Intelligent customer service method and device, electronic equipment and storage medium

Abstract

The invention relates to an intelligent customer service method, and aims to provide personalized and accurate customer service reply through advanced technical means. The method comprises the following steps: firstly, acquiring and analyzing a query request of a user regardless of the form of text, voice or image; thirdly, intention recognition is conducted through a preset response model, a user portrait is generated in combination with historical interaction data of the user, and customer service requirements are predicted according to the user portrait. And by correcting the prediction intention data and considering the user preference, an accurate customer service reply is generated. In addition, the system also collects subsequent behavior data of the user, constructs a service scoring data set, and continuously optimizes the service model by using a machine self-learning algorithm. According to the method, the user satisfaction can be effectively improved, the click rate and the conversion rate are increased, the response time is shortened, the problem solving rate is improved, and continuous service optimization and performance improvement are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent online customer service, and in particular to an intelligent customer service method, device, electronic equipment and storage medium. Background Art

[0002] In the digital age, intelligent customer service systems have become a key tool for companies to improve customer service quality and reduce costs.

[0003] Existing intelligent customer service systems generally provide users with choices through a preset question library, and provide answers to the selected questions based on the corresponding answer database. Existing technology requires users to make multiple choices to understand the answers they need, or when the question library is not updated, users cannot even select the corresponding questions to solve their needs, which reduces the efficiency of problem solving when users seek customer service help, thereby reducing the user's consulting user experience, and therefore needs to be improved. Summary of the invention

[0004] In order to improve the user's consulting user experience, the present application provides an intelligent customer service method, device, electronic device and storage medium.

[0005] In the first aspect, the above-mentioned invention objective of the present application is achieved through the following technical solutions:

[0006] An intelligent customer service method, the method comprising the steps of:

[0007] Obtaining customer service request and query request information of the user, parsing the query request information, and determining the type of request information, wherein the query request information includes text information, voice information, and image information;

[0008] The preset response model uses a corresponding information processing technology based on the type of the request information to perform intent recognition on the query request letter to generate first predicted intent data;

[0009] Obtaining historical interaction data of the user, and analyzing the historical interaction data to generate a user preference profile and a user behavior profile, wherein the historical interaction data includes the user's previous query records, purchase history, and preference settings;

[0010] generating predicted customer service needs based on the user preference profile and the user behavior profile, wherein the predicted customer service needs include the user's potential inquiry intention, potential inquiry scope, and the user's language habits;

[0011] modifying the first predicted intention data based on the predicted customer service demand to generate second predicted intention data;

[0012] generating a customer service response based on the second predicted intent data and the user preference profile, wherein the customer service response includes answers to questions and product recommendations;

[0013] Obtaining subsequent user behavior after the user receives the customer service reply, and generating basic service rating data based on the subsequent user behavior, wherein the basic service rating data includes satisfaction, click-through rate, conversion rate, response time, and problem resolution rate;

[0014] Record the service rating basic data corresponding to all customer service services and construct a service rating data set;

[0015] The pre-set service optimization model analyzes the service rating data set based on a machine self-learning algorithm, generates a service optimization strategy to optimize the response model, thereby improving accuracy and user satisfaction, and the optimization includes model training, parameter adjustment and feedback loop.

[0016] By adopting the above technical solution, more personalized and accurate customer service responses are provided. It can process multimodal user input including text, voice and images, ensuring the comprehensiveness and adaptability of the service. Combined with the user's historical interaction data and behavioral portraits, this method can provide highly personalized services to meet the unique needs of different users. In addition, through machine self-learning algorithms, the intelligent customer service system can continuously self-learn and optimize, thereby improving service accuracy and user satisfaction. Comprehensive data set construction and real-time feedback loops enable the system to respond quickly to market changes and adjust service strategies in a timely manner to adapt to the evolution of user needs. These advantages not only improve user satisfaction and customer service experience, increase click-through rate and conversion rate, shorten response time, and improve problem resolution rate, but also enable the intelligent customer service system to continuously adapt to changes in user needs and provide more efficient services.

[0017] In a preferred example, the present application can be further configured as follows: in the step of analyzing the service scoring data set based on the machine self-learning algorithm in the preset service optimization model to generate a service optimization strategy to optimize the response model, the service optimization model is preset with a service self-scoring formula:

[0018] S=α×F+β×C+γ×R+δ×T+ε×P+θ×E, α+β+γ+δ+ε+θ=1;

[0019] Where S is the self-scoring value, which is used to measure the overall performance of the intelligent customer service system;

[0020] F is the user satisfaction score, which represents the user's direct rating of the service response;

[0021] C is the click-through rate, which indicates the degree of user interaction with the recommended content;

[0022] R is the conversion rate, which indicates the proportion of users who take actions such as adding items to cart and paying for purchases after receiving recommendations.

[0023] T is the response time, which is the average time required for the system to respond to user requests;

[0024] P is the problem solving rate, which is the proportion of user problems that are successfully solved;

[0025] E is the result of sentiment analysis, which indicates the degree of positivity or negativity of the sentiment tendency in the user conversation.

[0026] By adopting the above technical solutions and collecting user feedback and evaluations, the intelligent customer service system can understand the satisfaction and dissatisfaction of the service, discover existing problems and room for improvement, and timely adjust and optimize the service strategy. Supervised learning technology can use user feedback data to train and optimize the model of the intelligent customer service system to improve its accuracy and efficiency. The self-learning scoring formula can continuously adjust and optimize services based on user behavior and feedback to provide more personalized services; by analyzing user behavior and optimizing replies, the system can allocate resources more effectively and improve service efficiency. The self-learning scoring formula adapts to changes in user needs through continuous self-improvement, provides more efficient and personalized services, and enhances user experience.

[0027] In a preferred example, the present application may be further configured as follows: after obtaining the subsequent user behavior of the user after receiving the customer service reply and generating the service rating basic data according to the subsequent user behavior, the following steps are included:

[0028] Classifying the subsequent user behavior to mark the type of the subsequent user behavior;

[0029] Analyze the subsequent user behavior to determine whether the current response meets the user's needs;

[0030] If the current reply is judged to not meet the user's needs, the current reply will be marked as an abnormal reply;

[0031] Associating the abnormal response with the corresponding subsequent user type to construct an abnormality-demand data set;

[0032] The preset abnormal response analysis model analyzes the abnormal-demand data set based on a machine self-learning algorithm to generate the difference between the response and the demand, thereby optimizing the current response to meet the actual needs of the user.

[0033] By adopting the above technical solution, the intelligent customer service system can automatically detect and optimize its services, improve the accuracy of responses and user satisfaction. This method can achieve continuous self-improvement, reduce users' problem-solving time, and reduce the company's customer service costs. In addition, by analyzing user behavior and optimizing responses, the system can provide more personalized services and enhance user experience. Compared with the prior art, the intelligent customer service method of this application can more accurately identify and respond to users' actual needs, and continuously optimize service strategies through machine self-learning algorithms.

[0034] In a preferred example, the present application can be further configured as follows: in the step of analyzing the subsequent user behavior to determine whether the current reply meets the user's needs, the steps are as follows:

[0035] Obtaining subsequent user behavior types, where the subsequent user behavior types include changing the topic, expressing gratitude, expressing dissatisfaction, and clicking on a recommended link;

[0036] Subsequent user behavior types are classified according to pre-set judgment criteria to determine whether the current response meets user needs.

[0037] By adopting the above technical solution, by analyzing the user's subsequent behavior, the intelligent customer service system can determine whether its reply is accurate, so as to adjust the reply strategy in time, improve the accuracy of the response, and promptly identify and solve the points of user dissatisfaction, which can significantly improve customer satisfaction and loyalty. Through continuous monitoring and analysis of user behavior, the intelligent customer service system can continuously optimize its service process and reduce invalid or unsatisfactory interactions; reduce operating costs: automated feedback analysis reduces the need for manual intervention, thereby reducing the company's operating costs; by understanding user behavior and preferences, the intelligent customer service system can provide more personalized services, thereby improving user experience. The intelligent customer service system can learn from each interaction and continuously improve its algorithms and knowledge base to provide better services.

[0038] In a preferred example, the present application can be further configured as follows: in the step where the preset abnormal response analysis model analyzes the abnormal-demand data set based on the machine self-learning algorithm to generate the difference between the response and the demand, thereby optimizing the current response to meet the actual needs of the user, the steps are as follows:

[0039] Obtaining the subsequent user behavior and current response corresponding to the abnormal-demand data set, and extracting the reply features of the current response and the demand features of the subsequent user behavior;

[0040] Comparing the response features with the requirement features to generate difference features;

[0041] Performing analysis based on the difference characteristics and the answer characteristics to generate answer data that meets the required characteristics;

[0042] The reply data is associated with the query request information to update a reply database, wherein the reply database is used by the response model to generate first predicted intent data based on the query request information.

[0043] By adopting the above technical solutions, the intelligent customer service system can more accurately understand the actual needs of users and provide responses that are more in line with these needs. The self-optimization process improves the accuracy of responses and user satisfaction, reduces users' dependence on manual customer service, and thus reduces the company's operating costs.

[0044] In a preferred example, the present application can be further configured to provide multi-language support to serve users of different languages, wherein the multi-language support includes automatic language recognition and translation.

[0045] In a preferred example, the present application can be further configured as follows: the machine self-learning algorithm includes a feature extraction algorithm and a clustering algorithm.

[0046] In the second aspect, the above invention objective of the present application is achieved through the following technical solutions:

[0047] An intelligent customer service device, the device comprising:

[0048] A user request parsing unit, used to obtain a user's customer service request and query request information, parse the query request information, and determine the type of request information, wherein the query request information includes text information, voice information, and image information;

[0049] A first predicted intention data generating unit, configured to pre-set a response model to perform intention recognition on the query request letter based on the type of the request information using a corresponding information processing technology to generate first predicted intention data;

[0050] A user profile generation unit, configured to obtain historical interaction data of a user and analyze the historical interaction data to generate a user preference profile and a user behavior profile, wherein the historical interaction data includes the user's previous query records, purchase history, and preference settings;

[0051] A predicted customer service demand generating unit, configured to generate predicted customer service demand based on the user preference profile and the user behavior profile, wherein the predicted customer service demand includes the user's potential inquiry intention, potential inquiry scope, and the user's language habits;

[0052] a second predicted intention data generating unit, configured to modify the first predicted intention data based on the predicted customer service demand to generate second predicted intention data;

[0053] A customer service response generating unit, configured to generate a customer service response based on the second predicted intention data and the user preference profile, wherein the customer service response includes an answer to a question and a product recommendation;

[0054] A service rating basic data generating unit, used to obtain the subsequent user behavior of the user after receiving the customer service reply, and generate service rating basic data according to the subsequent user behavior, wherein the service rating basic data includes satisfaction, click-through rate, conversion rate, response time and problem resolution rate;

[0055] A service score data set construction unit, used to record the service score basic data corresponding to all customer service services and construct a service score data set;

[0056] The service optimization strategy generating unit is used to analyze the service rating data set based on the machine self-learning algorithm with a pre-set service optimization model, and generate a service optimization strategy to optimize the response model.

[0057] On the third aspect, the above-mentioned purpose of the present application is achieved through the following technical solutions:

[0058] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned intelligent customer service method when executing the computer program.

[0059] Fourthly, the above-mentioned purpose of the present application is achieved through the following technical solutions:

[0060] A computer-readable storage medium stores a computer program, which implements the steps of the above-mentioned intelligent customer service method when executed by a processor.

[0061] In summary, the present application includes at least one of the following beneficial technical effects:

[0062] 1. It provides more personalized and accurate customer service responses. It can process multimodal user input including text, voice and images, ensuring the comprehensiveness and adaptability of the service. Combined with the user's historical interaction data and behavior portraits, this method can provide highly personalized services to meet the unique needs of different users. In addition, through machine self-learning algorithms, the intelligent customer service system can continuously self-learn and optimize, thereby improving the accuracy of services and user satisfaction. Comprehensive data set construction and real-time feedback loops enable the system to respond quickly to market changes and adjust service strategies in a timely manner to adapt to the evolution of user needs. These advantages not only improve user satisfaction and customer service experience, increase click-through rate and conversion rate, shorten response time, and improve problem resolution rate, but also enable the intelligent customer service system to continuously adapt to changes in user needs and provide more efficient services;

[0063] 2. By collecting user feedback and comments, the intelligent customer service system can understand the satisfaction and dissatisfaction of the service, discover existing problems and room for improvement, and adjust and optimize the service strategy in a timely manner. The supervised learning technology can use user feedback data to train and optimize the model of the intelligent customer service system to improve its accuracy and efficiency. The self-learning scoring formula can continuously adjust and optimize the service according to the user's behavior and feedback to provide more personalized services; by analyzing user behavior and optimizing replies, the system can allocate resources more effectively and improve service efficiency. The self-learning scoring formula adapts to changes in user needs through continuous self-improvement, provides more efficient and personalized services, and enhances user experience;

[0064] 3. The intelligent customer service system can automatically detect and optimize its services, improve the accuracy of responses and user satisfaction. This method can achieve continuous self-improvement, reduce user problem solving time, and reduce the customer service costs of enterprises. In addition, by analyzing user behavior and optimizing responses, the system can provide more personalized services and enhance user experience. Compared with the prior art, the intelligent customer service method of this application can more accurately identify and respond to users' actual needs, and continuously optimize service strategies through machine self-learning algorithms;

[0065] 4. By analyzing the user's subsequent behavior, the intelligent customer service system can determine whether its reply is accurate, so as to adjust the reply strategy in time, improve the accuracy of the response, and promptly identify and resolve points of user dissatisfaction, which can significantly improve customer satisfaction and loyalty. Through continuous monitoring and analysis of user behavior, the intelligent customer service system can continuously optimize its service process and reduce invalid or unsatisfactory interactions; reduce operating costs: automated feedback analysis reduces the need for manual intervention, thereby reducing the company's operating costs; by understanding user behavior and preferences, the intelligent customer service system can provide more personalized services, thereby improving the user experience. The intelligent customer service system can learn from each interaction and continuously improve its algorithms and knowledge base to provide better services. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flowchart of the steps of an intelligent customer service method in one embodiment of the present application;

[0067] Figure 2 This is a principle block diagram of an intelligent customer service device in one embodiment of the present application;

[0068] Figure 3 It is a schematic diagram of an electronic device in an embodiment of the present application.

[0069] Figure Number:

[0070] 1. User request parsing unit; 2. First prediction intention data generation unit; 3. User portrait generation unit; 4. Prediction customer service demand generation unit; 5. Second prediction intention data generation unit; 6. Customer service reply generation unit; 7. Service rating basic data generation unit; 8. Service rating data set construction unit; 9. Service optimization strategy generation unit. DETAILED DESCRIPTION

[0071] The present application is further described in detail below in conjunction with the accompanying drawings.

[0072] In one embodiment, if Figure 1 As shown, the present application discloses an intelligent customer service method, which specifically includes the following steps:

[0073] S10: Obtaining the customer service request and query request information of the user, parsing the query request information, and determining the type of the request information;

[0074] The query request information includes text information, voice information and image information;

[0075] Specifically, the user submitted a text query through the intelligent customer service system: "I want to buy a SLR camera suitable for beginners, do you have any recommendations?"

[0076] The system recognizes that this is a product recommendation request, which belongs to the "product consultation" type in the query request information.

[0077] It should be noted that multi-language support is provided to serve users of different languages, and the multi-language support includes automatic language recognition and translation.

[0078] S20: The preset response model uses a corresponding information processing technology based on the type of the request information to perform intent recognition on the query request letter to generate first predicted intent data;

[0079] Specifically, the response model uses natural language processing technology to parse the user query and recognizes that the user's intention is to find a SLR camera suitable for beginners, thereby generating the first predicted intent data, labeled as "SLR camera recommendation."

[0080] S30: Acquire historical interaction data of the user, and analyze the historical interaction data to generate a user preference profile and a user behavior profile;

[0081] The historical interaction data includes the user's previous query records, purchase history and preference settings. Specifically, by obtaining the user's historical purchase records, it is found that the user has purchased photography books and entry-level tripods, thus generating a user preference profile, such as interest in photography and preference for entry-level products; and a user behavior profile, such as frequently browsing photography-related products at night.

[0082] S40: generating a predicted customer service demand based on the user preference profile and the user behavior profile, wherein the predicted customer service demand includes the user's potential inquiry intention, potential inquiry scope, and the user's language habits;

[0083] Specifically, based on user preferences and behavioral portraits, we predict that users may be interested in photography skills and are accustomed to using professional terms. Predicted customer service needs include asking about photography skills, looking for entry-level photography equipment, and preferring consultation in the evening.

[0084] S50: modifying the first predicted intention data based on the predicted customer service demand to generate second predicted intention data;

[0085] Specifically, based on the predicted customer service needs, the first predicted intent data is modified, and the information that the user may be interested in photography skills is added, thereby generating the second predicted intent data, including "SLR camera recommendation" and "photography skills consultation."

[0086] S60: Generate a customer service response based on the second predicted intention data and the user preference profile, wherein the customer service response includes a question answer and a product recommendation;

[0087] Specifically, based on the second predicted intent data and the user preference profile, the intelligent customer service system replies to the user: “Hello, for beginners, I recommend this camera (with a purchase link), which is easy to operate and suitable for beginners. At the same time, here are some resources on photography skills, I hope they will be helpful to you.

[0088] S70: Obtaining subsequent user behavior of the user after receiving the customer service reply, and generating service rating basic data according to the subsequent user behavior;

[0089] The basic data for service rating include satisfaction, click-through rate, conversion rate, response time, problem solving rate and sentiment analysis results. Specifically, the user was satisfied with the reply, clicked on the recommended SLR camera link, and finally completed the purchase. By analyzing the customer's behavior, it was concluded that the basic data for service rating include high satisfaction, high click-through rate, conversion rate, response time and problem solving rate.

[0090] S80: Record the service score basic data corresponding to all customer service services and construct a service score data set;

[0091] Specifically, all the basic scoring data of this service are recorded and added to the service scoring dataset to construct a dataset containing multiple service instances for subsequent analysis.

[0092] S90: The preset service optimization model analyzes the service rating data set based on a machine self-learning algorithm to generate a service optimization strategy to optimize the response model;

[0093] Wherein, the optimization includes model training, parameter adjustment and feedback loop to improve accuracy and user satisfaction;

[0094] Specifically, the service optimization model analyzes the service rating dataset and finds that the user is less satisfied with the photography skills consultation, and generates a service optimization strategy for the user, such as enhancing the quality of responses to photography skills consultation and adjusting the product recommendation algorithm.

[0095] In summary, the intelligent customer service method of this embodiment provides more personalized and accurate customer service responses through advanced technical means compared to the prior art. It can process multimodal user input including text, voice and image, ensuring the comprehensiveness and adaptability of the service. Combined with the user's historical interaction data and behavior portrait, the method can provide highly personalized services to meet the unique needs of different users. In addition, through the machine self-learning algorithm, the intelligent customer service system can continuously self-learn and optimize, thereby improving the accuracy of the service and user satisfaction. Comprehensive data set construction and real-time feedback loops enable the system to respond quickly to market changes and adjust service strategies in a timely manner to adapt to the evolution of user needs. These advantages not only improve user satisfaction and customer service experience, increase click-through rate and conversion rate, shorten response time, and improve problem resolution rate, but also enable the intelligent customer service system to continuously adapt to changes in user needs and provide more efficient services.

[0096] It should be noted that in step S90: the preset service optimization model analyzes the service scoring data set based on the machine self-learning algorithm to generate a service optimization strategy to optimize the response model, the service optimization model presets a service self-scoring formula:

[0097] S=α×F+β×C+γ×R+δ×T+ε×P+θ×E, α+β+γ+δ+ε+θ=1;

[0098] Where S is the self-scoring value, which is used to measure the overall performance of the intelligent customer service system;

[0099] F is the user satisfaction score, which represents the user's direct rating of the service response;

[0100] C is the click-through rate, which indicates the degree of user interaction with the recommended content;

[0101] R is the conversion rate, which indicates the proportion of users who take actions such as adding items to cart and paying for purchases after receiving recommendations.

[0102] T is the response time, which is the average time required for the system to respond to user requests;

[0103] P is the problem solving rate, which is the proportion of user problems that are successfully solved;

[0104] E is the result of sentiment analysis, which indicates the degree of positivity or negativity of the sentiment tendency in the user conversation.

[0105] Among them, α is the weight value corresponding to the user satisfaction score, β is the weight value corresponding to the click rate, γ is the weight value corresponding to the conversion rate, δ is the weight value corresponding to the response time, ε is the weight value corresponding to the problem solving rate, and θ is the weight value corresponding to the sentiment analysis result.

[0106] Subsequent applications of the formula score include:

[0107] Performance evaluation: The self-learning scoring formula can be used to regularly evaluate the performance of the intelligent customer service system. By calculating the score, you can understand the performance of the system in different time periods and compare it with the preset performance goals.

[0108] Problem diagnosis: If the score is lower than expected, you can analyze the various influencing factors in depth to identify the reasons for the drop in score. For example, if the problem resolution rate P is low, the knowledge and skills of the customer service system may need to be updated or optimized.

[0109] Continuous optimization: Based on the analysis results of scores and influencing factors, the intelligent customer service system can be continuously optimized. For example, if the response time T is long, you can consider improving the system's processing capacity or optimizing the algorithm.

[0110] Resource allocation: Self-learning scoring can also help companies make smarter resource allocation decisions in marketing and customer service. For example, if the conversion rate R is high, it may be necessary to increase investment in this part of resources to further improve performance.

[0111] Model training and adjustment: In machine learning models, self-learning scores can be used as feedback signals to adjust and train models to better adapt to user needs and market changes.

[0112] Reporting and presentation: The results of self-learning scoring can be used to generate reports and presentations to provide decision support for management. They can also be used to demonstrate the efficiency and effectiveness of the intelligent customer service system to customers.

[0113] Furthermore, the calculation of the sentiment analysis result E can be achieved by a variety of methods, including methods based on sentiment dictionaries, machine learning models, and deep learning models in the embodiments of the present application. The following are several methods for calculating sentiment analysis results:

[0114] Method based on sentiment dictionary: This method uses a pre-defined sentiment dictionary to assign a sentiment value to each sentiment word. By analyzing the vocabulary in the text, the sentiment tendency of the entire text can be calculated. For example, the Dalian University of Technology Sentiment Dictionary is a commonly used sentiment dictionary, which contains sentiment words and their corresponding sentiment values. When calculating the sentiment score, you can traverse each word in the text. If the word is in the sentiment dictionary, the final sentiment score is adjusted according to its sentiment value and context (such as negative words and degree adverbs). The key to this method is how to accurately handle negative words and degree adverbs, and how to summarize the sentiment values ​​of individual words into the sentiment score of the entire text.

[0115] Machine learning models: Machine learning models, such as support vector machines (SVM), logistic regression (LR), naive Bayes (NB), and artificial neural networks (ANN), can be used for sentiment analysis. These models usually require a large amount of labeled data to train, and can then predict the sentiment of new text. For example, you can use TF-IDF vectorization to process text data, and then use the SVM model for sentiment classification.

[0116] Deep learning models: Deep learning models, such as convolutional neural networks (CNN) and recurrent neural networks (RNN), can be used for more complex sentiment analysis tasks. These models are able to capture deep features and contextual information in the text, thereby improving the accuracy of sentiment analysis.

[0117] VADER (Valence Aware Dictionary and Entity Reasoner): VADER is a dictionary-based sentiment analysis tool that assigns a fixed sentiment weight to each word. VADER takes into account the strength and context of words and is able to handle modern texts including emoticons and Internet slang. VADER's sentiment scores range from -1 (very negative) to +1 (very positive), with scores closer to 1 or -1 indicating stronger sentiment.

[0118] In practical applications, sentiment analysis results E can be used in many ways, such as:

[0119] Sentiment monitoring: Monitor customer sentiment towards products or services so you can respond in a timely manner.

[0120] Market Analysis: Analyze consumers’ emotional responses to market trends to guide product development and marketing strategies.

[0121] Customer Service: Automatically categorize and prioritize customer inquiries based on sentiment to improve customer satisfaction.

[0122] Content recommendation: Recommend relevant content based on user sentiment feedback to increase user engagement.

[0123] The weight coefficients α, β, γ, δ, ε, and θ represent the degree of influence of different feedback indicators on the total score. The selection of these weight coefficients can be based on a variety of criteria, including but not limited to:

[0124] Business goal: The weight should reflect the priority of the business goal. For example, if the business goal is to improve user satisfaction, the weight of the user satisfaction score may be set higher.

[0125] User behavior analysis: By analyzing user behavior data, you can determine which indicators are more relevant to user retention and conversion rate, and thus assign higher weights to these indicators.

[0126] Historical data and statistical analysis: By analyzing historical data, you can determine which indicators have had the greatest impact on the business in the past and assign higher weights to these indicators.

[0127] Expert opinion: The opinions of industry experts or data scientists can be used to determine the relative importance of different metrics.

[0128] Experiments and A / B testing: Through experiments and A / B testing, you can test the impact of different weight configurations on business indicators and optimize the weight settings.

[0129] Market research: Market research can provide information about user preferences and industry standards, which can be used to adjust weightings.

[0130] Regulation and Compliance: In some industries, regulations may require that certain metrics be given a higher weight to ensure compliance.

[0131] Resource availability: If certain data are more readily available or less expensive, there may be a tendency to give those metrics a higher weight.

[0132] Model performance: In machine learning models, techniques such as cross-validation can be used to determine the impact of different weight configurations on model performance.

[0133] Dynamic Adjustment: Weights can be dynamic and adjusted based on real-time data and market changes as well as user needs.

[0134] To sum up, by collecting user feedback and evaluations, the intelligent customer service system can understand the satisfaction and dissatisfaction of the service, discover existing problems and room for improvement, and timely adjust and optimize the service strategy. Supervised learning technology can use user feedback data to train and optimize the model of the intelligent customer service system to improve its accuracy and efficiency. The self-learning scoring formula can continuously adjust and optimize services according to user behavior and feedback to provide more personalized services; by analyzing user behavior and optimizing replies, the system can allocate resources more effectively and improve service efficiency. The self-learning scoring formula adapts to changes in user needs through continuous self-improvement, provides more efficient and personalized services, and enhances user experience.

[0135] Compared with existing technologies, this solution uses machine self-learning algorithms to analyze anomaly-demand data sets, which can more accurately identify and correct problems in responses, achieve more accurate anomaly detection and response optimization, and thus achieve technological leadership in the field of customer service.

[0136] After the step of S70: obtaining the subsequent user behavior of the user after receiving the customer service reply, and generating service rating basic data according to the subsequent user behavior, the following steps are included:

[0137] S701: Classifying the subsequent user behavior to mark the type of the subsequent user behavior;

[0138] Specifically, after receiving the customer service reply, the user performed the following actions: clicked on the recommended product link, asked questions again, and left without further action. These actions can be classified as "active interaction", "continued consultation", and "no follow-up action" respectively.

[0139] S702: Analyze the subsequent user behavior to determine whether the current reply meets the user's needs;

[0140] Specifically, by analyzing users' click-through rates and conversion rates, we found that 80% of users who clicked on the recommended product links completed the purchase, indicating that the response meets user needs. If the proportion of users who ask questions again is high, it may indicate that users are not satisfied with the initial response or still have doubts.

[0141] S703: If the current reply is determined to not meet the user's needs, the current reply is marked as an abnormal reply;

[0142] Specifically, if the proportion of users who ask the question again exceeds a preset threshold (eg, 30%), the reply is marked as an abnormal reply, and the system automatically marks such replies as differential replies for further analysis and optimization.

[0143] S704: Associating the abnormal response with the corresponding subsequent user type to construct an abnormality-demand data set;

[0144] The marked difference replies are associated with the user behavior type of "continue consultation" and recorded in the abnormal-demand dataset, which contains all cases marked as abnormal replies and their corresponding user behavior types.

[0145] S705: The preset abnormal response analysis model analyzes the abnormal-demand data set based on a machine self-learning algorithm to generate the difference between the response and the demand, thereby optimizing the current response to meet the actual needs of the user;

[0146] Specifically, machine learning algorithms such as isolation forests or autoencoders are used to analyze the anomaly-demand dataset to identify the differences between replies and demands, and the differences identified by the model are used to optimize the response model, for example, by adjusting the reply generation algorithm or enhancing relevant information in the knowledge base.

[0147] In summary, the intelligent customer service system can automatically detect and optimize its services, improve the accuracy of responses and user satisfaction. This method can achieve continuous self-improvement, reduce user problem solving time, and reduce the customer service costs of enterprises. In addition, by analyzing user behavior and optimizing responses, the system can provide more personalized services and enhance user experience. Compared with the prior art, the intelligent customer service method of this application can more accurately identify and respond to the actual needs of users, and continuously optimize service strategies through machine self-learning algorithms.

[0148] In step S702, the following steps are included:

[0149] S7021: Obtaining a subsequent user behavior type, where the subsequent user behavior type includes changing the topic, expressing gratitude, expressing dissatisfaction, and clicking a recommended link;

[0150] S7022: Classify subsequent user behavior types according to preset judgment criteria to determine whether the current reply meets user needs.

[0151] An example of steps S7021-7022 is as follows:

[0152] After receiving a reply from the intelligent customer service, the user may have the following behaviors:

[0153] Red herring: A user asks a question that is unrelated to the previous question.

[0154] Express gratitude: The user expresses satisfaction with the response, such as by sending a message like “Thanks, that’s helpful.”

[0155] Expressing dissatisfaction: The user is dissatisfied with the response, such as sending "This is not what I wanted."

[0156] Click on the recommended link: The user is interested in the solution or product recommended by the intelligent customer service and clicks on the provided link. The intelligent customer service system classifies user behavior based on the preset judgment criteria. For example:

[0157] If a user changes the subject, it may be a sign that the current response does not fully address the user's question.

[0158] If the user expresses gratitude, it usually means that the current response meets the user's needs.

[0159] If users express dissatisfaction, this is a clear indication that the current response does not meet user needs.

[0160] If a user clicks on the referral link, this indicates that the user is interested in the reply, which can be a positive sign.

[0161] To sum up, by analyzing the user's subsequent behavior, the intelligent customer service system can determine whether its reply is accurate, so as to adjust the reply strategy in time, improve the accuracy of the response, and promptly identify and resolve the points of user dissatisfaction, which can significantly improve customer satisfaction and loyalty. Through continuous monitoring and analysis of user behavior, the intelligent customer service system can continuously optimize its service process, reduce invalid or unsatisfactory interactions, and reduce operating costs. Automated feedback analysis reduces the need for manual intervention, thereby reducing the company's operating costs. By understanding user behavior and preferences, the intelligent customer service system can provide more personalized services, thereby improving the user experience. The intelligent customer service system can learn from each interaction and continuously improve its algorithms and knowledge base to provide better services.

[0162] In step S705: the preset abnormal response analysis model analyzes the abnormal-demand data set based on the machine self-learning algorithm to generate the difference between the response and the demand, thereby optimizing the current response to meet the actual needs of the user, including the following steps:

[0163] S7051: Obtain the subsequent user behavior and current reply corresponding to the abnormal-demand data set, and extract the reply features of the current reply and the demand features of the subsequent user behavior;

[0164] Specifically, suppose a user asks a question about SLR cameras and the intelligent customer service provides a recommendation, but the user's subsequent behavior shows that they clicked on a different camera model and ultimately purchased another product. The system extracts the features of the current reply (such as recommended product model, price range) and the demand features of the user behavior (such as other models clicked, purchase behavior).

[0165] S7052: Compare the response feature with the requirement feature to generate a difference feature;

[0166] Specifically, the system found that there were differences in function and price between the recommended camera model and the model actually purchased by the user. These differences were marked as feature differences, thereby generating difference features such as "price sensitivity" and "specific function preference".

[0167] S7053: Analyze based on the difference characteristics and the answer characteristics to generate answer data that meets the requirement characteristics;

[0168] Specifically, based on the difference characteristics, the system analyzes that users are more sensitive to price and have a preference for specific features of the camera (such as Wi-Fi connection), and generates new response data that includes product recommendations that better meet user needs.

[0169] S7054: Associating the reply data with the query request information to update a reply database, wherein the reply database is used by the response model to generate first predicted intent data according to the query request information;

[0170] Specifically, the system associates the new product recommendation with the user's original query request and updates the reply database so that more accurate recommendations can be provided in the future. The reply database updates include new product recommendations and other camera models that the user may be interested in.

[0171] Through this approach, the intelligent customer service system can more accurately understand the actual needs of users and provide responses that are more in line with those needs. The self-optimization process improves the accuracy of responses and user satisfaction, reduces users' reliance on manual customer service, and thus reduces the company's operating costs.

[0172] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0173] In one embodiment, an intelligent customer service device is provided, which corresponds one-to-one to the intelligent customer service method in the above embodiment. Figure 2 As shown, the intelligent customer service device includes a user request parsing unit 1, which is used to obtain the user's customer service request and query request information, parse the query request information, and determine the type of request information, wherein the query request information includes text information, voice information and image information;

[0174] A first predicted intention data generating unit 2, configured to pre-set a response model to perform intention recognition on the query request letter based on the type of the request information using a corresponding information processing technology to generate first predicted intention data;

[0175] A user profile generation unit 3 is used to obtain the user's historical interaction data and analyze the historical interaction data to generate a user preference profile and a user behavior profile. The historical interaction data includes the user's previous query records, purchase history and preference settings;

[0176] A predicted customer service demand generating unit 4, configured to generate predicted customer service demand based on the user preference profile and the user behavior profile, wherein the predicted customer service demand includes the user's potential inquiry intention, potential inquiry scope, and the user's language habits;

[0177] A second predicted intention data generating unit 5, configured to modify the first predicted intention data based on the predicted customer service demand to generate second predicted intention data;

[0178] A customer service response generation unit 6, configured to generate a customer service response based on the second predicted intention data and the user preference profile, wherein the customer service response includes an answer to a question and a product recommendation;

[0179] A service rating basic data generating unit 7 is used to obtain the subsequent user behavior of the user after receiving the customer service reply, and generate service rating basic data according to the subsequent user behavior, wherein the service rating basic data includes satisfaction, click-through rate, conversion rate, response time and problem resolution rate;

[0180] A service score data set construction unit 8, used to record the service score basic data corresponding to all customer service services and construct a service score data set;

[0181] The service optimization strategy generating unit 9 is configured to analyze the service rating data set based on a machine self-learning algorithm and generate a service optimization strategy to optimize the response model.

[0182] For the specific definition of the intelligent customer service device, please refer to the definition of the intelligent customer service method above, which will not be repeated here. Each module in the above-mentioned intelligent customer service device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the electronic device in the form of hardware, or can be stored in the memory of the electronic device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0183] In one embodiment, an electronic device is provided. The electronic device may be a server, and its internal structure diagram may be as follows: Figure 3 As shown. The electronic device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the electronic device is used to store a database. The network interface of the electronic device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an intelligent customer service method is implemented.

[0184] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0185] Obtaining customer service request and query request information of the user, parsing the query request information, and determining the type of request information, wherein the query request information includes text information, voice information, and image information;

[0186] The preset response model uses a corresponding information processing technology based on the type of the request information to perform intent recognition on the query request letter to generate first predicted intent data;

[0187] Obtaining historical interaction data of the user, and analyzing the historical interaction data to generate a user preference profile and a user behavior profile, wherein the historical interaction data includes the user's previous query records, purchase history, and preference settings;

[0188] generating predicted customer service needs based on the user preference profile and the user behavior profile, wherein the predicted customer service needs include the user's potential inquiry intention, potential inquiry scope, and the user's language habits;

[0189] modifying the first predicted intention data based on the predicted customer service demand to generate second predicted intention data;

[0190] generating a customer service response based on the second predicted intent data and the user preference profile, wherein the customer service response includes answers to questions and product recommendations;

[0191] Obtaining subsequent user behavior after the user receives the customer service reply, and generating basic service rating data based on the subsequent user behavior, wherein the basic service rating data includes satisfaction, click-through rate, conversion rate, response time, and problem resolution rate;

[0192] Record the service rating basic data corresponding to all customer service services and construct a service rating data set;

[0193] The pre-set service optimization model analyzes the service rating data set based on a machine self-learning algorithm, generates a service optimization strategy to optimize the response model, thereby improving accuracy and user satisfaction, and the optimization includes model training, parameter adjustment and feedback loop.

[0194] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0195] Obtaining customer service request and query request information of the user, parsing the query request information, and determining the type of request information, wherein the query request information includes text information, voice information, and image information;

[0196] The preset response model uses a corresponding information processing technology based on the type of the request information to perform intent recognition on the query request letter to generate first predicted intent data;

[0197] Obtaining historical interaction data of the user, and analyzing the historical interaction data to generate a user preference profile and a user behavior profile, wherein the historical interaction data includes the user's previous query records, purchase history, and preference settings;

[0198] generating predicted customer service needs based on the user preference profile and the user behavior profile, wherein the predicted customer service needs include the user's potential inquiry intention, potential inquiry scope, and the user's language habits;

[0199] modifying the first predicted intention data based on the predicted customer service demand to generate second predicted intention data;

[0200] generating a customer service response based on the second predicted intent data and the user preference profile, wherein the customer service response includes answers to questions and product recommendations;

[0201] Obtaining subsequent user behavior after the user receives the customer service reply, and generating basic service rating data based on the subsequent user behavior, wherein the basic service rating data includes satisfaction, click-through rate, conversion rate, response time, and problem resolution rate;

[0202] Record the service rating basic data corresponding to all customer service services and construct a service rating data set;

[0203] The pre-set service optimization model analyzes the service rating data set based on a machine self-learning algorithm, generates a service optimization strategy to optimize the response model, thereby improving accuracy and user satisfaction, and the optimization includes model training, parameter adjustment and feedback loop.

[0204] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0205] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0206] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An intelligent customer service method, characterized in that: The method comprises the steps of: obtaining a customer service request and query request information of a user, parsing the query request information, and determining the type of the request information, wherein the query request information comprises text information, voice information, and image information; The preset response model uses a corresponding information processing technology based on the type of the request information to perform intent recognition on the query request letter to generate first predicted intent data; Obtaining historical interaction data of the user, and analyzing the historical interaction data to generate a user preference profile and a user behavior profile, wherein the historical interaction data includes the user's previous query records, purchase history, and preference settings; generating predicted customer service needs based on the user preference profile and the user behavior profile, wherein the predicted customer service needs include the user's potential inquiry intention, potential inquiry scope, and the user's language habits; modifying the first predicted intention data based on the predicted customer service demand to generate second predicted intention data; generating a customer service response based on the second predicted intent data and the user preference profile, wherein the customer service response includes answers to questions and product recommendations; Obtaining subsequent user behavior after the user receives the customer service reply, and generating basic service rating data based on the subsequent user behavior, wherein the basic service rating data includes satisfaction, click-through rate, conversion rate, response time, and problem resolution rate; Record the service rating basic data corresponding to all customer service services and construct a service rating data set; The pre-set service optimization model analyzes the service rating data set based on a machine self-learning algorithm, generates a service optimization strategy to optimize the response model, thereby improving accuracy and user satisfaction, and the optimization includes model training, parameter adjustment and feedback loop.

2. An intelligent customer service method according to claim 1, characterized in that: In the step of analyzing the service scoring data set based on the machine self-learning algorithm and generating a service optimization strategy to optimize the response model, the service optimization model is preset with a service self-scoring formula: S=α×F+β×C+γ×R+δ×T+ε×P+θ×E, α+β+γ+δ+ε+θ=1; Where S is the self-scoring value, which is used to measure the overall performance of the intelligent customer service system; F is the user satisfaction score, which represents the user's direct rating of the service response; C is the click-through rate, which indicates the degree of user interaction with the recommended content; R is the conversion rate, which indicates the proportion of users who take actions such as adding items to cart and paying for purchases after receiving recommendations. T is the response time, which is the average time required for the system to respond to user requests; P is the problem solving rate, which is the proportion of user problems that are successfully solved; E is the result of sentiment analysis, which indicates the degree of positivity or negativity of the sentiment tendency in the user conversation. Among them, α is the weight value corresponding to the user satisfaction score, β is the weight value corresponding to the click rate, γ is the weight value corresponding to the conversion rate, δ is the weight value corresponding to the response time, ε is the weight value corresponding to the problem solving rate, and θ is the weight value corresponding to the sentiment analysis result.

3. An intelligent customer service method according to claim 1, characterized in that: After obtaining the subsequent user behavior of the user after receiving the customer service reply and generating the service rating basic data according to the subsequent user behavior, the following steps are included: Classifying the subsequent user behavior to mark the type of the subsequent user behavior; Analyze the subsequent user behavior to determine whether the current response meets the user's needs; If the current reply is judged to not meet the user's needs, the current reply will be marked as an abnormal reply; Associating the abnormal response with the corresponding subsequent user type to construct an abnormality-demand data set; The preset abnormal response analysis model analyzes the abnormal-demand data set based on a machine self-learning algorithm to generate the difference between the response and the demand, thereby optimizing the current response to meet the actual needs of the user.

4. An intelligent customer service method according to claim 3, characterized in that: The step of analyzing the subsequent user behavior to determine whether the current reply meets the user's needs includes the following steps: Obtaining subsequent user behavior types, where the subsequent user behavior types include changing the topic, expressing gratitude, expressing dissatisfaction, and clicking on a recommended link; Subsequent user behavior types are classified according to pre-set judgment criteria to determine whether the current response meets user needs.

5. An intelligent customer service method according to claim 4, characterized in that: In the step where the preset abnormal response analysis model analyzes the abnormal-demand data set based on a machine self-learning algorithm to generate the difference between the response and the demand, thereby optimizing the current response to meet the actual needs of the user, the following steps are included: Obtaining the subsequent user behavior and current response corresponding to the abnormal-demand data set, and extracting the reply features of the current response and the demand features of the subsequent user behavior; Comparing the response features with the requirement features to generate difference features; Performing analysis based on the difference characteristics and the answer characteristics to generate answer data that meets the required characteristics; The reply data is associated with the query request information to update a reply database, wherein the reply database is used by the response model to generate first predicted intent data based on the query request information.

6. An intelligent customer service method according to claim 1, characterized in that: Provide multi-language support to serve users of different languages, including automatic language recognition and translation.

7. An intelligent customer service method according to claim 1, characterized in that: The machine self-learning algorithm includes a feature extraction algorithm and a clustering algorithm.

8. An intelligent customer service device, applied to an intelligent customer service method according to any one of claims 1 to 7, characterized in that: The device comprises: a user request parsing unit (1), which is used to obtain a user's customer service request and query request information, parse the query request information, and determine the type of the request information, wherein the query request information includes text information, voice information, and image information; A first predicted intention data generating unit (2) is used to pre-set a response model to perform intention recognition on the query request letter based on the type of the request information using a corresponding information processing technology to generate first predicted intention data; A user profile generation unit (3), used to obtain historical interaction data of the user and analyze the historical interaction data to generate a user preference profile and a user behavior profile, wherein the historical interaction data includes the user's previous query records, purchase history and preference settings; A predicted customer service demand generating unit (4), configured to generate a predicted customer service demand based on the user preference profile and the user behavior profile, wherein the predicted customer service demand includes the user's potential inquiry intention, potential inquiry scope, and the user's language habits; A second predicted intention data generating unit (5), configured to modify the first predicted intention data based on the predicted customer service demand to generate second predicted intention data; A customer service response generation unit (6), configured to generate a customer service response based on the second predicted intention data and the user preference profile, wherein the customer service response includes an answer to a question and a product recommendation; A service rating basic data generating unit (7), used for acquiring subsequent user behavior after the user receives the customer service reply, and generating service rating basic data according to the subsequent user behavior, wherein the service rating basic data includes satisfaction, click-through rate, conversion rate, response time and problem resolution rate; A service rating data set construction unit (8), used to record the service rating basic data corresponding to all customer service services and construct a service rating data set; A service optimization strategy generating unit (9) is used to analyze the service scoring data set based on a machine self-learning algorithm with a pre-set service optimization model, and generate a service optimization strategy to optimize the response model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of an intelligent customer service method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of an intelligent customer service method as described in any one of claims 1 to 7 are implemented.

Images