Broadband complaint processing method and device based on large model and related equipment
By automating the processing of broadband complaint work orders through fault classification and maintenance models based on large language models, the problems of low processing efficiency and low accuracy in existing technologies are solved, efficient and accurate broadband fault processing is achieved, and user satisfaction is improved.
Patent Information
- Application Number
- CN202510816027.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-18
- Publication Date
- 2025-09-23
AI Technical Summary
Existing broadband complaint handling methods rely on manual classification and repair, resulting in low processing efficiency and accuracy, poor user experience, and easily triggering secondary complaints from users, further exacerbating user dissatisfaction.
A fault classification and maintenance model based on a large language model is used to automatically process broadband complaint work orders. The fault type is determined through the fault classification model, and a maintenance plan is generated based on the fault type and status information.
It has achieved automated processing of broadband complaints, reduced manual intervention, improved processing efficiency and accuracy, reduced operating costs, and increased user satisfaction.
Smart Images
Figure CN120689028A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of network broadband technology, and in particular to a broadband complaint processing method, apparatus and related equipment based on a large model. Background Art
[0002] In modern production and life, broadband is widely used. When users find problems such as low broadband speed, network disconnection or network delay, they will complain.
[0003] In the existing online processing process for broadband complaints, the staff first receives the user's description of the broadband failure through telephone or on-site recording, and the customer service staff records the information of the broadband failure to form a broadband complaint work order. The subsequent staff classifies the broadband failure according to the broadband complaint work order, and finally assigns maintenance personnel who can handle this type of broadband failure to repair the broadband failure.
[0004] In actual applications, first, the classification of broadband faults is limited by subjective factors of the staff, which may result in the assigned maintenance personnel being unsuitable to repair that type of broadband fault; second, the repair of broadband faults is limited by the professional knowledge and experience of the maintenance personnel, which may result in the provided solutions lacking pertinence and effectiveness. This not only fails to fundamentally solve the problem, but may also trigger secondary complaints from users, further exacerbating user dissatisfaction. Summary of the Invention
[0005] In view of this, the purpose of this application is to provide a broadband complaint processing method, apparatus and related equipment based on a large model to solve the technical problems of low processing efficiency and low processing accuracy in existing broadband complaint processing methods.
[0006] In a first aspect, the present application provides a method for handling broadband complaints based on a large model, the method comprising: Determining first information based on a current broadband complaint work order received from a broadband user; The current broadband complaint worksheet is used to record the current broadband fault of the faulty broadband complained by the broadband user; the first information indicates the current broadband status of the faulty broadband; Determining second information using a fault classification model based on the first information; Wherein, the second information indicates the current width fault type of the current broadband fault; Determining a current broadband fault repair plan using a fault repair model based on the first information and the second information; The fault classification model and the fault repair model are obtained by training a large language model respectively.
[0007] In a second aspect, the present application provides a broadband complaint handling device based on a large model, the device comprising: a first information determination module, a second information determination module, and a maintenance plan determination module; The first information determination module is configured to determine the first information based on the current broadband complaint work order received from the broadband user; The current broadband complaint worksheet is used to record the current broadband fault of the faulty broadband complained by the broadband user; the first information indicates the current broadband status of the faulty broadband; The second information determining module is configured to determine second information using a fault classification model based on the first information; Wherein, the second information indicates the current width fault type of the current broadband fault; The maintenance solution determination module is configured to determine a current broadband fault maintenance solution using a fault maintenance model based on the first information and the second information; The fault classification model and the fault repair model are obtained by training a large language model respectively.
[0008] In a third aspect, the present application provides an electronic device, comprising a processor and a memory, wherein the memory is used to store an application program, and the processor runs or executes a software program stored in the memory so that the electronic device implements the above-mentioned large model-based broadband complaint handling method.
[0009] In a fourth aspect, the present application provides a computer-readable storage medium, which is used to store program codes executed by a processor, and the program codes are used to implement the above-mentioned large model-based broadband complaint handling method.
[0010] In a fifth aspect, the present application provides a computer program product, which includes computer instructions. When the computer instructions are run on an electronic device, the electronic device implements the above-mentioned large model-based broadband complaint handling method.
[0011] Beneficial effects: The present application provides a method for intelligently processing broadband complaints, which can automatically determine first information based on the current broadband complaint work order received from the broadband user; can also automatically determine second information based on the first information through a fault classification model; can also automatically determine the current broadband fault repair plan based on the first information and the second information through a fault repair model; because the above-mentioned large model-based broadband complaint processing method realizes the automated processing of broadband complaints, it reduces the manual intervention link, thereby reducing operating costs and significantly improving the processing efficiency and accuracy of broadband complaint processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. The following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.
[0013] Figure 1 A flowchart of a method for handling broadband complaints based on a large model provided in an embodiment of the present application; Figure 2 This is a code example diagram for data loading and preprocessing during the training process of the fault classification model provided in an embodiment of the present application; Figure 3 A code example diagram for constructing a data encoder during the training process of the fault classification model provided in an embodiment of the present application; Figure 4 A code example diagram of the training process of the fault classification model provided in an embodiment of the present application; Figure 5 This is a code example diagram of an evaluation model during the training process of a fault classification model provided in an embodiment of the present application; Figure 6 A schematic structural diagram of a large-model-based broadband complaint handling device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0014] In the existing broadband service landscape, user complaints serve as a direct channel for providing feedback on service quality and service issues. Efficient and accurate handling of complaints is crucial for maintaining user satisfaction and enhancing brand loyalty. However, the traditional broadband complaint handling process is overly reliant on manual labor. For example, the classification of broadband faults and the determination of repair plans for broadband faults are both manually performed, as shown below: First, identifying and classifying broadband faults by staff is not only time-consuming and labor-intensive, but also susceptible to subjective factors, making classification efficiency and accuracy difficult to guarantee. Faced with a growing volume of user complaints, staff are often overwhelmed and unable to respond quickly to each user's needs, resulting in long wait times and a significantly reduced user experience.
[0015] Secondly, having service personnel repair broadband issues is not only time-consuming and laborious, but also limited by their expertise and experience. They may have difficulty accurately identifying the root cause of the problem, and the solutions they provide may lack specificity and effectiveness. This not only fails to fundamentally resolve the problem, but can also trigger secondary complaints from users, further exacerbating their dissatisfaction.
[0016] To solve the above technical problems, the embodiments of this application provide a broadband complaint handling solution based on a large model. To make the purpose, technical solution, and advantages of the embodiments of this application more clear, the technical solution of this application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments provided are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.
[0017] First, this application provides a broadband complaint processing method based on a large model, which is applied to broadband complaint processing equipment, such as Figure 1 As shown, Figure 1 A flowchart of a broadband complaint handling method based on a large model provided in an embodiment of the present application is provided. The method includes: S110 to S130, details of which are as follows: S110: Determine first information based on the received broadband user's current broadband complaint work order; Among them, the current broadband complaint work order is used to record the current broadband fault of the faulty broadband complained by the broadband user; the first information indicates the current broadband status of the faulty broadband.
[0018] Specifically, in an embodiment of the present application, the broadband complaint processing device is connected to a device or system such as a call center customer service system of a broadband operator that can receive broadband complaint work orders from broadband users; wherein, the broadband complaint work order should be a paper document or electronic record that records the content of the broadband complaint. When the broadband complaint work order is a paper document obtained through on-site complaint recording, the paper document can be converted into an electronic record by manual entry or scanning recognition.
[0019] In actual applications, there are many sources of broadband complaint work orders. The broadband operator's call center customer service system can answer the broadband user's broadband complaint call and record the content of the complaint call to form a broadband complaint work order. The application with broadband complaint service can also receive the broadband complaint content filled out by the broadband user on the application and form a broadband complaint work order based on the broadband complaint content. The broadband operator's offline broadband business premises can also meet with broadband users and record the broadband complaint content narrated by the broadband user to form a broadband complaint work order.
[0020] In actual operation, the call center customer service system is the main source of broadband complaint work orders. When the call center customer system receives a broadband user's broadband complaint call and generates a corresponding broadband complaint work order, the call center customer service system (or its staff) can send the broadband complaint work order to the broadband complaint processing equipment.
[0021] In an embodiment of the present application, the first information includes: the current broadband status information of the broadband user's faulty broadband, which indicates the current broadband status of the broadband user's faulty broadband; the current broadband status information is the status information of the broadband user's broadband queried in the call center customer service system at the time of receiving the broadband complaint or at the time of querying the broadband status information.
[0022] In one implementation, the first information includes: user identification, broadband installation address, current broadband fault information, and current broadband status information; S110 includes: steps (1) to (2), the details of which are as follows: Step (1): According to the current broadband complaint work order, determine the user ID of the broadband user, the broadband installation address of the faulty broadband, and the current broadband fault information of the faulty broadband provided by the broadband user.
[0023] Specifically, in actual operation, when the broadband complaint processing device receives the current broadband complaint work order, it can perform NPL (Natural Language Processing) analysis on the broadband complaint work order to extract the user ID, broadband installation address and current broadband fault information from the broadband complaint work order; wherein, the user ID is a unique ID that the user indicates the broadband user, which is equivalent to the "identity card" of the broadband user in the call center customer service system and the broadband complaint processing device; the data form of the user ID can be determined according to actual needs, and this application does not make specific restrictions on this; the broadband installation address is usually the address where the broadband is installed. If the broadband installation address of the broadband user has changed, the broadband installation address should be the changed broadband installation address; the broadband fault information is the information of the faulty broadband provided by the broadband user, such as "slow network speed", "internet speed is relatively slow", etc.
[0024] Step (2): According to the user ID, determine the current broadband status information of the faulty broadband complained by the broadband user.
[0025] Specifically, in this embodiment of the present application, the current broadband status information refers to the current status information of the faulty broadband, retrieved through the call center customer service system, such as the optical line terminal (OLT) received optical power, line attenuation, packet loss rate, latency, and bandwidth utilization. In practice, broadband users' broadband status information can be retrieved at any time based on their needs. However, the "current broadband status information" referred to in this embodiment of the present application specifically refers to the broadband status information retrieved for the purpose of handling broadband complaints from broadband users.
[0026] S120: Determine second information using a fault classification model based on the first information.
[0027] The second information indicates the current width fault type of the current broadband fault.
[0028] Specifically, in an embodiment of the present application, a fault classification model is used to classify the current broadband fault; wherein, the first information, namely the user identification, broadband installation address, current broadband fault information and current broadband status information, is the input data of the fault classification model, and the current broadband fault type corresponding to the current broadband fault is the output data of the fault classification model.
[0029] In actual operation, after obtaining the first information, the first information can be input into the fault classification model to determine the current broadband fault type of the current broadband fault recorded in the current broadband complaint work order, such as "connection problem", "speed problem", etc. The current broadband fault type can provide a reference for formulating the current broadband fault maintenance plan.
[0030] In actual applications, the current broadband fault type can directly indicate the possible cause of the broadband fault. For example, if the current broadband fault type is "connection problem", it indicates that the broadband fault complained by the broadband user may be a problem with the router connection or other connection issues; user information can directly indicate the broadband installation address. For example, if it is believed that the broadband user's current broadband fault requires offline repair by maintenance personnel, the broadband installation address in the user information needs to be indicated in the current broadband fault repair plan; the current broadband status information can directly indicate the operating status of the broadband.
[0031] In an embodiment of the present application, the fault classification model is a trained large language model (LLM). In actual operation, since the fault classification model is obtained by training with a large number of training samples, its accuracy in predicting the current broadband fault type only depends on the prediction ability learned during the training process. Since this prediction ability is stable, the fault classification model's prediction process for the current broadband fault type does not need to rely on manual labor and will not be affected by subjective factors of the staff, thereby effectively improving classification efficiency and accuracy.
[0032] In one implementation, the first information further includes: the current broadband complaint text; the second information includes: the current broadband fault type, the current broadband fault urgency, and the current user emotion type; before S120, the method further includes: step (3), the details of which are as follows: Step (3): determining a current broadband complaint text corresponding to the current broadband complaint voice according to the received current broadband complaint voice of the broadband user; The current broadband complaint voice is voice data when a broadband user makes a broadband complaint.
[0033] Specifically, in the embodiment of the present application, the current broadband complaint voice is usually obtained by the call center customer service system in conversation with the broadband user. In actual operation, the current broadband complaint voice can clearly show the emotions of the broadband user, such as "anxious" or "furious".
[0034] In actual operation, the current broadband complaint voice can be converted into the current broadband complaint text through voice recognition method.
[0035] In actual operation, after the current broadband complaint text is determined, the user ID, broadband installation address, current broadband fault information and current broadband status information, as well as the current broadband complaint text, are input as the first information into the fault classification model, so as to enable the fault classification model to output the previous broadband fault type, the current broadband fault urgency and the current user emotion type; wherein, the broadband fault type, the current broadband fault urgency and the current user emotion type user-assisted complaint maintenance model determines the current broadband fault maintenance plan.
[0036] In actual applications, if the urgency of the broadband fault is different, the corresponding broadband fault repair plan will also be different. For example, if the current broadband fault urgency is severe, the current broadband fault repair plan should indicate that experienced maintenance personnel who are close to the faulty broadband installation address need to go to the broadband installation address as soon as possible to perform offline repairs on the faulty broadband; the broadband fault repair plan will also be different depending on the user's emotional type. For example, if the broadband user's emotion is "furious", the broadband fault repair plan of the broadband user can be appropriately accelerated, so that efficient handling of user complaints can be achieved based on user perception.
[0037] In actual application, before executing S120, the method further includes: steps (4) to (6), the details of which are as follows: Step (4): Obtain a training sample set including a plurality of first training samples.
[0038] The first training sample includes a broadband complaint text indicating the faulty broadband.
[0039] Specifically, in the embodiment of the present application, before determining the current broadband fault type of the current broadband fault through the fault classification model, that is, before determining the second information through the fault classification model, it is necessary to first train the fault classification model.
[0040] In the process of training the fault classification model, it is necessary to first construct a training sample set including multiple first training samples; in the embodiment of the present application, the call center customer service system is used to export the historical broadband complaint business processing records of the past two years and the broadband complaint voices corresponding to each record, and about 50,000 data can be obtained. After obtaining the historical broadband complaint business processing records and broadband complaint voices, it is necessary to clear irrelevant content such as consultation data and reply information in the above data to obtain only the historical broadband complaint business processing records and broadband complaint voices related to broadband complaints; among them, the consultation data is a conversation initiated by the broadband user with a broadband complaint, but the content of the conversation is to consult the broadband business, so it needs to be removed; the reply information is the information responded to the broadband user's inquiry other than the broadband complaint, such as "Is there a discount for broadband renewal now?" Since the relevant information of "broadband renewal" is irrelevant to broadband complaints, it also needs to be removed.
[0041] After obtaining historical broadband complaint business processing records and broadband complaint voices only about broadband complaints, each historical broadband complaint business processing record is converted into a unified format and a Chinese word segmentation tool (such as Jieba) is used for word segmentation and stop words are removed to retain key information. The broadband complaint voice is also converted into broadband complaint text, and finally multiple first training samples for training the fault classification model are obtained.
[0042] Step (5): Based on the label, label each of the first training samples in the training sample set.
[0043] The labels include: a label for indicating the type of broadband fault, a label for indicating the urgency of the fault, and a label for indicating the type of user emotion.
[0044] Specifically, in the embodiment of the present application, after a training sample set including a plurality of first training samples is obtained, each first training sample needs to be labeled based on a preset label.
[0045] In an embodiment of the present application, the preset labels include at least three, namely, a label for indicating the type of broadband fault, a label for indicating the urgency of the fault, and a label for indicating the type of user emotion; wherein, the content of the label for indicating the type of broadband fault includes at least connection problems, speed problems, and billing problems, etc., the content of the label for indicating the urgency of the fault includes at least high, medium, and low, etc., and the content of the label for indicating the type of user emotion includes at least positive, neutral, and negative, etc.; in actual operation, the above three labels are marked in the form of sequence data, and the order of the three labels in the sequence is fixed, that is, the order of the three labels in the sequence of labels corresponding to each first training sample is the same, which is used to accurately extract the content of each label from the sequence of labels based on the established order of each label in the sequence of labels of the first training sample during the training process of the fault classification model.
[0046] For example, the sequence of labels of the first training sample A is [speed problem; high; neutral]; "speed problem" is the label used to label the broadband fault type corresponding to the first training sample A, "high" is the label used to label the fault urgency corresponding to the first training sample A, and "neutral" is the label used to label the user emotion type corresponding to the first training sample A.
[0047] In actual operation, in the process of determining the degree of broadband failure of the first training sample, the greater the urgency of the broadband failure, the more quickly the broadband failure needs to be repaired; in practical applications, the scope affected by the broadband failure can be regarded as a measure of urgency. For example, if the broadband user who makes a broadband complaint is an individual, the broadband failure complained by the broadband user can be considered to be of mild urgency; if the broadband user who makes a broadband complaint is an enterprise or institution, the broadband failure complained by the broadband user can be considered to be of moderate urgency or severe urgency; in actual operation, the scope affected by the broadband failure can be determined from the broadband complaint work order, and the broadband user will explain the scope affected by the broadband failure when making a broadband complaint.
[0048] After labeling each first training sample, multiple first training sample sets and corresponding labels are divided into training sample sets and validation sample sets in a ratio of 70% and 30% through train_test_split for subsequent model training and evaluation. In actual operation, the first training samples can be labeled and the sample sets can be divided through scripts, such as Figure 2 As shown, Figure 2 This is a code example diagram for data loading and preprocessing during the training process of the fault classification model provided in an embodiment of the present application.
[0049] In an embodiment of the present application, after labeling each first training sample, the training sample set needs to be encapsulated; in actual operation, the Dataset class (or custom subclass) of the PyTorch model is used to encapsulate the training sample set so that the training sample set can be easily accessed by the fault classification model; the Bert Tokenizer is used to segment and encode the text in the training sample set, including adding special tags, truncation and padding, and the encoded data (usually the input ID, attention mask, etc. processed by the BERT tokenizer) is converted into a Tensor Dataset so that it is compatible with the DataLoader of the PyTorch model.
[0050] In an embodiment of the present application, in addition to obtaining a training sample set including multiple first training samples and labeling each first training sample in the training sample set, it is also necessary to build a network model of the fault classification model. In an embodiment of the present application, the Chinese model (bert-base-chinese) of the BERT (Bidirectional Encoder Representations from Transformers) model is used as the network structure of the fault classification model, and then a custom PyTorch model including the BERT model and an additional classification layer is defined. In actual operation, it is also necessary to use DataLoader to create data loaders for the training sample set and the verification sample set to implement functions such as batch loading of data, shuffling of the order, and multi-process acceleration to improve training efficiency, such as Figure 3 As shown, Figure 3 This is a code example diagram for constructing a data encoder during the training process of the fault classification model provided in an embodiment of the present application.
[0051] Step (6): iteratively training the fault classification model using the training sample set and the labels corresponding to each of the first training samples in the training sample set.
[0052] Specifically, in the embodiment of the present application, during the iterative training of the fault classification model, a training cycle is first defined, including the calculation of the loss function value and the optimizer of the model parameters; wherein the AdamW optimizer is used as the optimizer of the model parameters; in the iterative training of the fault classification model, the training sample set is input in batches and the loss function value is calculated and the model parameters of the fault classification model are updated in reverse, such as Figure 4 As shown, Figure 4 This is a code example diagram of the training process of the fault classification model provided in an embodiment of the present application.
[0053] In the embodiments of this application, Focal Loss is used as the loss function for training the fault classification model. The training sample set of the fault classification model has the problem of class imbalance. Focal Loss can give greater weight to samples that are difficult to classify, making the fault classification model pay more attention to these samples, thereby improving overall performance.
[0054] In the embodiment of the present application, the model performance of the fault classification model is evaluated using a validation sample set, including the accuracy of the successful calculation of the complaint data analysis and comparison and other indicators, such as Figure 5 As shown, Figure 5 This is a code example diagram of an evaluation model during the training process of a fault classification model provided in an embodiment of the present application.
[0055] S130: Determine a current broadband fault repair plan using a fault repair model according to the first information and the second information.
[0056] The fault classification model and the fault repair model are obtained by training a large language model respectively.
[0057] Specifically, in an embodiment of the present application, the fault repair model is used to provide a repair plan for the current broadband fault; the first information and the second information are input data of the fault repair model, and the current broadband fault repair plan corresponding to the current broadband fault is output data of the fault classification model.
[0058] In practice, broadband fault repairs include both online and offline solutions. The online solution uses automated scripts based on fault diagnosis results to automatically perform repair operations, such as restarting the router and adjusting bandwidth configurations. If the online solution fails, the broadband complaint handling device notifies the user and implements an offline solution. The offline solution involves maintenance personnel visiting the broadband installation site for repairs. Using a fault classification model, broadband faults reported by users are classified and intelligently pre-processed for complaint tickets, reducing labor and intervention and improving ticket processing efficiency.
[0059] In the embodiment of the present application, the fault classification model and the fault repair model are obtained by training large language models (LLM) respectively.
[0060] In one implementation, before S130, the method further includes: steps (7) to (9), the details of which are as follows: Step (7): According to the user identifier, retrieve the first historical broadband complaint processing data corresponding to the user identifier from the database.
[0061] Specifically, in the embodiment of the present application, after obtaining the user ID, an attempt may be made to retrieve all historical broadband complaint processing data of the broadband user corresponding to the user ID from the database, namely, the first historical broadband complaint processing data. The first historical broadband complaint processing data can reflect information such as the broadband user's broadband usage habits and broadband failure tendencies. Therefore, if the current broadband failure repair plan is determined based on the first historical broadband complaint processing data, the accuracy of the determined current broadband failure repair plan can be effectively improved. Step (8): Based on the broadband installation address, retrieve the second historical broadband complaint processing data corresponding to the broadband installation address from the database.
[0062] Specifically, in an embodiment of the present application, after the broadband installation address is obtained, all historical broadband complaint processing data corresponding to the broadband installation address can be retrieved from the database, that is, the second historical broadband complaint processing data; the second historical broadband complaint processing data can reflect information such as the broadband failure tendency of the broadband installation address. Therefore, if the current broadband fault repair plan is determined based on the second historical broadband complaint processing data, the accuracy of the determined current broadband fault repair plan can be effectively improved.
[0063] It should be noted that since broadband users may change their broadband installation addresses due to reasons such as moving, one broadband user, i.e., user ID, may correspond to multiple broadband installation addresses. Correspondingly, if a broadband installation address has been rented to multiple broadband users in a time sequence, one broadband installation address may correspond to multiple broadband users. Therefore, when querying historical broadband complaint processing data, it is necessary to query separately according to the user ID and broadband installation address to obtain as much information as possible to assist the fault repair model in determining the current broadband fault repair plan.
[0064] Step (9): According to the broadband installation address, retrieve the third historical broadband complaint processing data of the broadband coverage area where the broadband installation address is located from the database.
[0065] Specifically, in an embodiment of the present application, after the broadband installation address is obtained, all historical broadband complaint processing data of the broadband coverage area where the broadband installation address is located can be retrieved from the database, that is, the third historical broadband complaint processing data. The broadband coverage area can be a certain community in the city, which can be determined according to actual needs. This application does not make specific limitations on this; the third historical broadband complaint processing data can reflect information such as the broadband failure tendency of the broadband coverage area. Therefore, if the current broadband fault repair plan is determined based on the third historical broadband complaint processing data, the accuracy of the determined current broadband fault repair plan can be effectively improved.
[0066] It should be noted that in the process of obtaining the first historical broadband complaint processing data, if all historical broadband complaint processing data of the broadband user corresponding to the user ID cannot be retrieved from the database, it means that this is the first time that the broadband user has made a broadband complaint, so the historical broadband complaint processing data corresponding to the user ID of the broadband user does not exist in the database.
[0067] In an embodiment of the present application, for the situation where "all historical broadband complaint processing data of the broadband user corresponding to the user identifier cannot be retrieved from the database", the embodiment of the present application determines the first historical broadband complaint processing data of the current broadband complaint of the broadband user as "none".
[0068] In actual operation, the first historical broadband complaint processing data, the second historical broadband complaint processing data, and the third historical broadband complaint processing data are input into the fault repair model in the form of sequence or matrix data.
[0069] For example, for broadband user A, if all historical broadband complaint processing data corresponding to the user ID can be retrieved from the database according to his user ID, the retrieved historical broadband complaint processing data will be determined as the first historical broadband complaint processing data, and the sequence input into the fault repair model will be [first historical broadband complaint processing data; second historical broadband complaint processing data; third historical broadband complaint processing data]; if the historical broadband complaint processing data corresponding to the user ID cannot be retrieved from the database according to his user ID, the sequence input into the fault repair model will be [none; second historical broadband complaint processing data; third historical broadband complaint processing data].
[0070] In actual operation, if the fault repair model is a neural network model with relatively low computing power, data padding should be performed at the “None” in “[None; Second historical broadband complaint processing data; Third historical broadband complaint processing data]” so that the fault repair model can accurately extract the “Second historical broadband complaint processing data and the third historical broadband complaint processing data” according to the order of each data in the sequence; if the fault repair model is a large language model with relatively high computing power, it can be directly input in the form of [None; Second historical broadband complaint processing data; Third historical broadband complaint processing data] without data padding. The large language model can accurately extract the “Second historical broadband complaint processing data and the third historical broadband complaint processing data” according to [None; Second historical broadband complaint processing data; Third historical broadband complaint processing data]. In one implementation method, S130 includes: step (10), the details of which are as follows: Step (10): determining a current broadband fault repair plan through a fault repair model based on the first information, the second information, and the third information; The third information includes first historical broadband complaint processing data, second historical broadband complaint processing data and third historical broadband complaint processing data.
[0071] Specifically, in the embodiment of the present application, the third information is also the input data of the fault repair model; the third information indicates the historical broadband complaint processing data, and the historical broadband complaint processing data is the record obtained by processing the broadband complaints that occurred before the current broadband complaint work order; in the embodiment of the present application, whenever the broadband complaint processing device handles a broadband complaint, the entire process of handling the broadband complaint is recorded in the database for query.
[0072] In practical applications, the data input to the fault repair model are all data that need to be used to assist in formulating the current broadband fault repair plan.
[0073] The user identification and broadband installation address in the first information are used to be added to the output current broadband fault repair plan, so that maintenance personnel can clearly know the broadband user and broadband installation address targeted by the current broadband fault repair method.
[0074] The current broadband fault information and current broadband status information in the first information, the current broadband fault type in the second information, and the first historical broadband complaint processing data, the second historical broadband complaint processing data, and the third historical broadband complaint processing data in the third information are used for the fault repair model to understand the current broadband fault and formulate the current broadband fault repair process and / or precautions.
[0075] The current broadband fault urgency and the current user emotion type in the second information are used by the fault repair model to determine the execution priority of the current broadband fault repair plan.
[0076] In actual operation, under the prompts of the first information, the second information and the third information, the fault repair model finally outputs the current broadband fault repair plan including the broadband user, the broadband safety address, the current broadband fault repair process and / or precautions and the execution priority of the current broadband fault repair plan.
[0077] In an embodiment of the present application, the fault repair model is a trained large language model. In actual operation, since the fault repair model is obtained by training with a large number of second training samples, its accuracy in predicting the current broadband fault repair plan depends only on the predictive ability learned during the training process. Since this predictive ability is stable, the fault repair model's prediction process for the current broadband fault repair plan does not need to rely on manual labor and will not be affected by subjective factors of the staff. Therefore, the current broadband fault repair plan can be formulated efficiently and accurately.
[0078] In one implementation, after S130, the method further includes: steps (11) to (14), the details of which are as follows: Step (11): According to the current broadband complaint type, the current broadband fault work order and the current broadband fault repair plan are assigned to the current maintenance personnel corresponding to the current broadband complaint type.
[0079] Specifically, in an embodiment of the present application, after determining the current broadband fault repair plan for the current complaint broadband work order, the current broadband complaint work order and its corresponding current broadband fault repair plan can be assigned to maintenance personnel with corresponding professional skills and experience to ensure that the assigned maintenance personnel can efficiently handle specific types of faults.
[0080] Step (12): In response to the current maintenance personnel completing the current broadband fault repair plan, determining whether the current broadband fault has been successfully repaired.
[0081] Specifically, if the current broadband fault repair plan is an offline repair plan, the maintenance personnel will go to the broadband installation address to perform troubleshooting and repair according to the assigned broadband fault repair plan. After the repair is completed, the maintenance personnel will confirm with the user whether the broadband has returned to normal. If the broadband fault has been resolved and the user confirms that they are satisfied, the repair task is completed. Step (13): If the repair is not successful, re-execute the step of determining the current broadband fault repair plan through the fault repair model based on the first information and the second information.
[0082] Specifically, if the maintenance personnel confirms that the broadband fault has not been resolved, it is necessary to re-execute S130 to determine a new current broadband fault repair plan for performing another broadband fault repair.
[0083] In actual operation, after a broadband complaint is handled, the broadband complaint handling equipment automatically sends a questionnaire to broadband users, asking about their satisfaction and experience with the broadband fault repair results, to collect additional comments and suggestions from broadband users, especially when the online broadband repair plan fails to successfully repair the broadband fault.
[0084] In actual operation, the broadband complaint handling equipment regularly analyzes data about the broadband complaint handling equipment reported by broadband users, such as repair success rate and response time. Subsequently, based on the data reported by broadband users, the fault classification model and fault repair model are regularly updated. By using technologies such as transfer learning, new data can be quickly integrated into the existing fault classification model and fault repair model, thereby improving the adaptability and accuracy of the fault classification model and fault repair model. Among them, "new data" refers to the broadband complaint handling data collected during the use of the broadband complaint handling equipment.
[0085] In actual operation, a real-time monitoring mechanism can also be established for the processing process of the broadband complaint processing equipment to adjust the processing process of the broadband complaint processing equipment according to the data obtained from real-time monitoring to deal with sudden failure modes.
[0086] In actual applications, before executing S130 , it is necessary to first train the fault repair model. The training process of the fault repair model can refer to the training process of the fault classification model, which will not be described in detail here.
[0087] In summary, the broadband complaint processing method based on a large model provided in the embodiment of the present application can perform real-time processing of broadband complaint work orders of broadband users, such as real-time classification of broadband faults, thereby improving the efficiency and accuracy of broadband fault classification, and real-time broadband fault repair of broadband faults, thereby improving the efficiency and accuracy of broadband fault repair.
[0088] Second, this application provides a broadband complaint processing device based on a large model, such as Figure 6 As shown, Figure 6 This is a schematic diagram of the structure of a broadband complaint processing device based on a large model provided in an embodiment of the present application. The device includes: a first information determination module 310, a second information determination module 320, and a maintenance plan determination module 330; A first information determining module 310 is configured to determine first information based on a current broadband complaint work order received from a broadband user; The current broadband complaint worksheet is used to record the current broadband fault of the faulty broadband complained by the broadband user; the first information indicates the current broadband status of the faulty broadband; A second information determination module 320 is configured to determine second information based on the first information using a fault classification model; The second information indicates the current width fault type of the current broadband fault; A maintenance plan determination module 330 is configured to determine a current broadband fault maintenance plan using a fault maintenance model based on the first information and the second information; The fault classification model and the fault repair model are obtained by training a large language model respectively.
[0089] In one implementation, the first information includes: a user identifier, a broadband installation address, current broadband fault information, and current broadband status information; the first information determination module 310 is further configured to determine, based on the current broadband complaint work order, the user identifier of the broadband user, the broadband installation address of the faulty broadband, and the current broadband fault information of the faulty broadband provided by the broadband user; The first information determining module 310 is further configured to determine, based on the user identifier, current broadband status information of the faulty broadband complained by the broadband user.
[0090] In one implementation, the first information also includes: the current broadband complaint text; the second information includes: the current broadband fault type, the current broadband fault urgency and the current user emotion type; the first information determination module 310 is also used to determine the current broadband complaint text corresponding to the current broadband complaint voice based on the current broadband complaint voice received from the broadband user; wherein the current broadband complaint voice is the voice data when the broadband user makes a broadband complaint.
[0091] In one implementation, the apparatus further includes: a training module; The training module is further configured to obtain a training sample set including a plurality of first training samples; Wherein, the first training sample includes a broadband complaint text indicating the faulty broadband; The training module is further configured to label each of the first training samples in the training sample set based on the label frame; The labels include: a label for indicating the type of broadband fault, a label for indicating the urgency of the fault, and a label for indicating the type of user emotion; The training module is further configured to iteratively train the fault classification model using the training sample set and the labels corresponding to each of the first training samples in the training sample set. In one implementation, the second information determination module 320 is further configured to retrieve first historical broadband complaint processing data corresponding to the user identifier from a database based on the user identifier; The second information determination module 320 is further configured to retrieve, from a database, second historical broadband complaint processing data corresponding to the broadband installation address based on the broadband installation address; The second information determining module 320 is further configured to retrieve, from a database, third historical broadband complaint processing data of the broadband coverage area where the broadband installation address is located, based on the broadband installation address.
[0092] In one implementation, the maintenance solution determination module 330 is further configured to determine a current broadband fault maintenance solution using a fault maintenance model based on the first information, the second information, and the third information; The third information includes first historical broadband complaint processing data, second historical broadband complaint processing data and third historical broadband complaint processing data.
[0093] In one implementation, the maintenance solution determination module 330 is further configured to assign the current broadband fault work order and the current broadband fault maintenance solution to the current maintenance personnel corresponding to the current broadband complaint type according to the current broadband complaint type; The maintenance plan determination module 330 is further configured to determine whether the current broadband fault has been successfully repaired in response to the current maintenance personnel completing the current broadband fault maintenance plan; The repair solution determination module 330 is further configured to, if the repair is unsuccessful, re-execute the step of determining the current broadband fault repair solution using the fault repair model based on the first information and the second information.
[0094] Third, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, steps S110 to S130 provided in the above embodiment are implemented.
[0095] Fourth, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, steps S110 to S130 of the above embodiment are executed.
[0096] Fifth, the computer program product provided in this application includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the method in the previous method embodiment. For specific implementation, please refer to steps S110 to S130 of the method embodiment, which will not be repeated here.
[0097] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments provided above are merely illustrative. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0098] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0099] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0100] It should be noted that if the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.
[0101] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.
[0102] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.
Claims
1. A broadband complaint handling method based on a large model, characterized in that: The method comprises: Determining first information based on a current broadband complaint work order received from a broadband user; The current broadband complaint worksheet is used to record the current broadband fault of the faulty broadband complained by the broadband user; the first information indicates the current broadband status of the faulty broadband; Determining second information using a fault classification model based on the first information; Wherein, the second information indicates the current width fault type of the current broadband fault; Determining a current broadband fault repair plan using a fault repair model based on the first information and the second information; The fault classification model and the fault repair model are obtained by training a large language model respectively.
2. The method according to claim 1, characterized in that The first information includes: user identification, broadband installation address, current broadband fault information and current broadband status information; the first information is determined based on the current broadband complaint work order received from the broadband user, including: Determining, based on the current broadband complaint work order, the user ID of the broadband user, the broadband installation address of the faulty broadband, and the current broadband fault information of the faulty broadband provided by the broadband user; The current broadband status information of the faulty broadband complained by the broadband user is determined according to the user identifier.
3. The method according to claim 2, characterized in that The first information further includes: the current broadband complaint text; the second information includes: the current broadband fault type, the current broadband fault urgency, and the current user emotion type; before determining the second information using the fault classification model based on the first information, the method further includes: Determining, based on the received current broadband complaint voice of the broadband user, a current broadband complaint text corresponding to the current broadband complaint voice; The current broadband complaint voice is the voice data of the broadband user making a broadband complaint.
4. The method according to claim 3, characterized in that Before determining the second information by using the fault classification model, the method further includes: Acquire a training sample set including a plurality of first training samples; Wherein, the first training sample includes a broadband complaint text indicating the faulty broadband; Based on the label, labeling each first training sample in the training sample set; The labels include: a label for indicating the type of broadband fault, a label for indicating the urgency of the fault, and a label for indicating the type of user emotion; The fault classification model is iteratively trained using the training sample set and the label corresponding to each first training sample in the training sample set.
5. The method according to claim 2, characterized in that Before determining a current broadband fault repair solution using a fault repair model based on the first information and the second information, the method further includes: According to the user identifier, first historical broadband complaint processing data corresponding to the user identifier is retrieved from the database; according to the broadband installation address, second historical broadband complaint processing data corresponding to the broadband installation address is retrieved from the database; According to the broadband installation address, third historical broadband complaint processing data of the broadband coverage area where the broadband installation address is located is retrieved from the database.
6. The method according to claim 5, characterized in that The determining, based on the first information and the second information, a current broadband fault repair plan using a fault repair model includes: Determining a current broadband fault repair plan using a fault repair model based on the first information, the second information, and the third information; The third information includes the first historical broadband complaint processing data, the second historical broadband complaint processing data and the third historical broadband complaint processing data.
7. The method according to claim 1, characterized in that After determining a current broadband fault repair plan using a fault repair model based on the first information and the second information, the method further includes: According to the current broadband complaint type, the current broadband fault work order and the current broadband fault repair plan are assigned to the current maintenance personnel corresponding to the current broadband complaint type; In response to the current maintenance personnel completing the execution of the current broadband fault repair plan, determining whether the current broadband fault has been successfully repaired; If the repair is unsuccessful, the step of determining a current broadband fault repair plan using a fault repair model based on the first information and the second information is performed again.
8. A broadband complaint processing device based on a large model, characterized in that: The device includes: a first information determination module, a second information determination module and a maintenance plan determination module; The first information determination module is configured to determine the first information based on the current broadband complaint work order received from the broadband user; The current broadband complaint worksheet is used to record the current broadband fault of the faulty broadband complained by the broadband user; the first information indicates the current broadband status of the faulty broadband; The second information determining module is configured to determine second information using a fault classification model based on the first information; Wherein, the second information indicates the current width fault type of the current broadband fault; The maintenance solution determination module is configured to determine a current broadband fault maintenance solution using a fault maintenance model based on the first information and the second information; The fault classification model and the fault repair model are obtained by training a large language model respectively.
9. An electronic device, characterized in that: The electronic device includes a processor and a memory, the memory is used to store an application program, and the processor runs or executes a software program stored in the memory so that the electronic device implements the broadband complaint processing method based on a large model as described in any one of claims 1 to 7.
10. A computer program product, characterized in that The computer program product includes computer instructions. When the computer instructions are run on an electronic device, the electronic device implements the broadband complaint handling method based on a large model as described in any one of claims 1 to 7.