Network intelligent operation and maintenance method, apparatus and storage medium

Abstract

Provided in the present disclosure are a network intelligent operation and maintenance method, an apparatus and a storage medium. The network intelligent operation and maintenance method comprises: according to a network operation and maintenance request command, acquiring a network operation and maintenance text result; according to the network operation and maintenance request command, generating or adapting an AI capability containing a text-to-speech (TTS) function; and, according to the AI capability containing the TTS function, converting the network operation and maintenance text result into a network operation and maintenance speech result.

Description

Network intelligent operation and maintenance method and device, and storage medium

[0001] This disclosure claims priority to Chinese patent application No. 202410087612.2, filed on January 22, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0002] The present disclosure relates to the field of network operation and maintenance technology, and in particular to a network intelligent operation and maintenance method, a network intelligent operation and maintenance device, and a computer-readable storage medium. Background Art

[0003] Currently used network operation and maintenance methods typically only offer simple text-based query capabilities when querying network operation and maintenance information. The text results directly obtained from network operation and maintenance request commands typically only reflect the status of the queried network devices and lack voice broadcast capabilities. Summary of the Invention

[0004] In a first aspect, the present disclosure provides a network intelligent operation and maintenance method, which is applied to a network management system. The network intelligent operation and maintenance method includes:

[0005] Obtain network operation and maintenance text results according to network operation and maintenance request commands;

[0006] Generate or adapt text-to-speech AI capabilities including TTS functionality based on the network operation and maintenance request command;

[0007] According to the AI ​​capability including the TTS function, the network operation and maintenance text result is converted into a network operation and maintenance voice result.

[0008] In some embodiments, obtaining a network operation and maintenance text result according to a network operation and maintenance request command includes:

[0009] In response to the network management system having a text result corresponding to the network operation and maintenance request command cached therein, the text result is obtained locally from the network management system; or,

[0010] In response to the network management system not caching the text result corresponding to the network operation and maintenance request command, an instruction is generated according to the network operation and maintenance request command, and the instruction is sent to the network device to obtain the text result corresponding to the network operation and maintenance request command from the network device.

[0011] In some embodiments, generating a text-to-speech AI capability including a TTS function according to the network operation and maintenance request command includes:

[0012] Add the initial models that have passed functional testing, performance testing, and security testing into the model library;

[0013] Selecting an initial model from the model library according to a network operation and maintenance request command;

[0014] Use pre-collected data to train and test the selected initial model so that the trained model meets the preset conditions;

[0015] Combine and connect the trained models to form an orchestration pipeline;

[0016] Register the orchestration pipeline as an AI capability including a TTS function corresponding to the network operation and maintenance request command.

[0017] In some embodiments, adapting the AI ​​capability of text-to-speech including TTS function according to the network operation and maintenance request command includes:

[0018] In response to querying the AI ​​capability including the TTS function according to the network operation and maintenance request command, obtain the AI ​​capability including the TTS function whose input data corresponds to the network operation and maintenance text result to be obtained by the network operation and maintenance request command, and whose output data corresponds to the network operation and maintenance voice result to be obtained by the network operation and maintenance request command.

[0019] In some embodiments, selecting an initial model from the model library according to a network operation and maintenance request command includes:

[0020] According to the network operation and maintenance text result corresponding to the network operation and maintenance request command and the network operation and maintenance voice result corresponding to the network operation and maintenance request command, an initial model is selected from the model library, and the network operation and maintenance request command and the selected initial model are stored in correspondence.

[0021] In some embodiments, the trained models are combined and connected to form an orchestration pipeline, including:

[0022] Combining and connecting multiple trained models to form the orchestration pipeline, wherein: multiple models are connected in chronological order and / or in parallel order, the input data of the first-ranked model corresponds to the network operation and maintenance text result of the network operation and maintenance request command, and the input data of the second-ranked model corresponds to the output data of the first-ranked model, and the orchestration is performed in this way until the output data of the last-ranked model corresponds to the network operation and maintenance voice result of the network operation and maintenance request command.

[0023] In some embodiments, registering the orchestration pipeline as an AI capability including a TTS function corresponding to the network operation and maintenance request command includes:

[0024] The performance of different orchestration pipelines used for the same type of network operation and maintenance request commands is obtained, and the orchestration pipelines and input data information, output data information, and performance information of the orchestration pipelines are recorded corresponding to the network operation and maintenance request commands.

[0025] In some embodiments, adapting the AI ​​capability of text-to-speech including TTS function according to the network operation and maintenance request command further includes:

[0026] In response to the network operation and maintenance request command carrying a performance requirement, an AI capability including a TTS function and having corresponding performance information is selected according to the performance requirement.

[0027] In some embodiments, based on the AI ​​capability including the TTS function, converting the network operation and maintenance text result into a network operation and maintenance voice result includes:

[0028] The network operation and maintenance text result of the first data attribute is input into the first model of the orchestration pipeline of the AI ​​capability including the TTS function, and all models of the orchestration pipeline are run in sequence according to the model arrangement order to complete the text digestion and text-to-speech conversion of the network operation and maintenance text result, until the network operation and maintenance voice result of the second data attribute is output from the last model of the orchestration pipeline, wherein the second data attribute is the result of text digestion based on the first data attribute.

[0029] In a second aspect, the present disclosure provides a network intelligent operation and maintenance device. The network intelligent operation and maintenance device includes:

[0030] The acquisition module is used to obtain the network operation and maintenance text results according to the network operation and maintenance request command;

[0031] a generation module for generating a text-to-speech (TTS) capability based on the network operation and maintenance request command; and / or an adaptation module for adapting the text-to-speech (TTS) capability based on the network operation and maintenance request command;

[0032] The AI ​​module is connected to the acquisition module, the generation module and / or the adaptation module, and is used to convert the network operation and maintenance text results into network operation and maintenance voice results based on the AI ​​capability including the TTS function.

[0033] In some embodiments, the acquisition module includes:

[0034] A local acquisition unit, configured to, in response to a network management system having a text result corresponding to the network operation and maintenance request command cached therein, acquire the text result locally from the network management system;

[0035] A network acquisition unit is used to generate an instruction according to the network operation and maintenance request command in response to the network management system not caching the text result corresponding to the network operation and maintenance request command, and send the instruction to the network device to obtain the text result corresponding to the network operation and maintenance request command from the network device.

[0036] In some embodiments, the generating module includes:

[0037] Build a model library unit to incorporate initial models that have passed functional testing, performance testing, and safety testing into the model library;

[0038] A model selection unit, connected to the model library construction unit, is used to select an initial model from the model library according to a network operation and maintenance request command;

[0039] The training model unit is connected to the selection model unit and is used to train and test the selected initial model using pre-collected data so that the trained model meets the preset conditions;

[0040] The model orchestration unit is connected to the training model unit and is used to combine and connect the trained models to form an orchestration pipeline;

[0041] The capability generation unit is connected to the model orchestration unit and is used to register the orchestration pipeline as an AI capability including a TTS function corresponding to the network operation and maintenance request command.

[0042] In some embodiments, the adaptation module is connected to the generation module and is configured to:

[0043] In response to querying the AI ​​capability including the TTS function according to the network operation and maintenance request command, obtain the AI ​​capability including the TTS function whose input data corresponds to the network operation and maintenance text result to be obtained by the network operation and maintenance request command, and whose output data corresponds to the network operation and maintenance voice result to be obtained by the network operation and maintenance request command.

[0044] In some embodiments, the selection model unit is configured to:

[0045] According to the network operation and maintenance text result corresponding to the network operation and maintenance request command and the network operation and maintenance voice result corresponding to the network operation and maintenance request command, an initial model is selected from the model library, and the network operation and maintenance request command and the selected initial model are stored in correspondence.

[0046] In some embodiments, the model arrangement unit is configured to:

[0047] Combining and connecting multiple trained models to form the orchestration pipeline, wherein: multiple models are connected in chronological order and / or in parallel order, the input data of the first-ranked model corresponds to the network operation and maintenance text result of the network operation and maintenance request command, and the input data of the second-ranked model corresponds to the output data of the first-ranked model, and the orchestration is performed in this way until the output data of the last-ranked model corresponds to the network operation and maintenance voice result of the network operation and maintenance request command.

[0048] In some embodiments, the capability generation unit is configured to:

[0049] The performance of different orchestration pipelines used for the same type of network operation and maintenance request commands is obtained, and the orchestration pipelines and input data information, output data information, and performance information of the orchestration pipelines are recorded corresponding to the network operation and maintenance request commands.

[0050] In some embodiments, the adaptation module is further configured to:

[0051] In response to the network operation and maintenance request command carrying a performance requirement, an AI capability including a TTS function and having corresponding performance information is selected according to the performance requirement.

[0052] In some embodiments, the AI ​​module is configured to:

[0053] The network operation and maintenance text result of the first data attribute is input into the first model of the orchestration pipeline of the AI ​​capability including the TTS function, and all models of the orchestration pipeline are run in sequence according to the model arrangement order to complete the text digestion and text-to-speech conversion of the network operation and maintenance text result, until the network operation and maintenance voice result of the second data attribute is output from the last model of the orchestration pipeline, wherein the second data attribute is the result of text digestion based on the first data attribute.

[0054] In a third aspect, the present disclosure provides a computer-readable storage medium having a computer program stored therein, which, when executed by a processor, implements the network intelligent operation and maintenance method described in the first aspect.

[0055] In a fourth aspect, the present disclosure provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the network intelligent operation and maintenance method involved in the above aspects.

[0056] In a fifth aspect, the present disclosure provides a computer program comprising computer instructions, which, when executed by a processor, implement the network intelligent operation and maintenance method as described in the above aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] FIG1 is a flow chart of a network intelligent operation and maintenance method according to an embodiment of the present disclosure.

[0058] FIG2 is an interactive flow chart of a network intelligent operation and maintenance method according to an embodiment of the present disclosure.

[0059] FIG3 is a schematic structural diagram of a network intelligent operation and maintenance device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0060] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the embodiments of the present disclosure will be described in further detail below with reference to the accompanying drawings.

[0061] It should be understood that some embodiments and drawings described herein are only used to explain the present disclosure rather than to limit the present disclosure.

[0062] It can be understood that, in the absence of conflict, the various embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0063] It will be understood that, for the convenience of description, the drawings of the present disclosure only show parts related to the present disclosure, while parts irrelevant to the present disclosure are not shown in the drawings.

[0064] It can be understood that each unit and module involved in the embodiments of the present disclosure may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0065] It will be understood that, without conflict, the functions and operations marked in the flowcharts and block diagrams of the present disclosure may occur in an order different from that marked in the drawings.

[0066] It is understood that the flowcharts and block diagrams of the present disclosure illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present disclosure. Each box in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or by a combination of hardware and computer instructions.

[0067] It can be understood that the units and modules involved in the embodiments of the present disclosure can be implemented by software or hardware, for example, the units and modules can be located in a processor.

[0068] In order to facilitate understanding of the present disclosure, the inventive concept of the present disclosure is first described.

[0069] The digital economy has experienced rapid growth in recent years. Networks are the fundamental infrastructure of this digital economy, and their stable and secure operation is crucial for ensuring its continued success. Technological advancements and service expansion have led to increasingly complex networks. For example, the commercial launch of 5G (fifth-generation mobile communication technology) networks has connected an increasing number of users, significantly improving bandwidth, low latency, and hyperconnectivity. Computing networks are integrating computing power and networking, evolving towards a converged computing and networking architecture. These changes have increased the complexity of network structures, which in turn complicates network operations and maintenance. Complex networks also present new security challenges.

[0070] In this context, current network operation and maintenance methods typically only offer simple text-based query capabilities when querying network operation and maintenance information. The textual results obtained directly from network operation and maintenance request commands typically only reflect the status of the queried network devices, but not the network operation and maintenance results that operators desire. Furthermore, since these network operation and maintenance methods lack voice broadcast capabilities, they are inconvenient for operators and reduce operational efficiency.

[0071] Based on this, network operation and maintenance has at least two new problems: the first is how to accurately obtain the network operation and maintenance results that operators want to know from the large amount of device information parameters obtained from complex networks; the second is in certain specific scenarios, such as when the operation and maintenance operator's hands are operating the equipment, and both hands and eyes need to focus on the operation, how to feedback the network operation and maintenance results to the operator without delaying the operation.

[0072] As an enabling technology, AI (Artificial Intelligence) has significantly impacted network evolution. AI has become a key feature of future networks, effectively improving network operational efficiency and possessing significant economic and social value. Therefore, applying AI to network operations and maintenance can help address both of the aforementioned challenges.

[0073] The present disclosure proposes a network intelligent operation and maintenance method, which converts the network operation and maintenance text results into voice feedback to the network operator through the AI ​​capability of text-to-speech including TTS (Text To Speech) function. The method has the following characteristics: 1) The network operation and maintenance results are fed back using intelligent voice, which frees the operator's hands and eyes, thereby improving the efficiency of operation and maintenance operations; 2) The model corresponding to the AI ​​capability required for the network operation and maintenance request command is arranged to form an AI capability including TTS function (also called intelligent voice capability), and the AI ​​capability including TTS function adapts to the input data and the operator's needs to ensure the accuracy of the voice results; 3) A local AI model library is established, and a trusted AI model is used to ensure the security of the network system. The network intelligent operation and maintenance method and network intelligent operation and maintenance device provided in the present disclosure are introduced below through some embodiments.

[0074] Example 1:

[0075] As shown in Figure 1, the present disclosure provides a network intelligent operation and maintenance method, which is applied to a network management system. The network intelligent operation and maintenance method includes S1-S3.

[0076] In S1, a network operation and maintenance text result is obtained according to a network operation and maintenance request command.

[0077] In S2, according to the network operation and maintenance request command, an AI capability including a TTS function of text to speech is generated or adapted.

[0078] In S3, the network operation and maintenance text result is converted into a network operation and maintenance voice result according to the AI ​​capability including the TTS function.

[0079] In this embodiment, in order to improve the efficiency of network operation and maintenance, an intelligent network operation and maintenance method is provided. The method creates or selects an appropriate AI capability containing a TTS function for a network operation and maintenance request command, and uses the adapted AI capability containing a TTS function to convert the network operation and maintenance text results into network operation and maintenance voice results. The conversion includes text digestion (extraction and analysis of the text results) and text-to-speech. The AI ​​capability containing the TTS function includes an orchestration pipeline formed by a combination of AI models, that is, the AI ​​capability containing the TTS function performs text digestion and text-to-speech on the network operation and maintenance text results through the orchestration pipeline formed by the combination of AI models to finally output the network operation and maintenance voice results, so that the network operation and maintenance text results can be correctly deconstructed to obtain the network operation and maintenance voice results required by the operator, thereby ensuring the accuracy of the voice results. At the same time, the network operation and maintenance text results are converted into voice and broadcast to the operation and maintenance operator, freeing the operator's hands and eyes, and improving the query efficiency and operation efficiency of network operation and maintenance.

[0080] In one embodiment, as shown in FIG2 , the network management system may be an AI-enhanced network operation and management system (e.g., AITOM (AI enhanced Telecom Operation and Management)). The network management system connects operators and network devices. Operators can establish a connection with the network management system through an operation and maintenance terminal. The network devices may be hardware devices or software network functions. A network intelligent operation and maintenance process includes S101-S104.

[0081] In S101, the operator sends a network operation request command (Request) to the network management system. In some embodiments, the operator uses the operation terminal to send the network operation request command to an application of the network management system, which is previously developed based on the AI ​​capability of the network management system including the TTS function.

[0082] In S102, the network management system obtains a network operation and maintenance text result (text data) based on the network operation and maintenance request command. In some embodiments, after receiving the network operation and maintenance request command sent by the operator, the application of the network management system performs corresponding operations based on the network operation and maintenance request command to obtain the network operation and maintenance text result. The network operation and maintenance text result may come from two possible sources: one is that the network management system itself has a local text result (local text) cached, such as caching the network operation and maintenance text result in a local database, so that the local text result can be directly obtained; the other is that the network management system sends a network query instruction (instruction) to the network device to obtain the network query text result (text reply) returned by the network device, and then obtains the network operation and maintenance text result based on the network query text result.

[0083] In S103, the network management system generates or adapts a text-to-speech AI capability (TTS-AI capability) that includes a TTS function based on the network operation and maintenance request command. In some embodiments, if a network operation and maintenance request command appears for the first time, the network management system generates an AI capability that includes a TTS function that converts text data into speech; otherwise, the network management system selects an AI capability that includes a TTS function that converts text data into speech for the request (for example, a network operation and maintenance request command). That is, the network management system has an AI capability that includes a TTS function that is adapted to the network operation and maintenance request command. The AI ​​capability that includes a TTS function can be created based on a model library and historical network operation and maintenance data. After receiving the network operation and maintenance request command, the application of the network management system selects an appropriate AI capability that includes a TTS function. The AI ​​capability including the TTS function includes multiple arranged models (an orchestration pipeline is formed by combining the AI ​​models), the input data attribute of the AI ​​capability including the TTS function is adapted to the first data attribute of the network operation and maintenance text result (the original text information obtained by the query), and the output data attribute is adapted to the second data attribute of the network operation and maintenance voice result (the voice converted after digestion and processing of the original text information), the input data attribute is the attribute of the input data of the first model among the multiple arranged models, and the output data attribute is the attribute of the output data of the last model among the multiple arranged models.

[0084] In S104, the application of the network management system converts the network operation and maintenance text results into network operation and maintenance voice results (Speech) based on the AI ​​capability including the TTS function, and broadcasts the network operation and maintenance voice results to the operator.

[0085] In one embodiment, obtaining a network operation and maintenance text result according to a network operation and maintenance request command includes:

[0086] In response to the network management system having a text result corresponding to the network operation and maintenance request command cached therein, the text result is obtained locally from the network management system; or,

[0087] In response to the network management system not caching the text result corresponding to the network operation and maintenance request command, an instruction is generated according to the network operation and maintenance request command, and the instruction is sent to the network device to obtain the text result corresponding to the network operation and maintenance request command from the network device.

[0088] In this embodiment, in order to improve the efficiency of obtaining network operation and maintenance text results, the network management system is provided with a caching mechanism, which can cache the network device query text results corresponding to some network operation and maintenance query commands locally in the network management system, eliminating the need to query the network device for its corresponding text results every time.

[0089] In the implementation example shown in FIG2 , corresponding to S102, after the network management system receives the network operation and maintenance request command, it may need to translate the network operation and maintenance request command into a network query instruction so that it can be provided to the network device. In some embodiments, if the network management system has a network operation and maintenance text result (such as a network operation management text result) corresponding to the network operation and maintenance request command, the command is not translated. If there is no corresponding network operation management text result, the network management system translates the network operation and maintenance request command into a network query instruction for the network device, and sends the network query instruction to the corresponding network device, so that the network device returns the network operation management text result according to the network query instruction. That is, if the network management system does not translate the network operation and maintenance request command into a query instruction, the network management system directly obtains the corresponding text result locally; otherwise, the network management system sends the transferred query instruction to the network device, receives and organizes the result returned by the network device, and then obtains the text result.

[0090] In one embodiment, generating a text-to-speech AI capability including a TTS function according to the network operation and maintenance request command includes:

[0091] Add the initial models that have passed functional testing, performance testing, and security testing into the model library;

[0092] Selecting an initial model from the model library according to a network operation and maintenance request command;

[0093] Use pre-collected data to train and test the selected initial model so that the trained model meets the preset conditions;

[0094] Combine and connect the trained models to form an orchestration pipeline;

[0095] Register the orchestration pipeline as an AI capability including a TTS function corresponding to the network operation and maintenance request command.

[0096] In the implementation example shown in FIG2 , corresponding to S103 , the network management system may create an AI capability including a TTS function, including: building a model library, selecting a model, training a model, model orchestration, and capability generation, the specific contents of which are described below.

[0097] Build a model library: The network management system establishes a local AI model library. The AI ​​models in this library come from various sources, including independent development, online application markets, and external application markets. Before inclusion in the model library, these AI models must be tested for functionality, performance, and security. By using local, trusted AI models, network security is ensured.

[0098] Model selection: Based on the network operation and maintenance request command, the required AI model is selected from the local AI model library, and a table of correspondence between the network operation and maintenance request command and the selected AI model is recorded. The network operation and maintenance request command here can be a historical network operation and maintenance request command, a network operation and maintenance request command designed by an operation and maintenance expert, or a network operation and maintenance request command initiated by the current operator. However, these types of network operation and maintenance request commands may have undergone certain processing, and the purpose of the processing is to determine the model combination required for a specific type of network operation and maintenance request command.

[0099] Training model: Use the annotated data in the network (network operation and maintenance data for training obtained by collecting and annotating historical network operation and maintenance request commands, including historical network operation and maintenance request commands and corresponding historical network operation and maintenance text results and expected historical network operation and maintenance voice results, etc.) to train the selected AI model, adjust the parameters of the model to achieve the expected results, and then complete the adaptation of the input side and the output side. That is, the AI ​​model can process the text results (historical network operation and maintenance text results) returned by the network management system, and the AI ​​model can return the query results (expected historical network operation and maintenance voice results) required by the network operator, and finally test these models. The model training process may include: obtaining historical network operation and maintenance commands, historical network operation and maintenance text results and expected historical network operation and maintenance voice results, as well as the first historical data attribute of the historical network operation and maintenance text results, and the second historical data attribute of the expected historical network operation and maintenance voice results; using the historical network operation and maintenance text results as input, training the selected multiple initial models until the expected historical network operation and maintenance query voice results are obtained.

[0100] Orchestration models: Trained and tested models are arranged and combined into an orchestration pipeline, enabling them to collaborate and accomplish specific tasks. The models in an orchestration pipeline form a workflow from text data to speech. For example, an orchestration pipeline might include models M1 and M2. M1 is used to digest log data from network devices. The output of M1 can be input to M2, which then converts the text to speech. In other words, the network operation text results are input to the first model in the orchestration pipeline, while the network operation speech results are output from the last model in the pipeline.

[0101] Generate an AI capability that includes a TTS function: After the trained AI models are spliced ​​and combined to form an orchestration pipeline suitable for a specific task, the orchestration pipeline is registered in the network operation and maintenance system as an AI capability that includes a TTS function corresponding to the network operation and maintenance request command, and the information about the AI ​​capability that includes a TTS function is recorded and stored. The information about the AI ​​capability that includes a TTS function includes network operation and maintenance request commands, input data, output data, and performance tags, etc. Therefore, the above-mentioned process of generating an AI capability that includes a TTS function can be, for example: recording the historical network operation and maintenance commands and the multiple orchestration models as corresponding AI capabilities that include a TTS function, and recording the first historical data attribute as the input data attribute of the AI ​​capability that includes a TTS function, and recording the second historical data attribute as the output data attribute of the AI ​​capability that includes a TTS function.

[0102] For each network operation and maintenance request command, the generated AI capability including the TTS function is recorded and stored for use in subsequent processes. In subsequent use, the corresponding AI capability including the TTS function can be obtained by comparing the network operation and maintenance request command with the recorded historical network operation and maintenance request commands.

[0103] In one embodiment, adapting the AI ​​capability of text-to-speech including TTS function according to the network operation and maintenance request command includes:

[0104] In response to querying the registered AI capability including the TTS function according to the network operation and maintenance request command, the AI ​​capability including the TTS function is obtained, the input data of which corresponds to the network operation and maintenance text result to be obtained by the network operation and maintenance request command, and the output data of which corresponds to the network operation and maintenance voice result to be obtained by the network operation and maintenance request command.

[0105] In this embodiment, an appropriate text-to-speech AI capability including a TTS function can be selected for the network operation and maintenance request command based on the network operation and maintenance request command, the input data attribute, and the output data. When adapting the AI ​​capability including the TTS function, the first data attribute of the network operation and maintenance text result to be obtained by the network operation and maintenance request command and the second data attribute of the network operation and maintenance voice result to be obtained are obtained; and the AI ​​capability including the TTS function whose input data attribute is the first data attribute and whose output data attribute is the second data attribute is selected. For example, the above process can be implemented as follows: pre-recording multiple AI capabilities including the TTS function as shown in Table 1, comparing the network operation and maintenance text result information to be obtained by the network operation and maintenance request command (e.g., the attribute of the network operation and maintenance text result is the device log) with the information in the input data column of Table 1 (e.g., the attribute of the input data is also the device log), comparing the network operation and maintenance voice result information to be obtained by the network operation and maintenance request command (e.g., the attribute of the network operation and maintenance voice result is the device status) with the information in the output data column of Table 1 (e.g., the attribute of the output data is also the device status), obtaining the corresponding AI capability including the TTS function, and then obtaining the corresponding model combination according to Table 1.

[0106] Table 1 Examples of AI capabilities including TTS functionality

[0107] An AI capability that includes a TTS function generally includes multiple arranged models. The input data attribute of the AI ​​capability that includes the TTS function adapts to the first data attribute of the network operation and maintenance text result, and the output data attribute adapts to the second data attribute of the network operation and maintenance voice result. The input data attribute is the attribute of the input data of the first model in the arranged multiple models, and the output data attribute is the attribute of the output data of the last model in the arranged multiple models. The second data attribute (e.g., device status) is the result of extraction and analysis (also known as text digestion) based on the first data attribute (e.g., device log), and the network operation and maintenance voice result is the result of text digestion and text-to-speech conversion (e.g., conversion to speech after text digestion) of the network operation and maintenance text result. A model combination method is designed for a specific network operation and maintenance request command (i.e., a specific task), so that multiple models are interconnected to form a model pipeline or a combined model, and the interconnection of models is used to complete all tasks corresponding to the network operation and maintenance request command. By designing a combination of models, in addition to completing specific tasks, processing speed and accuracy can also be improved. For example, in machine learning, there are two methods for combining algorithms: bagging and boosting. These methods can improve accuracy through model combination. As shown in Table 1, for a network operation and maintenance request command 1, the AI ​​capability that can be selected and adapted, including the TTS function, is TTS-AI capability 1. TTS-AI capability 1 sequentially arranges Model 1 and Model 3. The input data attribute of Model 1 is the device log, and the input data attribute of Model 3 is connected to the output data attribute of Model 1. The output data attribute of Model 3 is the device status in voice form. In other words, for request command 1, the network operation and maintenance text result obtained is the device log of the network device. However, the operator needs to query the device status through request command 1. Therefore, the device log obtained by the query is input into Model 1. Model 1 processes the device log and obtains the device status data in text form. Then, it outputs the corresponding data to Model 3. Correspondingly, Model 3 processes the received input data to obtain the device status data in voice form.

[0108] In one embodiment, selecting an initial model from the model library according to a network operation and maintenance request command includes:

[0109] Based on the network operation and maintenance text result corresponding to the network operation and maintenance request command and the network operation and maintenance voice result corresponding to the network operation and maintenance request command, an initial model is selected from the model library, and the network operation and maintenance request command and the selected initial model are stored in correspondence. For example, based on the network operation and maintenance text result corresponding to the network operation and maintenance request command and the network operation and maintenance voice result corresponding to the network operation and maintenance request command, as well as the process of converting the network operation and maintenance text result into the network operation and maintenance voice result, an initial model can be selected from the model library, and the network operation and maintenance request command and the selected initial model can be stored in correspondence.

[0110] In one embodiment, the trained models are combined and connected to form an orchestration pipeline, including:

[0111] Combining and connecting multiple trained models to form the orchestration pipeline, wherein: multiple models are connected in chronological order and / or in parallel order, the input data of the first-ranked model corresponds to the network operation and maintenance text result of the network operation and maintenance request command, and the input data of the second-ranked model corresponds to the output data of the first-ranked model, and the orchestration is performed in this way until the output data of the last-ranked model corresponds to the network operation and maintenance voice result of the network operation and maintenance request command.

[0112] In this embodiment, AI models are selected from the local AI model library to form an AI model combination. Examples of AI model combinations can be shown in Table 2. For different historical network operation and maintenance request commands (request commands 1, 2, etc.), different model combinations are selected according to the specific tasks targeted by the request commands. For example, some request commands only require simple information extraction of text results and then conversion into speech, while others require more complex processing of text results, so the selected model combinations are not the same. In some embodiments, text digestion can be completed by the first several models, and the last model completes text-to-speech. Even if the last model is a text-to-speech model, the final text-to-speech model may be different due to different content to be broadcast (for example, in some scenarios, the text-to-speech model needs to be able to accurately recognize some professional terms). In addition, the text-to-speech model can also be a combination of multiple models.

[0113] Table 2 Examples of AI model combinations

[0114] In one embodiment, registering the orchestration pipeline as an AI capability including a TTS function corresponding to the network operation and maintenance request command includes:

[0115] The performance of different orchestration pipelines used for the same type of network operation and maintenance request commands is obtained, and the orchestration pipelines and input data information, output data information, and performance information of the orchestration pipelines are recorded corresponding to the network operation and maintenance request commands.

[0116] In one embodiment, adapting the AI ​​capability of text-to-speech including TTS function according to the network operation and maintenance request command further includes:

[0117] In response to the network operation and maintenance request command carrying a performance requirement, an AI capability including a TTS function and having corresponding performance information is selected according to the performance requirement.

[0118] In this embodiment, as shown in Table 1, the AI ​​capabilities with TTS functionality may include multiple AI capabilities that can complete request command 1. For example, in addition to TTS-AI capability 1, Table 1 may also record TTS-AI capability 2, and TTS-AI capability 2 can also complete request command 1. Different AI capabilities with TTS functionality (TTS-AI capability 1, TTS-AI capability 2, etc.) have different performance characteristics. For example, some AI capabilities with TTS functionality focus on efficiency and fast response speed; some AI capabilities with TTS functionality emphasize high accuracy; while others are more secure, etc. Therefore, Table 1 needs to record both each AI capability with TTS functionality and its specific performance. When selecting an AI capability with TTS functionality to adapt, in addition to selecting an AI capability with TTS functionality that adapts to input and output data, it is also necessary to select an AI capability with TTS functionality that adapts to performance.

[0119] In one embodiment, the network operation and maintenance text result is converted into a network operation and maintenance voice result based on the AI ​​capability including the TTS function, including:

[0120] The network operation and maintenance text result of the first data attribute is input into the first model of the orchestration pipeline of the AI ​​capability including the TTS function, and all models of the orchestration pipeline are run in sequence according to the model arrangement order to complete the text digestion and text-to-speech conversion of the network operation and maintenance text result, until the network operation and maintenance voice result of the second data attribute is output from the last model of the orchestration pipeline, wherein the second data attribute is the result of text digestion based on the first data attribute.

[0121] In this embodiment, the orchestration pipeline corresponding to the AI ​​capability including the TTS function can be utilized, and according to the model arrangement order of the orchestration pipeline, the text digestion model is first used to convert the network operation and maintenance text result of the first data attribute into the text digestion result of the second data attribute, and then the text-to-speech model is used to convert the text digestion result of the second data attribute into the network operation and maintenance voice result of the second data attribute. That is, the AI ​​capability including the TTS function includes: a text digestion model for digesting the network operation and maintenance text result and a text-to-speech model for voice-converting the text digestion result output by the text digestion model, and finally obtaining the network operation and maintenance voice result. For example, the intelligent TTS capability may only include a text information extraction model and a text-to-speech model, and the text result of the query may be the reply information of one or more network device systems, and these information need to be processed to extract the required text information. The text-to-speech model broadcasts this text information. For example, for the network operation and maintenance request command "Query the IP address of port 0 / 0 / 1", the network device returns the text result "ip address 192.168.3.1 255.255.255.0". The text information extraction model extracts 192.168.3.1 and then combines it into the text for voice broadcast: "The IP address of port 0 / 0 / 1 is 192.168.3.1". The text-to-speech model converts this text into voice broadcast.

[0122] In one embodiment, the text digestion model includes a text information extraction model and / or a text information analysis model arranged in chronological order.

[0123] In one embodiment, the text information extraction model includes one text information extraction model or a plurality of text information extraction sub-models arranged in parallel order;

[0124] In one embodiment, the text information analysis model includes one text information analysis model or a plurality of text information analysis sub-models arranged in parallel order.

[0125] In this embodiment, for complex network operation and maintenance request commands, the text digestion model may be more complex. For example, request command 2 in Table 2 may be "query whether device D (router) is healthy", which is translated into the query instruction "router command: display health". The results returned by device D include: device temperature information, power supply information, fan information, power information, CPU occupancy information, memory occupancy information and storage medium usage information; in the corresponding AI model combination, model 2 can be a text extraction model of various fields (which may include multiple parallel text information extraction sub-models), which extracts values ​​such as temperature, power supply, and fan; model 3 can be an analysis model (text information analysis model), which analyzes the values ​​of various fields and finally obtains the health status: "Device D is good, but the CPU utilization is too high"; model 7 can be a text-to-speech model, which broadcasts the above text "Device D is good, but the CPU utilization is too high".

[0126] In one embodiment, the method further comprises:

[0127] Before obtaining the network device query text result according to the network operation and maintenance query command, receiving the network operation and maintenance query command from the operation and maintenance terminal through the intranet system;

[0128] After converting the network device query text result into a network operation and maintenance query voice result based on the AI ​​capability including the TTS function, the network operation and maintenance query voice result is sent to the operation and maintenance terminal through the intranet system, so that the operation and maintenance terminal broadcasts the network operation and maintenance query voice result.

[0129] In the implementation example shown in FIG2 , in S101 , the network operation and maintenance request command initiated by the network operator may include various operation and maintenance-related queries, such as querying the operating status of a certain device, querying the alarm information of a certain device, etc. The operator may send the network operation and maintenance request command to the network management system through the operation and maintenance terminal. The network management system initiates the network intelligent operation and maintenance method process in response to receiving the network operation and maintenance request command. Similarly, the network management system will also send the voice broadcast information obtained by converting the text results using the AI ​​capability including the TTS function to the operation and maintenance terminal. The operation and maintenance terminal connects to the network management system through the intranet system to ensure the connection security.

[0130] A complete example process is described as follows in conjunction with Figure 2:

[0131] In S101, the network operator initiates a network operation and maintenance request command, and operation and maintenance query is one of the most basic network operation and maintenance request operations, and various network information is obtained through query. The network operation and maintenance request is initiated by the network operator. For example, the network operator can initiate a network operation and maintenance request through an operation and maintenance terminal (for example, a computer or handheld device). After that, the operation and maintenance terminal interacts with the network management system through the network, and can use SNMP (Simple Network Management Protocol), NETCONF, RESTCONF protocols, etc. to perform network operation and maintenance request operations such as operation and maintenance query. The network connection for performing these operations is based on the intranet system of the network management system, so these network connections are logically or physically isolated from the public network.

[0132] In S102, the network management system receives the network operation and maintenance request command and obtains the network operation and maintenance text results. If the network management system has the query results corresponding to the network operation and maintenance request command, it may not translate the command to the network device. If the network management system does not have the query results corresponding to the network operation and maintenance request command, it will translate the network operation and maintenance request command into a query instruction for the network device and send the query instruction to the corresponding network device. For example, the network management system already stores some network management data, which can be directly obtained from the SNMP MIB (Management Information Base). If this data contains the query results corresponding to the network operation and maintenance request command, the network management system does not need to translate the network operation and maintenance request command. If the network management system does not store the corresponding data, it will need to translate the network operation and maintenance request command into a query instruction for the network device and distribute the query instruction to the network device via the CLI (command-line interface), NETCONF (Network Configuration) protocol, RESTCONF (Representational State Transfer Configuration) protocol, or other methods. That is, if the network management system does not translate the network operation and maintenance request command into a query command, the network management system directly obtains the local text result; otherwise, the network management system sends the transferred query instruction to the network device, receives and organizes the results returned by the network device, and then obtains the text result.

[0133] In S103, the network management system generates or adapts an AI capability that includes a TTS function. The network management system can generate a new AI capability that includes a TTS function in response to a new network operation and maintenance request command, or it can select and adapt an existing AI capability that includes a TTS function. The existing AI capability that includes a TTS function is pre-generated based on historical data. The basic operations for generating an AI capability that includes a TTS function include: building a model library, selecting a model, training a model, orchestrating a model, and generating an AI capability that includes a TTS function. For details about these operations, please refer to the following description.

[0134] Build a model library: The network management system establishes a local AI model library. The models in the local model library come from sources including independent development, online application markets, and external application markets. Before being included in the model library, these models must be tested for functionality, performance, and security.

[0135] Select the model: According to the network operation and maintenance request command, select the AI ​​model to be used from the local AI model library, and establish a corresponding relationship table between the network operation and maintenance request command and the AI ​​model. For example, as described in Table 2, the relationship table records the AI ​​model combination used for each request command (such as request command 1 and request command 2). When it is necessary to select an AI model later, if a similar request command is encountered, the model (or model combination) can be directly selected according to the relationship table (for example, Table 2), which is more convenient and quick.

[0136] Training models: Use historical data annotated in the network to train the selected AI model combination, adjusting the model parameters to achieve the desired results. This completes the adaptation of the input and output sides, that is, the AI ​​model can process the text results returned by the network management system and return the query results required by network operators. Finally, these models are tested. Training models requires training all included AI models, using different training data for different models. For example, if Model 1 processes device log data, it needs to use log data as input. After training, this model can filter out the required information from the log data.

[0137] Orchestration models: Arrange and combine trained and tested models so that they collaborate to complete specific tasks. A request command is processed by a combination of multiple models. For example, if request command 1 instructs to query the device's operating status, model 1 processes the network device logs and filters out the device's operating status. The output of model 1 serves as the input for model 2, which filters out keywords and converts them into speech. Furthermore, to increase processing speed, multiple models may be used to execute commands in parallel. For example, if the logs of 10 devices need to be processed simultaneously, 10 models 1 can be used for parallel processing, and the results can be sent to model 2 for sequential broadcast.

[0138] Generating AI capabilities with TTS functionality: Trained AI models are combined to create AI capabilities with TTS functionality suitable for specific tasks. The process of generating AI capabilities with TTS functionality includes adapting the input side to specific network device data and the output side to specific voice broadcast content. During this step, the correspondence between request commands and AI capabilities with TTS functionality is recorded, along with input and output data, model orchestration, and performance. Table 1 shows this information.

[0139] In S104, the network management system uses AI capabilities including TTS functions to convert the network operation and maintenance text results, and broadcasts the converted network operation and maintenance voice results to the network operator.

[0140] Example 2:

[0141] As shown in FIG3 , the present disclosure provides a network intelligent operation and maintenance device, comprising:

[0142] Acquisition module 1, used to obtain network operation and maintenance text results according to the network operation and maintenance request command;

[0143] A generation module 2 for generating a text-to-speech AI capability including a TTS function according to the network operation and maintenance request command; and / or an adaptation module 3 for adapting the text-to-speech AI capability including a TTS function according to the network operation and maintenance request command;

[0144] The AI ​​module 4 is connected to the acquisition module 1 and the generation module 2 and / or the adaptation module 3, and is used to convert the network operation and maintenance text results into network operation and maintenance voice results based on the AI ​​capability including the TTS function.

[0145] In one embodiment, the device is a network management system, which connects the operation and maintenance terminal and the network equipment.

[0146] In one embodiment, the acquisition module 1 includes:

[0147] A local acquisition unit, configured to, in response to a network management system having a text result corresponding to the network operation and maintenance request command cached therein, acquire the text result locally from the network management system;

[0148] A network acquisition unit is used to generate an instruction according to the network operation and maintenance request command in response to the network management system not caching the text result corresponding to the network operation and maintenance request command, and send the instruction to the network device to obtain the text result corresponding to the network operation and maintenance request command from the network device.

[0149] In one embodiment, the generating module 2 includes:

[0150] Build a model library unit to incorporate initial models that have passed functional testing, performance testing, and safety testing into the model library;

[0151] A model selection unit, connected to the model library construction unit, is used to select an initial model from the model library according to a network operation and maintenance request command;

[0152] The training model unit is connected to the selection model unit and is used to train and test the selected initial model using pre-collected data so that the trained model meets the preset conditions;

[0153] The model orchestration unit is connected to the training model unit and is used to combine and connect the trained models to form an orchestration pipeline;

[0154] The capability generation unit is connected to the model orchestration unit and is used to register the orchestration pipeline as an AI capability including a TTS function corresponding to the network operation and maintenance request command.

[0155] In one embodiment, the adaptation module 3 is connected to the generation module 2 and is configured to:

[0156] In response to querying the AI ​​capability including the TTS function according to the network operation and maintenance request command, obtain the AI ​​capability including the TTS function whose input data corresponds to the network operation and maintenance text result to be obtained by the network operation and maintenance request command, and whose output data corresponds to the network operation and maintenance voice result to be obtained by the network operation and maintenance request command.

[0157] In one embodiment, the model selection unit is configured to:

[0158] According to the network operation and maintenance text result corresponding to the network operation and maintenance request command and the network operation and maintenance voice result corresponding to the network operation and maintenance request command, an initial model is selected from the model library, and the network operation and maintenance request command and the selected initial model are stored in correspondence.

[0159] In one embodiment, the model arrangement unit is configured to:

[0160] Combining and connecting multiple trained models to form the orchestration pipeline, wherein: multiple models are connected in chronological order and / or in parallel order, the input data of the first-ranked model corresponds to the network operation and maintenance text result of the network operation and maintenance request command, and the input data of the second-ranked model corresponds to the output data of the first-ranked model, and the orchestration is performed in this way until the output data of the last-ranked model corresponds to the network operation and maintenance voice result of the network operation and maintenance request command.

[0161] In one embodiment, the capability generation unit is configured to:

[0162] The performance of different orchestration pipelines used for the same type of network operation and maintenance request commands is obtained, and the orchestration pipelines and input data information, output data information, and performance information of the orchestration pipelines are recorded corresponding to the network operation and maintenance request commands.

[0163] In one embodiment, the adaptation module 3 is further configured to:

[0164] In response to the network operation and maintenance request command carrying a performance requirement, an AI capability including a TTS function and having corresponding performance information is selected according to the performance requirement.

[0165] In one embodiment, the AI ​​module 4 is configured to:

[0166] The network operation and maintenance text result of the first data attribute is input into the first model of the orchestration pipeline of the AI ​​capability including the TTS function, and all models of the orchestration pipeline are run in sequence according to the model arrangement order to complete the text digestion and text-to-speech conversion of the network operation and maintenance text result, until the network operation and maintenance voice result of the second data attribute is output from the last model of the orchestration pipeline, wherein the second data attribute is the result of text digestion based on the first data attribute.

[0167] Example 3:

[0168] Embodiment 3 of the present disclosure provides a computer-readable storage medium having a computer program stored therein, which, when executed by a processor, implements the network intelligent operation and maintenance method described in embodiment 1 or the network intelligent operation and maintenance device described in embodiment 2.

[0169] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0170] In addition, the present disclosure may further provide a computer device comprising a memory and a processor. The memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the network intelligent operation and maintenance method described in Example 1 or implements the network intelligent operation and maintenance device described in Example 2.

[0171] The memory is connected to the processor. The memory may be a flash memory, a read-only memory or other memory. The processor may be a central processing unit or a single-chip microcomputer.

[0172] Embodiments 1-3 of the present disclosure provide a network intelligent operation and maintenance method, a network intelligent operation and maintenance device, and a computer-readable storage medium, which obtain network operation and maintenance text results according to a network operation and maintenance request command, obtain an adapted AI capability including a TTS function according to the network operation and maintenance request command, and convert the network operation and maintenance text results into the network operation and maintenance voice results required by the network operation and maintenance request command through the AI ​​capability including the TTS function, so as to obtain the network operation and maintenance results that the operator wants to know and perform voice broadcasting, thereby improving the accuracy of the network operation and maintenance results, freeing the operator's hands and eyes, and thus improving the operation and maintenance efficiency.

[0173] It is understood that the above embodiments are merely exemplary embodiments for illustrating the principles of the present disclosure, and the present disclosure is not limited thereto. Those skilled in the art may make various modifications and improvements without departing from the essence of the present disclosure, and such modifications and improvements are also considered to be within the scope of protection of the present disclosure.

Claims

1. A network intelligent operation and maintenance method, applied to a network management system, wherein, The method includes: Obtaining a network operation and maintenance text result according to a network operation and maintenance request command; Generating or adapting an artificial intelligence (AI) capability that includes a text-to-speech (TTS) function from text to speech according to the network operation and maintenance request command; Converting the network operation and maintenance text result into a network operation and maintenance voice result according to the AI capability that includes the TTS function.

2. The method according to claim 1, wherein, Obtaining a network operation and maintenance text result according to a network operation and maintenance request command, including: In response to the network management system caching the text result corresponding to the network operation and maintenance request command, obtaining the text result locally from the network management system; or, In response to the network management system not caching the text result corresponding to the network operation and maintenance request command, generating an instruction according to the network operation and maintenance request command, and sending the instruction to a network device to obtain the text result corresponding to the network operation and maintenance request command from the network device.

3. The method according to claim 1 or 2, wherein Generating an AI capability that includes the TTS function from text to speech according to the network operation and maintenance request command, including: Including an initial model that has passed functional testing, performance testing, and security testing into a model library; Selecting an initial model from the model library according to the network operation and maintenance request command; Training and testing the selected initial model using pre-collected data so that the trained model meets preset conditions; Combining and connecting the trained models to form an orchestration pipeline; Registering the orchestration pipeline as an AI capability that includes the TTS function corresponding to the network operation and maintenance request command.

4. The method according to claim 1 or 2, wherein Adapting an AI capability that includes the TTS function from text to speech according to the network operation and maintenance request command, including: In response to querying an AI capability that includes the TTS function according to the network operation and maintenance request command, obtaining an AI capability that includes the TTS function of the network operation and maintenance text result to be obtained corresponding to the input data of the network operation and maintenance request command and the network operation and maintenance voice result to be obtained corresponding to the output data of the network operation and maintenance request command.

5. The method according to claim 3, wherein Selecting an initial model from the model library according to a network operation and maintenance request command, including: Selecting an initial model from the model library according to the network operation and maintenance text result corresponding to the network operation and maintenance request command to the network operation and maintenance voice result corresponding to the network operation and maintenance request command, and storing the network operation and maintenance request command and the selected initial model in correspondence.

6. The method according to claim 3, wherein, Combining and connecting the trained models to form an orchestration pipeline, including: Combining and connecting multiple trained models to form the orchestration pipeline. In the orchestration pipeline: multiple models are connected in sequence and / or in parallel. The input data of the first model corresponds to the network operation and maintenance text result of the network operation and maintenance request command, and the input data of the second model corresponds to the output data of the first model, and so on for orchestration until the output data of the last model corresponds to the network operation and maintenance voice result of the network operation and maintenance request command.

7. The method according to claim 3, wherein Registering the orchestration pipeline as an AI capability that includes the TTS function corresponding to the network operation and maintenance request command, including: Obtain the performance of the same type of network operation and maintenance request commands using different orchestration pipelines, and record the orchestration pipelines and the input data information, output data information, and performance information of the orchestration pipelines corresponding to the network operation and maintenance request commands.

8. The method according to claim 4, wherein, According to the network operation and maintenance request command, adapt the AI capability including the TTS function of text-to-speech, and further include: In response to the performance requirement carried in the network operation and maintenance request command, select the AI capability including the TTS function with corresponding performance information according to the performance requirement.

9. The method according to claim 1 or 2, wherein, According to the AI capability including the TTS function, convert the network operation and maintenance text result into a network operation and maintenance voice result, including: Input the network operation and maintenance text result of the first data attribute into the model ranked first in the orchestration pipeline of the AI capability including the TTS function, and sequentially run all the models of the orchestration pipeline in the model orchestration order to complete the text digestion and text-to-speech of the network operation and maintenance text result until the network operation and maintenance voice result of the second data attribute in the form of voice is output from the model ranked last in the orchestration pipeline, where the second data attribute is the result of text digestion based on the first data attribute.

10. A network intelligent operation and maintenance device, including: An acquisition module, configured to obtain a network operation and maintenance text result according to a network operation and maintenance request command; A generation module, configured to generate an artificial intelligence (AI) capability including a text-to-speech (TTS) function of text-to-speech according to the network operation and maintenance request command; and / or an adaptation module, configured to adapt the AI capability including the TTS function of text-to-speech according to the network operation and maintenance request command; An AI module, connected to the acquisition module, the generation module, and / or the adaptation module, configured to convert the network operation and maintenance text result into a network operation and maintenance voice result according to the AI capability including the TTS function.

11. The apparatus according to claim 10, wherein, The acquisition module includes: A local acquisition unit, configured to, in response to the network management system caching the text result corresponding to the network operation and maintenance request command, locally acquire the text result from the network management system; A network acquisition unit, configured to, in response to the network management system not caching the text result corresponding to the network operation and maintenance request command, generate an instruction according to the network operation and maintenance request command, and send the instruction to a network device to acquire the text result corresponding to the network operation and maintenance request command from the network device.

12. The device according to claim 10 or 11, wherein, The generation module includes: A model library construction unit, configured to incorporate the initial models that have passed functional testing, performance testing, and security testing into the model library; A model selection unit, connected to the model library construction unit, configured to select an initial model from the model library according to a network operation and maintenance request command; A model training unit, connected to the model selection unit, configured to train and test the selected initial model using pre-collected data so that the trained model meets preset conditions; A model orchestration unit, connected to the model training unit, configured to combine and connect the trained models to form an orchestration pipeline; An ability generation unit, connected to the model orchestration unit, is configured to register the orchestration pipeline as an AI ability including TTS function corresponding to the network operation and maintenance request command.

13. The device according to claim 10 or 11, wherein The adaptation module is connected to the generation module and is configured to: In response to querying an AI ability including TTS function according to the network operation and maintenance request command, obtain the AI ability including TTS function of the network operation and maintenance text result that the input data is to obtain corresponding to the network operation and maintenance request command, and the network operation and maintenance voice result that the output data is required to obtain corresponding to the network operation and maintenance request command.

14. The device according to claim 12, wherein, The model selection unit is configured to: Select an initial model from the model library according to the network operation and maintenance text result corresponding to the network operation and maintenance request command to the network operation and maintenance voice result corresponding to the network operation and maintenance request command, and store the network operation and maintenance request command and the selected initial model correspondingly.

15. The device according to claim 12, wherein, The model orchestration unit is configured to: Combine and connect multiple trained models to form the orchestration pipeline. In the orchestration pipeline: multiple models are connected in sequence and / or in parallel. The input data of the first model corresponds to the network operation and maintenance text result of the network operation and maintenance request command, and the input data of the second model corresponds to the output data of the first model. Orchestration is performed in this way until the output data of the last model corresponds to the network operation and maintenance voice result of the network operation and maintenance request command.

16. The apparatus according to claim 12, wherein, The ability generation unit is configured to: Obtain the performance performance of different orchestration pipelines for the same type of network operation and maintenance request command, and record the orchestration pipeline, as well as the input data information, output data information, and performance performance information of the orchestration pipeline corresponding to the network operation and maintenance request command.

17. The apparatus according to claim 13, wherein, The adaptation module is further configured to: In response to the network operation and maintenance request command carrying a performance requirement, select an AI ability including TTS function with corresponding performance performance information according to the performance requirement.

18. The device according to claim 10 or 11, wherein, The AI module is configured to: Input the network operation and maintenance text result with the first data attribute into the first model of the orchestration pipeline of the AI ability including TTS function, and sequentially run all the models of the orchestration pipeline in the model orchestration order to complete the text digestion and text-to-speech of the network operation and maintenance text result until the network operation and maintenance voice result of the voice with the second data attribute is output from the last model of the orchestration pipeline, where the second data attribute is the result of text digestion based on the first data attribute.

19. A computer-readable storage medium, wherein, [[ID= ​ ​

Images