An Internet of Things information technology customer service system based on emotional energy perception

By introducing emotional energy sensing technology into the IoT customer service system, and using cloud servers and artificial intelligence to analyze users' emotional states and generate personalized response templates, the problem of existing systems being unable to accurately identify emotions and combine them with emotional needs is solved, thereby improving communication quality and user satisfaction.

CN118350824BActive Publication Date: 2024-11-15上饶丫牛软件中心(个人独资)
View PDF 13 Cites 0 Cited by

Patent Information

Application Number
CN202410586991.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-13
Publication Date
2024-11-15
Estimated Expiration
2044-05-13

AI Technical Summary

Technical Problem

Existing IoT-based customer service systems that rely on emotion energy perception cannot accurately identify users' emotional states and fail to provide effective responses that combine emotional states with technical needs, resulting in poor communication quality.

Method used

It employs a customer service module, a text collection module, a preliminary emotion state analysis module, an in-depth emotion analysis module, a response template generation module, and an emotion perception service module. By analyzing dialogue text through a cloud server, it utilizes transfer learning, sentiment analysis tools, long short-term memory networks, and fuzzy logic control to generate personalized response templates that combine emotional state and technical requirements.

Benefits of technology

It improved user experience and satisfaction, enhanced communication quality, reduced misunderstandings and conflicts, ensured data security and accessibility, and enabled personalized and accurate service responses.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118350824B_ABST
    Figure CN118350824B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of information technology services, and discloses an Internet of Things information technology customer service system based on emotional energy perception, including a customer service module, a text collection module, an emotional state preliminary analysis module, an emotional in-depth analysis module, a reply template generation module, and an emotional perception service module. The present invention can identify the emotional changes of users, thereby providing personalized customer service, enhancing user experience and satisfaction, and replying content according to the emotional state and technical needs of users, which helps to improve the communication quality between customer service personnel and users and reduce misunderstandings and conflicts; at the same time, the cloud server stores all the dialogue texts of the interaction between users and customer service, effectively manages a large amount of data, and ensures the security and accessibility of the data. The cloud server can ensure the real-time transmission and processing of data, which is crucial for real-time monitoring of user emotions and timely response.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of information technology services, and in particular to an Internet of Things information technology customer service system based on emotion energy perception. Background Art

[0002] Emotional energy sensing is the technology of identifying and processing human emotional states. The perceived emotional energy can be used to improve interactions with humans, communicate with people more empathetically, or adjust services and responses to adapt to users' emotional changes. IoT information technology involves technologies that enable these devices to interconnect and process data, including sensors, communication protocols, data storage, cloud computing, edge computing, etc.

[0003] By analyzing the text, voice or other physiological signals provided by users when interacting with customer service, we can judge the user's emotions. We can use this information to reply to users or adjust service strategies to provide a more caring and personalized service experience, thereby enhancing user satisfaction and loyalty.

[0004] For example, Chinese patent 202210791783.4 discloses an Internet of Things information technology customer service system based on emotional energy perception, which includes an artificial intelligence service module, a customer service staff allocation module, and a customer service staff processing module, which can recommend corresponding consulting results based on the user's emotions. However, the above system still has the following shortcomings: when identifying the user's emotions, the keyboard behavior and mouse behavior are used to judge whether the user is impatient, but the identification method may be affected by the user's operating habits, and the user's true emotional state cannot be judged. At the same time, the emotional state and technical requirements are not combined to respond, so the quality of communication cannot be guaranteed.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In view of this, the purpose of the embodiments of the present application is to provide an Internet of Things information technology customer service system based on emotional energy perception to overcome the above-mentioned technical problems existing in the existing related technologies.

[0007] To this end, the specific technical solution adopted by the present invention is as follows:

[0008] An Internet of Things information technology customer service system based on emotion energy perception includes a customer service module, a text collection module, an emotion state preliminary analysis module, an emotion in-depth analysis module, a reply template generation module and an emotion perception service module.

[0009] Among them, the customer service module is used to connect with users and provide an interface for customer service personnel to respond to users' questions about IoT information technology.

[0010] The text collection module is used to collect the conversation text during the interaction between the user and the customer service, and send the conversation text to the cloud server through the hypertext transfer protocol.

[0011] The emotional state preliminary analysis module is used to analyze the conversation text on the cloud server, identify emotional fluctuations, and screen out users.

[0012] The emotional in-depth analysis module is used to analyze the corresponding conversation text data of the screened users using artificial intelligence algorithms on the cloud server, identify the user's emotional state, and synchronize the analysis results to the reply template generation module.

[0013] The reply template generation module is used to generate corresponding reply templates according to emotional states and technical requirements.

[0014] The emotion perception service module is used to provide services based on the user's emotional state. At the same time, the service content is based on the user's technical needs and matches the user's emotional state.

[0015] Optionally, when collecting the conversation text during the interaction between the user and the customer service and sending the conversation text to the cloud server via the Hypertext Transfer Protocol, the Hypertext Transfer Protocol Security Protocol is used to encrypt the data during the transmission process to protect the security of the conversation text and desensitize the conversation text data.

[0016] Optionally, the emotional state preliminary analysis module includes a feature extraction module, an emotional fluctuation quantification module, a threshold setting module, a user emotional fluctuation calculation module, a user screening module and a real-time monitoring module.

[0017] Among them, the feature extraction module is used to clean the conversation text data and decompose the conversation text into separate words. At the same time, transfer learning is used to extract text features using a model pre-trained on a large corpus.

[0018] The sentiment fluctuation quantification module is used to use sentiment analysis tools to evaluate the sentiment tendency and intensity of text and convert the results of sentiment analysis into numerical values.

[0019] The threshold setting module is used to set the threshold value for judging the degree of emotional fluctuation.

[0020] The user emotion fluctuation calculation module is used to calculate the average emotion fluctuation value of the user's conversation text for each user, and determine whether the user is in an emotion fluctuation state by comparing the user's average emotion fluctuation value with a set threshold.

[0021] The user screening module is used to screen out users whose average emotion fluctuation value exceeds a threshold.

[0022] The real-time monitoring module is used to monitor the real-time conversation text on the cloud server and mark the screened users.

[0023] Optionally, when setting the threshold, set the objective function:

[0024] J(θ) = -F1_Score(θ);

[0025] In the formula, θ represents the threshold parameter, and J(θ) represents the objective function;

[0026] F1_Score represents the harmonic mean of precision and recall.

[0027] Optionally, after setting the objective function, a prior distribution is selected for the threshold parameter, and an initial threshold parameter set is selected, the objective function is calculated, and the calculation results and the initial threshold parameter set are collected as initial data points;

[0028] Use the initial data points to train the Gaussian process model, and use the probability model to calculate the posterior probability and improvement probability of each point in the search space, and select the point with the highest improvement probability as the next threshold parameter;

[0029] Calculate the objective function using the selected threshold parameter and update the Gaussian process;

[0030] Repeat the selection of threshold parameters and the update of Gaussian process until the stopping condition is met;

[0031] After the stopping condition is met, the threshold parameter that optimizes the objective function is selected as the final threshold.

[0032] Optionally, the emotion in-depth analysis module includes a numeration module, a long short-term memory network module, a fuzzy logic control module, a deployment module, a real-time analysis module and a result output module.

[0033] Among them, the digitization module is used to convert text into numerical form using the bag-of-words model.

[0034] The LSTM module is used to train LSTM networks using datasets with labeled emotional states.

[0035] A fuzzy logic control module is used to classify emotional states based on the output of the long short-term memory network.

[0036] The deployment module is used to deploy the trained long short-term memory network and fuzzy logic control module to the cloud server and create an application programming interface to receive the results of the preliminary analysis of text data and emotional states.

[0037] The real-time analysis module is used to receive the filtered text data of the user's real-time conversation through the application programming interface, and pre-process and extract features of the text data received in real time, and input the processed data into the trained long short-term memory network and fuzzy logic control module for analysis to obtain the emotional state.

[0038] The result output module is used to synchronize the analysis results of the emotional state in JSON format to the reply template generation module.

[0039] Optionally, when training the LSTM network using a training dataset with labeled emotional states, construct a LSTM network structure, and select a loss function, an optimizer, and an evaluation metric;

[0040] The emotion probability acquisition model built using the long short-term memory network architecture is trained, verified and evaluated using a dataset with labeled emotional states, and the hyperparameters are adjusted based on the performance of the emotion probability acquisition model.

[0041] Optionally, based on the output of the long short-term memory network, when classifying the emotional state, the input variables and output variables of the fuzzy logic control are defined, and fuzzy sets and membership functions are defined for each input variable and output variable, and rules are defined based on emotional probability and domain knowledge, and a rule base is constructed at the same time;

[0042] The output of the long short-term memory network is input into the membership function and converted into the membership value of the fuzzy set;

[0043] Each rule in the rule base is activated according to the input fuzzy set membership value, and the comprehensive membership of each output fuzzy set is calculated through the reasoning mechanism of fuzzy logic, and the membership value of the inferred fuzzy set is converted into an accurate output.

[0044] Optionally, the reply template generation module includes a matrix construction module, a clustering module and a reply template classification module.

[0045] Among them, the matrix construction module is used to construct an emotional demand matrix, and the rows of the emotional demand matrix represent different emotional states, and the columns represent different categories of user technical needs.

[0046] The clustering module is used to classify users' technical requirements using clustering algorithms.

[0047] The reply template classification module is used to collect a number of reply templates for each technical requirement, and the number of reply templates covers different emotional states.

[0048] Optionally, when using a clustering algorithm to classify the technical requirements of users, a clustering algorithm and the number of clusters are selected, and a clustering operation is performed on the feature vectors of the technical requirements, and each cluster is labeled.

[0049] Embodiments of the present invention include the following benefits:

[0050] (1) The present invention provides an Internet of Things information technology customer service system based on emotional energy perception, which can identify the emotional changes of users, thereby providing personalized customer service, enhancing user experience and satisfaction, and replying to content based on the user's emotional state and technical needs, which helps to improve the communication quality between customer service personnel and users and reduce misunderstandings and conflicts. At the same time, the cloud server stores all the dialogue texts of the interaction between users and customer service, effectively manages a large amount of data, and ensures the security and accessibility of the data. The cloud server can ensure the real-time transmission and processing of data, which is crucial for real-time monitoring of user emotions and timely response.

[0051] (2) The present invention processes and analyzes emotional data at different levels through the coordinated use of an emotional state preliminary analysis module and an emotional in-depth analysis module, thereby more effectively understanding the emotional state of the customer and making corresponding responses. Through the preliminary analysis in the first stage, possible emotional fluctuations can be quickly screened out, helping customer service personnel focus on customer situations that require special attention, thereby improving work efficiency. Through the in-depth analysis and optimization in the second stage, artificial intelligence algorithms are used to further improve the accuracy of emotional state judgment, thereby gaining a more detailed understanding of the emotional state of the customer, laying the foundation for providing personalized and accurate services. Phased processing can make the system more flexible, and the processing methods of different stages can be adjusted and optimized according to actual conditions to adapt to customer needs and service requirements in different scenarios.

[0052] (3) The present invention generates corresponding reply templates based on emotional states and technical requirements. Combining the analysis of emotions and requirements, the system can more accurately understand the actual needs of users, thereby providing more relevant and accurate solutions. By considering the emotional state of users, the system can communicate in a more empathetic way, which can significantly improve user satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 It is a principle block diagram of an Internet of Things information technology customer service system based on emotional energy perception according to an embodiment of the present invention.

[0055] In the figure:

[0056] 1. Customer service module; 2. Text collection module; 3. Emotional state preliminary analysis module; 4. Emotional in-depth analysis module; 5. Reply template generation module; 6. Emotional perception service module. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0060] According to an embodiment of the present invention, an Internet of Things information technology customer service system based on emotion energy perception is provided.

[0061] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an embodiment of the present invention, the Internet of Things information technology customer service system based on emotion energy perception includes a customer service module 1, a text collection module 2, an emotional state preliminary analysis module 3, an emotional in-depth analysis module 4, a reply template generation module 5 and an emotion perception service module 6.

[0062] Among them, the customer service module 1 is used to connect with users and provide an interface for customer service personnel to reply to users' questions about IoT information technology. It enables customer service personnel to communicate with users effectively and answer users' various questions about IoT information technology. Customer service personnel have a special interface to view questions submitted by users and provide corresponding replies.

[0063] For example, a user encounters a connection problem with his smart home device. The user submits the corresponding question through the customer service page of the IoT platform, and the customer service staff responds with the steps to solve the problem.

[0064] The text collection module 2 is used to collect the conversation text during the interaction between the user and the customer service, and send the conversation text to the cloud server through the hypertext transfer protocol.

[0065] In a further embodiment, when the conversation text between the user and the customer service is collected and sent to the cloud server via the Hypertext Transfer Protocol, the Hypertext Transfer Protocol Secure (HTTPS) is used to encrypt the data during the transmission process to protect the security of the conversation text, and the conversation text data is desensitized to remove information that may identify the user, such as name, address, telephone number, etc.

[0066] HTTPS encrypted transmission: Using the Hypertext Transfer Protocol Secure (HTTPS), the conversation text will be encrypted during transmission to the cloud server. HTTPS is a secure version of HTTP, which uses the SSL / TLS protocol at the transport layer to encrypt data packets to protect data from being eavesdropped, tampered with, or forged when it is transmitted on the Internet.

[0067] The emotional state preliminary analysis module 3 is used to analyze the conversation text in the cloud server, identify emotional fluctuations, and screen out users.

[0068] In a further embodiment, the emotional state preliminary analysis module includes a feature extraction module, an emotional fluctuation quantification module, a threshold setting module, a user emotional fluctuation calculation module, a user screening module and a real-time monitoring module;

[0069] Among them, the feature extraction module is used to clean the conversation text data, such as deleting special characters, punctuation marks, useless stop words, and decomposing the conversation text into separate words. At the same time, transfer learning is used to extract text features using a model pre-trained on a large corpus. For example, BERT is a pre-trained deep learning model that uses the Transformer architecture to understand the bidirectional context in the text.

[0070] The sentiment fluctuation quantification module is used to evaluate the sentiment tendency and intensity of text using sentiment analysis tools (such as VADER, TextBlob, etc.), and convert the results of sentiment analysis into numerical values, such as the intensity scores of positive and negative sentiments; when the preprocessed text is input into the sentiment analysis tool, for example, VADER pays special attention to the intensity and polarity of sentiment words, while considering word order, word form changes and sentiment modifiers, and outputs a sentiment intensity score.

[0071] The threshold setting module is used to set one or more thresholds for determining the degree of emotional fluctuation.

[0072] The user emotion fluctuation calculation module is used to calculate the average emotion fluctuation value of the user's conversation text for each user, and determine whether the user is in an emotion fluctuation state by comparing the user's average emotion fluctuation value with a set threshold.

[0073] The user screening module is used to screen out users whose average emotional fluctuation values ​​exceed the threshold, so that customer service can first focus on those users who may need extra care.

[0074] The real-time monitoring module is used to monitor the real-time conversation text on the cloud server and mark the screened users for further processing or intervention.

[0075] In a further embodiment, when setting the threshold, the objective function is set:

[0076] J(θ)=-F1_Score(threshold=θ);

[0077] Where θ represents the threshold parameter, J(θ) represents the objective function, and the threshold parameter θ is mapped to a value for optimization;

[0078] F1_Score represents the harmonic mean of precision and recall, which is used to evaluate the performance of emotion fluctuation detection.

[0079] In Bayesian optimization, we want to maximize the F1 score to get the best model performance. But Bayesian optimization algorithms (especially those using Gaussian processes) usually deal with minimization problems more directly. Therefore, we take the opposite of the F1 score, convert the problem of maximizing the F1 score into the problem of minimizing J(θ), and convert the original maximization problem into a minimization problem, directly using the properties of the Gaussian process to guide the search for the optimal threshold. This means finding a threshold θ that can achieve the highest balance between precision and recall at this threshold.

[0080] In a further embodiment, after setting the objective function, a prior distribution is selected for the threshold parameter. Typically, if there is no prior knowledge, a uniform distribution can be selected, that is, all possible parameter values ​​are considered equally likely, and an initial set of threshold parameters is selected, the objective function is calculated, and the calculation results and the initial set of threshold parameters are collected as initial data points;

[0081] The initial data points are used to train the Gaussian process model. The Gaussian process is a non-parametric Bayesian method suitable for uncertainty modeling of functions. A probability model is used to calculate the posterior probability and improvement probability of each point in the search space, and the point with the highest improvement probability is selected as the next threshold parameter. The improvement probability refers to the probability that a point will bring about an improvement in the objective function.

[0082] The target function is calculated using the selected threshold parameter and the Gaussian process is updated, that is, the target function is calculated using the newly selected threshold parameter and the result is used to update the Gaussian process model;

[0083] The selection of threshold parameters and the updating of the Gaussian process are repeated until the stopping condition is met, such as reaching a predetermined number of iterations, the probability of improvement is lower than a certain threshold, or computing resource limitations.

[0084] After the stop condition is met, the threshold parameter that optimizes the objective function (i.e., the highest F1_Score) is selected as the final threshold.

[0085] The emotional in-depth analysis module 4 is used to use an artificial intelligence algorithm to analyze the corresponding conversation text data of the screened users on the cloud server, identify the user's emotional state, and synchronize the analysis results to the reply template generation module 5.

[0086] In a further embodiment, the emotion in-depth analysis module includes a numeration module, a long short-term memory network module, a fuzzy logic control module, a deployment module, a real-time analysis module, and a result output module;

[0087] Among them, the digitization module is used to convert text into numerical form, word frequency or TF-IDF vector using the bag-of-words model.

[0088] The LSTM module is used to train LSTM networks using datasets with labeled emotional states.

[0089] The fuzzy logic control module is used to classify emotional states based on the output of the long short-term memory network (emotional probability); define fuzzy logic rules for classifying emotional states. The emotion probability output by the LSTM network is used as input. Fuzzy logic rules are applied to classify emotional states.

[0090] The deployment module is used to deploy the trained long short-term memory network and fuzzy logic control module to the cloud server and create an application programming interface to receive the results of the preliminary analysis of text data and emotional states.

[0091] The real-time analysis module is used to receive the filtered text data of the user's real-time conversation through the application programming interface, and pre-process and extract features of the text data received in real time, and input the processed data into the trained long short-term memory network and fuzzy logic control module for analysis to obtain the emotional state.

[0092] The result output module is used to synchronize the analysis results of the emotional state to the reply template generation module in JSON or other formats.

[0093] In a further embodiment, when a training data set with labeled emotional states is used to train a long short-term memory network, a long short-term memory network structure is constructed, and a loss function, an optimizer, and an evaluation index are selected; wherein, constructing the long short-term memory network structure includes defining an LSTM layer, adding a fully connected (Dense) layer as an output layer, and the number of units of the layer is equal to the number of categories of emotional states. A loss function is selected, such as categorical_crossentropy (multi-classification problem) or binary_crossentropy (binary classification problem). An optimizer is selected, such as Adam or SGD. An evaluation index is selected, such as accuracy.

[0094] The emotion probability acquisition model built using the long short-term memory network architecture is trained, verified and evaluated using a dataset with labeled emotional states. At the same time, hyperparameters such as the number of units in the LSTM layer and the learning rate are adjusted according to the performance of the emotion probability acquisition model.

[0095] In a further embodiment, when classifying emotional states based on the output of the long short-term memory network, the input variables (emotional probability) and output variables (emotional states) of the fuzzy logic control are defined, and fuzzy sets and membership functions are defined for each input variable and output variable, and if-then rules are defined based on the emotional probability and domain knowledge, and a rule base is constructed at the same time;

[0096] The output of the LSTM network (emotion probability) is input into the membership function and converted into the membership value of the fuzzy set. For example, if the probability of the LSTM outputting the emotion "happy" is 0.6, this probability value will be converted into a membership value of a fuzzy set according to the membership function corresponding to the emotion "happy";

[0097] For each rule in the rule base, it is activated according to the input fuzzy set membership value. If the input membership value satisfies the condition part (if part) of the rule, the conclusion part of the rule is activated, and the comprehensive membership of each output fuzzy set is calculated through the reasoning mechanism of fuzzy logic (such as Mamdani reasoning or Larsen reasoning), and the membership value of the fuzzy set obtained by reasoning is converted into an accurate output. Defuzzification is to map the membership of a fuzzy set to a specific output value. Commonly used defuzzification methods are as follows.

[0098] Maximum membership method: select the output value corresponding to the fuzzy set with the highest membership.

[0099] Median method: Calculate the weighted average of all fuzzy set membership values ​​to obtain the final output value.

[0100] Among them, fuzzy sets and membership functions provide a way to deal with uncertainty and ambiguity. The rule base defines the relationship between input and output based on expert knowledge or empirical data. The reasoning mechanism uses fuzzy sets and rule bases to make decisions. Fuzzy logic control can process the output of the LSTM model and provide a more robust and explainable emotional state classification result.

[0101] The reply template generation module 5 is used to generate corresponding reply templates according to emotional states and technical requirements.

[0102] In a further embodiment, the reply template generation module includes a matrix construction module, a clustering module, and a reply template classification module;

[0103] Among them, the matrix construction module is used to construct an emotional demand matrix, and the rows of the emotional demand matrix represent different emotional states (such as satisfaction, disappointment, anger, anxiety, etc.), and the columns represent different categories of user technical needs (such as setting problems, network failures, functional abnormalities, account problems, etc.).

[0104] The clustering module is used to classify users' technical requirements using clustering algorithms.

[0105] The reply template classification module is used to collect a number of reply templates for each technical requirement, and the number of reply templates covers different emotional states.

[0106] In addition, the templates are optimized by combining business knowledge and customer service experience to ensure that they can effectively solve user problems. When new user needs come in, the system automatically extracts features, identifies the need category, and maps it to the corresponding template. Based on the mapping results, the corresponding technical need template is called to quickly generate a response or solution.

[0107] In a further embodiment, when using a clustering algorithm to classify the technical requirements of users, a clustering algorithm and the number of clusters are selected, and a clustering operation is performed on the feature vectors of the technical requirements. At the same time, each cluster is labeled to identify the specific technical requirements they represent.

[0108] Among them, choose a suitable clustering algorithm, such as K-Means, DBSCAN, hierarchical clustering. Determine the optimal number of clusters, including the elbow rule: calculate the compactness of internal clusters when the number of clusters changes from 1 to N, and select the point where the compactness gain is minimized as the number of clusters. Silhouette coefficient: evaluate the quality of clustering. A silhouette coefficient close to 1 indicates a good clustering effect. DB index: When the index is lower, it means that the clusters are better separated and the clusters are tighter.

[0109] Use the selected algorithm to cluster the feature vectors. This process will group similar technical requirements into the same cluster. Analyze each cluster to understand the specific technical requirements it represents. This includes keyword extraction, topic modeling, manual review, and creation of labels, such as "network issues", "hardware failures", "software updates", etc. These labels should reflect the common points of user requirements in the cluster. Through an iterative process, clustering, analysis, and adjustments are repeated to optimize the quality of clusters and the corresponding technical requirement labels.

[0110] The emotion perception service module 6 is used to provide services according to the user's emotional state. At the same time, the service content is based on the user's technical needs and matches the user's emotional state. That is, according to the emotion recognition results and the user's technical needs, the corresponding service is selected, such as providing comfort, advice, technical support, etc. According to the matched service content, the corresponding response is provided to the user. A reply in text, voice or other forms is generated to meet the user's needs.

[0111] In order to facilitate understanding of the above technical solutions of the present invention, the working principle or operation mode of the present invention in the actual process is described in detail below.

[0112] In actual application, user A encounters a connection problem with his smart home device. For example, he sends a message through the customer service system: "My smart bulb cannot connect to the home network. The installation is not smooth." User A's message is received by the customer service system and displayed on the customer service interface. The message text is encrypted and transmitted to the cloud server via HTTPS.

[0113] The emotion fluctuation quantification module uses sentiment analysis tools to find negative emotion tendencies in the text. The user emotion fluctuation calculation module calculates that the emotion fluctuation value of user A exceeds the set threshold.

[0114] At this point, the digitization module converts the text into numerical form, and the long short-term memory network module analyzes the text and determines that the emotional state of user A is "frustrated." The fuzzy logic control module further classifies the emotional state.

[0115] The reply template generation module generates a reply template containing comforting statements and solutions based on the emotional state of "frustration" and technical requirements. You can reply: "Mr. or Ms. A, we understand your frustration. Please follow the steps below to reset your device...". Result output module: Send the reply content back to the customer service interface in JSON format, which the customer service staff can use directly or adjust as needed and send to user A. Use emotional energy perception to enhance the customer service experience and make it more personalized and effective. Identify the customer's emotional state and technical requirements, and provide corresponding service content to improve customer satisfaction and problem-solving efficiency.

[0116] In summary, the present invention provides an Internet of Things information technology customer service system based on emotional energy perception, which can identify the emotional changes of users, thereby providing personalized customer service, enhancing user experience and satisfaction, and replying content according to the emotional state and technical needs of users, which helps to improve the communication quality between customer service personnel and users and reduce misunderstandings and conflicts. At the same time, the cloud server stores all the dialogue texts of the interaction between users and customer service, effectively manages a large amount of data, and ensures the security and accessibility of the data. The cloud server can ensure the real-time transmission and processing of data, which is very important for real-time monitoring of user emotions and timely response. The present invention processes and analyzes emotional data at different levels through the coordinated use of the emotional state preliminary analysis module and the emotional in-depth analysis module, so as to more effectively understand the emotional state of the customer and make corresponding responses. Through the preliminary analysis of the first stage, possible emotional fluctuations are quickly screened out, helping customer service personnel focus on the customer situation that needs to be focused on, and improving work efficiency. Through the in-depth analysis and optimization of the second stage, the artificial intelligence algorithm is used to further improve the accuracy of emotional state judgment, so as to understand the emotional state of the customer more finely, and lay the foundation for providing personalized and accurate services. Phased processing can make the system more flexible, and the processing methods of different stages can be adjusted and optimized according to actual conditions to adapt to customer needs and service requirements in different scenarios. The present invention generates corresponding reply templates based on emotional state and technical requirements, and combines the analysis of emotions and needs to enable the system to understand the actual needs of users more accurately, thereby providing more relevant and accurate solutions. By considering the emotional state of users, the system can communicate in a more empathetic way, which can significantly improve user satisfaction and loyalty.

[0117] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional charging modules / units in the systems and devices may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0118] 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 sequence. 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 of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising 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.

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An Internet of Things information technology customer service system based on emotional energy perception, characterized in that: The Internet of Things information technology customer service system based on emotion energy perception includes a customer service module, a text collection module, an emotion state preliminary analysis module, an emotion in-depth analysis module, a reply template generation module and an emotion perception service module; The customer service module is used to connect with users and provide an interface for customer service personnel to respond to users' questions about IoT information technology; The text collection module is used to collect the conversation text during the interaction between the user and the customer service, and send the conversation text to the cloud server via the hypertext transfer protocol; Among them, when collecting the conversation texts during the interaction between users and customer service and sending them to the cloud server via the Hypertext Transfer Protocol, the Hypertext Transfer Protocol Security Protocol is used to encrypt the data during the transmission process to protect the security of the conversation texts and desensitize the conversation text data; The emotional state preliminary analysis module is used to analyze the conversation text on the cloud server, identify emotional fluctuations, and screen out users; The emotional state preliminary analysis module includes a feature extraction module, an emotional fluctuation quantification module, a threshold setting module, a user emotional fluctuation calculation module, a user screening module and a real-time monitoring module; The feature extraction module is used to clean the conversation text data and decompose the conversation text into individual words, while using transfer learning to extract text features using a model pre-trained on a large corpus; The emotional fluctuation quantification module is used to use the emotional analysis tool to evaluate the emotional tendency and intensity of the text, and convert the results of the emotional analysis into numerical values; The threshold setting module is used to set a threshold value for determining the degree of emotional fluctuation; Among them, when setting the threshold, set the objective function: ; In the formula, θ represents the threshold parameter, J ( θ ) represents the objective function; F 1_ Score Represents the harmonic mean of precision and recall; After setting the objective function, a prior distribution is selected for the threshold parameter, and an initial threshold parameter set is selected, the objective function is calculated, and the calculation results and the initial threshold parameter set are collected as initial data points; Use the initial data points to train the Gaussian process model, and use the probability model to calculate the posterior probability and improvement probability of each point in the search space, and select the point with the highest improvement probability as the next threshold parameter; Calculate the objective function using the selected threshold parameter and update the Gaussian process; Repeat the selection of threshold parameters and the update of Gaussian process until the stopping condition is met; After the stop condition is met, the threshold parameter that optimizes the objective function is selected as the final threshold; The user emotion fluctuation calculation module is used to calculate the average emotion fluctuation value of the user's conversation text for each user, and determine whether the user is in an emotion fluctuation state by comparing the user's average emotion fluctuation value with a set threshold; The user screening module is used to screen out users whose average emotion fluctuation value exceeds a threshold; The real-time monitoring module is used to monitor the real-time conversation text on the cloud server and mark the screened users; The emotional in-depth analysis module is used to analyze the corresponding conversation text data of the screened users using an artificial intelligence algorithm on the cloud server, identify the emotional state of the users, and synchronize the analysis results to the reply template generation module; The emotion in-depth analysis module includes a numerical module, a long short-term memory network module, a fuzzy logic control module, a deployment module, a real-time analysis module and a result output module; The digitization module is used to convert text into a numerical form using a bag-of-words model; The long short-term memory network module is used to train the long short-term memory network using a dataset with labeled emotional states; When using a dataset with labeled emotional states to train a LSTM network, construct the LSTM network structure and select the loss function, optimizer, and evaluation metric; Use a dataset with labeled emotional states to train, validate, and evaluate the emotion probability acquisition model built using the long short-term memory network architecture, and adjust the hyperparameters based on the performance of the emotion probability acquisition model; The fuzzy logic control module is used to classify emotional states based on the output of the long short-term memory network; Among them, based on the output of the long short-term memory network, when classifying the emotional state, the input variables and output variables of the fuzzy logic control are defined, and the fuzzy sets and membership functions are defined for each input variable and output variable. The rules are defined based on the emotional probability and domain knowledge, and the rule base is constructed at the same time; The output of the long short-term memory network is input into the membership function and converted into the membership value of the fuzzy set; Each rule in the rule base is activated according to the input fuzzy set membership value, and the comprehensive membership of each output fuzzy set is calculated through the reasoning mechanism of fuzzy logic, and the membership value of the fuzzy set obtained by reasoning is converted into precise output; The deployment module is used to deploy the trained long short-term memory network and fuzzy logic control module to the cloud server, and create an application programming interface to receive the results of the preliminary analysis of text data and emotional state; The real-time analysis module is used to receive the text data of the screened real-time conversation of the user through the application programming interface, and pre-process and extract features of the text data received in real time, and input the processed data into the trained long short-term memory network and fuzzy logic control module for analysis to obtain the emotional state; The result output module is used to synchronize the analysis results of the emotional state to the reply template generation module in JSON format; The reply template generation module is used to generate a corresponding reply template according to the emotional state and technical requirements; The reply template generation module includes a matrix construction module, a clustering module and a reply template classification module; The matrix construction module is used to construct an emotional demand matrix, wherein the rows of the emotional demand matrix represent different emotional states, and the columns represent different categories of user technical needs; The clustering module is used to classify the technical requirements of users using a clustering algorithm; The reply template classification module is used to collect a number of reply templates for each technical requirement, and the number of reply templates covers different emotional states; The emotion perception service module is used to provide services according to the user's emotional state, and the service content is based on the user's technical needs and matches the user's emotional state.

2. According to claim 1, an Internet of Things information technology customer service system based on emotional energy perception is characterized in that: When the clustering algorithm is used to classify the technical requirements of the users, the clustering algorithm and the number of clusters are selected, and the characteristic vectors of the technical requirements are clustered, and each cluster is labeled.

Citation Information

Patent Citations

  • Text emotion recognition method and device and storage medium

    CN111723198A

  • Voice recognition system for intelligent customer service conversation

    CN111739516A

  • Emotion recognition method and device based on user session analysis

    CN112580347A

  • Text emotion recognition method and device, equipment and storage medium

    CN112966106A

  • Emotion recognition method and device, equipment, storage medium and program product

    CN114298019A