A method and system for controlling the network element access process based on a large natural language model

Abstract

This invention relates to the field of communication resource technology, specifically to a method and system for controlling network element access processes based on a natural language large model. The method includes the following steps: constructing a network element access process; constructing a multi-turn dialogue dataset that satisfies a fine-tuning format, and predicting the next step of the process based on the user's historical dialogues; constructing a physical table to store process control form information; when the user enters a message in the chat input box or interacts with the process step form, calling the fine-tuned large model to predict the unique identifier and parameter information of the form that should be triggered in the next process. The beneficial effects are: the network element access process control method and system based on a natural language large model proposed in this invention integrates the network element access process into the natural interaction with the chatbot. With the help of advanced natural language large model technology, the chatbot can not only deeply understand the historical conversation content between the user and the robot, but also enable the robot to accurately and effectively control the network element access process.

Description

Technical Field

[0001] This invention relates to the field of communication resource technology, specifically to a method and system for controlling the network element access process based on a large natural language model. Background Technology

[0002] The network element access process control method can be briefly summarized as follows: controlling the network element access process according to the principle of "resources first, network second". Specifically, it involves multiple stages such as resource access triggering, network element control policy setting, network access control, and network element access management, to ensure that network elements can access the network safely and efficiently, and to achieve intelligent control and management of network resources.

[0003] In existing technologies, existing chatbot platforms allow users to interact with the robot via text to achieve functions such as knowledge Q&A and text dialogue; however, users need to perform cumbersome manual management, which complicates business processes and brings management difficulties.

[0004] Therefore, we need a network element access process control method and system based on a large natural language model to solve the problems of cumbersome manual management required by existing control methods, which complicates business processes and brings management difficulties; it can simplify the management difficulties brought about by complex business processes and provide users with a more convenient and efficient user experience. Summary of the Invention

[0005] The purpose of this invention is to provide a network element access process control method and system based on a large natural language model, so as to solve the problem that the existing control methods mentioned in the background art require cumbersome manual management, which complicates the business process and brings management difficulty; it can simplify the management difficulty brought about by complex business processes and provide users with a more convenient and efficient user experience.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a network element access process control method based on a large natural language model, the method comprising the following steps:

[0007] Establish network element access process;

[0008] Construct a multi-turn dialogue dataset that satisfies a fine-tuning format, and predict the next step in the process based on the user's historical dialogue.

[0009] Build physical tables to store flow control form information;

[0010] When a user enters a message in the chat input box or interacts with a form in the process flow, the finely tuned large model is invoked to predict the unique identifier and parameter information of the form that should be triggered in the next process flow.

[0011] The access address retrieved from the physical table and the filtered parameters are assembled into a request message and the request is invoked to obtain the process step form data of the current process and present it on the chatbot interface.

[0012] Preferably, the specific operations for constructing the network element access process include:

[0013] On-site resource acquisition, optical path scheduling, compliance verification, information collection and improvement, network access information confirmation, alarm verification, and asset synchronization.

[0014] Preferably, constructing a multi-turn dialogue dataset that satisfies the fine-tuning format specifically includes:

[0015] Each dataset consists of three parts: History, Input, and Output. History identifies the user's historical dialogue with the system, including user input, process forms provided by the robot, and user interaction data with the forms. Input identifies user input, and Output indicates the next process to be processed.

[0016] Using the llama-factory platform, the open-source large model ChatGLM3-6B was fine-tuned with a pre-constructed dataset. The fine-tuned model then predicted the next step in the user's conversation based on their historical dialogue.

[0017] Preferably, the specific operations for constructing a physical table to store flow control form information include:

[0018] The physical table contains a unique identifier (id), a name, an access address (url), and parameter information (params) for the access form. To visualize each step of the process in the chatbot, each form in the process is developed on the front end, and the form information is registered in the physical table.

[0019] Preferably, the unique identifier of the form can be matched with the physical table to obtain the form's name, access address, and parameter information.

[0020] A network element access process control system based on a natural language large model, the system consists of an access process construction module, a process prediction module, a physical table construction module, a parameter prediction module, and a data display module;

[0021] The network access process construction module is used to construct the network element access process;

[0022] The process prediction module is used to construct a multi-turn dialogue dataset that meets the fine-tuning format and predict the next step of the process based on the user's historical dialogue.

[0023] The physical table building module is used to build physical tables to store flow control form information;

[0024] The parameter prediction module is used to predict the unique identifier and parameter information of the form that should be triggered in the next process when the user enters a message in the chat input box or interacts with the form in the process step.

[0025] The data display module assembles the access address retrieved from the physical table and the filtered parameters into a request message and makes a request call to obtain the process step form data of the current process and present it on the chatbot interface.

[0026] Preferably, the network access process construction module includes on-site resource acquisition, optical path scheduling, compliance verification, information collection improvement, network access information confirmation, alarm verification, and asset synchronization.

[0027] Preferably, in the process prediction module, each dataset consists of three parts: History, Input, and Output. The History identifies the user's historical dialogue with the system, including user input, process step forms provided by the robot, and user interaction data with the forms. The Input identifies the user input, and the Output indicates the process that should be processed next.

[0028] Using the llama-factory platform, the open-source large model ChatGLM3-6B was fine-tuned with a pre-constructed dataset. The fine-tuned model then predicted the next step in the user's conversation based on their historical dialogue.

[0029] Preferably, the physical table construction module contains a unique identifier (id), a name, an access address (url), and parameter information (params) of the access form. In order to visualize each step of the process in the chatbot, each form in the process is developed on the front end and the form information is registered in the physical table.

[0030] Preferably, in the parameter prediction module, the unique identifier of the form can be matched with the physical table to obtain the form's name, access address, and parameter information.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] This invention proposes a network element access process control method and system based on a natural language processing model. It integrates the network element access process into natural interaction with a chatbot. Leveraging advanced natural language processing model technology, the chatbot can deeply understand the historical conversations between the user and the chatbot, including natural language communication, process form steps, and user interactions with forms. This allows the chatbot to accurately and effectively control the network element access process. Breaking through the limitations of traditional chatbots that rely solely on pure text interaction, this invention innovatively integrates text dialogue and form information seamlessly into the chatbot, providing users with a completely new interactive experience and making each step of the process more intuitive and understandable. The management of the network element access process is optimized and upgraded through natural language processing. Users no longer need to manually trigger the process flow, which is particularly convenient when managing complex processes. Through intelligent guidance from the natural language processing model, the process flow becomes more natural and smooth, greatly improving the convenience of user operation and management. It further solves the problem of existing control methods requiring cumbersome manual management, complicating business processes and increasing management difficulty. It simplifies the management difficulty brought about by complex business processes, providing users with a more convenient and efficient user experience. Attached Figure Description

[0033] Figure 1 This is a control diagram illustrating the network access process of the network access element in this invention;

[0034] Figure 2 This is a flowchart of the network access process of this invention;

[0035] Figure 3 This is a schematic diagram illustrating an example of a multi-turn dialogue dataset format that satisfies the fine-tuning format of this invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Example 1

[0038] Please see Figures 1 to 3 This invention provides a technical solution: a method for controlling the network element access process based on a large natural language model, the method comprising the following steps:

[0039] This method is based on the open-source large model ChatGLM3-6B, the llama-factory fine-tuning framework, and the chatbot platform of the resource capability center. The following section details how this method utilizes a large natural language model to control the network element entry process. The relevant technical details and principles are as follows. See Figure (1) for a detailed process description.

[0040] (1) First, construct the network element access process, as shown in Figure (2).

[0041] (2) Next, construct a multi-turn dialogue dataset that satisfies the fine-tuning format, as shown in Figure (3). Each dataset consists of three parts: History, Input, and Output. History identifies the user's historical dialogue with the system, including user input, process steps provided by the robot, and user interaction data with the forms. Input identifies user input, and Output indicates the next process to be processed.

[0042] (3) Using the llama-factory platform, the open-source large model ChatGLM3-6B was fine-tuned using the dataset constructed above. The fine-tuned large model can predict the next step based on the user's historical dialogue.

[0043] (4) Next, construct a physical table to store the flow control form information. The physical table needs to contain a unique identifier (id), name (name), access address (url), and access form parameter information (params), etc. In order to visualize each step of the process in the chatbot, each form in the process needs to be developed on the front end and the form information needs to be registered in the physical table.

[0044] (5) When a user enters a message in the chat input box or interacts with a form in the workflow, the fine-tuned large model is invoked to predict the unique identifier and parameter information of the form that should be triggered in the next workflow. The unique identifier of the form can be matched with the physical table to obtain the form's name, access address, and parameter information. Since the form parameters predicted by the large model may be missing, redundant, or extra, it is necessary to filter and select the predicted form parameters to meet the requirements.

[0045] (6) Assemble the access address retrieved from the physical table and the filtered parameters into a request message and make a request call to obtain the process step form data of the current process and present it on the chatbot interface.

[0046] Example 2

[0047] Based on Example 1, a network element access process control system based on a natural language large model is proposed. The system consists of an access process construction module, a process prediction module, a physical table construction module, a parameter prediction module, and a data display module.

[0048] The network access process construction module is used to construct the network element access process, including on-site resource acquisition, optical path scheduling, compliance verification, information collection improvement, network access information confirmation, alarm verification, and asset synchronization.

[0049] The process prediction module is used to construct a multi-turn dialogue dataset that meets the fine-tuning format and predict the next step of the process based on the user's historical dialogue. Each dataset consists of three parts: History, Input, and Output. The History identifies the user's historical dialogue with the system, including user input, process step forms provided by the robot, and user interaction data with the forms. The Input identifies the user input, and the Output indicates the next step of the process to be processed.

[0050] Using the llama-factory platform, the open-source large model ChatGLM3-6B was fine-tuned with a pre-constructed dataset. The fine-tuned model then predicted the next step in the user's conversation based on their historical dialogue.

[0051] The physical table construction module is used to build physical tables to store flow control form information. The physical table contains a unique identifier (id), a name, an access address (url), and access form parameter information (params). In order to visualize each step of the process in the chatbot, each form in the process is developed on the front end and the form information is registered in the physical table.

[0052] The parameter prediction module is used to predict the unique identifier and parameter information of the form that should be triggered in the next process when the user enters a message in the chat input box or interacts with the form in the process. The unique identifier of the form can be matched with the physical table to obtain the form's name, access address and parameter information.

[0053] The data display module assembles the access address retrieved from the physical table and the filtered parameters into a request message and makes a request call to obtain the process step form data of the current process and present it on the chatbot interface.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the network element access process based on a large natural language model, characterized in that: The method includes the following steps: Establish network element access procedures; Construct a multi-turn dialogue dataset that satisfies a fine-tuning format, and predict the next step in the process based on the user's historical dialogue. Build physical tables to store flow control form information; When a user enters a message in the chat input box or interacts with a form in the process flow, the finely tuned large model is invoked to predict the unique identifier and parameter information of the form that should be triggered in the next process flow. The access address retrieved from the physical table and the filtered parameters are assembled into a request message and the request is invoked to obtain the process step form data of the current process and present it on the chatbot interface.

2. The method and system for controlling the network element access process based on a large natural language model according to claim 1, characterized in that: The specific steps for constructing the network element access process include: On-site resource acquisition, optical path scheduling, compliance verification, information collection and improvement, network access information confirmation, alarm verification, and asset synchronization.

3. The method and system for controlling the network element access process based on a large natural language model according to claim 1, characterized in that: Constructing a multi-turn dialogue dataset that satisfies the fine-tuning format specifically includes: Each dataset consists of three parts: History, Input, and Output. History identifies the user's historical dialogue with the system, including user input, process forms provided by the robot, and user interaction data with the forms. Input identifies user input, and Output indicates the next process to be processed. Using the llama-factory platform, the open-source large model ChatGLM3-6B was fine-tuned with a pre-constructed dataset. The fine-tuned model then predicted the next step in the user's conversation based on their historical dialogue.

4. The method and system for controlling the network element access process based on a large natural language model according to claim 1, characterized in that: The specific operations for building a physical table to store flow control form information include: The physical table contains a unique identifier (id), a name, an access address (url), and parameter information (params) for the access form. To visualize each step of the process in the chatbot, each form in the process is developed on the front end, and the form information is registered in the physical table.

5. The method and system for controlling the network element access process based on a large natural language model according to claim 1, characterized in that: The unique identifier of the form can be matched with the physical table to obtain the form's name, access address, and parameter information.

6. A network element access process control system based on a natural language large model, as described in any one of claims 1-5, characterized in that: The system consists of an access process construction module, a process prediction module, a physical table construction module, a parameter prediction module, and a data display module. The network access process construction module is used to construct the network element access process; The process prediction module is used to construct a multi-turn dialogue dataset that meets the fine-tuning format and predict the next step of the process based on the user's historical dialogue. The physical table building module is used to build physical tables to store flow control form information; The parameter prediction module is used to predict the unique identifier and parameter information of the form that should be triggered in the next process when the user enters a message in the chat input box or interacts with the form in the process step. The data display module assembles the access address retrieved from the physical table and the filtered parameters into a request message and makes a request call to obtain the process step form data of the current process and present it on the chatbot interface.

7. The network element access process control system based on a natural language large model according to claim 6, characterized in that: The network access process construction module includes on-site resource acquisition, optical path scheduling, compliance verification, information collection improvement, network access information confirmation, alarm verification, and asset synchronization.

8. The network element access process control system based on a large natural language model according to claim 6, characterized in that: The process prediction module consists of three parts for each dataset: History, Input, and Output. History identifies the user's historical dialogue with the system, including user input, process step forms provided by the robot, and user interaction data with the forms. Input identifies user input, and Output indicates the next process to be processed. Using the llama-factory platform, the open-source large model ChatGLM3-6B was fine-tuned with a pre-constructed dataset. The fine-tuned model then predicted the next step in the user's conversation based on their historical dialogue.

9. A network element access process control system based on a large natural language model according to claim 6, characterized in that: The physical table construction module contains a unique identifier (id), a name, an access address (url), and parameter information (params) of the access form. In order to visualize each step of the process in the chatbot, each form in the process is developed on the front end and the form information is registered in the physical table.

10. A network element access process control system based on a natural language large model according to claim 6, characterized in that: The parameter prediction module can match the unique identifier of the form with the physical table to obtain the form's name, access address, and parameter information.

Images