A customer service robot model training system, method and storage medium

Abstract

The present application discloses a model training system, method and storage medium for a customer service robot, the system comprising: a data acquisition module for collecting interaction data and preprocessing the interaction data; a data processing module for constructing a dialogue data set with intent labels based on the interaction data; a model training module for training a customer service robot model using the dialogue data set, generating a corresponding model response data set, optimizing the parameters and structure of the customer service robot model based on the model response data set, interaction data and dialogue data set, generating a corresponding dialogue strategy adjustment data set, and iteratively optimizing the customer service robot model using the dialogue strategy adjustment data set. The present application can realize automatic learning and optimization of the customer service robot model through machine learning and natural language processing technology, improve the accuracy and efficiency of question responses, greatly improve service quality and user satisfaction, and can be widely used in the field of customer service robot technology.

Description

Technical Field

[0001] The present application relates to the technical field of customer service robots, and in particular to a model training system, method and storage medium for a customer service robot. Background Art

[0002] With the development of technology, intelligent customer service robots have been widely used in various industries. However, traditional customer service robots are often unable to dynamically adapt to changes in user needs, resulting in a decline in service quality and user satisfaction.

[0003] In related technologies, customer service robots mainly rely on pre-set rules and static data to respond, lack the ability to self-learn and optimize, and find it difficult to accurately respond to user inquiries, with low response accuracy and efficiency. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose a model training system, method and storage medium for a customer service robot, which can realize automatic learning and optimization of the customer service robot model and improve the accuracy and efficiency of question responses.

[0005] On the one hand, the present application implements a model training system for a customer service robot, the system comprising:

[0006] A data acquisition module is used to collect interaction data and pre-process the interaction data; the interaction data includes at least user voice input data, user text input data, user interaction feedback data, and conversation record data between the user and the customer service robot;

[0007] A data processing module, configured to construct a conversation dataset with intent labels based on the interaction data;

[0008] A model training module is used to train a customer service robot model using the dialogue dataset to generate a corresponding model response dataset, perform parameter optimization and structural optimization on the customer service robot model based on the model response dataset, the interaction data, and the dialogue dataset, and generate a corresponding dialogue strategy adjustment dataset, and use the dialogue strategy adjustment dataset to iteratively optimize the customer service robot model.

[0009] In some embodiments, the system further comprises:

[0010] A knowledge base module, configured to store the dialogue strategy adjustment dataset;

[0011] The model interaction module is used to provide a model calling interface, through which the customer service robot model is applied to the target customer service robot.

[0012] In some embodiments, the data acquisition module includes a data collection unit and a data pre-processing unit;

[0013] The data collection unit is used to dynamically collect the interaction data;

[0014] The data preprocessing unit is used to perform data preprocessing on the interaction data, wherein the data preprocessing includes data cleaning and data formatting.

[0015] In some embodiments, the conversation dataset includes a plurality of user query data and an intent label corresponding to each of the user query data, robot response data, and user feedback data; the model training module includes an intent recognition training unit and a response generation unit;

[0016] The intention recognition training unit is configured to sequentially input each user query data into the customer service robot model, perform intent recognition using the customer service robot model, learn intent recognition parameters and generate intent recognition data corresponding to each user query data, and adjust the intent recognition parameters corresponding to the customer service robot model so that the intent recognition data of each user query data corresponds to the intent label;

[0017] The response generation unit is used to use the customer service robot model to respond according to each user query data and the intention recognition data corresponding to each user query data, and generate model response data corresponding to each user query data.

[0018] In some embodiments, the model training module further includes a first model optimization unit and a second model optimization unit;

[0019] The first model optimization unit is configured to perform model performance evaluation based on the model response data, the robot response data, and the user feedback data corresponding to each of the user query data, generate a model performance evaluation result, and optimize the hyperparameters and model structure corresponding to the customer service robot model based on the model performance evaluation result;

[0020] The second optimization unit of the model is used to optimize the dialogue strategy parameters and response generation logic parameters of the customer service robot model based on the model performance evaluation results and the user feedback data corresponding to each user query data.

[0021] In some embodiments, the first optimization unit of the model is used to obtain multiple performance evaluation indicators, determine the specific indicator value corresponding to each performance evaluation indicator according to the model response data, the robot response data and the user feedback data corresponding to each user query data, and adjust the hyperparameter according to the specific indicator value corresponding to each performance evaluation indicator;

[0022] The first optimization unit of the model is also used to dynamically adjust the model structure according to the specific indicator values ​​corresponding to each performance evaluation indicator and the hyperparameters, wherein the dynamic adjustment of the model structure at least includes dynamically adjusting the number of model layers and the number of neurons.

[0023] In some embodiments, the dialogue strategy parameters include at least response speed, intent recognition confidence, sentiment analysis sensitivity, and attention mechanism parameters;

[0024] The second optimization unit of the model is used to dynamically adjust the dialogue strategy parameters based on the performance evaluation results and the user feedback data corresponding to each user query data, and adjust the template matching logic of the customer service robot model in response to each user query data based on the user feedback data and the dialogue strategy parameters corresponding to each user query data.

[0025] In some embodiments, the model training module further includes an optimization response unit and an iterative optimization unit;

[0026] The optimized response unit is configured to respond to each user query data using the customer service robot model according to the optimized dialogue strategy parameters, response generation logic parameters, hyperparameters, and model structure, generate dialogue strategy adjustment data corresponding to each user query data, and obtain the dialogue strategy adjustment dataset, wherein the dialogue strategy adjustment dataset includes the dialogue strategy adjustment data corresponding to each user query data;

[0027] The iterative optimization unit is used to use the dialogue strategy adjustment data set to train the customer service robot model and iteratively optimize the model parameters and structure of the customer service robot model.

[0028] Another aspect of the present invention provides a model training method for a customer service robot, the method comprising the following steps:

[0029] Collect interaction data and pre-process the interaction data; the interaction data includes at least user voice input data, user text input data, user interaction feedback data, and conversation record data between the user and the customer service robot;

[0030] Constructing a conversation dataset with intent labels based on the interaction data;

[0031] The customer service robot model is trained using the dialogue dataset to generate a corresponding model response dataset. Parameters and structure of the customer service robot model are optimized based on the model response dataset, the interaction data, and the dialogue dataset. A corresponding dialogue strategy adjustment dataset is generated, and the customer service robot model is iteratively optimized using the dialogue strategy adjustment dataset.

[0032] Another aspect of an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the model training method described above.

[0033] The embodiments of the present application include at least the following beneficial effects: The present application provides a customer service robot model training system, method, and storage medium, which collects interaction data through a data acquisition module, pre-processes the interaction data, constructs a conversation dataset with intent labels based on the interaction data through a data processing module, trains a customer service robot model using the conversation dataset through a model training module, generates a corresponding model response dataset, optimizes the parameters and structure of the customer service robot model based on the model response dataset, interaction data, and conversation dataset, generates a corresponding conversation strategy adjustment dataset, and iteratively optimizes the customer service robot model using the conversation strategy adjustment dataset. The present application can realize automatic learning and optimization of the customer service robot model through machine learning and natural language processing technologies, enabling the customer service robot to continuously learn and adapt, provide more accurate responses and services, improve the accuracy and efficiency of question responses, and greatly enhance service quality and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a structural diagram of a model training system for a customer service robot provided in an embodiment of the present application;

[0035] Figure 2 This is another structural diagram of a model training system for a customer service robot provided in an embodiment of the present application;

[0036] Figure 3 This is a flowchart of a model training method for a customer service robot provided in an embodiment of the present application;

[0037] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0039] It should be noted that although the system diagrams illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the system or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0041] Before explaining the embodiments of the present application in detail, some of the nouns and terms involved in the embodiments of the present application are first explained. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0042] 1) A / B testing: This is an "a priori" experimental system that generates predictive conclusions, which differ significantly from "post-hoc" inductive conclusions. The goal of A / B testing is to obtain representative experimental conclusions through scientific experimental design, representative sampling, traffic segmentation, and small-volume testing, ensuring that these conclusions are reliable when generalized to the entire traffic flow.

[0043] 2) LLM: Large Language Model (LLM) refers to a deep learning model trained using large amounts of text data. It can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question-answering, and conversation, and are an important path to artificial intelligence. Currently, large language models use a similar Transformer architecture and pre-training objectives (such as language modeling) as small models. The difference from small models is the increased model size, training data, and computing resources.

[0044] Reference Figure 1 , Figure 1 This is an optional structural diagram of a model training system for a customer service robot provided in an embodiment of the present application. The device may include:

[0045] A data collection module is used to collect and pre-process interaction data, where the interaction data includes at least user voice input data, user text input data, user interaction feedback data, and conversation record data between the user and the customer service robot;

[0046] The data processing module is used to construct a conversation dataset with intent labels based on the interaction data;

[0047] The model training module is used to train the customer service robot model using the dialogue dataset, generate the corresponding model response dataset, optimize the parameters and structure of the customer service robot model based on the model response dataset, interaction data and dialogue dataset, and generate the corresponding dialogue strategy adjustment dataset. The dialogue strategy adjustment dataset is used to iteratively optimize the customer service robot model.

[0048] This application proposes a model training system for a customer service robot. This system aims to use machine learning to enable the robot to evolve, continuously optimize its conversational strategies and problem-solving capabilities, and provide more intelligent, efficient, and personalized customer service. The system includes key modules, including a data acquisition module, a data processing module, and a model training module. Through a cyclical and iterative model workflow, the system achieves continuous improvement and performance enhancements for the customer service robot.

[0049] In some embodiments, the data acquisition module includes a data collection unit and a data preprocessing unit.

[0050] A data collection unit, used for dynamically collecting interaction data;

[0051] The data preprocessing unit is used to perform data preprocessing on the interactive data, wherein the data preprocessing includes data cleaning and data formatting.

[0052] In some embodiments, optionally, the data collection unit collects each conversation between the customer service robot and the user, including user questions, robot responses, user subsequent behavior and user feedback, to obtain interaction data.

[0053] In some embodiments, the interaction data is cleaned by a data preprocessing unit to remove invalid or redundant data, and then the cleaned interaction data is formatted, including formatting operations such as standardization, encoding or dimensionality reduction, to convert the unstructured interaction data into an analyzable structured form for subsequent analysis of the interaction data.

[0054] In some embodiments, the conversation dataset includes multiple user query data and intent labels corresponding to each user query data, robot response data and user feedback data. The intent labels cover all intent categories that may appear in customer service scenarios, such as querying product information, seeking technical support, complaint feedback, etc.

[0055] Optionally, the data processing module uses natural language processing technology, including word embedding, intent recognition, sentiment analysis, etc., to process each user query data in the interaction data (including user voice data and user text data, etc.), identify the language pattern, intent and emotion corresponding to each user query data and generate a robot response, obtain the corresponding robot response data of each user query data and generate an intent label corresponding to each user query data.

[0056] In some embodiments, the model training module includes an intent recognition training unit and a response generation unit.

[0057] An intent recognition training unit is used to sequentially input each user's query data into the customer service robot model, perform intent recognition using the customer service robot model, learn intent recognition parameters and generate intent recognition data corresponding to each user's query data, and adjust the intent recognition parameters corresponding to the customer service robot model so that the intent recognition data of each user's query data corresponds to the intent label;

[0058] The response generation unit is used to use the customer service robot model to reply based on each user's query data and the intention recognition data corresponding to each user's query data, and generate model response data corresponding to each user's query data.

[0059] In some embodiments, the large model can significantly improve the accuracy and coverage of intent recognition due to its powerful generalization ability and ability to process large amounts of language data. Optionally, a pre-trained large model is selected as a basis, such as GPT-3, BERT, T5 or other LLM architectures. The selected large model can be further fine-tuned according to the corpus of a specific field to enable it to better understand user intent in customer service scenarios and obtain the above-mentioned customer service robot model.

[0060] In some embodiments, the model training module adopts a deep learning method and uses a conversation data set to train the customer service robot, learn and update model parameters, and improve the service quality of the customer service robot model and the accuracy of response to user input.

[0061] Specifically, each user query data is input into the customer service robot model in sequence through the above-mentioned intention recognition training unit, and the customer service robot model is used to perform intent recognition, learn and update the intention recognition parameters, so that each user query data is mapped to the corresponding intent label, and generate intent recognition data corresponding to the intent label of each user query data.

[0062] In some embodiments, the response generation unit uses the customer service robot model to respond based on each user's query data and the intention recognition data corresponding to each user's query data, generate model response data corresponding to each user's query data, and compare the model response data of each user's query data with the robot response data, and adjust other model parameters of the customer service robot model in combination with the preset loss function.

[0063] In some embodiments, the model training module further includes a first model optimization unit and a second model optimization unit;

[0064] The first model optimization unit is used to evaluate the model performance based on the model response data, robot response data, and user feedback data corresponding to each user query data, generate a model performance evaluation result, and optimize the hyperparameters and model structure corresponding to the customer service robot model based on the model performance evaluation result;

[0065] The second optimization unit of the model is used to optimize the dialogue strategy parameters and response generation logic parameters of the customer service robot model based on the model performance evaluation results and the user feedback data corresponding to each user query data.

[0066] In some embodiments, the model training module automatically adjusts the response strategy of the customer service robot model based on user feedback data and model performance evaluation results, thereby improving the conversation quality, problem-solving efficiency and user satisfaction of the customer service robot.

[0067] In some embodiments, optionally, the first optimization unit of the model is configured to obtain a plurality of performance evaluation indicators, determine a specific indicator value corresponding to each performance evaluation indicator based on the model response data, the robot response data, and the user feedback data corresponding to each user query data, and adjust the hyperparameters based on the specific indicator value corresponding to each performance evaluation indicator;

[0068] The first optimization unit of the model evaluates the effectiveness of the robot's current response strategy based on the model response data, robot response data, user feedback data and multiple performance evaluation indicators corresponding to each user query data. The performance evaluation indicators may include but are not limited to calculating the response accuracy, user satisfaction score, problem-solving speed and conversation length.

[0069] In some embodiments, the first optimization unit of the model adjusts the hyperparameters according to the specific values ​​of the indicators corresponding to each performance evaluation indicator. The hyperparameters may include but are not limited to learning rate and regularization parameters.

[0070] Exemplarily, the first optimization unit of the model adjusts the learning rate according to the specific values ​​of the indicators corresponding to each performance evaluation indicator so that the model converges more effectively. If the model performance evaluation results show that the loss function of the model decreases slowly during the training process, the learning rate of the model is increased; conversely, if the model performance evaluation results reflect that the model is overfitting, the learning rate of the model is reduced; the first optimization unit of the model can also adjust the regularization parameters (such as the coefficients of the L1 or L2 regularization terms) according to the model performance evaluation results to control the complexity of the model and prevent overfitting.

[0071] In some embodiments, the first optimization unit of the model is also used to dynamically adjust the model structure according to the specific indicator values ​​and hyperparameters corresponding to each performance evaluation indicator, wherein the dynamic adjustment of the model structure at least includes dynamically adjusting the number of model layers and the number of neurons.

[0072] Optionally, the first model optimization unit dynamically adjusts the number of model layers and the number of neurons according to the model performance evaluation results to find the optimal model complexity.

[0073] In some embodiments, the dialogue strategy parameters include at least response speed, intent recognition confidence, sentiment analysis sensitivity and attention mechanism parameters; the second optimization unit of the model is used to dynamically adjust the dialogue strategy parameters based on the performance evaluation results and the user feedback data corresponding to each user query data, and adjust the template matching logic of the customer service robot model in response to each user query data based on the user feedback data and dialogue strategy parameters corresponding to each user query data.

[0074] Optionally, the second optimization unit of the model can dynamically adjust the confidence of intent recognition. When the false positive rate is high, the confidence of intent recognition can be improved; conversely, if the missed detection rate is high, the confidence of intent recognition can be appropriately lowered. At the same time, the second optimization unit of the model can dynamically adjust the sensitivity of sentiment analysis to more accurately capture user emotions and improve the conversation experience. The second optimization unit of the model can also optimize the attention mechanism parameters to enhance the model's ability to capture important information and improve the pertinence and accuracy of replies. In addition, the second optimization unit of the model can also optimize the template matching logic used in response generation so that it can better adapt to the language habits and expressions of different users.

[0075] Exemplarily, the second optimization unit of the model modifies the dialogue strategy parameters based on the performance indicator of reply accuracy and the corresponding specific value of the indicator. When the specific value of the reply accuracy indicator is lower than the set threshold, the response strategy adjustment rule is triggered to improve the specific value of the reply accuracy indicator by adjusting the intent recognition confidence, sentiment analysis sensitivity, attention mechanism parameters or template matching logic. After adjusting the response strategy, the reply accuracy of the new and old strategies can be compared through A / B testing to verify whether the adjustment of the response strategy has improved the reply accuracy.

[0076] In some embodiments, the model training module performs incremental training using conversation data sets in small batches to gradually optimize the model performance and avoid performance fluctuations caused by one-time large-scale adjustments. Optionally, the model first optimization unit and the model second optimization unit process user feedback in real time and adjust the model parameters and model structure immediately without waiting for periodic retraining. Strategy adjustment is a continuous process. Based on the model learning results of each iteration, the dialogue strategy parameters are continuously optimized through the model second optimization unit, so that the customer service robot can provide more efficient, accurate and personalized services, while improving efficiency and reducing costs, thereby improving user satisfaction.

[0077] In some embodiments, the model training module further includes an optimization response unit and an iterative optimization unit.

[0078] An optimized response unit is configured to respond to each user query data using the customer service robot model according to the optimized dialogue strategy parameters, response generation logic parameters, hyperparameters, and model structure, generate dialogue strategy adjustment data corresponding to each user query data, and obtain a dialogue strategy adjustment dataset, wherein the dialogue strategy adjustment dataset includes the dialogue strategy adjustment data corresponding to each user query data;

[0079] The iterative optimization unit is used to use the dialogue strategy to adjust the data set to train the customer service robot model, and iteratively optimize the model parameters and structure of the customer service robot model.

[0080] In some embodiments, the above model training system may further include:

[0081] Knowledge base module, used to store dialogue strategy adjustment dataset;

[0082] The model interaction module is used to provide a model calling interface, through which the customer service robot model is applied to the target customer service robot.

[0083] In some embodiments, the knowledge base module stores information that can improve service quality. Optionally, the knowledge base module automatically updates the knowledge base content, stores each user's query data and corresponding dialogue strategy adjustment data to reflect newly learned information and solutions to common problems, for subsequent iterative optimization of the customer service robot model.

[0084] In some embodiments, reference Figure 2 , Figure 2This is another optional structural diagram of a model training system for a customer service robot provided in an embodiment of the present application, which includes a data acquisition module, a data processing module, a model training module, a knowledge base module and a model interaction module. The data acquisition module is used to collect interaction data and pre-process the interaction data. The data processing module is used to process the interaction data using natural language processing technology and construct a dialogue data set with intent labels. The model training module is used to use the dialogue data set to train and iteratively optimize the customer service robot model, adjust model parameters and model structure, and generate a dialogue strategy adjustment data set. The knowledge base module is used to store the dialogue strategy adjustment data set and return small batches of dialogue strategy adjustment data sets to the model training module. The model training module uses the dialogue strategy adjustment data set to continuously train and optimize the customer service robot model. The model interaction module is used to provide a model call interface and apply the optimized customer service robot model to the target customer service robot.

[0085] Reference Figure 3 , Figure 3 This is an optional flowchart of a model training method for a customer service robot provided in an embodiment of the present application. The method may include but is not limited to steps S101 to S103:

[0086] Step S101: Collect interaction data and pre-process the interaction data, wherein the interaction data at least includes user voice input data, user text input data, user interaction feedback data, and conversation record data between the user and the customer service robot;

[0087] Step S102: constructing a conversation dataset with intent labels based on the interaction data;

[0088] Step S103: Use the dialogue dataset to train the customer service robot model and generate a corresponding model response dataset. Based on the model response dataset, interaction data, and dialogue dataset, perform parameter optimization and structural optimization on the customer service robot model, generate a corresponding dialogue strategy adjustment dataset, and use the dialogue strategy adjustment dataset to iteratively optimize the customer service robot model.

[0089] It can be understood that the contents of the above system embodiments are applicable to the present method embodiments, the functions specifically implemented by the present method embodiments are the same as those of the above system embodiments, and the beneficial effects achieved are also the same as those achieved by the above system embodiments.

[0090] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned model training method when executing the computer program. The electronic device can be any smart terminal including a tablet computer.

[0091] It can be understood that the contents of the above method embodiments are applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0092] See also Figure 4 , Figure 4 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0093] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0094] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the model training method of the embodiment of the present application;

[0095] Input / output interface 903, used to implement information input and output;

[0096] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0097] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0098] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0099] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned model training method is implemented.

[0100] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiment, the functions specifically implemented by the present storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0101] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0102] The embodiments of the present application provide a customer service robot model training system, method, and storage medium. The system collects interaction data through a data acquisition module, pre-processes the interaction data, constructs a conversation dataset with intent labels based on the interaction data through a data processing module, and trains a customer service robot model using the conversation dataset through a model training module to generate a corresponding model response dataset. The system then optimizes the parameters and structure of the customer service robot model based on the model response dataset, interaction data, and conversation dataset, generates a corresponding conversation strategy adjustment dataset, and iteratively optimizes the customer service robot model using the conversation strategy adjustment dataset. The embodiments of the present application can achieve automatic learning and optimization of the customer service robot model through machine learning and natural language processing technologies, enabling the customer service robot to continuously learn and adapt, provide more accurate responses and services, improve the accuracy and efficiency of question responses, and significantly enhance service quality and user satisfaction.

[0103] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0104] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0105] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0106] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0107] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0108] It should be appreciated that embodiments of the present invention may be implemented or practiced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The methods may be implemented in a computer program using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program may be implemented in assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed application-specific integrated circuit for this purpose.

[0109] If the integrated unit is implemented in the form of a software functional unit 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 is essentially or the part that contributes to the prior art or all or 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 multiple 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 methods of various embodiments 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 programs.

[0110] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A model training system for a customer service robot, characterized in that: The system comprises: A data collection module is used to collect interaction data and pre-process the interaction data; the interaction data at least includes user voice input data, user text input data, user interaction feedback data and conversation record data between the user and the customer service robot; A data processing module, used to construct a dialogue data set with intent labels based on the interaction data; A model training module is used to train a customer service robot model using the dialogue data set to generate a corresponding model response data set, perform parameter optimization and structural optimization on the customer service robot model according to the model response data set, the interaction data and the dialogue data set, and generate a corresponding dialogue strategy adjustment data set, and iteratively optimize the customer service robot model using the dialogue strategy adjustment data set; The dialogue data set includes multiple user query data and intent labels corresponding to each of the user query data, robot response data and user feedback data; the model training module includes an intent recognition training unit and a response generation unit; The intention recognition training unit is used to input each of the user query data into the customer service robot model in sequence, use the customer service robot model to perform intention recognition, learn intention recognition parameters and generate intention recognition data corresponding to each of the user query data, and adjust the intention recognition parameters corresponding to the customer service robot model so that the intention recognition data of each of the user query data corresponds to the intention label; The response generation unit is used to use the customer service robot model to respond according to each user query data and the intention recognition data corresponding to each user query data, and generate model response data corresponding to each user query data; The model training module also includes a first model optimization unit and a second model optimization unit; The first model optimization unit is used to perform model performance evaluation according to the model response data, the robot response data and the user feedback data corresponding to each of the user query data, generate a model performance evaluation result, and optimize the hyperparameters and model structure corresponding to the customer service robot model according to the model performance evaluation result; The second model optimization unit is used to optimize the dialogue strategy parameters and response generation logic parameters of the customer service robot model according to the model performance evaluation result and the user feedback data corresponding to each of the user query data; The first optimization unit of the model is used to obtain multiple performance evaluation indicators, determine the specific value of the indicator corresponding to each performance evaluation indicator according to the model response data, the robot response data and the user feedback data corresponding to each user query data, and adjust the hyperparameter according to the specific value of the indicator corresponding to each performance evaluation indicator; The first optimization unit of the model is further used to dynamically adjust the model structure according to the specific value of the indicator corresponding to each of the performance evaluation indicators and the hyperparameter, wherein the dynamic adjustment of the model structure at least includes dynamically adjusting the number of model layers and the number of neurons; The dialogue strategy parameters include at least response speed, intent recognition confidence, sentiment analysis sensitivity and attention mechanism parameters; The second optimization unit of the model is used to dynamically adjust the dialogue strategy parameters according to the performance evaluation result and the user feedback data corresponding to each user query data, and adjust the template matching logic of the customer service robot model in response to each user query data according to the user feedback data corresponding to each user query data and the dialogue strategy parameters; The model training module also includes an optimization response unit and an iterative optimization unit; The optimized response unit is used to respond to each of the user query data using the customer service robot model according to the optimized dialogue strategy parameters, response generation logic parameters, hyperparameters and model structure, generate dialogue strategy adjustment data corresponding to each of the user query data and obtain the dialogue strategy adjustment data set, wherein the dialogue strategy adjustment data set includes the dialogue strategy adjustment data corresponding to each of the user query data; The iterative optimization unit is used to use the dialogue strategy to adjust the data set to train the customer service robot model, and iteratively optimize the model parameters and structure of the customer service robot model.

2. The model training system for a customer service robot according to claim 1, characterized in that: The system further comprises: A knowledge base module, used to store the dialogue strategy adjustment data set; The model interaction module is used to provide a model calling interface, through which the customer service robot model is applied to the target customer service robot.

3. The model training system for a customer service robot according to claim 1, characterized in that: The data acquisition module includes a data collection unit and a data preprocessing unit; The data collection unit is used to dynamically collect the interaction data; The data preprocessing unit is used to perform data preprocessing on the interaction data, wherein the data preprocessing includes data cleaning and data formatting.

4. A model training method for a customer service robot, characterized in that: The method comprises the following steps: Collect interaction data and pre-process the interaction data; the interaction data at least includes user voice input data, user text input data, user interaction feedback data and conversation record data between the user and the customer service robot; Constructing a conversation dataset with intent labels based on the interaction data; Using the dialogue data set to train a customer service robot model, generating a corresponding model response data set, performing parameter optimization and structural optimization on the customer service robot model according to the model response data set, the interaction data, and the dialogue data set, and generating a corresponding dialogue strategy adjustment data set, and using the dialogue strategy adjustment data set to iteratively optimize the customer service robot model; The dialogue data set includes multiple user query data and intent labels corresponding to each of the user query data, robot response data and user feedback data; the model training module includes an intent recognition training unit and a response generation unit; The intention recognition training unit is used to input each of the user query data into the customer service robot model in sequence, use the customer service robot model to perform intention recognition, learn intention recognition parameters and generate intention recognition data corresponding to each of the user query data, and adjust the intention recognition parameters corresponding to the customer service robot model so that the intention recognition data of each of the user query data corresponds to the intention label; The response generation unit is used to use the customer service robot model to respond according to each user query data and the intention recognition data corresponding to each user query data, and generate model response data corresponding to each user query data; The model training module also includes a first model optimization unit and a second model optimization unit; The first model optimization unit is used to perform model performance evaluation according to the model response data, the robot response data and the user feedback data corresponding to each of the user query data, generate a model performance evaluation result, and optimize the hyperparameters and model structure corresponding to the customer service robot model according to the model performance evaluation result; The second model optimization unit is used to optimize the dialogue strategy parameters and response generation logic parameters of the customer service robot model according to the model performance evaluation result and the user feedback data corresponding to each of the user query data; The first optimization unit of the model is used to obtain multiple performance evaluation indicators, determine the specific value of the indicator corresponding to each performance evaluation indicator according to the model response data, the robot response data and the user feedback data corresponding to each user query data, and adjust the hyperparameter according to the specific value of the indicator corresponding to each performance evaluation indicator; The first optimization unit of the model is further used to dynamically adjust the model structure according to the specific value of the indicator corresponding to each of the performance evaluation indicators and the hyperparameter, wherein the dynamic adjustment of the model structure at least includes dynamically adjusting the number of model layers and the number of neurons; The dialogue strategy parameters include at least response speed, intent recognition confidence, sentiment analysis sensitivity and attention mechanism parameters; The second optimization unit of the model is used to dynamically adjust the dialogue strategy parameters according to the performance evaluation result and the user feedback data corresponding to each user query data, and adjust the template matching logic of the customer service robot model in response to each user query data according to the user feedback data corresponding to each user query data and the dialogue strategy parameters; The model training module also includes an optimization response unit and an iterative optimization unit; The optimized response unit is used to respond to each of the user query data using the customer service robot model according to the optimized dialogue strategy parameters, response generation logic parameters, hyperparameters and model structure, generate dialogue strategy adjustment data corresponding to each of the user query data and obtain the dialogue strategy adjustment data set, wherein the dialogue strategy adjustment data set includes the dialogue strategy adjustment data corresponding to each of the user query data; The iterative optimization unit is used to use the dialogue strategy to adjust the data set to train the customer service robot model, and iteratively optimize the model parameters and structure of the customer service robot model.

5. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the model training method described in claim 4 is implemented.

Images