Early warning analysis method and device for intelligent service hotline work order, equipment and medium

Abstract

The invention discloses an early warning analysis method and device for an intelligent service hotline work order, equipment and a medium. The method comprises the following steps: acquiring voice content in an intelligent service hotline work order; converting the voice content into characters to obtain a work order to be processed; inputting the work order to be processed into a pre-trained AI large model for classification to obtain different fault types; whether the work order early warning score of a single fault type is an abnormal value or not is judged, or whether the work order early warning total value of all fault types is an abnormal value or not is judged, if the work order early warning score or the work order early warning total value is an abnormal value, early warning is carried out, and therefore early warning analysis of the intelligent service hotline work order is completed. The method is based on the AI large model, and can timely and comprehensively identify and respond to changes of customer demands, thereby effectively performing early warning on the intelligent service hotline work order, and meeting high-standard requirements of modern enterprises on customer services.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and medium for early warning analysis of intelligent service hotline work orders. Background Art

[0002] Customer service tickets are tools used by the customer service system to create and handle customer issues. When a customer's needs cannot be simply solved through online Q&A, the customer service system will create a form based on the customer's problem and then forward the ticket to the relevant department of the company for resolution. This can better understand customer needs and improve customer satisfaction and loyalty. With the continuous expansion of corporate scale and business, customer satisfaction and loyalty have become necessary indicators for corporate development, and an important manifestation of customer satisfaction and loyalty is the quality of customer service ticket processing.

[0003] A large number of customer service tickets may lead to a decrease in customer satisfaction, which may cause the company to lose customers. At the same time, timed and unprocessed tickets may also cause customer complaints and negative comments, affecting the company's reputation and image. In the prior art, the early warning methods for customer service tickets or service hotline tickets mainly include the following methods:

[0004] 1. Historical data analysis warning: Use data mining and machine learning technology to analyze historical work order data to predict which types of work orders may lead to a decline in customer satisfaction or substandard service levels. Such analysis can help companies identify potential high-risk work orders in advance and take preventive measures. 2. Customer importance warning: Some customers may be classified as VIP customers by the company due to reasons such as high package UPPER value, long cooperation time or important strategic position. For the work orders of such customers, the system can set automatic priority upgrades and warnings to ensure that their needs can be responded to and processed quickly. 3. Queue length warning: Monitor the length of the work order queue of each customer service team. When the queue length exceeds the preset threshold, the system issues a warning, prompting management or the customer service team to adjust resources or optimize processes to avoid a decline in service quality. All three methods are traditional warning mechanisms. Although they can help companies identify and respond to potential problems to a certain extent, they are not perfect solutions. They each have limitations and cannot fully meet the high standards of modern companies for customer service.

[0005] First, historical data analysis and early warning rely on past data patterns. However, in a rapidly changing market environment, the relevance and timeliness of historical data may be greatly reduced. This makes predictions based on historical data often unable to accurately reflect current customer needs and service levels, resulting in companies being slow to respond to customer issues.

[0006] Secondly, although customer importance warning can help companies prioritize VIP customers’ work orders, this method is often subjective. Over-reliance on certain customers may lead to unfair resource allocation and ignore the reasonable demands and experiences of other customers. In addition, over-concentrating resources on a few important customers may reduce overall customer service efficiency, affect the service quality of other customers, and even increase the risk of dependence on VIP customers in long-term relationships.

[0007] Finally, although the queue length warning can monitor the progress of work order processing, it focuses too much on the number of work orders to be processed and ignores the urgency and complexity of the work orders. This simplistic processing method may lead to unscientific and unreasonable resource allocation and fail to fundamentally optimize the workflow or improve processing efficiency. At the same time, in order to reduce the queue length, customer service staff may tend to process work orders quickly or even close work orders rashly, thereby sacrificing service quality.

[0008] However, the above methods only rely on historical data or a single indicator, and fail to fully consider the multi-dimensional needs of customers and comprehensive factor analysis. Therefore, they are unable to timely and comprehensively identify and respond to changes in customer needs, and are unable to effectively issue early warnings for smart service hotline work orders. Summary of the invention

[0009] The technical problem to be solved by the present invention is to address the above-mentioned deficiencies in the prior art and propose a warning analysis method, device, equipment and medium for intelligent service hotline work orders. The method can timely and comprehensively identify and respond to changes in customer needs, thereby effectively warning intelligent service hotline work orders to meet the high standards of modern enterprises for customer service.

[0010] In a first aspect, the present invention provides a method for early warning analysis of an intelligent service hotline work order, the method comprising the following steps:

[0011] Step S1: Obtain the voice content in the smart service hotline work order;

[0012] Step S2: converting the voice content into text to obtain a work order to be processed;

[0013] Step S3: input the work order to be processed into a pre-trained AI model for classification to obtain different fault types;

[0014] Step S4: Determine whether the work order warning score of a single fault type is an abnormal value, or determine whether the total work order warning score of all fault types is an abnormal value:

[0015] If the work order warning score or the work order warning total value is an abnormal value, a warning is issued, thereby completing the warning analysis of the smart service hotline work order.

[0016] Furthermore, before the step S3, the method further includes step S0: training the AI ​​big model;

[0017] The step S0 specifically includes the following steps:

[0018] Obtain historical work order data; and select the initial AI learning model;

[0019] Wherein, the AI ​​learning model includes GPT, and / or CHATGPT, and / or Glaude, and / or Wenxinyiyan, and / or Yuanjingda model;

[0020] The initial AI learning model is trained according to the historical work order data to obtain an AI learning model, thereby completing the training of the AI ​​large model.

[0021] Furthermore, after step S4, the method further includes step S5: pushing the abnormal value to a fault handler so that the fault handler handles the abnormal value.

[0022] Furthermore, the work order warning score f(x i ) is calculated as follows:

[0023]

[0024] in,

[0025] f(x i ) is the work order warning score of the i-th fault type;

[0026] J i is the initial warning level of the i-th fault type; the initial warning level is a preset value;

[0027] x i It is the importance weight coefficient preset for different fault types;

[0028] len is the length of the user's unique attribute type;

[0029] U n is the user attribute coefficient of the nth unique attribute;

[0030] T n The frequency of user failures in the current time period for the nth unique attribute;

[0031] L n The frequency of user failures in the last time period for the nth unique attribute;

[0032] Δ n Adjust the value of the factor for the nth unique attribute;

[0033] The calculation formula of the total value f(x) of the work order warning of all fault types is as follows:

[0034]

[0035] in,

[0036] K is the total category of fault types.

[0037] In a second aspect, the present invention provides an early warning analysis device for an intelligent service hotline work order, the device comprising:

[0038] An acquisition unit, used to acquire voice content in the smart service hotline work order;

[0039] A conversion unit, connected to the acquisition unit, for converting the voice content into text to obtain a work order to be processed;

[0040] A classification unit, connected to the conversion unit, for inputting the work order to be processed into a pre-trained AI large model for classification to obtain different fault types;

[0041] A first determination unit, connected to the classification unit, is used to determine whether the work order warning score of a single fault type is an abnormal value;

[0042] A second determination unit, connected to the classification unit, is used to determine whether the total value of work order warnings of all fault types is an abnormal value;

[0043] An early warning unit is connected to the first determination unit and the second determination unit respectively, and is used to issue an early warning when the first determination unit determines that the work order early warning score of a single fault type is an abnormal value, or when the second determination unit determines that the total work order early warning value of all fault types is an abnormal value, thereby completing the early warning analysis of the smart service hotline work order.

[0044] Furthermore, the device also includes:

[0045] A training unit, connected to the classification unit, for training the AI ​​big model;

[0046] The training unit comprises:

[0047] Acquisition module, used to obtain historical work order data;

[0048] Selection module, used to select the initial AI learning model;

[0049] Wherein, the AI ​​learning model includes GPT, and / or CHATGPT, and / or Glaude, and / or Wenxinyiyan, and / or Yuanjingda model;

[0050] The training module is connected to the acquisition module and the selection module respectively, and is used to train the initial AI learning model according to the historical work order data to obtain the AI ​​learning model, thereby completing the training of the AI ​​large model.

[0051] Furthermore, the device also includes:

[0052] The push unit is connected to the early warning unit and is used to push the abnormal value to the fault handler so that the fault handler can handle the abnormal value.

[0053] Further, the first determination unit includes a first calculation module;

[0054] The first calculation module is used to calculate the work order warning score of a single fault type. The first calculation module stores the work order warning score f(x i ), the specific calculation formula is as follows:

[0055]

[0056] in,

[0057] f(x i ) is the work order warning score of the i-th fault type;

[0058] J i is the initial warning level of the i-th fault type; the initial warning level is a preset value;

[0059] x i It is the importance weight coefficient preset for different fault types;

[0060] len is the length of the user's unique attribute type;

[0061] U n is the user attribute coefficient of the nth unique attribute;

[0062] T n The frequency of user failures in the current time period for the nth unique attribute;

[0063] L n The frequency of user failures in the last time period for the nth unique attribute;

[0064] Δ n Adjust the value of the factor for the nth unique attribute;

[0065] The second determination unit includes a second calculation module, which is connected to the first calculation module and is used to calculate the total value of work order warnings of all fault types;

[0066] The second calculation module stores a calculation formula for the total value f(x) of work order warnings of all fault types, and the calculation formula is as follows:

[0067]

[0068] K is the total category of fault types.

[0069] In a third aspect, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the early warning analysis method for the intelligent service hotline work order according to the first aspect.

[0070] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor executes the early warning analysis method for the intelligent service hotline work order according to the first aspect.

[0071] The present invention is based on the AI ​​big model, has comprehensive coverage and dynamic learning capabilities, can timely and comprehensively identify and respond to changes in customer needs, thereby effectively issuing early warnings for smart service hotline work orders to meet the high standards of modern enterprises for customer service.

[0072] The specific beneficial effects are as follows:

[0073] 1. Comprehensively analyze customers' multi-dimensional needs and various factors

[0074] The present invention can comprehensively analyze the multi-dimensional needs and related factors of customers, and fully identify the real intentions and potential problems of customers, thereby ensuring the pertinence and effectiveness of work order processing.

[0075] 2. Improve response speed

[0076] Through automated processing and real-time analysis, the present invention significantly shortens customer waiting time, improves response efficiency, enables enterprises to quickly meet customer needs, and thus improves customer satisfaction.

[0077] 3. Accurately identify customer needs

[0078] The present invention uses natural language processing technology to accurately understand customers' problems and emotions and provide more effective solutions, thereby enhancing customer experience and satisfaction.

[0079] 4. Dynamic learning and optimization

[0080] With the help of AI big models, the present invention can continuously learn and optimize based on customer feedback and historical data, improve service quality and processing capabilities, and ensure that the system always remains efficient and accurate.

[0081] 5. Intelligent anomaly detection and early warning

[0082] The present invention can timely identify potential problems by judging the warning score and total value of the work order in real time, help enterprises take measures in advance, reduce the risk of customer loss, and improve service reliability.

[0083] 6. Reduce operating costs

[0084] Through automated and intelligent processing, the present invention reduces reliance on manual customer service, reduces the operating costs of the enterprise, and improves resource utilization efficiency, enabling the enterprise to achieve cost savings while maintaining high-quality services. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 Schematic diagram of a warning analysis method for an intelligent service hotline work order in an embodiment of the present invention;

[0086] Figure 2 This is a warning analysis framework diagram of the smart service hotline work order in an embodiment of the present invention;

[0087] Figure 3 Schematic diagram of an early warning analysis device for an intelligent service hotline work order in an embodiment of the present invention;

[0088] Among them, the figure numbers are: 10, acquisition unit, 20, conversion unit, 30, classification unit, 40, first determination unit, 50, second determination unit, 60, early warning unit. DETAILED DESCRIPTION

[0089] In order to enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0090] It should be understood that the specific embodiments and drawings described herein are only used to explain the present invention rather than to limit the present invention.

[0091] It can be understood that, in the absence of conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other.

[0092] It can be understood that, for the convenience of description, the drawings of the present invention only show the parts related to the present invention, while the parts irrelevant to the present invention are not shown in the drawings.

[0093] It can be understood that each unit and module involved in the embodiments of the present invention 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.

[0094] It can be understood that, in the absence of conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0095] It is understood that the flowcharts and block diagrams of the present invention illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present invention. 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 may be implemented by a combination of hardware and computer instructions.

[0096] It can be understood that the units and modules involved in the embodiments of the present invention can be implemented by software or hardware. For example, the units and modules can be located in a processor.

[0097] Embodiment 1:

[0098] like Figure 1 As shown, this embodiment provides a warning analysis method for intelligent service hotline work orders. The method is based on an AI big model and is widely used in customer service centers, technical support, complaint handling, product quality monitoring, marketing, operations management, smart homes, and financial services. The method can analyze customer feedback in real time, identify potential problems, and issue warnings, thereby improving service quality, optimizing resource allocation, enhancing customer satisfaction, and effectively managing risks. The method includes the following steps:

[0099] Step S0: training the AI ​​big model;

[0100] The step S0 specifically includes the following steps:

[0101] Obtain historical work order data; and select the initial AI learning model;

[0102] Wherein, the AI ​​learning model includes GPT, and / or CHATGPT, and / or Glaude, and / or Wenxinyiyan, and / or Yuanjingda model;

[0103] The initial AI learning model is trained according to the historical work order data to obtain an AI learning model, thereby completing the training of the AI ​​large model.

[0104] The historical work order data obtained in this embodiment is marked. These marks are used to provide additional information so that the AI ​​learning model can better understand and process the data. The marked historical work order data includes the following:

[0105] (1) Work order ID: uniquely identifies each work order.

[0106] (2) Customer information: such as customer name, contact information, etc.

[0107] (3) Problem description: The specific content of the question or request raised by the customer.

[0108] (4) Fault category: classification of problems, such as technical failures, service requests, complaints, etc.

[0109] (5) Priority: Mark the urgency of the work order (high, medium, low).

[0110] (6) Processing result: The status of the work order after it is processed, such as "solved", "to be resolved", "unresolved", etc.

[0111] (7) Processing time: The time it takes for a work order to be created and resolved.

[0112] (8) Customer feedback: Customer satisfaction ratings or comments on the solution.

[0113] (9) Processing Personnel: Information of the customer service personnel responsible for handling the work order.

[0114] GPT, ChatGPT, Glaude, Wenxinyiyan and Yuanjingda models are all natural language processing models based on artificial intelligence technology, each of which has unique characteristics and application scenarios. GPT (Generative Pre-trained Transformer) is a deep learning model developed by OpenAI, focusing on generating coherent text, and is widely used in dialogue systems, text generation, and content creation. ChatGPT is a variant of GPT, which is specially optimized for dialogue scenarios and aims to provide a more natural and humane communication experience. Glaude is another natural language processing model that may perform well in certain specific tasks. Wenxinyiyan and Yuanjingda models are also AI language models designed to meet the needs of Chinese users and provide services such as intelligent dialogue, text generation, and information retrieval. These models have demonstrated strong capabilities in different fields and applications, promoting the development and application of artificial intelligence technology.

[0115] Step S0 is the key link in training the AI ​​big model, which specifically includes multiple sub-steps. First, the system needs to obtain labeled historical work order data, which usually contains the customer's consultation content, problem description, and corresponding processing results and classification labels. These labeled data provide a basis for the model's learning, enabling it to understand the characteristics and processing methods of different types of work orders. Next, selecting the initial AI learning model is a crucial step. The selected model can be a variety of advanced natural language processing models, such as GPT, CHATGPT, Glaude, Wenxin Yiyan or Yuanjing big model. These models have their own characteristics, can handle complex language tasks, and have good learning ability. Subsequently, the selected initial AI learning model is trained using the acquired labeled historical work order data. In this process, the model learns how to extract useful information from the input work order data by continuously adjusting its internal parameters, and classifies and predicts according to the tags in the historical data. Finally, after multiple rounds of iterative training, the model will gradually optimize to form a target AI learning model that can accurately understand and process customer service work orders, thereby completing the training of the AI ​​big model. This process not only improves the accuracy and reliability of the model, but also provides a solid foundation for subsequent work order analysis and early warning.

[0116] The specific steps of training the AI ​​big model in this embodiment are as follows:

[0117] (1) Data preprocessing: After obtaining the labeled historical work order data, the data must first be cleaned and preprocessed. This includes removing irrelevant information, processing missing values, and standardizing text formats to ensure data quality and consistency.

[0118] (2) Text vectorization: Convert text data into a numerical format that the model can process. Common methods include the bag-of-words model, TF-IDF (term frequency-inverse document frequency), and word embedding (such as Word2Vec or GloVe). For example, word embedding methods are used to convert each word into a high-dimensional vector so that similar words are close in the vector space.

[0119] (3) Model selection and initialization: Select a suitable initial AI learning model, such as GPT, ChatGPT, Glaude, Wenxinyiyan, or Yuanjingda model. Depending on the task requirements, you can choose a pre-trained model for fine-tuning, or start training from scratch.

[0120] (4) Training process:

[0121] Input data: The processed and labeled historical work order data is input into the selected model. The content of each work order (such as customer consultation, problem description) is used as input, and the corresponding processing results and classification labels are used as target outputs.

[0122] Loss function: Define a loss function to measure the gap between the model output and the actual label. Commonly used loss functions include the cross entropy loss function.

[0123] Optimization algorithm: Use an optimization algorithm (such as Adam or SGD) to adjust the model parameters to minimize the loss function. The gradient is calculated through the back-propagation algorithm and the model weights are updated.

[0124] (5) Model evaluation: During the training process, the validation set is used regularly to evaluate the performance of the model. Indicators such as accuracy, recall, and F1-score can be used to measure the performance of the model on different types of work orders.

[0125] (6) Fine-tuning and iteration: Based on the evaluation results, fine-tune the model, adjust hyperparameters (such as learning rate, batch size, etc.), and perform multiple rounds of training to improve the performance of the model.

[0126] (7) Deployment and application: After training is completed, the model is deployed to actual applications for real-time work order processing and early warning analysis.

[0127] Specific examples are as follows:

[0128] Suppose that the GPT model is used to process customer service hotline tickets. First, obtain a batch of labeled historical ticket data, which contains the customer's consultation content (such as "I can't log in to my account"), problem description (such as "The user entered the wrong password"), processing results (such as "The password has been reset") and classification labels (such as "Account problem").

[0129] In the data preprocessing stage, the data is cleaned, irrelevant symbols and stop words are removed, and the text format is ensured to be consistent. Next, the text data is vectorized and each word is converted into a vector using Word2Vec.

[0130] Then, GPT is selected as the initial learning model, and the processed work order data is input into the model for training. The model gradually learns how to identify different types of work orders and their processing methods by learning the relationship between the input text and the corresponding processing results.

[0131] During the training process, the cross entropy loss function is used to evaluate the gap between the model output and the actual label, and the model parameters are adjusted through the Adam optimization algorithm. After multiple rounds of training and evaluation, the performance of the model continues to improve, and it can eventually accurately identify and process new customer work orders and realize intelligent services.

[0132] Step S1: Obtain the voice content in the smart service hotline work order.

[0133] This process usually involves using speech recognition technology to convert the customer's voice input into text format in real time or after recording. First, the system receives the customer's voice signal through a microphone or telephone line, and then uses advanced speech recognition algorithms (such as deep learning models) to analyze these signals and identify the words and sentences in the voice. During the recognition process, the system will consider factors such as the tone, speed, and accent of the voice to improve the accuracy of recognition. Finally, the processed voice content will be converted into text to form the work order data to be processed.

[0134] Step S2: convert the voice content into text to obtain a work order to be processed.

[0135] Step S3: Input the work order to be processed into a pre-trained AI model for classification to obtain different fault types.

[0136] As a specific implementation method, the work order warning score f(x i ) is calculated as follows:

[0137]

[0138] in,

[0139] f(x i ) is the work order warning score of the i-th fault type;

[0140] J i is the initial warning level of the i-th fault type; the initial warning level is a preset value;

[0141] x i It is the importance weight coefficient preset for different fault types;

[0142] len is the length of the user's unique attribute type;

[0143] U n is the user attribute coefficient of the nth unique attribute;

[0144] T n The frequency of user failures in the current time period for the nth unique attribute;

[0145] L n The frequency of user failures in the last time period for the nth unique attribute;

[0146] Δ n Adjust the value of the factor for the nth unique attribute;

[0147] The calculation formula of the total value f(x) of the work order warning of all fault types is as follows:

[0148]

[0149] in,

[0150] K is the total category of fault types.

[0151] Step S4: Determine whether the work order warning score of a single fault type is an abnormal value, or determine whether the total work order warning score of all fault types is an abnormal value:

[0152] If the work order warning score or the work order warning total value is an abnormal value, a warning is issued, thereby completing the warning analysis of the smart service hotline work order.

[0153] In step S3, the work orders to be processed are input into the pre-trained AI big model for classification to identify different types of faults. In specific implementation, the calculation formula for the work order warning score for each fault type takes into account multiple factors, including the initial alert level, the importance weight coefficient, the user's unique attributes and their related fault occurrence frequency. The initial alert level is a preset value used to set the basic alert standard for the fault type; the importance weight coefficient reflects the impact of different fault types on the business. The user's unique attribute type length and the attribute coefficient of each unique attribute, combined with the fault occurrence frequency in the current and previous time periods, help the model more accurately assess the urgency and importance of the fault. In addition, the factor adjustment value is used to dynamically adjust the warning score to adapt to different business environments and user needs. Then, in step S4, the system will determine whether the work order warning score of a single fault type is an abnormal value, and whether the total value of the work order warning for all fault types is an abnormal value. If any warning score or total value is found to be abnormal, the system will trigger the warning mechanism, thereby completing the warning analysis of the smart service hotline work order, ensuring that the enterprise can respond to potential risks and problems in a timely manner.

[0154] The specific example is as follows, assuming the following data:

[0155] Fault type i=1,2;

[0156] Initial alert level J1 = 50, J2 = 60;

[0157] Weight coefficients x1 = 1.5, x2 = 2.0;

[0158] User attribute type length len = 3;

[0159] User attribute coefficient U = [1.2, 1.5, 1.1];

[0160] Failure frequency in the current time period T = [20, 30, 10];

[0161] The fault frequency in the last time period L = [10, 15, 5];

[0162] Factor adjustment value Δ = [1.1, 1.2, 1.0];

[0163] Then the warning score of the first fault type is calculated according to the formula;

[0164] f(x1)=62.66;

[0165] Then calculate the warning score of the second fault type;

[0166] f(x2)=76.88;

[0167] Then calculate the total value of work order warnings for all fault types:

[0168] f(x)=f(x1)+f(x2)=139.54

[0169] Finally, the warning score or the total warning value is used to determine whether a warning is needed. If a single warning score exceeds the set warning threshold, or the total warning value exceeds the set total value threshold, the system will automatically send the warning information to the warning push module to notify relevant personnel in a timely manner.

[0170] Step S5: Push the abnormal value to the fault handler so that the fault handler can handle the abnormal value.

[0171] After receiving the early warning information, the fault handling personnel will conduct on-site inspection and analysis based on the content of the customer's complaint to determine whether it is a high-frequency group fault problem, develop a solution and provide feedback on the handling results.

[0172] Generally speaking, after receiving the early warning information, the fault handling personnel will first carefully review the specific content of the customer's complaint. Then, they will go to the site for investigation and conduct in-depth analysis to see if the problem is a high-frequency problem of mass failure. After confirming the nature of the problem, the handling personnel will formulate a corresponding solution and promptly feedback the handling results to the customer. This process forms a closed loop of problem solving, which not only ensures the effective solution of the problem, but also enhances the customer's trust and satisfaction.

[0173] like Figure 2 As shown, the processing flow of this embodiment is as follows:

[0174] (1) Customer service hotline staff answer user complaint calls.

[0175] (2) The speech recognition module analyzes the content of the complaint call and converts the complaint content into text.

[0176] (3) The complaint ticket is transferred to the big voice model analysis module to analyze the ticket type and compare it with the basic attribute information of the database user.

[0177] (4) Calculate the warning threshold through the warning threshold algorithm analysis module.

[0178] (5) When the estimated threshold exceeds the warning red line value, the warning data is sent to the warning push module, and the warning information is sent to the group fault handler and the enterprise WeChat group.

[0179] (6) The handler responds to the warning processing progress and closes the customer service warning ticket.

[0180] This embodiment is based on the AI ​​big model, has the ability of comprehensive coverage and dynamic learning, can timely and comprehensively identify and respond to changes in customer needs, so as to effectively warn the intelligent service hotline work order to meet the high standards of modern enterprises for customer service. Specifically, this embodiment can comprehensively analyze the multi-dimensional needs and related factors of customers, comprehensively identify the true intentions and potential problems of customers, and ensure the pertinence and effectiveness of work order processing. In addition, through automated processing and real-time analysis, the system significantly shortens the waiting time of customers, improves response efficiency, and enables enterprises to quickly meet customer needs, thereby improving customer satisfaction. Using natural language processing technology, this embodiment can accurately understand the problems and emotions of customers, provide more effective solutions, and further enhance customer experience and satisfaction. With the advantages of the AI ​​big model, this embodiment can also continuously learn and optimize according to customer feedback and historical data, improve service quality and processing capabilities, and ensure that the system always remains efficient and accurate. By judging the warning score and total value of the work order in real time, the system can timely identify potential problems, help enterprises take measures in advance, reduce the risk of customer churn, and improve the reliability of services. At the same time, the application of automation and intelligent processing reduces dependence on manual customer service, reduces the company's operating costs, improves resource utilization efficiency, and enables companies to achieve cost savings while maintaining high-quality services.

[0181] Embodiment 2:

[0182] like Figure 3 As shown, this embodiment provides an early warning analysis device for intelligent service hotline work orders, the device comprising:

[0183] An acquisition unit 10 is used to acquire the voice content in the smart service hotline work order;

[0184] The conversion unit 20 is connected to the acquisition unit 10 and is used to convert the voice content into text to obtain a work order to be processed;

[0185] The classification unit 30 is connected to the conversion unit 20 and is used to input the work order to be processed into a pre-trained AI large model for classification to obtain different fault types;

[0186] A first determination unit 40, connected to the classification unit 30, is used to determine whether the work order warning score of a single fault type is an abnormal value;

[0187] The second determination unit 50 is connected to the classification unit 30 and is used to determine whether the total value of the work order warnings of all fault types is an abnormal value;

[0188] The early warning unit 60 is connected to the first determination unit 40 and the second determination unit 50 respectively, and is used to issue an early warning when the first determination unit determines that the work order early warning score of a single fault type is an abnormal value, or when the second determination unit determines that the total work order early warning value of all fault types is an abnormal value, thereby completing the early warning analysis of the smart service hotline work order.

[0189] As a specific implementation, the device further includes:

[0190] A training unit, connected to the classification unit, for training the AI ​​big model;

[0191] The training unit comprises:

[0192] Acquisition module, used to obtain historical work order data;

[0193] Selection module, used to select the initial AI learning model;

[0194] Wherein, the AI ​​learning model includes GPT, and / or CHATGPT, and / or Glaude, and / or Wenxinyiyan, and / or Yuanjingda model;

[0195] The training module is connected to the acquisition module and the selection module respectively, and is used to train the initial AI learning model according to the historical work order data to obtain the AI ​​learning model, thereby completing the training of the AI ​​large model.

[0196] As a specific implementation, the device further includes:

[0197] The push unit is connected to the early warning unit and is used to push the abnormal value to the fault handler so that the fault handler can handle the abnormal value.

[0198] As a specific implementation, the first determination unit includes a first calculation module;

[0199] The first calculation module is used to calculate the work order warning score of a single fault type. The first calculation module stores the work order warning score f(x i ), the specific calculation formula is as follows:

[0200]

[0201] in,

[0202] f(x i ) is the work order warning score of the i-th fault type;

[0203] J i is the initial warning level of the i-th fault type; the initial warning level is a preset value;

[0204] x i It is the importance weight coefficient preset for different fault types;

[0205] len is the length of the user's unique attribute type;

[0206] U n is the user attribute coefficient of the nth unique attribute;

[0207] T n The frequency of user failures in the current time period for the nth unique attribute;

[0208] L n The frequency of user failures in the last time period for the nth unique attribute;

[0209] Δ n Adjust the value of the factor for the nth unique attribute;

[0210] The second determination unit includes a second calculation module, which is connected to the first calculation module and is used to calculate the total value of work order warnings of all fault types;

[0211] The second calculation module stores a calculation formula for the total value f(x) of work order warnings of all fault types, and the calculation formula is as follows:

[0212]

[0213] K is the total category of fault types.

[0214] The device in this embodiment can execute the method in embodiment 1.

[0215] Embodiment 3:

[0216] This embodiment provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the early warning analysis method for the smart service hotline work order according to Embodiment 1.

[0217] Embodiment 4:

[0218] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the processor executes the early warning analysis method for the smart service hotline work order according to Embodiment 1.

[0219] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A warning analysis method for intelligent service hotline work orders, characterized in that: The method comprises the following steps: Step S1: Obtain the voice content in the smart service hotline work order; Step S2: converting the voice content into text to obtain a work order to be processed; Step S3: input the work order to be processed into a pre-trained AI model for classification to obtain different fault types; Step S4: Determine whether the work order warning score of a single fault type is an abnormal value, or determine whether the total work order warning score of all fault types is an abnormal value: If the work order warning score or the work order warning total value is an abnormal value, a warning is issued, thereby completing the warning analysis of the smart service hotline work order.

2. The early warning analysis method for intelligent service hotline work orders according to claim 1 is characterized in that: Before step S3, the method further includes step S0: training an AI big model; The step S0 specifically includes the following steps: Obtain historical work order data; and select the initial AI learning model; Wherein, the AI ​​learning model includes GPT, and / or CHATGPT, and / or Glaude, and / or Wenxinyiyan, and / or Yuanjingda model; The initial AI learning model is trained according to the historical work order data to obtain an AI learning model, thereby completing the training of the AI ​​large model.

3. The early warning analysis method for intelligent service hotline work orders according to claim 1 is characterized in that: After step S4, the method further includes: Step S5: Push the abnormal value to the fault handler so that the fault handler can handle the abnormal value.

4. The early warning analysis method for intelligent service hotline work orders according to any one of claims 1 to 3, characterized in that: The work order warning score f(x i ) is calculated as follows: in, f(x i ) is the work order warning score of the i-th fault type; J i is the initial warning level of the i-th fault type; the initial warning level is a preset value; x i It is the importance weight coefficient preset for different fault types; len is the length of the user's unique attribute type; U n is the user attribute coefficient of the nth unique attribute; T n The frequency of user failures in the current time period for the nth unique attribute; L n The frequency of user failures in the last time period for the nth unique attribute; Δ n Adjust the value of the factor for the nth unique attribute; The calculation formula of the total value f(x) of the work order warning of all fault types is as follows: in, K is the total category of fault types.

5. An early warning analysis device for intelligent service hotline work orders, characterized in that: include: An acquisition unit, used to acquire voice content in the smart service hotline work order; A conversion unit, connected to the acquisition unit, for converting the voice content into text to obtain a work order to be processed; A classification unit, connected to the conversion unit, for inputting the work order to be processed into a pre-trained AI large model for classification to obtain different fault types; A first determination unit, connected to the classification unit, is used to determine whether the work order warning score of a single fault type is an abnormal value; A second determination unit, connected to the classification unit, is used to determine whether the total value of work order warnings of all fault types is an abnormal value; An early warning unit is connected to the first determination unit and the second determination unit respectively, and is used to issue an early warning when the first determination unit determines that the work order early warning score of a single fault type is an abnormal value, or when the second determination unit determines that the total work order early warning value of all fault types is an abnormal value, thereby completing the early warning analysis of the smart service hotline work order.

6. The early warning analysis device for intelligent service hotline work orders according to claim 5, characterized in that: Also includes: A training unit, connected to the classification unit, for training the AI ​​big model; The training unit comprises: Acquisition module, used to obtain historical work order data; Selection module, used to select the initial AI learning model; Wherein, the AI ​​learning model includes GPT, and / or CHATGPT, and / or Glaude, and / or Wenxinyiyan, and / or Yuanjingda model; The training module is connected to the acquisition module and the selection module respectively, and is used to train the initial AI learning model according to the historical work order data to obtain the AI ​​learning model, thereby completing the training of the AI ​​large model.

7. The early warning analysis device for intelligent service hotline work orders according to claim 5, characterized in that: Also includes: A push unit is connected to the early warning unit and is used to push the abnormal value to the fault handler so that the fault handler can handle the abnormal value.

8. The early warning analysis device for intelligent service hotline work orders according to any one of claims 5 to 7, characterized in that: The first determination unit includes a first calculation module; The first calculation module is used to calculate the work order warning score of a single fault type. The first calculation module stores the work order warning score f(x i ), the specific calculation formula is as follows: in, f(x i ) is the work order warning score of the i-th fault type; J i is the initial warning level of the i-th fault type; the initial warning level is a preset value; x i It is the importance weight coefficient preset for different fault types; len is the length of the user's unique attribute type; U n is the user attribute coefficient of the nth unique attribute; T n The frequency of user failures in the current time period for the nth unique attribute; L n The frequency of user failures in the last time period for the nth unique attribute; Δ n Adjust the value of the factor for the nth unique attribute; The second determination unit includes a second calculation module, which is connected to the first calculation module and is used to calculate the total value of work order warnings of all fault types; The second calculation module stores a calculation formula for the total value f(x) of work order warnings of all fault types, and the calculation formula is as follows: K is the total category of fault types.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein a computer program is stored in the memory, and when the processor runs the computer program stored in the memory, the processor executes the early warning analysis method for the smart service hotline work order according to any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor executes the early warning analysis method for the intelligent service hotline work order according to any one of claims 1 to 4.

Images