Network operation and maintenance scene recognition method, electronic equipment and medium

Abstract

The embodiment of the invention provides a network operation and maintenance scene recognition method, electronic equipment and a medium. The method comprises the steps that operation and maintenance word vectors are obtained and determined according to an operation and maintenance demand text; performing vector retrieval in a preset vector database according to the operation and maintenance word vector to obtain at least one candidate word vector, and obtaining a candidate operation and maintenance scene according to the candidate word vector; determining a first prompt text according to the operation and maintenance demand text, the candidate operation and maintenance scene and a preset first prompt text template; and inputting the first prompt text into a pre-trained first recognition model, so that the first recognition model outputs first operation and maintenance demand recognition information at least comprising the target operation and maintenance scene. The candidate operation and maintenance scenes are screened by vectorizing the operation and maintenance demand text to reduce the recognition range of the operation and maintenance scenes, and the semantic understanding capability of the first recognition model is combined, so that the recognition accuracy of the target operation and maintenance scene is improved by reducing the interference of the non-related operation and maintenance scenes on the semantic understanding of the first recognition model.

Description

Technical Field

[0001] This application relates to the field of network operation and maintenance technology, and particularly relates to a method for identifying network operation and maintenance scenarios, an electronic device, and a medium. Background Art

[0002] Currently, before some major events, in order to ensure the normal operation of the network at the event site, it is often necessary for operation and maintenance personnel to take corresponding safeguard measures for the network devices in the area. The traditional method is that operation and maintenance personnel take some safeguard measures based on their existing operation and maintenance experience before the event. This method is too dependent on the ability of operation and maintenance personnel, and it is impossible to ensure the timeliness of operation and maintenance in this way.

[0003] Based on this, in related technologies, some methods for semi-automatically performing operation and maintenance by identifying operation and maintenance scenarios through natural language processing technology have emerged, such as through the method of intention. This method pre-sets a template containing common operation and maintenance scenarios and solutions. When the user inputs the activity type, the solution for the most matching activity will be executed. This method is more automatic and standardized compared to the manual method.

[0004] With the development of society, operation and maintenance scenarios have gradually increased and become more refined, while the operation and maintenance scenarios that natural language processing technology can support for identification are limited, and it is not easy to distinguish the increasing operation and maintenance scenarios, making the identification of operation and maintenance scenarios relatively inaccurate. Summary of the Invention

[0005] This application proposes a method for identifying network operation and maintenance scenarios, an electronic device, and a medium, aiming to improve the accuracy of identifying operation and maintenance scenarios.

[0006] To achieve the above object, the first aspect of this application provides a method for identifying network operation and maintenance scenarios, and the method includes:

[0007] Obtain an operation and maintenance requirement text, and determine an operation and maintenance word vector according to the operation and maintenance requirement text;

[0008] Perform vector retrieval in a preset vector database according to the operation and maintenance word vector to obtain at least one candidate word vector, and obtain candidate operation and maintenance scenarios according to the candidate word vector;

[0009] Determine a first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and a preset first prompt text template;

[0010] Input the first prompt text into a pre-trained first recognition model, so that the first recognition model outputs first operation and maintenance requirement recognition information, where the first operation and maintenance requirement recognition information at least includes a target operation and maintenance scenario.

[0011] To achieve the above object, a second aspect of the embodiments of the present application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect above is implemented.

[0012] To achieve the above object, a third aspect of the embodiments of the present application provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect above is implemented.

[0013] A network operation and maintenance scenario recognition method, an electronic device, and a medium provided by the embodiments of the present application include: obtaining an operation and maintenance requirement text, and determining an operation and maintenance word vector according to the operation and maintenance requirement text; performing vector retrieval in a preset vector database according to the operation and maintenance word vector to obtain at least one candidate word vector, and obtaining a candidate operation and maintenance scenario according to the candidate word vector; determining a first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenario, and a preset first prompt text template; inputting the first prompt text into a pre-trained first recognition model, so that the first recognition model outputs first operation and maintenance requirement recognition information, where the first operation and maintenance requirement recognition information at least includes a target operation and maintenance scenario. By performing vectorization processing on the operation and maintenance requirement text to screen candidate operation and maintenance scenarios to narrow the recognition range of operation and maintenance scenarios, and combining with the semantic understanding ability of the first recognition model, the recognition accuracy of the target operation and maintenance scenario is improved by reducing the interference of non-related operation and maintenance scenarios on the semantic understanding of the first recognition model. Description of the Drawings

[0014] Figure 1 is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by an embodiment of the present application;

[0015] Figure 2 is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by another embodiment of the present application;

[0016] Figure 3 is Figure 1 a schematic diagram of the steps of a sub-step embodiment of step S101 in

[0017] Figure 4 is Figure 1 a schematic diagram of the steps of another sub-step embodiment of step S101 in

[0018] Figure 5 is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by another embodiment of the present application;

[0019] Figure 6 is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by another embodiment of the present application;

[0020] Figure 7 It is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by another embodiment of the present application;

[0021] Figure 8 It is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by another embodiment of the present application;

[0022] Figure 9 It is a schematic diagram of the structure of a network operation and maintenance scenario recognition device provided by an embodiment of the present application;

[0023] Figure 10 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0024] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0025] It should be noted that unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0026] Currently, before some major events, in order to ensure the normal operation of the network at the event site, it is often necessary for operation and maintenance personnel to take corresponding safeguard measures for the network equipment in the area. The traditional method is that operation and maintenance personnel take some safeguard measures based on the existing operation and maintenance experience of the operation and maintenance personnel before the event. This method is too dependent on the ability of the operation and maintenance personnel, but it is impossible to ensure the timeliness of operation and maintenance in this way.

[0027] Based on this, in the related art, some methods for identifying operation and maintenance scenarios through natural language processing technology for semi-automatic operation and maintenance have emerged, such as through the method of intent. This method pre-sets a template containing common operation and maintenance scenarios and solutions. When the user inputs the event type, the solution of the most matching event will be executed. This method is more automatic and standardized than the manual method.

[0028] With the development of society, operation and maintenance scenarios have gradually increased and become more refined, while the operation and maintenance scenarios that natural language processing technology can support for identification are limited, and it is not easy to distinguish the gradually increasing operation and maintenance scenarios, making the identification of operation and maintenance scenarios relatively inaccurate.

[0029] Based on this, the embodiments of the present application propose a network operation and maintenance scenario recognition method, an electronic device and a medium, aiming to improve the accuracy of operation and maintenance scenario recognition.

[0030] First, the terms appearing in this application are explained as follows:

[0031] Large Language Model (LLM): A model trained based on deep learning technology and a large amount of text data, aiming to understand and generate human language. They can perform a wide range of tasks, including text summarization, translation, sentiment analysis, etc. Large language models are trained with a large amount of data to enable them to understand and generate natural language and other types of content to perform various tasks.

[0032] A network operation and maintenance scenario recognition method, an electronic device, and a medium provided by an embodiment of this application will be specifically described through the following embodiments. First, a network operation and maintenance scenario recognition method provided in the first aspect of the embodiments of this application is described.

[0033] Please refer to Figure 1 , Figure 1 , which is a schematic diagram of the steps of a network operation and maintenance scenario recognition method provided by an embodiment of this application. In this embodiment, the method includes but is not limited to the following steps.

[0034] Step S101: Obtain the operation and maintenance requirement text, and determine the operation and maintenance word vector according to the operation and maintenance requirement text.

[0035] Step S102: Perform vector retrieval in a preset vector database according to the operation and maintenance word vector to obtain at least one candidate word vector, and obtain candidate operation and maintenance scenarios according to the candidate word vector.

[0036] Step S103: Determine the first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and a preset first prompt text template.

[0037] Step S104: Input the first prompt text into a pre-trained first recognition model, so that the first recognition model outputs the first operation and maintenance requirement recognition information.

[0038] It should be noted that the first operation and maintenance requirement recognition information here includes at least the target operation and maintenance scenario.

[0039] Due to the high timeliness requirements of network operation and maintenance, when a network operation and maintenance requirement appears, it is necessary to quickly feedback operation and maintenance information such as the operation and maintenance scenario corresponding to the network operation and maintenance requirement to perform corresponding operation and maintenance operations. Therefore, this application uses a large language model to determine the target operation and maintenance scenario by asking questions in the form of operation and maintenance requirement text.

[0040] In actual network operation and maintenance applications, different operation and maintenance scenarios are set for different operation and maintenance situations, which results in a large number of actually set operation and maintenance scenarios. When a large number of operation and maintenance scenarios are input into the large language model, the large language model needs to perform semantic understanding on a large number of operation and maintenance scenarios, and it is difficult to accurately distinguish the target operation and maintenance scenario from multiple operation and maintenance scenarios through semantic understanding. Therefore, it is necessary to reduce the number of operation and maintenance scenarios input into the large language model to ensure that the semantic understanding ability of the large language model can support the accurate identification of the target operation and maintenance scenario. In order to ensure the correctness of the identification of the target operation and maintenance scenario, the operation and maintenance scenarios input into the large language model need to be related to the operation and maintenance requirement text. At this time, the operation and maintenance scenarios input into the large language model need to be screened before the large language model can be used.

[0041] Specifically, when the device obtains the operation and maintenance requirement text, it first needs to screen the operation and maintenance scenarios according to the operation and maintenance requirement text. In order to associate the operation and maintenance requirement text with the operation and maintenance scenarios, it is first necessary to vectorize the operation and maintenance requirement text, convert the operation and maintenance requirement text into a vector that can be calculated, and construct a connection between the operation and maintenance requirement text and the operation and maintenance scenarios through the feature vector. Therefore, the device first performs word segmentation on the operation and maintenance requirement text, decomposes the entire operation and maintenance requirement text into multiple tokens, and then extracts features for each token to map the entire operation and maintenance requirement text to a multi-dimensional feature space, thereby obtaining the operation and maintenance word vector corresponding to the operation and maintenance requirement text.

[0042] Since network operation and maintenance will change according to the actual network situation, the operation and maintenance scenarios will also change accordingly, which greatly reduces the applicability rate of the neural network model for screening operation and maintenance scenarios in the current usage scenario. If the operation and maintenance scenarios change, the generality of the neural network model for screening operation and maintenance scenarios may not meet the requirements of the timeliness of network operation and maintenance. At this time, it may be necessary to retrain the neural network model.

[0043] Therefore, a preset vector database is set up. The preset vector database can store multiple word vectors, which are converted from a large number of existing operation and maintenance statements. Each operation and maintenance statement corresponds to one of the existing operation and maintenance scenarios (that is, one word vector corresponds to one operation and maintenance scenario). The approximate vector search function of the preset vector database enables the preset vector database to perform vector search according to the input search conditions. When the operation and maintenance scenarios change, increase or decrease, only the binding between the word vectors in the preset vector database and the operation and maintenance scenarios needs to be changed, thereby reducing the impact of the essential attributes of the neural network model on the screening of operation and maintenance scenarios.

[0044] Based on this, input query conditions including operation and maintenance word vectors and the screening quantity into a preset vector database. The preset vector database calculates the vector similarity between the operation and maintenance word vector and each stored word vector respectively, and obtains the vector similarity between each stored word vector and the operation and maintenance word vector. Then, arrange the multiple vector similarities, determine one or more vector similarities corresponding to the screening quantity from the multiple vector similarities according to the screening quantity, and determine the corresponding one or more candidate word vectors based on the one or more screened vector similarities. Since the stored word vectors are obtained by vectorizing existing operation and maintenance statements, when there is one candidate word vector, determine one candidate operation and maintenance scenario according to the candidate word vector, and when there are multiple candidate word vectors, one or more candidate operation and maintenance scenarios can be determined according to the multiple candidate word vectors.

[0045] Next, perform text construction according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and a preset first prompt text template, so as to determine the first prompt text. The first prompt text includes example corpus formed according to the candidate operation and maintenance scenarios and question corpus formed according to the operation and maintenance requirement text, and the question corpus includes a part for questioning the target operation and maintenance scenario.

[0046] Input the first prompt text into a pre-trained first recognition model to determine the target operation and maintenance scenario. The first recognition model is a large language model that can generate natural language and other information through the input text. Since the first prompt text includes example corpus, the first recognition model first performs semantic understanding according to the example corpus to obtain semantic understanding information, and then, performs semantic understanding on the question corpus according to the obtained semantic understanding information, so as to generate an answer corresponding to the question corpus (i.e., the first operation and maintenance requirement recognition information), and output the first operation and maintenance requirement recognition information outward. And since the question corpus includes a part for questioning the target operation and maintenance scenario, the first operation and maintenance requirement recognition information will correspondingly include the target operation and maintenance scenario.

[0047] In the embodiment of the present application, by performing vectorization processing on the operation and maintenance requirement text to screen candidate operation and maintenance scenarios to narrow the recognition scope of operation and maintenance scenarios, and combining with the semantic understanding ability of the first recognition model, the recognition accuracy of the target operation and maintenance scenario is improved by reducing the interference of non-related operation and maintenance scenarios on the semantic understanding of the first recognition model.

[0048] It should be noted that the specific form of the first recognition model here is diverse. Exemplarily, such as the Baichuan model, the chatGLM model, etc., and the embodiment of the present application does not limit this.

[0049] It should be noted that the specific manner of vectorizing the operation and maintenance requirement text to determine the operation and maintenance word vector here is diverse, which can be the following embodiments or other embodiments, and the embodiment of the present application does not limit this.

[0050] In one embodiment, the sbert-base-chinese-nli model is used to tokenize the operation and maintenance requirement text to obtain multiple tokens. Each token is subjected to word embedding processing, and the operation and maintenance word vectors are obtained through positional encoding and the transformer encoding layer.

[0051] In one embodiment, the operation and maintenance requirement text is tokenized to obtain multiple tokens. Each token is individually mapped to the feature space by means such as one-hot encoding, bag of words (BOW), word2vec, etc. to obtain a word vector. Then, according to the method of speech extraction based on multiple tokens, the semantic vector is incorporated into the obtained word vector to obtain the operation and maintenance word vector.

[0052] It should be noted that the acquisition method of the screening quantity in the query condition is diverse. Exemplarily, a screening quantity is set according to the semantic understanding ability of the first recognition model; or, within the allowable range of the semantic understanding ability of the first recognition model, it is determined according to the vector similarity between each stored word vector and the operation and maintenance word vector, etc. The embodiments of the present application do not limit this.

[0053] It should be noted that the number of candidate operation and maintenance scenarios determined by the candidate word vectors here may or may not correspond. The present application does not limit this. Exemplarily, if there is only one operation and maintenance scenario corresponding to the multiple retrieved candidate word vectors, then there is only one candidate operation and maintenance scenario obtained according to the multiple candidate word vectors.

[0054] Exemplarily, if there are multiple operation and maintenance scenarios corresponding to the multiple retrieved candidate word vectors, then there are multiple candidate operation and maintenance scenarios obtained according to the multiple candidate word vectors.

[0055] It should be noted that the specific method of inputting the first prompt text into the first recognition model here is diverse, which may be other embodiments or other embodiments. The embodiments of the present application do not limit this.

[0056] In one embodiment, there is only one first recognition model set. The first prompt text is directly input into the first recognition model to obtain the first operation and maintenance requirement recognition information output by the first recognition model, so as to determine the target operation and maintenance scenario.

[0057] In one embodiment, there are multiple large language models set in the device. The first recognition model is determined from the multiple large language models according to the satisfied conditions. Exemplarily, since the semantic understanding abilities of different large language models are different, the first recognition model is determined from the multiple large language models according to the number of obtained candidate operation and maintenance scenarios.

[0058] In the actual operation and maintenance process, the operation and maintenance requirement text may also carry information related to operation and maintenance other than the target operation and maintenance scenario. In order to facilitate operation and maintenance personnel to obtain accurate and comprehensive operation and maintenance information and improve the information utilization rate of the operation and maintenance requirement text, it is necessary to identify the information related to operation and maintenance other than the target operation and maintenance scenario.

[0059] Please refer to Figure 2 , Figure 2 which is a schematic diagram of the steps of the network operation and maintenance scenario identification method provided by another embodiment of this application. In one embodiment, after step S104, the method further includes but is not limited to the following steps.

[0060] Step S201: Determine the second prompt text according to the operation and maintenance requirement text and the second preset text template.

[0061] Step S202: Input the second prompt text into the pre-trained second recognition model so that the second recognition model outputs operation and maintenance element information.

[0062] For the same type of information, when the operation and maintenance requirement text is input into different large language models, due to the different semantic understanding degrees of each large language model, the results of extracting this type of information may be different; moreover, for large language models with the same architecture, different tasks are executed, which will cause two large language models with the same architecture to have different biases, and there may be different accuracies for extracting different types of information. In order to reduce the impact of the differences in semantic understanding ability and bias problems between large language models on the extraction of operation and maintenance information, it is necessary to use additional large language models to identify different types of information.

[0063] Specifically, text construction is carried out according to the operation and maintenance requirement text, the candidate operation and maintenance scenario, and the preset second prompt text template, and a second prompt text different from the first prompt text is reconstructed. The second prompt text includes example corpus about operation and maintenance element information and question corpus formed according to the operation and maintenance requirement text, and the question corpus includes the part of questioning operation and maintenance element information.

[0064] Input the second prompt text into the pre-trained second recognition model to determine the operation and maintenance element information. The second recognition model is a large language model that can generate natural language and other information through the input text. And the second prompt text includes example corpus. The second recognition model will first perform semantic understanding according to the example corpus to obtain semantic understanding information, and then perform semantic understanding on the question corpus according to the obtained semantic understanding information, so as to generate the answer corresponding to the question corpus (i.e., operation and maintenance element information), and output the operation and maintenance element information outward.

[0065] It should be noted that the second recognition model here can be a model with the same architecture as the first recognition model, or it can be a different model. The embodiments of the present application do not limit this. The specific form of the second recognition model here is diverse. Exemplarily, such as the Baichuan model, the chatGLM model, etc. The embodiments of the present application do not limit this.

[0066] It should be noted that the specific manner of inputting the second prompt text into the second recognition model here is diverse. It can be other embodiments or other embodiments. The embodiments of the present application do not limit this.

[0067] In one embodiment, there is only one second recognition model set. The second prompt text is directly input into the second recognition model to obtain the operation and maintenance element information output by the second recognition model.

[0068] In one embodiment, there are multiple large language models set in the device. The first recognition model is determined from the multiple large language models according to the satisfied conditions. Exemplarily, since the semantic understanding capabilities of different large language models are different, the second recognition model is determined from the multiple large language models according to the number of obtained candidate operation and maintenance scenarios.

[0069] It should be noted that the operation and maintenance element information here includes at least one of time information and location information. Due to human subjective initiative, the operation and maintenance requirement texts input into the device in various ways are diverse. The second recognition type model can only recognize the operation and maintenance element information contained in the operation and maintenance requirement text.

[0070] Exemplarily, if the operation and maintenance requirement text only contains time information, then only time information will be included in the operation and maintenance element information finally output by the second recognition model to the operation and maintenance personnel, and the location information will be displayed as "none" or not displayed.

[0071] Exemplarily, if the operation and maintenance requirement text only contains location information, then only location information will be included in the operation and maintenance element information finally output by the second recognition model to the operation and maintenance personnel, and the time information will be displayed as "none" or not displayed.

[0072] Exemplarily, if the operation and maintenance requirement text contains time information and location information, then the operation and maintenance element information finally output by the second recognition model to the operation and maintenance personnel will include time information and location information.

[0073] In an embodiment of the present application, the second recognition model is used to obtain operation and maintenance element information, which improves the information utilization rate in the operation and maintenance requirement text and enhances the accuracy, comprehensiveness, and convenience for operation and maintenance personnel to obtain operation and maintenance information. Moreover, since the acquisition of operation and maintenance element information and the acquisition of the target operation and maintenance scenario adopt additional recognition models, the influence of the semantic understanding ability and bias of the first recognition model on the operation and maintenance element information for obtaining the operation and maintenance element information is reduced, and the accuracy of obtaining the operation and maintenance element information is improved.

[0074] In one embodiment, the first operation and maintenance requirement recognition information further includes operation and maintenance element information. It should be noted that the operation and maintenance element information here includes at least one of time information and location information.

[0075] Due to the limited computing power of the device (enabling two models requires a considerable part of the device's computing resources) and storage capacity (a large language model will generate cache data and store some data locally during the calculation process, and these data have high dimensions and large amounts, consuming a considerable part of the device's storage resources, and two large language models will consume even more storage resources), some devices can only compromise and set the first recognition model to identify the two types of information.

[0076] Specifically, in addition to the example corpus formed according to the candidate operation and maintenance scenarios, the first prompt text also includes an example corpus regarding operation and maintenance element information. In addition to the part that questions the target operation and maintenance scenario, the question corpus also includes the part that questions the operation and maintenance element information.

[0077] When the first prompt text is input into the first recognition model, the first recognition model will not only perform semantic understanding of the example corpus formed according to the candidate operation and maintenance scenarios to obtain operation and maintenance scenario semantic understanding information, but also perform semantic understanding of the example corpus regarding operation and maintenance element information to obtain operation and maintenance element semantic understanding information.

[0078] Then, according to the operation and maintenance scenario semantic understanding information, semantic understanding is performed on the part that questions the target operation and maintenance scenario to obtain the target operation and maintenance scenario. According to the operation and maintenance element semantic understanding information, semantic understanding is performed on the part that questions the operation and maintenance element information to obtain the operation and maintenance element information. Finally, the first operation and maintenance requirement recognition information is output according to the target operation and maintenance scenario and the operation and maintenance element information.

[0079] It should be noted that the order in which the first recognition model performs semantic understanding on the part that questions the target operation and maintenance scenario and the part that questions the operation and maintenance element information is diverse. It can first perform semantic understanding on the part that questions the target operation and maintenance scenario and then perform semantic understanding on the part that questions the operation and maintenance element information, or it can first perform semantic understanding on the part that questions the operation and maintenance element information and then perform semantic understanding on the part that questions the target operation and maintenance scenario, etc. The present application does not limit this.

[0080] It should be noted that since the operation and maintenance requirement texts are diverse, the first recognition model performs semantic understanding on the part of the questioned operation and maintenance element information according to the semantic information of the operation and maintenance elements, and the specific forms of the obtained operation and maintenance element information are also diverse. It can be the following embodiments or other embodiments, and the embodiments of the present application do not limit this.

[0081] In one embodiment, if the operation and maintenance requirement text only contains time information, then only the time information will be included in the operation and maintenance element information finally output by the first recognition model to the operation and maintenance personnel, and the location information will be displayed as "none" or not displayed.

[0082] In one embodiment, if the operation and maintenance requirement text only contains location information, then only the location information will be included in the operation and maintenance element information finally output by the first recognition model to the operation and maintenance personnel, and the time information will be displayed as "none" or not displayed.

[0083] In one embodiment, if the operation and maintenance requirement text contains time information and location information, then the operation and maintenance element information finally output by the first recognition model to the operation and maintenance personnel will include time information and location information.

[0084] The embodiments of the present application extract the target operation and maintenance scenario and the operation and maintenance element information simultaneously through the first recognition model, which improves the output speed of the first operation and maintenance requirement recognition information, thereby improving the efficiency of the operation and maintenance personnel to obtain the first operation and maintenance requirement recognition information; and reduces the consumption of device resources for the extraction of operation and maintenance information, thereby making the subsequent extraction of operation and maintenance information more efficient.

[0085] It should be noted that the specific method for obtaining the operation and maintenance requirement text is diverse. It can be the following embodiments or other embodiments, and the embodiments of the present application do not limit this.

[0086] Please refer to Figure 3 , Figure 3 is Figure 1 a schematic diagram of the steps of a sub-step embodiment of step S101 in. In one embodiment, step S101 includes but is not limited to the following sub-steps.

[0087] Step S301, obtain the operation and maintenance requirement information input by the target object.

[0088] Step S302, in the case where the operation and maintenance requirement information is text information, obtain the operation and maintenance requirement text according to the text information.

[0089] Step S303, in the case where the operation and maintenance requirement information is non-text information, perform text conversion on the non-text information to obtain the operation and maintenance requirement text.

[0090] Since the device does not limit the type of input information, the operation and maintenance requirement information that the device can obtain is diverse. That is to say, some operation and maintenance requirement information is not necessarily text, and the device cannot perform text vectorization processing based on non-text information. It is necessary to perform text conversion processing on non-text information before text vectorization processing can be carried out.

[0091] Specifically, after obtaining the operation and maintenance requirement information input by the target object, text vectorization processing will not be directly started. Instead, the type of the operation and maintenance requirement information is judged to obtain the type information of the operation and maintenance requirement information. Depending on the content represented by the type information, different functions are executed.

[0092] In the case where the operation and maintenance requirement information is text information, there is no need for text conversion, and the operation and maintenance requirement text can be obtained based on the text information.

[0093] In the case where the operation and maintenance requirement information is non-text information, text conversion is performed on the non-text information to obtain the operation and maintenance requirement text.

[0094] In the case where the operation and maintenance requirement information includes both text information and non-text information, first, the text information part and the non-text information part in the operation and maintenance requirement information are separated. Then, text conversion is performed on the non-text information part to obtain the text information corresponding to the non-text information part. The operation and maintenance requirement text is generated based on the text information part and the text information corresponding to the non-text information part.

[0095] It should be noted that the specific form of obtaining the operation and maintenance requirement text based on the text information is diverse. Exemplarily, the text information can be directly determined as the operation and maintenance requirement text; or the text information can be pre-processed to remove useless characters and other useless corpus in the text information to obtain the operation and maintenance requirement text, etc. The embodiments of the present application do not limit this.

[0096] It should be noted that the type of non-text information here is diverse. Exemplarily, when the non-text information is picture-like information, the picture is recognized by means of scene recognition, etc., and the picture-like information is converted into an operation and maintenance requirement text expressed by text; when the non-text information is voice information, the operation and maintenance requirement text is obtained by means of speech-to-text conversion. It should be understood that when the operation and maintenance requirement information is video, it will include but not be limited to picture information and voice information at this time. At this time, it is necessary to perform picture recognition and speech-to-text conversion and combine the text information obtained by the two methods as the operation and maintenance requirement text.

[0097] It should be noted that the target object here refers to the person who inputs the operation and maintenance requirement text to the device, and it is specifically diverse. Exemplarily, it can be an operation and maintenance personnel, a device management personnel, etc. For the convenience of narration in the embodiments of the present application, the operation and maintenance personnel are used to describe the embodiments.

[0098] It should be noted that the specific methods for obtaining the operation and maintenance requirement information input by the target object here are diverse, and there can be but are not limited to the following embodiments.

[0099] In one embodiment, the device is provided with an interaction device, and the operation and maintenance personnel directly input the operation and maintenance requirement information locally on the device through the interaction device.

[0100] In one embodiment, the device is a remote device, and an interface is provided for the terminal of the operation and maintenance personnel. The operation and maintenance personnel can upload the operation and maintenance requirement information to the device through the terminal, and the device can obtain the operation and maintenance requirement text according to the uploaded operation and maintenance requirement information. Exemplarily, for example, the operation and maintenance personnel upload a certain operation and maintenance requirement email in their mailbox; or input a paragraph of text, etc.

[0101] In the embodiment of the present application, by obtaining the operation and maintenance requirement information input by the target object, the target operation and maintenance scenario recognition process can be carried out according to the actual operation and maintenance requirements of the target object, improving the degree of fit between the target operation and maintenance scenario recognition and the actual operation and maintenance tasks of the target object; and, since the first recognition module is used to perform target operation and maintenance scenario recognition on the input operation and maintenance requirement information, the target operation and maintenance scenario recognition can be completed in the interaction process with the target object, improving the interactivity between the target operation and maintenance scenario recognition and the target object.

[0102] Please refer to Figure 4 , Figure 4 For Figure 1 the schematic diagram of the steps of another sub-step embodiment of step S101 in. In one embodiment, step S101 includes but is not limited to the following sub-steps.

[0103] Step S401, obtain the event information pushed by the operation and maintenance system.

[0104] Step S402, convert the event information pushed by the operation and maintenance system into an operation and maintenance requirement text.

[0105] In addition to obtaining the operation and maintenance requirement information actively input by the target object to meet the operation and maintenance requirements of the target object, the device can also obtain the information that is not actively input and send an operation and maintenance warning to the operation and maintenance personnel, so that the operation and maintenance personnel can better prepare for network operation and maintenance.

[0106] Specifically, obtain the event information pushed by the operation and maintenance system and convert the event information pushed by the operation and maintenance system into an operation and maintenance requirement text. The event information pushed by the operation and maintenance system is generated by the operation and maintenance system. The operation and maintenance system can monitor some information sources. When these information sources generate specific information, the operation and maintenance system generates the event information pushed by the operation and maintenance system according to these specific information and pushes it to the device.

[0107] It should be noted that the operation and maintenance system directly pushes these specific information to the device, so that the event information pushed by the operation and maintenance system will essentially include information related to operation and maintenance and other unnecessary information. After the device receives the event information pushed by the operation and maintenance system, it needs to perform information conversion on the event information pushed by the operation and maintenance system to obtain the operation and maintenance requirement text.

[0108] Exemplarily, monitor the emails of operation and maintenance personnel. When the email is from a specific contact or a specific word appears in the full text of the email, the entire email is directly pushed to the device as the event information pushed by the operation and maintenance system, and the device converts this email into the operation and maintenance requirement text.

[0109] Exemplarily, monitor a specific website through a website monitoring tool. When a tweet with relevant words appears on the website, the tweet or the URL of the tweet is sent to the device as the event information pushed by the operation and maintenance system. When the tweet is the event information pushed by the operation and maintenance system, the device needs to extract information from the tweet to convert and obtain the operation and maintenance requirement text; when the website is the URL of the tweet, the device needs to use a crawler tool to crawl the tweet and extract information from the tweet to convert and obtain the operation and maintenance requirement text.

[0110] Exemplarily, the original function of the operation and maintenance system is to perform security monitoring on some network devices, record security logs or issue alarms. By setting the operation and maintenance system to generate event information pushed by the operation and maintenance system according to the security logs at regular intervals, or triggered by a certain event, generate event information pushed by the operation and maintenance system according to the recorded security logs and / or network device alarm information and send it to the device. After receiving these information, the device performs text conversion to obtain the operation and maintenance requirement text.

[0111] The embodiment of the present application generates the operation and maintenance requirement text through the event information pushed by the operation and maintenance system, improves the intelligent level of target operation and maintenance scenario recognition, enables operation and maintenance personnel to understand the objectively existing operation and maintenance requirements in the communication network faster and more comprehensively, and thus improves the comprehensiveness of network operation and maintenance.

[0112] It should be noted that there are multiple modules in the first prompt text template here. The content in each module can be fixed or filled by the device according to certain information, so that determining the first prompt text according to the operation and maintenance requirement text, candidate operation and maintenance scenarios and the preset first prompt text template can have but are not limited to the following embodiments.

[0113] In one embodiment, the first prompt text template includes a conflict decoupling module, and step S103 includes but is not limited to the following sub-steps.

[0114] Step S501, obtain conflict decoupling text according to multiple candidate operation and maintenance scenarios.

[0115] Step S502, fill the conflict decoupling text into the conflict decoupling module.

[0116] It should be noted that the conflict decoupling text here decouples multiple candidate operation and maintenance scenarios from the operation and maintenance requirement text.

[0117] For operation and maintenance personnel, the target operation and maintenance scenario determined by the first recognition model according to the operation and maintenance requirement text in multiple candidate operation and maintenance scenarios may conflict with the operation and maintenance experience of the operation and maintenance personnel; moreover, there are some relatively similar operation and maintenance scenarios among multiple candidate operation and maintenance scenarios, which makes the identified target operation and maintenance scenario not necessarily the operation and maintenance scenario considered by the operation and maintenance personnel. Therefore, it is necessary to decouple it through the conflict decoupling text to make the identified target operation and maintenance scenario more matched with the cognition of the operation and maintenance personnel and reduce the conflict between similar operation and maintenance scenarios.

[0118] Specifically, a text library for storing conflict decoupling text is set in the device. According to multiple candidate operation and maintenance scenarios, conflict decoupling text related to the candidate operation and maintenance scenarios, conflict decoupling text with recognition conflicts between multiple candidate operation and maintenance scenarios, etc. are retrieved from the conflict decoupling text library, and these retrieved conflict decoupling text are filled into the conflict decoupling module in the first prompt text template.

[0119] Exemplarily, if the candidate operation and maintenance scenarios include "sudden high load", "major domestic event", "tidal area monitoring", "video guarantee", the conflict decoupling text related to these 4 candidate operation and maintenance scenarios is obtained from the conflict decoupling text library and filled into the conflict decoupling module:

[0120] 1. If the input clearly mentions "video guarantee", then it is preferentially judged as "video guarantee";

[0121] 2. When the number of people exceeds 10,000, it is a "major domestic event", and when it is less than that, it is a "sudden high load";

[0122] 3. If it causes blockage to traffic and public transportation, it is a "sudden high load", and if there is no blockage, it is a "major domestic event".

[0123] When the above situations exist in the operation and maintenance requirement text, decoupling is performed according to the content in the conflict decoupling text to determine the target operation and maintenance scenario.

[0124] In the embodiment of the present application, by obtaining the conflict decoupling text and filling it into the conflict decoupling module according to the conflict decoupling text, the first prompt text has the decoupling function, so that the recognition process of the target operation and maintenance scenario can reduce the influence of similar candidate operation and maintenance scenarios on the recognition accuracy; moreover, since the conflict decoupling text needs to be set by operation and maintenance personnel in advance, the first operation and maintenance requirement recognition information is more biased, improving the fitting degree between the first operation and maintenance requirement recognition information and the actual experience of operation and maintenance personnel.

[0125] In one embodiment, the first prompt text template includes a scenario example filling module, and step S103 includes but is not limited to the following sub-steps.

[0126] Step S601: For each candidate operation and maintenance scenario, obtain the scenario example text corresponding to the candidate operation and maintenance scenario.

[0127] Step S602: Fill multiple scenario example texts into the scenario example filling module.

[0128] Specifically, a text library for storing scenario example texts is set in the device. For each candidate operation and maintenance scenario, retrieve one or more corresponding scenario example texts from the scenario example text library according to the candidate operation and maintenance scenario; fill the one or more scenario example texts corresponding to each candidate operation and maintenance scenario into the scenario example filling module to form example corpus.

[0129] Exemplarily, still taking the four candidate operation and maintenance scenarios exemplified above as an example, assuming that each of the four candidate operation and maintenance scenarios has only one scenario example text in the scenario example text library, obtain these four scenario example texts and fill them into the scenario example filling module:

[0130] ## Input: At location Q, a fireworks display event will be held at U o'clock tonight, and many tourists come to watch.

[0131] ## Output: "Domestic major event"

[0132] ## Input: An event is held at location X, and the location reaches the maximum number of people on that day.

[0133] ## Output: "Sudden high load"

[0134] ## Input: Next Wednesday, it is necessary to monitor the traffic trends in the morning and evening in town Y.

[0135] ## Output: "Tidal area monitoring"

[0136] ## Input: A dragon boat open competition will be held at location P tomorrow, and XX TV station will conduct live reports.

[0137] ## Output: "Video guarantee"

[0138] In the embodiment of the present application, by obtaining the scenario example text corresponding to each candidate operation and maintenance scenario, the first prompt text does not need to include the scenario example texts corresponding to all preset operation and maintenance scenarios, reducing the size of the first prompt text, reducing the amount of data that the first recognition model needs to perform semantic understanding and the influence brought by other irrelevant scenario example texts, thereby improving the accuracy and efficiency of target operation and maintenance scenario recognition.

[0139] In one embodiment, the first prompt text template includes a candidate item filling module, and step S103 includes but is not limited to the following sub-steps.

[0140] Step S701, filling multiple candidate operation and maintenance scenarios into the candidate item filling module.

[0141] The candidate item filling module is used to construct partial question corpus to limit the recognition scope of the first recognition model, which can prevent the first recognition model from outputting texts other than multiple candidate operation and maintenance scenarios as the first operation and maintenance requirement information.

[0142] Exemplarily, still taking the four candidate operation and maintenance scenarios exemplified above as an example, filling the four candidate operation and maintenance scenarios into the candidate item filling module to form partial question corpus:

[0143] Please judge which of the following options the following input belongs to: ["Sudden high load", "Major domestic events", "Tidal area monitoring", "Video guarantee"]

[0144] When the first recognition model outputs the first operation and maintenance requirement information, the target operation and maintenance scenario included therein can only be one of these four.

[0145] In the embodiment of the present application, by filling multiple candidate operation and maintenance scenarios into the candidate item filling module, the recognition scope of the first recognition model is limited, and the accuracy of target operation and maintenance scenario recognition is improved.

[0146] In one embodiment, the first prompt text template includes an input filling module, and step S103 includes but is not limited to the following sub-steps.

[0147] Step S801, filling the operation and maintenance requirement text into the input filling module.

[0148] The input filling module is used to construct partial question corpus to form a question statement. Therefore, an operation and maintenance requirement text matching the format of the scenario example text can be constructed to improve the matching degree between the semantic understanding information and the operation and maintenance requirement text, so that the first recognition model can perform semantic understanding on the operation and maintenance requirement text by obtaining semantic understanding information according to the scenario example text, and improve the accuracy of target operation and maintenance scenario recognition.

[0149] Exemplarily, still taking the four candidate operation and maintenance scenarios exemplified above as an example, filling the operation and maintenance requirement text into the input filling module to form partial question corpus:

[0150] ## Input: The opening ceremony of the Dth sports meeting will be held at Stadium E at D o'clock in the evening on A year B month C day. Please make good guarantee.

[0151] ## Output:

[0152] In one embodiment, the first prompt text template includes a background filling module, and step S103 includes but is not limited to the following sub-steps.

[0153] Step S901, fill the background filling module according to multiple candidate operation and maintenance scenarios.

[0154] In one embodiment, directly fill multiple candidate operation and maintenance scenarios into the background filling module to form the background of the first prompt text.

[0155] Exemplarily, still taking the four candidate operation and maintenance scenarios exemplified above as an example, fill multiple candidate operation and maintenance scenarios into the background filling module to form the background of the first prompt text:

[0156] You are an excellent operation and maintenance assistant robot. Please identify which type of guarantee the user's input belongs to according to the following list information:

[0157] The list of guarantee types is as follows:

[0158] ## Sudden high load.

[0159] ## Major domestic events.

[0160] ## Tidal area monitoring.

[0161] ## Video guarantee.

[0162] In one embodiment, for each candidate operation and maintenance scenario, obtain the corresponding interpretation text according to the candidate operation and maintenance scenario; fill each candidate operation and maintenance scenario and its corresponding interpretation text into the background filling module to form the background of the first prompt text.

[0163] Still taking the four candidate operation and maintenance scenarios exemplified above as an example, fill multiple candidate operation and maintenance scenarios into the background filling module to form the background of the first prompt text:

[0164] You are an excellent operation and maintenance assistant robot. Please identify which type of guarantee the user's input belongs to according to the following list information:

[0165] The list of guarantee types is as follows:

[0166] ## Sudden high load: It means that due to a large number of tourists or people gathering in a certain area in a short period of time, serious traffic jams or other pressures have occurred in that area.

[0167] ## Major domestic events: It refers to some large-scale activities or competitions, usually requiring advance preparations, involving a large number of people, and may also affect the surrounding traffic and living order.

[0168] ##Tidal area monitoring: The main concern is the change in traffic flow at a specific location during a specific period.

[0169] ##Video guarantee: It mainly involves the requirements for video transmission and guarantee. For example, providing on-site video transmission and ensuring the quality of live broadcasts. Such examples are usually related to various activities, events or important events, such as the Dragon Boat Open, or when natural disasters occur and videos are needed for reporting.

[0170] In the embodiment of the present application, by obtaining the corresponding interpretation text for each candidate operation and maintenance scenario, the background of the first prompt text is constructed, so as to enhance the semantic understanding information obtained by the first recognition model according to the example corpus, thereby improving the accuracy of target operation and maintenance scenario recognition.

[0171] It should be noted that the specific method of obtaining the candidate operation and maintenance scenario according to the candidate word vector in step S102 is diverse, and it can be the following embodiments or other embodiments, and the embodiments of the present application do not limit this.

[0172] In one embodiment, the preset vector database stores multiple word vectors in the form of a table, and an operation and maintenance scenario item is set in the table. That is to say, in addition to storing a word vector in each row of the table, a corresponding operation and maintenance scenario will also be stored. For each of a retrieved candidate word vector and multiple candidate word vectors, the corresponding operation and maintenance scenario is extracted from the row to which the candidate word vector belongs as the candidate operation and maintenance scenario.

[0173] In one embodiment, the preset vector database includes multiple operation and maintenance scenario identifiers, and each of one or more candidate word vectors corresponds to one of the multiple operation and maintenance scenario identifiers. Step S102 includes but is not limited to the following sub-steps.

[0174] Step S1001, perform index retrieval in the preset relational database according to the corresponding operation and maintenance scenario identifier to determine the candidate operation and maintenance scenario corresponding to the operation and maintenance scenario identifier.

[0175] It should be noted that the preset relational database here includes multiple operation and maintenance scenario identifiers and multiple preset operation and maintenance scenarios, and the multiple preset operation and maintenance scenarios correspond to the multiple operation and maintenance scenario identifiers one by one, and the candidate operation and maintenance scenario is one of the multiple preset operation and maintenance scenarios.

[0176] Specifically, operation and maintenance scenarios are stored in different types of databases with word vectors. The operation and maintenance scenarios are queried through an associated query method. The preset vector database stores multiple word vectors in the form of a table, and an operation and maintenance scenario identifier item is set in the table. That is to say, in each row of the table, in addition to storing a word vector, a corresponding operation and maintenance scenario identifier is also stored. When the retrieved candidate word vector is one, the corresponding operation and maintenance scenario identifier is extracted from the row where the candidate word vector belongs; when the retrieved candidate word vectors are multiple, for each retrieved word vector, the corresponding operation and maintenance scenario identifier is extracted from the row where the candidate word vector belongs.

[0177] The preset relational database stores multiple operation and maintenance scenario identifiers and their corresponding multiple operation and maintenance scenarios in the form of a table. According to the candidate word vector, the candidate operation and maintenance scenario can be determined from the table. Since the operation and maintenance scenarios and word vectors are stored separately, when the operation and maintenance scenarios actually change, only the content in the preset relational database needs to be changed, without changing the preset vector database, thus ensuring the data integrity of the preset vector database, reducing the change of word vectors caused by problems with database operation statements, and improving the accuracy of target operation and maintenance scenario recognition; moreover, due to the separate database storage, the two types of data can be distributedly stored using dedicated database devices, thereby reducing the time for querying and invoking device resources and improving the efficiency of target operation and maintenance scenario recognition.

[0178] Please refer to Figure 5 and Figure 6 , Figure 5 and Figure 6 are respectively the schematic diagrams of the steps of the network operation and maintenance scenario recognition method provided by another embodiment of the present application. In one embodiment, after obtaining the operation and maintenance element information, the method further includes but is not limited to the following steps.

[0179] Step S1101, obtain the target operation and maintenance operation information corresponding to the target operation and maintenance scenario according to the target operation and maintenance scenario.

[0180] Step S1102, output the target operation and maintenance content according to the operation and maintenance element information and the target operation and maintenance operation information.

[0181] Specifically, each operation and maintenance scenario is set with corresponding operation and maintenance operation information. In the scenario of manual operation and maintenance, when performing network maintenance, operation and maintenance personnel need to perform operation and maintenance on network devices according to the operation and maintenance operation information. Therefore, the corresponding target operation and maintenance operation information can be obtained according to the target operation and maintenance operation information, so as to form a complete target operation and maintenance content with the target operation and maintenance operation information to prompt the operation and maintenance personnel.

[0182] Exemplarily, taking the "domestic major event" in the above embodiment as an example of the target operation and maintenance operation information, the operation and maintenance operation information corresponding to this operation and maintenance scenario is as follows:

[0183] Before the event occurs: Backup of network element data, identification of network hidden dangers, deployment of high-traffic capacity, and summary of pre-inspection;

[0184] During the event occurs: Network monitoring, early warning emergency handling, high-load optimization, and video service guarantee;

[0185] After the event occurs: Guarantee summary and restoration of network element data.

[0186] For the operation and maintenance requirement text "The opening ceremony of the Dth sports meeting will be held at E Stadium at D o'clock in the evening on C day of B month in A year. Please make good guarantees.", the target operation and maintenance scenario obtained is "Domestic major event", and the operation and maintenance element information includes "D o'clock in the evening on C day of B month in A year" and "E Stadium". Then the formed target operation and maintenance content is "The operations corresponding to the opening ceremony of the Dth sports meeting are to perform data backup, identification of network hidden dangers, and deployment of high-traffic capacity within the area of E Stadium before D o'clock in the evening on C day of B month in A year. During the event, perform network monitoring, early warning emergency handling, high-load optimization, and video service guarantee. After the event, notify the user whether it is necessary to restore the data to the state before the event."

[0187] It should be noted that step S1102 here can be generated through a large language model or in the mode of template filling. The embodiments of the present application do not limit this.

[0188] The embodiments of the present application output the target operation and maintenance content, so that the operation and maintenance personnel do not need to query the target operation and maintenance operation information after obtaining the target operation and maintenance scenario, which improves the comprehensiveness of the operation and maintenance information and the convenience for the operation and maintenance personnel to obtain comprehensive operation and maintenance information.

[0189] Please refer to Figure 7 and Figure 8 , Figure 7 and Figure 8 are respectively the step schematic diagrams of the network operation and maintenance scenario recognition method provided by another embodiment of the present application. In one embodiment, after step S1101, the method further includes but is not limited to the following steps.

[0190] Step S1201, determine one or more target operation and maintenance devices according to the target operation and maintenance operation information.

[0191] Step S1202, send operation and maintenance instructions to one or more target operation and maintenance devices according to the target operation and maintenance operation information, so that one or more target operation and maintenance devices perform corresponding operation and maintenance operations.

[0192] Specifically, the target operation and maintenance operation information is used to instruct operation and maintenance personnel to perform operation and maintenance on specific network devices. That is, the corresponding network devices (i.e., target operation and maintenance devices) can be determined according to the target operation and maintenance operation information. For network operation and maintenance, within the scope of automated operation and maintenance, network intelligent operation and maintenance has a faster response speed compared to manual operation and maintenance. Therefore, operation and maintenance instructions can be sent to one or more target operation and maintenance devices according to the target operation and maintenance operation information, directly causing one or more target operation and maintenance devices to perform the corresponding operation and maintenance operations, completing network operation and maintenance, and thus better meeting the timeliness requirements of network operation and maintenance.

[0193] In the embodiment of the present application, by determining one or more target operation and maintenance devices and sending operation and maintenance instructions to one or more target operation and maintenance devices to cause one or more target operation and maintenance devices to perform the corresponding operation and maintenance operations, the intelligence and automation of network operation and maintenance are improved, thereby improving the response speed to network operation and maintenance.

[0194] Please refer to Figure 9 , Figure 9 which is a schematic structural diagram of a network operation and maintenance scenario recognition device provided by an embodiment of the present application, used to execute the above various embodiments. The network operation and maintenance scenario recognition device 900 includes but is not limited to the following modules.

[0195] The service proxy module 910 includes a task management module 911, a text vectorization module 912, a vector database 913, a text template library 914, a text construction module 915, and a scenario-operation correspondence library 916;

[0196] The large language model module 920 includes a large language model storage module 921 and a model management module 922;

[0197] The event trigger module 930 is used to obtain operation and maintenance requirement information input by the target object or event information pushed by the operation and maintenance system;

[0198] The operation execution module 940 is used to send operation and maintenance instructions to one or more of the target operation and maintenance devices, so that one or more of the target operation and maintenance devices perform the corresponding operation and maintenance operations.

[0199] The visualization module 950 is used to output information obtained by one or more recognition models.

[0200] Specifically, the task management module 911 is used to manage operation and maintenance tasks in the device, communication between modules, and coordinate the execution steps of a single task.

[0201] The text vectorization module 912 is used to vectorize the operation and maintenance requirement text to obtain operation and maintenance word vectors.

[0202] The vector database 913 is used to obtain candidate operation and maintenance scenarios according to the operation and maintenance word vectors.

[0203] The text template library 914 is used to provide preset text templates.

[0204] The text construction module 915 is used to construct the first prompt text according to the candidate operation and maintenance scenarios, operation and maintenance requirement texts, and the first prompt text template; or, construct the second prompt text according to the operation and maintenance requirement texts and the second preset text template.

[0205] The scenario-operation correspondence library 916 is used to manage the operation and maintenance operation information corresponding to each operation and maintenance scenario.

[0206] The large language model storage module 921 is used to store one or more recognition models.

[0207] The model management module 922 is used to manage the recognition models in the large language model storage module 921.

[0208] The specific implementation manner of the network operation and maintenance scenario recognition device 900 is basically the same as the specific embodiments of the above network operation and maintenance scenario recognition method, and will not be elaborated here.

[0209] An embodiment of the present application further provides an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above network operation and maintenance scenario recognition method is implemented. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0210] Please refer to Figure 10 , Figure 10 which shows the hardware structure of an electronic device in another embodiment. The electronic device includes:

[0211] A processor 1001, which can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided by the embodiments of the present application;

[0212] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 1002 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1002 and are called by the processor 1001 to execute the network operation and maintenance scenario recognition method of the embodiments of this application;

[0213] The input / output interface 1003 is used to implement information input and output;

[0214] The communication interface 1004 is used to implement communication and interaction between this device and other devices. It can achieve communication through a wired manner (such as USB, network cable, etc.) or through a wireless manner (such as mobile network, WI FI, Bluetooth, etc.);

[0215] The bus 1005 transmits information between the various components of the device (such as the processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004);

[0216] Among them, the processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004 achieve communication connections with each other inside the device through the bus 1005.

[0217] The embodiments of this application also provide a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned network operation and maintenance scenario recognition method.

[0218] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory can optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above-mentioned network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and their combinations.

[0219] The embodiments described in the embodiments of this application are for more clearly explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are equally applicable to similar technical problems.

[0220] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown in the figures, or combine certain steps, or different steps.

[0221] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0222] In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent the situation where A exists alone, A and B exist simultaneously, or B exists alone. Where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one of the following" and its similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a exists alone, b exists alone, c exists alone, a and b exist simultaneously, a and c exist simultaneously, b and c exist simultaneously, or a, b, and c exist simultaneously, where a, b, and c can be single or multiple.

[0223] In the embodiments of the present application, "indicating" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. If the information indicated by a certain piece of information is called the information to be indicated, then in the specific implementation process, there are many ways to indicate the information to be indicated. For example, but not limited to, the information to be indicated can be directly indicated, such as indicating the information to be indicated itself or the index of the information to be indicated, etc. It is also possible to indirectly indicate the information to be indicated by indicating other information, where there is an association relationship between the other information and the information to be indicated. It is also possible to only indicate a part of the information to be indicated, while the other parts of the information to be indicated are known or pre-agreed. For example, it is also possible to use the arrangement order of each piece of information pre-agreed (such as stipulated in a protocol) to indicate specific information, thereby reducing the indication overhead to a certain extent.

[0224] In the embodiments of the present application, the terms and English abbreviations are all exemplary examples given for convenience of description, and should not constitute any limitation on the present application. The present application does not exclude the possibility of defining other terms that can achieve the same or similar functions in existing or future protocols.

[0225] In the embodiments of the present application, the terms "first", "second", and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", and "third" may explicitly or implicitly include one or more of such features.

[0226] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several variations and improvements can still be made, and these all fall within the protection scope of the present application.

Claims

1. A method for identifying network operation and maintenance scenarios, characterized in that The method includes: Obtaining an operation and maintenance requirement text, and determining an operation and maintenance word vector according to the operation and maintenance requirement text; Performing vector retrieval in a preset vector database according to the operation and maintenance word vector to obtain at least one candidate word vector, and obtaining candidate operation and maintenance scenarios according to the candidate word vector; Determining a first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and a preset first prompt text template; Inputting the first prompt text into a pre-trained first recognition model, so that the first recognition model outputs first operation and maintenance requirement recognition information, where the first operation and maintenance requirement recognition information at least includes a target operation and maintenance scenario.

2. The method according to claim 1, characterized in that, After obtaining the operation and maintenance requirement text, the method further includes: Determining a second prompt text according to the operation and maintenance requirement text and a second preset text template; Inputting the second prompt text into a pre-trained second recognition model, so that the second recognition model outputs operation and maintenance element information, where the operation and maintenance element information includes at least one of time information and location information.

3. The method according to claim 1, wherein The first operation and maintenance requirement recognition information further includes operation and maintenance element information, where the operation and maintenance element information includes at least one of time information and location information.

4. The method according to claim 1, wherein The obtaining of the operation and maintenance requirement text includes: Obtaining operation and maintenance requirement information input by a target object; When the operation and maintenance requirement information is text information, obtaining the operation and maintenance requirement text according to the text information; When the operation and maintenance requirement information is non-text information, performing text conversion on the non-text information to obtain the operation and maintenance requirement text.

5. The method according to claim 1, wherein The obtaining of the operation and maintenance requirement text includes: Obtaining operation and maintenance system push event information; Converting the operation and maintenance system push event information into the operation and maintenance requirement text.

6. The method according to claim 1, wherein The first prompt text template includes a conflict decoupling module; The determining of the first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and the preset first prompt text template includes: Obtaining conflict decoupling text according to multiple candidate operation and maintenance scenarios, where the conflict decoupling text decouples between the multiple candidate operation and maintenance scenarios and the operation and maintenance requirement text; Filling the conflict decoupling text into the conflict decoupling module.

7. The method according to claim 1, characterized in that, The first prompt text template includes a scenario example filling module; The determining of the first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and the preset first prompt text template includes: For each candidate operation and maintenance scenario, obtaining scenario example text corresponding to the candidate operation and maintenance scenario; Filling multiple pieces of the scenario example text into the scenario example filling module.

8. The method according to claim 1, wherein The first prompt text template includes a candidate item filling module; The determining of the first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and the preset first prompt text template includes: Filling multiple candidate operation and maintenance scenarios into the candidate item filling module.

9. The method according to claim 1, wherein The first prompt text template includes an input filling module; The determining of the first prompt text according to the operation and maintenance requirement text, the candidate operation and maintenance scenarios, and the preset first prompt text template includes: Filling the operation and maintenance requirement text into the input filling module.

10. The method according to claim 1, characterized in that, The preset vector database includes multiple operation and maintenance scenario identifiers, and each of the one or more candidate word vectors corresponds to one of the multiple operation and maintenance scenario identifiers; The obtaining of the candidate operation and maintenance scenarios according to the candidate word vectors includes: Performing index retrieval in a preset relationship database according to the corresponding operation and maintenance scenario identifier to determine the candidate operation and maintenance scenario corresponding to the operation and maintenance scenario identifier; Wherein, the preset relationship database includes multiple operation and maintenance scenario identifiers and multiple preset operation and maintenance scenarios, the multiple preset operation and maintenance scenarios are in one-to-one correspondence with the multiple operation and maintenance scenario identifiers, and the candidate operation and maintenance scenario is one of the multiple preset operation and maintenance scenarios.

11. The method according to any one of claims 2 or 3, characterized in that, After obtaining the target operation and maintenance scenario and the operation and maintenance element information, the recognition method includes: Obtaining target operation and maintenance operation information corresponding to the target operation and maintenance scenario according to the target operation and maintenance scenario; Outputting target operation and maintenance content according to the operation and maintenance element information and the target operation and maintenance operation information.

12. The method according to claim 11, wherein After obtaining the target operation and maintenance operation information corresponding to the target operation and maintenance scenario according to the target operation and maintenance scenario, the recognition method further includes: Determining one or more target operation and maintenance devices according to the target operation and maintenance operation information; Sending an operation and maintenance instruction to one or more of the target operation and maintenance devices according to the target operation and maintenance operation information, so that one or more of the target operation and maintenance devices perform corresponding operation and maintenance operations.

13. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 12 when executing the computer program.

14. A computer-readable storage medium storing a computer program, characterized in that, The computer program implements the method according to any one of claims 1 to 12 when executed by the processor.

Images