and maintenance problems for the target system;

[0030] Optionally, the initial question in this embodiment is a question in the field of operation and maintenance entered by the user for the first time, which may contain grammatical errors, unclear expressions or ambiguity. The initial question can be in the following form: "Analyze the causes of the failure."

[0031] Optionally, in this embodiment, the subject performing the question conversion operation includes but is not limited to a large language model, a graph neural network, and question conversion based on a preset rule engine.

[0032] Optionally, the target question in this embodiment is a question that is more standardized, easier to understand, and more accurately expresses the user's intention after the question conversion operation.

[0033] Step S204: Using the target question, searching the operation and maintenance vector database of the target system for N first text vectors that meet a first preset threshold, and determining N first texts corresponding to the N first text vectors, wherein the first text vectors include a first question vector and a first summary vector, one text vector corresponds to one first text, and N is a natural number greater than or equal to 1.

[0034] Optionally, the operation and maintenance vector database in this embodiment is a vector database containing operation and maintenance knowledge, and the information stored therein has been vectorized.

[0035] Optionally, the first text vector in this embodiment is a vector stored in the operation and maintenance vector database, which is used to represent a specific question or document summary, so as to facilitate retrieval through similarity comparison.

[0036] Optionally, the first preset threshold in this embodiment is a minimum similarity threshold set when searching in the operation and maintenance vector database. Only vectors whose cosine similarity between vectors exceeds this threshold will be selected. For example, if the first preset threshold is set to 0.8, then only those text vectors whose similarity with the target question vector is higher than 0.8 will be selected.

[0037] Step S206, performing a screening operation on the N texts to obtain M target texts, where M is a natural number less than or equal to N;

[0038] Optionally, the screening operation in this embodiment is used to screen out texts that are more relevant to the target question from N texts.

[0039] Step S208: Generate a target operation and maintenance answer based on the M target texts and the target questions.

[0040] Optionally, in this embodiment, the initial question obtained can be a question input by the user. The question input by the user is first converted into a question to obtain a target question. The target question is then used to search the operation and maintenance vector database for question information of the target text information after segmentation that meets the conditions. Finally, the corresponding text block is determined based on the above question information, that is, N first texts are determined.

[0041] Through the above steps, the initially obtained problem is subjected to a problem transformation operation to become a more standardized target problem. Then, the transformed problem is used to find N first text vectors in the operation and maintenance vector database whose relevance meets a preset threshold, and the corresponding N first texts are determined. A screening operation is performed on these N texts to select the most relevant M target texts. Finally, based on the M target texts and the target problem, a final operation and maintenance answer is generated. This solves the problem that the answers to problems in the operation and maintenance field in the related art are not accurate enough, and thus achieves the effect of improving the accuracy of the answers to problems in the operation and maintenance field.

[0042] Among them, the execution subject of the above steps can be a terminal, a server, a specific processor set in the terminal or the server, or a processor or processing device set relatively independently of the terminal or the server, but is not limited thereto. For example: edge computing devices, dedicated servers, etc.

[0043] In an exemplary embodiment, performing a problem transformation operation on the obtained initial problem to obtain a target problem includes: performing a first text cleaning operation on the above initial problem to obtain a first problem, where the above first text cleaning operation includes an operation of removing abnormal symbols and stop words in the above initial problem, and an operation of converting the above initial problem into a problem in a target format; inputting the above first problem into a pre-trained first target model to obtain target information output by the above first target model, where the above target information includes the application field, keywords, and semantic structure of the above first problem; based on the above target information and a selected preset prompt word, performing a rewriting operation on the above first problem to obtain the above target problem.

[0044] Optionally, the abnormal symbols in this embodiment are symbols that have an interfering effect on the semantic understanding of the initial problem, including but not limited to special characters similar to "***".

[0045] Optionally, the stop words in this embodiment are words without actual meaning, including but not limited to "de", "děi", "wǒ".

[0046] Optionally, the problem in the target format in this embodiment refers to converting the initial problem into a template-like expression such as "Why...?", "How...?", or "What is the reason for...?", and clearly pointing out the key information in the problem. The key information includes but not limited to domain-specific keywords, phrases, or terms. The problem in the target format will also perform a classification operation on the initial problem and clarify its type, such as fault troubleshooting, configuration suggestions, best practices, etc. Suppose the initial problem is: "The computer I use always crashes. I don't know what's going on?" After converting the above initial problem into a problem in the target format, the first problem may become: "Please analyze in detail the reasons for the frequent crashes of a personal computer and give possible solutions."

[0047] Optionally, the first target model in this embodiment is a model that can be used to rewrite questions, including but not limited to a large language model (LLM).

[0048] Optionally, the first target model in this embodiment can be Qwen2-7B, which adopts a Transformer architecture, including but not limited to the following components: an input embedding layer (Input Embedding Layer), which is used to convert each word in the first question into a vector representation of a fixed dimension to obtain several vector words; then, the several vector words are input into a positional encoding layer (Positional Encoding Layer), and position information is added to each vector word to further understand the relative position of the above words in the first question; then, the several vector words are input into a multi-head attention layer (Multi-head Attention Layers), and different attention heads focus on different parts of the input to capture the mutual dependence between words in the input first question; then, a feed forward neural network layer (Feed Forward Neural Network Layer) is used to perform a nonlinear transformation on the vector processed by the attention mechanism for further feature extraction and transformation; finally, layer normalization is used to normalize the output after the residual connection of each layer to obtain target information; then, the Qwen2-7B model rewrites the first question in combination with the preset prompt word and outputs the target question through the output layer (Output Layer).

[0049] For example, suppose the user's initial question is: "The server suddenly crashed. How do I find the cause?" After the first text cleaning operation, the first question is standardized as: "The server crashed. How do I find the cause?" The first question is then input into the Qwen2-7B model. The Qwen2-7B model parses and obtains the target information: the application field is IT operation and maintenance, the keywords are server, crash, troubleshooting, and cause, and the semantic structure is that the user is asking how to troubleshoot the server crash and hopes to get specific operation steps or suggestions; the Qwen2-7B model then combines the target information and the selected preset prompt words to perform a rewrite operation on the first question and output the target question.

[0050] Optionally, the first target model in this embodiment can rewrite the question based on understanding, converting it into a more precise, detailed, and standardized form. The principle of rewriting the first target model is to maintain the original meaning of the question while adding detailed information to make it more specific and easier to retrieve and answer later. For example, if the original question is "How do I solve the database connection timeout problem?", the rewritten question might be "How can I solve the database connection timeout problem by optimizing the database connection configuration, especially in a high-concurrency environment?"

[0051] Optionally, in this embodiment, for subsequent analysis and optimization, the historical records of question conversion can be saved, including the initial question, conversion instructions, target question, and model decision information during the conversion process; user feedback can also be collected to understand whether the rewritten question more accurately guides subsequent answers and whether the answers meet user needs; then, based on user feedback and actual results, the first target model is continuously optimized to make it more accurate and efficient in question conversion operations.

[0052] Through the above steps, the initial question is cleaned and formatted, abnormal symbols and stop words are removed, and it is converted into a standard format. Then, the pre-trained first target model is used to analyze the domain, keywords and semantic structure of the initial question to obtain the target information. Then, the target information and preset prompt words are combined to rewrite the question to obtain the target question. This provides precise guidance for subsequent operation and maintenance vector database retrieval, ensuring that the recalled text information is highly relevant to the user question, thereby improving the accuracy and efficiency of the entire operation and maintenance knowledge question and answer system.

[0053] In an exemplary embodiment, before using the target problem to search for N first text vectors that meet a first preset threshold in the operation and maintenance vector database of the target system, and determining the N texts corresponding to the N first text vectors, the method further includes: obtaining original text information from multiple data sources, and performing a preprocessing operation on the original text information to obtain target text information, wherein the preprocessing operation includes a second text cleaning operation, a text deduplication operation, and a text format conversion operation; performing a block operation on the target text information to obtain P text blocks, wherein the P text blocks include N first texts, and the P is a positive integer greater than or equal to the N; determining the second problem and the second summary corresponding to the P text blocks; and storing the text information in the P text blocks, the second problem corresponding to the P text blocks, and the second summary in the operation and maintenance vector database.

[0054] Optionally, the data source in this embodiment is used to provide original text information in the operation and maintenance field, including but not limited to operation and maintenance documents, log files, technical articles, manuals, the company's internal knowledge base, public information technology (IT) operation and maintenance forums, operation and maintenance guides on professional websites, etc.

[0055] Optionally, the original text information in this embodiment is unprocessed text data obtained directly from the data source, which may contain Hypertext Markup Language (HTML) tags, irrelevant advertising content, repeated information, etc., such as the original text of a post captured from an IT operation and maintenance forum.

[0056] Optionally, the text format conversion operation in this embodiment is used to convert the original text into a unified format, for example, converting all HTML documents into a text format to facilitate subsequent processing.

[0057] Optionally, the block segmentation operation in this embodiment refers to dividing the pre-processed long text into multiple smaller text blocks for subsequent processing, with a certain overlap between adjacent text blocks to maintain semantic coherence. The basis for block segmentation includes but is not limited to fixed characters and block segmentation based on semantic subject similarity. For example, a long operation and maintenance document is divided into multiple 500-word text blocks, with an overlap of 100 words between each block.

[0058] Optionally, in this embodiment, the text information corresponding to the P text blocks can be input into a pre-trained third target model to summarize and ask questions about the P text information through the third target model to obtain a second question and a second summary corresponding to each text information.

[0059] Optionally, the third target model used in this embodiment includes but is not limited to a large model, such as the above-mentioned Qwen2-7B model.

[0060] For example, Figure 3 is a schematic diagram of a method for processing original text information according to an embodiment of the present application, such as Figure 3 As shown, the original text information obtained from multiple data sources after the preprocessing operation is divided into blocks to obtain N text blocks; then the Qwen2-7B model is used to summarize and ask questions about the N text blocks to obtain corresponding summary information and preset questions.

[0061] Through the above steps, raw text is obtained from multiple data sources, preprocessed, and divided into P text blocks. A summary and questions are generated for each text block. Finally, the text block information, questions, and summary vectors are stored in the operation and maintenance vector database. This improves the quality and diversity of the data, makes the operation and maintenance vector database storage more reasonable, speeds up retrieval, improves system response efficiency, establishes a knowledge base in professional fields, and provides a solid foundation for intelligent operation and maintenance question and answer.

[0062] In an exemplary embodiment, the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries are all stored in an operation and maintenance vector database, including: performing a vector conversion operation on the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries to obtain P second text vectors, P second question vectors, and P second summary vectors; the text information in the P text blocks, the P second text vectors, the P second question vectors, and the P second summary vectors are all stored in the operation and maintenance vector database, wherein the P second question vectors include the first question vector in the N first text vectors, and the P second summary vectors include the first summary vector in the N first text vectors.

[0063] Optionally, the vector conversion operation in this embodiment is used to convert text information into a vector form. The vector conversion operation can be performed by a pre-trained fourth target model, and the fourth target model can be a model using bge-large-zh-v1.5.

[0064] Optionally, in the present embodiment, when the vector conversion operation is performed using the bge-large-zh-v1.5 model, after the text information is input into the bge-large-zh-v1.5 model, the word embedding layer is first used to convert each word or character in the input text information into a vector representation of a fixed dimension to obtain a number of word embedding vectors; then the position encoding layer adds a position encoding to the word embedding vector to further determine the position of each word in the sentence in the text information to obtain a fourth text vector, wherein the position encoding can be in the form of sine and cosine functions, or use other learning-based encoding methods. formula; then use the multi-head attention layer to focus on different parts of the text information and capture more complex dependencies, where the multi-head attention layer consists of multiple parallel attention heads, each head independently calculates the attention weight to help the model understand the association between words within the sentence and the logical relationship between sentences, and finally outputs the fifth text vector that can reflect the deep semantics of the text; then the fifth text vector is input into the pooling layer to convert it into a sixth text vector of fixed length, where the pooling methods used include but are not limited to average pooling, maximum pooling and classification pooling; finally, the output layer of the model generates the final seventh text vector.

[0065] Optionally, the operation and maintenance vector database in this embodiment can store the text information in P text blocks and P second text vectors in the following form: the storage of the operation and maintenance vector database includes three fields, namely ID, embedding, and metadata, where embedding represents the second text vector, and metadata corresponds to the text information in the text block. These two pieces of information are stored one-to-one in the vector database.

[0066] Through the above steps, the text information in the text block and its corresponding second question and second summary are converted into vectors, and these vector information are stored in the operation and maintenance vector database, which facilitates subsequent efficient retrieval and matching in the operation and maintenance vector database based on the target question, as well as subsequent management and use.

[0067] In an exemplary embodiment, the target problem is used to search for N first text vectors that meet a first preset threshold in the operation and maintenance vector database of the target system, and N texts corresponding to the N first text vectors are determined, including: performing a vector conversion operation on the target problem to obtain a target problem vector; using the target problem vector to search for a text vector that meets a first preset threshold in the operation and maintenance vector database to obtain N first text vectors; querying the text vectors corresponding to the N first problem vectors and the N first summary vectors included in the N first text vectors from a mapping relationship table to obtain N third text vectors, wherein the mapping relationship table is used to store the mapping relationship between the first problem vector and the third text vector, and the mapping relationship between the first summary vector and the third text vector; and determining the N texts corresponding to the N first text vectors based on the N third text vectors.

[0068] Optionally, the vector conversion operation in this embodiment is used to convert text information into a vector form. The vector conversion operation can be performed by a pre-trained fourth target model, and the fourth target model can be a model using bge-large-zh-v1.5.

[0069] Optionally, in this embodiment, the first text vector searched from the operation and maintenance vector database is a text vector whose similarity with the target problem vector meets a first preset threshold. For example, if the first preset threshold is set to 0.7, then only text vectors whose similarity with the target problem vector is greater than or equal to 0.7 will be retrieved.

[0070] Optionally, in this embodiment, the step of searching for the first text vector from the operation and maintenance vector database includes but is not limited to: sorting the text vectors in the operation and maintenance vector library according to the similarity between the text vectors in the operation and maintenance vector library and the target problem vector, and determining the first text vector according to the condition of the top K digits of similarity, where K is a positive integer greater than or equal to 1.

[0071] Optionally, the mapping relationship table in this embodiment stores the mapping relationship between the first question vector and the third text vector, the mapping relationship between the first summary vector and the third text vector, the storage location of the third text vector in the operation and maintenance vector database, and the storage location of the text corresponding to the third text vector in the operation and maintenance vector database.

[0072] Optionally, the N third text vectors in this embodiment are text vectors converted from the original text information, and the N third text vectors correspond to the N first text vectors, that is, the N first question vectors and N first summary vectors included in the N first text vectors are all for the N third text vectors.

[0073] Optionally, the step of determining N texts corresponding to N first text vectors based on N third text vectors in this embodiment includes but is not limited to: when the text vectors and the original text information in the operation and maintenance vector database are not stored one-to-one, querying the storage positions of the N original texts corresponding to the N third text vectors in the operation and maintenance vector database from the above-mentioned mapping relationship table, and obtaining the corresponding N texts from the operation and maintenance vector database based on the queried storage positions; or, when the text vectors and the original text information in the operation and maintenance vector database are stored one-to-one, querying the storage positions of the N third text vectors in the operation and maintenance vector database from the above-mentioned mapping relationship table, and then obtaining the corresponding N stored texts from the operation and maintenance vector database based on the queried storage positions.

[0074] Through the above steps, the target question is converted into a vector representation, and then N first text vectors whose similarity with the question vector meets a first preset threshold are retrieved from the operation and maintenance vector database, and the actual text corresponding to the N first text vectors is found through the mapping relationship table. Converting the target question into a vector helps to improve the accuracy and speed of retrieval. The mapping relationship table ensures accurate conversion from vectors to actual documents, thereby improving the work efficiency of the system.

[0075] In an exemplary embodiment, before querying the text vectors corresponding to the N first question vectors and the N first summary vectors included in the N first text vectors from the mapping relationship table to obtain N third text vectors, the method further includes: establishing a first mapping relationship between the first text information in the N first texts and the N first text vectors; establishing a second mapping relationship between the N first text vectors and the N first question vectors; establishing a third mapping relationship between the N first text vectors and the N first summary vectors; and storing the first mapping relationship, the second mapping relationship, and the third mapping relationship to obtain the mapping relationship table.

[0076] Optionally, in this embodiment, the first text information in the first text is text information obtained after preprocessing and blocking operations and used to construct an operation and maintenance vector database. The first text information includes professional knowledge and problem-solving steps in the operation and maintenance field, such as server troubleshooting, network configuration, software installation guide, etc.

[0077] Optionally, the first mapping relationship in this embodiment refers to a one-to-one mapping relationship between the first text information and the first text vector, that is, the text information in each text block has a corresponding vector representation. For example, the system records the first mapping relationship between the text information in the text block "Server Hardware Troubleshooting Steps" and its corresponding vector.

[0078] Optionally, the second mapping relationship in this embodiment refers to the mapping relationship between the first text vector and the first question vector. The first question vector may be a vector representation of a question generated by the large model for the first text information. For example, the question related to the above-mentioned "Server Hardware Troubleshooting Steps" text block may be "How to troubleshoot server hardware failures?" A second mapping relationship is established between this second question vector and the first text vector.

[0079] Optionally, the third mapping relationship in this embodiment refers to the mapping relationship between the first text vector and the first summary vector. The first summary vector may be a vector representation of the summary obtained by summarizing the first text information through a large model. For example, a summary related to the above text block may be "This article introduces the troubleshooting steps and common failure types of server hardware failures." A third mapping relationship is established between the vector representation of this summary and the first text vector.

[0080] Optionally, the mapping relationship table in this embodiment further stores the location where the first text vector is stored in the operation and maintenance vector database, and the location where the first text information is stored in the operation and maintenance vector database.

[0081] Through the above steps, mapping relationships are established between the first text information and the first text vector, the first text vector and the first question vector, and the first text vector and the first summary vector, and a mapping relationship table is constructed for storage. The construction of the mapping relationship closely associates the various components of the text, thereby improving the consistency and accuracy of answer generation. The mapping relationship is stored in the mapping relationship table, which provides a basis for retrieval and matching, and ensures the stability and reliability of the system.

[0082] In an exemplary embodiment, a target operation and maintenance answer is generated based on the M target texts and the target questions, including: determining the matching degrees between the M target texts and the target questions respectively to obtain M matching degrees; sorting the M target texts according to the M matching degrees; splicing the sorted M target texts to obtain target context information of the target questions; inputting the target context information and the target questions into a pre-trained second target model to obtain the target operation and maintenance answer output by the second target model.

[0083] Optionally, the target text in this embodiment refers to a collection of text related to the user's target question, retrieved through the operation and maintenance vector database. The target text typically contains knowledge or solutions in the field of operation and maintenance. For example, if the user's question is "How do I fix a server memory leak?", the target text might include multiple documents or knowledge items describing the cause of the memory leak, troubleshooting steps, and repair methods.

[0084] Optionally, the matching degree in this embodiment is used to represent the matching degree or correlation between the target text and the target question, including but not limited to using a reranking model, such as the bge-rerank-v2-m3 model.

[0085] Optionally, in this embodiment, the M target texts are sorted according to the M matching degrees, including but not limited to rearranging the M target texts from highest to lowest order to ensure that the most relevant texts are ranked first. For example, after calculation, the system sorts 10 documents according to their relevance to "How to solve the server memory leak problem?" to ensure that the user sees the most relevant document information first.

[0086] Optionally, the second target model in this embodiment includes but is not limited to a pre-trained LLM model, which is used to generate the final operation and maintenance answer. The second target model can be obtained after specific prompt word engineering optimization. For example, using pre-trained models such as Qwen2-7B, by inputting target context information and target questions, the model can integrate the knowledge in the context information and generate a detailed operation and maintenance answer.

[0087] Optionally, the target operation and maintenance answer in this embodiment refers to the system's final answer to the target problem. This answer is generated by the second target model based on the target context and the target problem, aiming to provide accurate, professional, and targeted solutions. For example, the target operation and maintenance answer to the question "How do I fix a server memory leak?" might include common causes of memory leaks, troubleshooting steps, optimization suggestions, and specific repair commands.

[0088] For example, suppose an operation and maintenance personnel encounters the problem of frequent server restarts and suspects that it may be caused by memory leaks. So he enters the question "How to solve the server memory leak problem?" into the operation and maintenance knowledge question and answer system and rewrites the user-entered question to obtain the target question. Then, he first searches through the vector database to find M operation and maintenance documents related to memory leaks, such as system log analysis, memory management software instructions, hardware fault detection guides, etc.; then uses the reranking model (such as bge-rerank-v2-m3) to calculate the matching degree between the M documents and the target question respectively, and obtains M matching degrees; then sorts the M documents according to the matching degree, puts the document with the highest matching degree at the front, and splices the first M documents after sorting to generate a target context information containing multiple document information; finally, the second target model (such as Qwen2-7B) receives the target context information and the target question as input to generate a detailed operation and maintenance answer.

[0089] Through the above steps, the matching degree between M target texts and the target question is calculated. The texts sorted by matching degree are concatenated to form the target context information. The context information and question are then input into the pre-trained second target model to generate the final operation and maintenance answer. The matching degree calculation and sorting ensure that the most important information is given priority, providing the model with rich historical problem-solving experience, helping to generate more detailed and accurate answers, and enhancing the logic and professionalism of the answers.

[0090] Let's take a specific example to illustrate this. Suppose the operation and maintenance department of a large data center is facing frequent server downtime. The operation and maintenance personnel need to quickly find and understand relevant knowledge to determine the cause of the failure and take appropriate measures. Figure 4 This is a process for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 1 ,like Figure 4 As shown, the process includes the following steps:

[0091] Step S402: Perform a question conversion operation on the initial question input by the operation and maintenance personnel to obtain a target question. The operation and maintenance personnel enter a vague initial question into the system, "What should I do if the server frequently crashes?" The initial question is input into a pre-trained language model (e.g., Qwen2-7B) to extract the application domain, keywords, and semantic structure of the question. The question is then rewritten according to the preset prompt "The problem description needs to include specific scenarios and possible causes of the failure" to obtain the target question "In a data center environment, servers frequently crash. Please list possible causes and troubleshooting steps."

[0092] Step S404: Using the target question, search the operation and maintenance vector database of the target system for N first text vectors that meet a first preset threshold, and determine N first texts corresponding to the N first text vectors, where the first text vector includes a first question vector and a first summary vector, one text vector corresponds to one first text, and N is a natural number greater than or equal to 1. First, use the bge-large-zh-v1.5 model to perform a vector conversion operation on the target question to obtain a target question vector. Then, use the target question vector to search the operation and maintenance vector database of the target system for N first text vectors that meet the first preset threshold. Then, determine the storage locations of the N first texts corresponding to the N first text vectors in the operation and maintenance vector database from a mapping relationship table, and read the N first texts from the operation and maintenance vector database.

[0093] Step S406: Filter the N texts to obtain M target texts, where M is a natural number less than or equal to N. The bge-rerank-v2-m3 model is used to rerank the N recalled texts to select the M target texts most relevant to the target issue. These M target texts contain detailed information about potential causes of the server downtime and troubleshooting steps.

[0094] Step S408: Calculate the matching degrees between the M target texts and the target questions, sort the M target texts according to the M matching degrees, obtain target context information, and generate a target operation and maintenance answer based on the target context information and the target question.

[0095] Through the above steps, operation and maintenance personnel can quickly obtain the most relevant and detailed operation and maintenance knowledge related to the problem, improving the efficiency and accuracy of problem solving. At the same time, by using vector databases and mapping relationship tables, the system can process large amounts of text data, provide high concurrency and low-latency retrieval capabilities, and meet the real-time needs in operation and maintenance scenarios.

[0096] Let's take another specific example. Suppose the operations department of a large data center is facing frequent server outages. The operations staff needs to quickly find and understand relevant knowledge to determine the cause of the failure and take appropriate measures. Figure 5 This is a process for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 2 ,like Figure 5 As shown, the process includes the following steps:

[0097] Step S502: The user inputs a question into the first LLM model to obtain a rewritten target question output by the first LLM model, wherein the first LLM model includes but is not limited to the Qwen2-7B model;

[0098] Step S504: searching, based on the target problem, for N first text vectors that meet a first preset threshold from the operation and maintenance vector database, where N is a positive integer greater than or equal to 1;

[0099] Step S506, obtaining N original texts corresponding to the found N first text vectors;

[0100] Step S508: performing a screening operation on the N original texts to obtain M original texts, and concatenating the M original texts to obtain target context information;

[0101] Step S510: Input the target context information and the target question into the second LLM model to obtain the target operation and maintenance answer, wherein the second LLM model includes but is not limited to the Qwen2-7B model. The second LLM model may be a model with a different architecture and composition from the first LLM model.

[0102] Through the above steps, accurate operation and maintenance knowledge and operating guidelines can be provided, and the time required to find and understand solutions can be significantly shortened. To a certain extent, it can make up for the shortcomings of large models in specific field knowledge, reduce the "hallucination" phenomenon, and improve the reliability and professionalism of answers. It can achieve the technical effect of improving the accuracy of large language models in the operation and maintenance field with less computing power consumption.

[0103] It should be noted that, through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0104] In this embodiment, a device for generating an operation and maintenance answer for a target system is also provided. The device is used to implement the above-mentioned embodiments and preferred implementation methods. The details that have been described will not be repeated here. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.

[0105] Figure 6 This is a structural block diagram of a device for generating an operation and maintenance answer for a target system according to an embodiment of the present application. Figure 6 As shown, the device includes:

[0106] The conversion module 602 is configured to perform a problem conversion operation on the obtained initial problem to obtain a target problem, wherein the target problem and the initial problem are both operation and maintenance problems for the target system;

[0107] A search module 604 is configured to use the target question to search for N first text vectors that meet a first preset threshold in the operation and maintenance vector database of the target system, and determine N first texts corresponding to the N first text vectors, wherein the first text vectors include a first question vector and a first summary vector, one text vector corresponds to one first text, and N is a natural number greater than or equal to 1;

[0108] A screening module 606 is configured to perform a screening operation on the N texts to obtain M target texts, where M is a natural number less than or equal to N.

[0109] The generation module 608 is used to generate a target operation and maintenance answer based on the M target texts and the target questions.

[0110] In an exemplary embodiment, the conversion module 602 includes: a first cleaning unit, configured to perform a first text cleaning operation on the initial question to obtain a first question, wherein the first text cleaning operation includes removing abnormal symbols and stop words in the initial question, and converting the initial question into a question in a target format; a first output unit, configured to input the first question into a pre-trained first target model to obtain target information output by the first target model, wherein the target information includes the application field, keywords and semantic structure of the first question; a first rewriting unit, configured to perform a rewriting operation on the first question based on the target information and the selected preset prompt words to obtain the target question.

[0111] In an exemplary embodiment, the search module 604 includes: a first acquisition unit, used to acquire original text information from multiple data sources, and perform a preprocessing operation on the original text information to obtain target text information, wherein the preprocessing operation includes a second text cleaning operation, a text deduplication operation, and a text format conversion operation; a first blocking unit, used to perform a blocking operation on the target text information to obtain P text blocks, wherein the P text blocks include N of the first texts, and P is a positive integer greater than or equal to N; determining the second question and the second summary corresponding to the P text blocks; a first storage unit, used to store the text information in the P text blocks, the second question and the second summary corresponding to the P text blocks in the operation and maintenance vector database.

[0112] In an exemplary embodiment, the search module 604 includes: a first conversion unit, configured to perform a vector conversion operation on the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries to obtain P second text vectors, P second question vectors, and P second summary vectors; and a second storage unit, configured to store the text information in the P text blocks, the P second text vectors, the P second question vectors, and the P second summary vectors in the operation and maintenance vector database, wherein the P second question vectors include the first question vector in the N first text vectors, and the P second summary vectors include the first summary vector in the N first text vectors.

[0113] In an exemplary embodiment, the above-mentioned search module 604 includes: a second conversion unit, used to perform a vector conversion operation on the above-mentioned target problem to obtain a target problem vector; a first search unit, used to use the above-mentioned target problem vector to search for a text vector that meets a first preset threshold in the above-mentioned operation and maintenance vector database to obtain N above-mentioned first text vectors; a first query unit, used to query from a mapping relationship table the text vectors corresponding to the N above-mentioned first problem vectors and the N above-mentioned first summary vectors included in the N above-mentioned first text vectors to obtain N third text vectors, wherein the above-mentioned mapping relationship table is used to store the mapping relationship between the above-mentioned first problem vector and the above-mentioned third text vector, and the mapping relationship between the above-mentioned first summary vector and the above-mentioned third text vector; a first determination unit, used to determine the N texts corresponding to the N above-mentioned first text vectors based on the N above-mentioned third text vectors.

[0114] In an exemplary embodiment, the search module 604 includes: a first establishing unit for establishing a first mapping relationship between the first text information in the N first texts and the N first text vectors; a second establishing unit for establishing a second mapping relationship between the N first text vectors and the N first question vectors; a third establishing unit for establishing a third mapping relationship between the N first text vectors and the N first summary vectors; and a third storage unit for storing the first mapping relationship, the second mapping relationship, and the third mapping relationship to obtain the mapping relationship table.

[0115] In an exemplary embodiment, the above-mentioned generation module 608 includes: a second determination unit, used to respectively determine the matching degrees between the M above-mentioned target texts and the above-mentioned target questions, and obtain M matching degrees; a first sorting unit, used to sort the M above-mentioned target texts according to the M above-mentioned matching degrees; a first splicing unit, used to splice the sorted M target texts, and obtain the target context information of the above-mentioned target question; a second output unit, used to input the above-mentioned target context information and the above-mentioned target question into the pre-trained second target model, and obtain the above-mentioned target operation and maintenance answer output by the above-mentioned second target model.

[0116] It should be noted that the above modules can be implemented through software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0117] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above method embodiments when run.

[0118] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0119] An embodiment of the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0120] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0121] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0122] An embodiment of the present application further provides another computer program product, comprising a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above method embodiments are implemented.

[0123] An embodiment of the present application also provides a computer program, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any of the above method embodiments.

[0124] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail here.

[0125] Obviously, those skilled in the art should understand that the modules or steps of the present application described above can be implemented using a general-purpose computing device, they can be concentrated on a single computing device, or distributed across a network composed of multiple computing devices, they can be implemented using program code executable by the computing device, and thus, they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be performed in a different order than herein, or they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0126] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for generating an operation and maintenance answer for a target system, characterized in that: include: Performing a problem conversion operation on the obtained initial problem to obtain a target problem, wherein the target problem and the initial problem are both operation and maintenance problems for the target system; Using the target question, searching for N first text vectors that meet a first preset threshold in an operation and maintenance vector database of the target system, and determining N first texts corresponding to the N first text vectors, wherein the first text vectors include a first question vector and a first summary vector, one text vector corresponds to one first text, and N is a natural number greater than or equal to 1; Performing a screening operation on the N texts to obtain M target texts, wherein M is a natural number less than or equal to N; Generate a target operation and maintenance answer based on the M target texts and the target question.

2. The method according to claim 1, characterized in that Perform question conversion on the obtained initial question to obtain the target question, including: Performing a first text cleaning operation on the initial question to obtain a first question, wherein the first text cleaning operation includes removing abnormal symbols and stop words in the initial question and converting the initial question into a question in a target format; Inputting the first question into a pre-trained first target model to obtain target information output by the first target model, wherein the target information includes the application field, keywords, and semantic structure of the first question; Based on the target information and the selected preset prompt word, a rewriting operation is performed on the first question to obtain the target question.

3. The method according to claim 1, characterized in that Before searching the operation and maintenance vector database of the target system for N first text vectors that meet a first preset threshold using the target problem and determining N texts corresponding to the N first text vectors, the method further includes: Obtaining original text information from multiple data sources and performing preprocessing operations on the original text information to obtain target text information, wherein the preprocessing operations include a second text cleaning operation, a text deduplication operation, and a text format conversion operation; Performing a block operation on the target text information to obtain P text blocks, wherein the P text blocks include N first texts, and P is a positive integer greater than or equal to N; Determine the second questions and second summaries corresponding to the P text blocks; The text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summary are all stored in the operation and maintenance vector database.

4. The method according to claim 3, characterized in that Storing the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries in an operation and maintenance vector database includes: Performing a vector conversion operation on the text information in the P text blocks, the second questions corresponding to the P text blocks, and the second summaries to obtain P second text vectors, P second question vectors, and P second summary vectors; The text information in the P text blocks, the P second text vectors, the P second question vectors, and the P second summary vectors are all stored in the operation and maintenance vector database, wherein the P second question vectors include the first question vector in the N first text vectors, and the P second summary vectors include the first summary vector in the N first text vectors.

5. The method according to claim 1, wherein Searching for N first text vectors that meet a first preset threshold in an operation and maintenance vector database of the target system using the target problem, and determining N texts corresponding to the N first text vectors, including: Performing a vector conversion operation on the target problem to obtain a target problem vector; Using the target problem vector, searching the operation and maintenance vector database for text vectors that meet a first preset threshold, to obtain N first text vectors; querying a mapping relationship table for text vectors corresponding to both the N first question vectors and the N first summary vectors included in the N first text vectors to obtain N third text vectors, wherein the mapping relationship table is used to store mapping relationships between the first question vectors and the third text vectors, and mapping relationships between the first summary vectors and the third text vectors; N texts corresponding to the N first text vectors are determined based on the N third text vectors.

6. The method according to claim 5, characterized in that Before obtaining N third text vectors by querying a mapping relationship table for text vectors corresponding to the N first question vectors and the N first summary vectors included in the N first text vectors, the method further includes: Establishing a first mapping relationship between the first text information in the N first texts and the N first text vectors; Establishing a second mapping relationship between the N first text vectors and the N first question vectors; Establishing a third mapping relationship between the N first text vectors and the N first summary vectors; The first mapping relationship, the second mapping relationship, and the third mapping relationship are stored to obtain the mapping relationship table.

7. The method according to claim 1, characterized in that Generating a target operation and maintenance answer based on the M target texts and the target question includes: Determine the matching degrees between the M target texts and the target questions respectively, and obtain M matching degrees; Sorting the M target texts according to the M matching degrees; Splicing the sorted M target texts to obtain the target context information of the target question; The target context information and the target question are both input into a pre-trained second target model to obtain the target operation and maintenance answer output by the second target model.

8. A device for generating an operation and maintenance answer for a target system, characterized in that: include: A conversion module, configured to perform a problem conversion operation on the obtained initial problem to obtain a target problem, wherein both the target problem and the initial problem are operation and maintenance problems for the target system; a search module, configured to use the target problem to search for N first text vectors that meet a first preset threshold in an operation and maintenance vector database of the target system, and determine N first texts corresponding to the N first text vectors, wherein the first text vectors include a first problem vector and a first summary vector, one text vector corresponds to one first text, and N is a natural number greater than or equal to 1; a screening module, configured to perform a screening operation on the N texts to obtain M target texts, wherein M is a natural number less than or equal to N; A generation module is used to generate a target operation and maintenance answer based on the M target texts and the target question.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images