System

Abstract

A system is provided.SOLUTION: A system, comprising: means for collecting historical inquiry data; means for analyzing the historical inquiry data to build a prediction model; means for monitoring user behavior data in real time; means for predicting a potential question of a user using the prediction model; and means for notifying the user of an appropriate solution to the predicted question.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] With conventional customer support systems, responses to customer inquiries tended to be slow, leading to customer dissatisfaction. In addition, the high cost of responding to inquiries placed a heavy burden on the management side. Furthermore, there was a lack of mechanisms to reduce the time and effort required for customers to make inquiries. [Means for solving the problem]

[0005] The present invention relates to a system that collects past inquiry data, analyzes it, and builds a predictive model. This system monitors user behavior data in real time and predicts potential user questions using a predictive model. By notifying users of appropriate solutions to predicted questions, the system can resolve customer concerns in advance and reduce inquiries. The system also includes a means for sorting out duplicate and incomplete data from past inquiry data and a means for periodic retraining to improve the accuracy of the predictive model. In this way, it is possible to increase customer satisfaction and reduce inquiry response costs.

[0006] "Inquiry Data" means records of information provided by customers regarding their questions or problems.

[0007] A "predictive model" is a mathematical model for predicting future events based on past data.

[0008] "User behavior data" refers to records of user behavior, such as operations and selections made within the system and access logs.

[0009] "Solutions" are specific solutions or information that can be applied to anticipated questions or problems.

[0010] "Real-time monitoring" means that the system immediately detects user behavior and processes the data immediately.

[0011] "Duplicate data" refers to data in which the same content is recorded multiple times.

[0012] "Incomplete data" refers to data that contains missing or incorrect information.

[0013] "Retraining" refers to the process of retraining a machine learning model with new or revised data in order to improve its accuracy.

[0014] "Notifying the user" refers to the system's action of notifying the user of the predicted results and solutions. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0016] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0017] First, the terms used in the following description will be explained.

[0018] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0019] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0020] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0021] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0022] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0023] [First embodiment]

[0024] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0025] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0026] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0028] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0031] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0032] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0033] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0034] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0035] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0036] The system of the present invention aims to collect and analyze past inquiry data, predict potential questions from users, and provide appropriate solutions.

[0037] Data collection

[0038] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[0039] Data analysis

[0040] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[0041] Model Building

[0042] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[0043] Predictions and Recommendations

[0044] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and suggest solutions. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts that the user might ask, "I want to check the delivery status of my order."

[0045] notification

[0046] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions, and the device notifies the user of the relevant information. By allowing users to receive information in real time, inquiries can be reduced. For example, if a user wants to know the "current delivery status" three days after placing an order, the server will immediately provide that information, and the device will notify the user as a pop-up message.

[0047] Specific examples

[0048] Example 1: Online shopping site

[0049] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. On the third day after an order is placed, the terminal notifies the user that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0050] Example 2: IT Support Help Desk

[0051] The server collects and organizes inquiry data about "VPN connection problems" at the IT support help desk. Anticipating that new employees will frequently ask questions about VPN connections at the beginning of the week, the server prepares a "VPN connection guide" in advance. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[0052] In this way, the system of the present invention can improve the efficiency of customer support and also significantly improve user convenience.

[0053] The processing flow will be explained below.

[0054] Step 1:

[0055] The server connects to the customer support system and collects past inquiry data, including inquiry content, occurrence time, resolution method, and user attribute information.

[0056] Step 2:

[0057] The server removes duplicates and incomplete data from the collected data, and also normalizes the data, for example, correcting spelling errors and using the same case.

[0058] Step 3:

[0059] The server performs text analysis on past inquiry data, automatically classifying subjects (topics), and statistically analyzing the frequency and occurrence patterns of each topic.

[0060] Step 4:

[0061] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[0062] Step 5:

[0063] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[0064] Step 6:

[0065] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[0066] Step 7:

[0067] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time.

[0068] Step 8:

[0069] The server uses a trained model to predict potential questions based on the data it monitors, for example, if a user spends a lot of time on a particular page, it predicts questions related to that page.

[0070] Step 9:

[0071] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc.

[0072] Step 10:

[0073] The server converts the generated solution into a message format and prepares it for notification to the user.

[0074] Step 11:

[0075] The device will notify users of anticipated questions and solutions in real time. For example, if a user is viewing the "Order History" page, a pop-up window will display the current delivery status.

[0076] In this way, users can obtain the information they need in a timely manner, reducing the time and effort required to make inquiries and reducing support costs.

[0077] Example 1

[0078] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0079] Conventional customer support systems have difficulty presenting appropriate solutions to user inquiries in real time. Furthermore, the analysis of inquiry data and the construction of predictive models are inefficient, making it difficult to make accurate predictions. This results in poor user convenience and a lack of improvement in the efficiency of support operations.

[0080] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0081] In this invention, the server includes means for connecting to a customer support system and collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing the text data, means for performing text analysis on the past inquiry data to automatically classify themes and extract features, means for building a predictive model using a machine learning algorithm based on the features, means for monitoring user behavior logs in real time and predicting questions, and means for generating appropriate solutions to the predicted questions and notifying the user. This makes it possible to predict potential user questions with high accuracy and provide appropriate solutions in real time.

[0082] A "customer support system" is a system for responding to user inquiries and assisting in resolving problems.

[0083] "Inquiry Data" means data containing information relating to a question or problem that a user submits to our customer support system.

[0084] "Duplicate data" refers to data in which the same inquiry or information is recorded multiple times.

[0085] "Incomplete data" refers to data where necessary information is missing or recorded incompletely.

[0086] "Text data normalization" is a process that removes special characters and unnecessary spaces in order to standardize the format of text data and improve the quality of the data.

[0087] "Text analysis" is the process of analyzing the content of text data using natural language processing technology and extracting themes and keywords.

[0088] "Automatic subject categorization" is the process of automatically categorizing each piece of data into a specific category or topic based on the content of the inquiry data.

[0089] "Features" refer to important attributes or keywords that machine learning algorithms use to learn from data.

[0090] A "predictive model" is a model built using machine learning algorithms to forecast future trends or outcomes based on past data.

[0091] "Machine learning algorithms" refer to mathematical models and methods for learning patterns from data and making predictions or classifications.

[0092] An "action log" is data that records the history of operations and actions performed by a user on a system.

[0093] "Real-time monitoring" is the process by which a system tracks user behavior in real time and analyzes it immediately.

[0094] "Question prediction" is the process of inferring the questions or problems a user may have in the future based on their behavior and past data.

[0095] "Solution generation" is the process of automatically creating appropriate answers or solutions to anticipated questions.

[0096] "Notification" is the process of presenting solutions to anticipated questions to the user.

[0097] The system of the present invention is primarily intended to improve the efficiency of customer support and enhance user convenience. Each component of the present invention and a specific implementation method thereof will be described below.

[0098] Data collection and preprocessing

[0099] The server connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. When collecting data, the data is obtained from the database via an API connection. The collected data is preprocessed using programming languages ​​such as Python and R. Specifically, libraries such as Pandas and Numpy are used to remove duplicate and incomplete data and normalize text data.

[0100] Text analysis and feature extraction

[0101] The server uses natural language processing libraries (such as NLTK or spaCy) to perform text analysis of the collected inquiry data. It automatically classifies each inquiry by topic and extracts features, such as keywords in the inquiry, the time of occurrence, and user attributes.

[0102] Building a machine learning model

[0103] The server selects an appropriate machine learning algorithm based on the extracted features and builds a predictive model. Examples of algorithms used include random forests, support vector machines, and deep learning. The model is trained using libraries such as Scikit-learn, TensorFlow, and PyTorch. Cross-validation is performed to evaluate the model's performance, and the model is periodically retrained as necessary.

[0104] Monitoring user behavior logs and predicting questions

[0105] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time. Apache Kafka and RabbitMQ are used to capture real-time data. Potential questions that users may have are predicted based on the collected behavior logs.

[0106] Solution generation and notification

[0107] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions. This allows the device to notify the user of the relevant information, reducing the number of inquiries. For example, if a user wants to know the "current delivery status" three days after placing an order, the device will display a pop-up message saying, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0108] Specific examples

[0109] online shopping site

[0110] The server collects and organizes inquiry data about the "delivery status of an order" from the online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. For example, if a user wants to know the "current delivery status" three days after placing an order, the terminal will notify them that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0111] IT Support Help Desk

[0112] The server collects inquiry data from the IT support help desk about the problem of "VPN connection not working" and predicts that new employees will often ask this question at the beginning of the week. It prepares a "VPN connection guide" in advance, and when a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[0113] As described above, by effectively combining each component, the system of the present invention is able to predict potential questions from users with high accuracy and provide appropriate solutions in real time.

[0114] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0115] Step 1:

[0116] The server connects to the customer support system and collects past inquiry data. It uses an API to access the database as input, obtaining information such as the inquiry content, time of occurrence, solution, and user attribute information. The output is a collection of raw inquiry data. Specifically, it sends an HTTP request to obtain the raw data and saves the obtained data in JSON format.

[0117] Step 2:

[0118] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. The input is the raw query data obtained in step 1. The output is the preprocessed, clean data. Specifically, it uses the Pandas library to remove duplicate rows and rows containing missing values. It also removes special characters and excess whitespace to normalize the text.

[0119] Step 3:

[0120] The server performs text analysis on past inquiry data, automatically classifying topics and extracting features. The input is the clean text data obtained in step 2. The output is the analyzed topics and their corresponding features. Specifically, it uses a natural language processing library (e.g., NLTK or spaCy) to tokenize the text and apply a topic model (e.g., LDA) to perform topic classification.

[0121] Step 4:

[0122] The server selects an appropriate machine learning algorithm based on the extracted features and constructs a predictive model. The input is the features obtained in step 3. The output is the constructed predictive model. Specifically, it uses libraries such as Scikit-learn, TensorFlow, and PyTorch to train the model using a machine learning algorithm (e.g., random forest or support vector machine).

[0123] Step 5:

[0124] The server evaluates the performance of the predictive model using cross-validation and performs optimization. The input is the predictive model constructed in step 4 and the evaluation data. The output is the performance score of the evaluated model and the optimized model. Specifically, it applies the cross-validation method, calculates evaluation indicators such as the precision and recall of the model, and performs parameter tuning.

[0125] Step 6:

[0126] The server monitors user behavior logs in real time and predicts potential questions. The input is the user behavior log data. The output is the predicted potential questions. Specifically, it uses Apache Kafka or RabbitMQ to capture real-time data and applies a predictive model to predict questions.

[0127] Step 7:

[0128] The server generates an appropriate solution for the predicted question and sends a notification to the device. The input is the question predicted in step 6 and the solution database. The output is the solution provided to the user. Specifically, the server refers to FAQs, manuals, and video guides, selects an appropriate solution, and sends an automatically generated notification message to the device.

[0129] Step 8:

[0130] The terminal notifies the user of the solution received from the server. The input is the notification message from the server. The output is the solution displayed to the user. As a specific operation, the solution is presented to the user as a pop-up message or an alert.

[0131] Through these steps, the system is able to predict users' potential questions with high accuracy and provide appropriate solutions in real time.

[0132] (Application example 1)

[0133] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0134] When shopping online, users often make inquiries seeking specific information, which increases the burden on customer support and reduces user satisfaction. To solve this problem, it is necessary to quickly provide appropriate information before users make inquiries. Furthermore, there is a need for a system that can predict users' potential questions and notify them of appropriate solutions in real time.

[0135] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0136] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, and means for monitoring user behavior data in real time, thereby making it possible to predict potential user questions using the prediction model, predict inquiries in real time based on user behavior logs, and provide appropriate information via push notifications.

[0137] "Past inquiry data" refers to data such as questions and complaints received from users in the past, the content of the inquiry, the time of occurrence, the solution, and user attribute information.

[0138] A "predictive model" is a machine learning algorithm built to predict potential user questions based on patterns and features derived from collected and analyzed past inquiry data.

[0139] "User behavior data" refers to behavioral logs such as the actions a user takes on a website or application, click patterns, input content, and viewing time.

[0140] "Push notifications" are a feature that allows applications or systems to automatically send information or messages to a user's device in real time.

[0141] "Duplicate data" refers to data in which the same content is recorded multiple times, and is data that can cause a decrease in the accuracy of analysis.

[0142] "Incomplete data" is data that lacks necessary information or has insufficient content, and is a factor that hinders accurate analysis and prediction.

[0143] "Retraining" is the process of repeatedly learning an existing predictive model using new data in order to improve its accuracy.

[0144] "Providing relevant information through push notifications" means sending solutions and related information to a user's device in real time in response to questions that are inferred based on the user's behavior.

[0145] This invention is a system that collects and analyzes past inquiry data, predicts potential user questions, and provides appropriate solutions. This system is realized based on server, terminal, and user behavior data.

[0146] Data Collection and Cleansing

[0147] The server first connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data. This process can be performed using data processing tools such as Python's pandas and nltk libraries.

[0148] Data analysis and feature extraction

[0149] The server then performs text analysis on the collected inquiry data and automatically classifies the subject matter (topics). It then statistically analyzes the frequency and occurrence patterns of each topic and extracts inquiry features (e.g., keywords, occurrence time, user attributes, etc.). Machine learning libraries such as sklearn can be used for this process.

[0150] Building and retraining the model

[0151] The server selects a machine learning algorithm (e.g., random forest or support vector machine) based on these features and builds a predictive model. The model's performance is evaluated using cross-validation, and accuracy is improved by periodic retraining.

[0152] Real-time monitoring and user prediction

[0153] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and provide solutions. For example, if a user spends a long time viewing the "Order Delivery Status" page, the server predicts the question, "I would like to check the current delivery status."

[0154] Notification of Solution

[0155] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The device provides this information to the user via push notifications. These notifications are displayed on the user's smartphone or in a web application. Push notifications can be delivered using notification services such as Firebase Cloud Messaging (FCM) or Apple Push Notification Service (APNs).

[0156] Specific examples

[0157] For example, in an application on an online shopping site, the server collects and analyzes past inquiry data about delivery status and predicts that there will be many inquiries about "delivery status" a few days after an order is placed. When a user tries to check their order history, the server sends a notification such as, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0158] Examples of prompt statements

[0159] "I have a request. I would like to train an AI model that predicts that when a user searches for "I want to know the delivery status of my order," the user has a question about delivery. Based on the past inquiry data below, please extract features based on the inquiry content, inquiry time, user attributes, and resolution status, and build a model using a random forest classifier."

[0160] In this way, the system of the present invention can effectively reduce the burden on customer support and improve user convenience.

[0161] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0162] Step 1:

[0163] The server connects to the customer support system and collects past inquiry data. This data includes the inquiry content, time of occurrence, solution method, and user attribute information. The server stores this data in a database, which is used for subsequent data analysis.

[0164] Input: Inquiry data from the customer support system

[0165] Output: Query data stored in a database

[0166] Step 2:

[0167] The server removes duplicate and incomplete data from the collected data and normalizes the text data. Specifically, it uses a data cleaning algorithm to remove unnecessary whitespace and special characters and maintain data consistency, resulting in a clean dataset.

[0168] Input: Collected inquiry data

[0169] Output: A clean dataset

[0170] Step 3:

[0171] The server performs text analysis on the clean data to automatically classify the subject matter. Next, it statistically analyzes the frequency and occurrence patterns of each topic to extract query features. This process uses natural language processing libraries to extract keywords and perform topic modeling.

[0172] Input: A clean dataset

[0173] Output: Feature-extracted data

[0174] Step 4:

[0175] The server selects a machine learning algorithm based on the feature values ​​and builds a predictive model. The model's performance is evaluated through cross-validation, and regular retraining is performed to improve accuracy. This results in a model that can predict potential user questions with high accuracy.

[0176] Input: Feature-extracted data

[0177] Output: Highly accurate predictive model

[0178] Step 5:

[0179] When a user uses a website, the server monitors the user's activity log in real time. Specifically, the pages the user accesses, their click patterns, and the content of their input are collected and analyzed as logs. Based on this data, the user's current behavioral patterns can be understood.

[0180] Input: Real-time user activity log

[0181] Output: Parsed behavior log

[0182] Step 6:

[0183] The server uses real-time behavior logs to run a predictive model to predict potential questions users may have. For example, if a user spends a long time viewing a particular page, questions related to that page are predicted. This results in predicted questions.

[0184] Input: Parsed behavior log

[0185] Output: Prediction results for the question

[0186] Step 7:

[0187] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The generated information is sent to the user's device as a push notification. This notification provides the user with the information they need to quickly solve their problem.

[0188] Input: Predicted result of question

[0189] Output: Push notification to user device

[0190] Step 8:

[0191] Users can check the push notification sent to their device and obtain the appropriate solution or information, which allows them to quickly resolve the problem without having to contact the company.

[0192] Input: Push notification

[0193] Output: The solution or information provided to the user

[0194] In this way, through a series of steps, a system is realized that predicts potential questions from users and provides appropriate information in real time.

[0195] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0196] The system of the present invention aims to collect and analyze past inquiry data, predict users' potential questions, and provide appropriate solutions, as well as combine an emotion engine to recognize users' emotions and provide more appropriate and customized support.

[0197] Data collection

[0198] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[0199] Data analysis

[0200] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[0201] Model Building

[0202] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[0203] emotion recognition

[0204] The server combines an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, thereby identifying the user's current emotion (e.g., joy, anger, sadness, surprise, etc.).

[0205] Predictions and Recommendations

[0206] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts the question, "I'd like to check the delivery status of my order." If the user's emotions indicate irritability, the server provides a particularly fast and courteous response.

[0207] notification

[0208] The server generates appropriate solutions for predicted questions. Solutions can include FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state and provides information at the appropriate time. For example, if the user is dissatisfied, the server will send a notification using particularly polite language and with a prompt response.

[0209] Specific examples

[0210] Example 1: Online shopping site

[0211] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0212] Example 2: IT Support Help Desk

[0213] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[0214] In this way, the system of the present invention can improve the efficiency of customer support and significantly improve user convenience. Furthermore, by combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[0215] The processing flow will be explained below.

[0216] Step 1:

[0217] The server connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[0218] Step 2:

[0219] The server removes duplicates and incomplete data from the collected data and normalizes the text data, for example, by standardizing uppercase and lowercase letters and correcting spelling errors.

[0220] Step 3:

[0221] The server performs text analysis on past inquiry data and automatically classifies subjects (topics), allowing for statistical analysis of the frequency and occurrence patterns of each topic.

[0222] Step 4:

[0223] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[0224] Step 5:

[0225] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[0226] Step 6:

[0227] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[0228] Step 7:

[0229] The server uses an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, including joy, anger, sadness, surprise, etc.

[0230] Step 8:

[0231] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time viewing a particular page, it predicts questions related to that page.

[0232] Step 9:

[0233] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state.

[0234] Step 10:

[0235] The server converts the generated solution into a message format and prepares it for notification to the user, including particularly polite language and prompt responses depending on the user's emotional state.

[0236] Step 11:

[0237] The device will notify the user of anticipated questions and solutions in real time. For example, if the user is viewing the "Order History" page, the device will notify the user, "The current delivery status is 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0238] In this way, users can obtain the information they need in a timely manner, saving them the trouble of making inquiries and reducing support response costs.By combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[0239] Example 2

[0240] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0241] Modern customer support systems are required to improve the speed and appropriateness of responses to user inquiries, but conventional systems lack the ability to take into account the user's emotional state. Therefore, to improve user satisfaction, a system that recognizes the user's emotions in real time and provides more customized responses is needed.

[0242] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0243] In this invention, the server includes means for collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing text data, means for analyzing the past inquiry data to classify topics and extract features, means for building a predictive model using a machine learning algorithm based on the extracted features, means for evaluating the performance of the predictive model and optimizing its accuracy, means for monitoring user behavior data in real time and recognizing the user's emotional state using an emotion engine, means for predicting potential user questions using the predictive model, and means for generating appropriate solutions to the predicted questions, customizing notification content based on the user's emotional state, and providing information at an appropriate time, thereby enabling appropriate and prompt customized support according to the user's emotional state.

[0244] "Past inquiry data" refers to historical information such as the content of inquiries made by the user, the time of occurrence, the solution, and user attribute information.

[0245] "Duplicate data" refers to data in which the same content is recorded multiple times.

[0246] "Incomplete data" refers to data that is missing necessary information.

[0247] "Text data normalization" refers to the process of converting text data into a unified format to make it easier to analyze.

[0248] "Topic" refers to the themes or categories used to classify the content of text data.

[0249] "Features" refer to important elements or parameters that machine learning algorithms use to analyze data.

[0250] A "machine learning algorithm" refers to a computational method for learning patterns and rules from data and making predictions and classifications.

[0251] "Predictive model" refers to a statistical or machine learning-based model built to forecast future events based on historical data.

[0252] "Cross-validation" refers to a method of dividing data and performing cross-validation to evaluate the performance of a model.

[0253] An "emotion engine" refers to algorithms or software that automatically recognize a user's emotional state from their input and behavioral data.

[0254] "Potential questions" refer to questions that are anticipated before a user explicitly asks them.

[0255] "Solutions" refer to answers or solutions to potential user questions or problems.

[0256] "Customizing notification content" refers to changing the content and presentation of the information to be notified according to the user's situation and emotions.

[0257] The system of the present invention aims to collect and analyze past inquiry data, predict potential user questions, and provide appropriate solutions. Furthermore, it uses an emotion engine to recognize user emotions and provide customized support.

[0258] Hardware and software configuration

[0259] The server is used to implement the following functions:

[0260] Data collection and cleansing: We use a dedicated script to connect to your customer support system and collect inquiry data.

[0261] Natural Language Processing (NLP) engine: Used to analyze the query content and classify the subject (topic).

[0262] Machine learning model: Based on the extracted features, a predictive model is constructed using an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.).

[0263] Emotion recognition engine: Recognizes the user's emotional state in real time based on user behavioral data and input information.

[0264] The terminal is used to implement the following functions:

[0265] Notifications: Generate appropriate solutions to predicted questions and provide customized notification content based on the user's emotional state.

[0266] System operation flow

[0267] 1. Data Collection

[0268] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[0269] Duplicate and incomplete data is removed from the collected data, and the text data is normalized.

[0270] 2. Data analysis

[0271] The server performs text analysis on past inquiry data to classify subjects (topics), and then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry.

[0272] 3. Model Building

[0273] The server selects and trains an appropriate machine learning algorithm based on the extracted features. It evaluates the model's performance using cross-validation and optimizes its accuracy. It then periodically retrains the model to improve its accuracy.

[0274] 4. Emotion recognition

[0275] The server combines an emotion engine to recognize the user's emotional state in real time from the user's behavioral data and input information.

[0276] 5. Predictions and Recommendations

[0277] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing the "Order History" page on a website, it predicts the question, "I want to check the delivery status of my order."

[0278] 6. Notification

[0279] The server generates appropriate solutions for predicted questions, such as FAQs, manuals, video guides, etc. It customizes notifications based on the user's emotional state and provides information at the right time.

[0280] Specific examples

[0281] Example 1: Online shopping site

[0282] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0283] Example 2: IT Support Help Desk

[0284] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[0285] Examples of prompt statements

[0286] Here are some example prompts to input to a generative AI model:

[0287] "How can I implement an algorithm that predicts potential questions users might have based on past inquiry data?"

[0288] "Describe the design of a system that recognizes a user's emotions and responds appropriately."

[0289] This system will significantly improve the efficiency of customer support, enhance user convenience, and provide more accurate and courteous support that reflects the user's emotional state.

[0290] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0291] System program processing flow

[0292] Step 1: Data collection

[0293] Input: Connection information for customer support system

[0294] Output: Collected historical inquiry data

[0295] What happens:

[0296] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, time of occurrence, resolution method, and user attribute information. Data collection is generally performed using APIs or database queries.

[0297] Step 2: Cleanse the data

[0298] Input: Collected inquiry data

[0299] Output: Cleansed data

[0300] What happens:

[0301] The server analyzes the collected data and removes duplicates and incomplete data. Specifically, it performs the following processes:

[0302] If there are multiple identical queries in the database, they are merged into one.

[0303] Detect rows with missing data or incomplete information and filter them out or complete them.

[0304] Step 3: Normalize the text data

[0305] Input: Cleansed data

[0306] Output: Normalized text data

[0307] What happens:

[0308] The server normalizes the collected text data, which includes lowercasing all text, removing special characters and extra spaces, and standardizing the text according to the user's language and region.

[0309] Step 4: Text Analysis

[0310] Input: normalized text data

[0311] Output: Categorized topic data

[0312] What happens:

[0313] The server uses a natural language processing (NLP) engine to classify the topics of the inquiry data by extracting keywords from the text data and categorizing them into categories (e.g., order status, returns, technical support, etc.).

[0314] Step 5: Statistical analysis

[0315] Input: Categorized topic data

[0316] Output: Thematic frequency and pattern analysis results

[0317] What happens:

[0318] The server statistically analyzes the frequency and occurrence patterns of each topic. For example, by analyzing whether a particular topic occurs frequently during a particular time period, it can identify problem occurrence patterns.

[0319] Step 6: Feature extraction

[0320] Input: Categorized topic data and analysis results

[0321] Output: Extracted features

[0322] What happens:

[0323] The server extracts features (keywords, occurrence time, user attributes, etc.) for each topic, allowing for a detailed understanding of the characteristics of each topic, which can be used in the next machine learning process.

[0324] Step 7: Select and train a machine learning model

[0325] Input: Extracted features

[0326] Output: A trained machine learning model

[0327] What happens:

[0328] The server selects an appropriate machine learning algorithm based on the extracted features and trains the model, using algorithms such as random forest, SVM, and deep learning to build a predictive model based on the query content.

[0329] Step 8: Evaluate and optimize the model

[0330] Input: A trained machine learning model

[0331] Output: Evaluated and optimized model

[0332] What happens:

[0333] The server evaluates the model's performance using cross-validation and adjusts hyperparameters as needed, for example, by performing 5-fold cross-validation and optimizing for accuracy above 90%.

[0334] Step 9: Emotion Recognition

[0335] Input: User behavior data and input information

[0336] Output: Classified emotion data

[0337] What happens:

[0338] The server uses an emotion engine to recognize the user's emotional state in real time based on their behavioral data and input information. For example, if a user sends messages in rapid succession, it can detect irritation.

[0339] Step 10: Anticipate potential questions

[0340] Input: User behavior log and emotional state

[0341] Output: Predicted potential questions

[0342] What happens:

[0343] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing their "order history," it predicts that they will "check the delivery status."

[0344] Step 11: Generate a solution

[0345] Input: predicted potential questions

[0346] Output: The generated solution

[0347] What happens:

[0348] The server generates appropriate solutions for predicted questions, providing specific solutions based on FAQs, manuals, video guides, etc.

[0349] Step 12: Notification

[0350] Input: Generated solution and user's emotional state

[0351] Output: Customized notification content

[0352] What happens:

[0353] The server customizes the notification content based on the user's emotional state and provides information at the appropriate time, such as "Your order has now been shipped and is expected to arrive tomorrow. If you have any questions, please feel free to contact us."

[0354] The above are the specific processing steps of the system program.

[0355] (Application example 2)

[0356] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0357] There is a need for a support system that can quickly and accurately respond to any problems or questions that users of autonomous vehicles may encounter while using the vehicle. In particular, it is necessary to improve user satisfaction by recognizing the user's emotional state and responding flexibly accordingly.

[0358] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0359] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, means for monitoring user behavior data in real time, means for predicting potential questions from users using the prediction model, means for notifying users of appropriate solutions to the predicted questions, and means for recognizing the user's emotional state and customizing a response.

[0360] This will enable us to respond quickly and flexibly to the various questions and problems that users of autonomous vehicles may encounter.Furthermore, by providing customized responses based on the user's emotional state, we can significantly improve user satisfaction.

[0361] "Past inquiry data" refers to records of questions and trouble reports submitted by users in the past, including the content of the inquiry, the date and time of the occurrence, the solution, and user attribute information.

[0362] "Analysis" is the act of extracting meaning and trends from collected data using statistical or computational techniques.

[0363] A "predictive model" is a collection of mathematical and statistical methods for predicting future conditions or outcomes based on past data.

[0364] "User behavior data" refers to data that records the actions and movements of users when using a system or device, including clicks, viewing time, and operation sequences.

[0365] "Real-time monitoring" refers to the system's ability to instantly collect and analyze user operations and situations.

[0366] "Means for predicting potential questions" refers to a technology that uses a predictive model to predict in advance what questions a user will ask in the future.

[0367] "Means for notifying appropriate solutions" refers to methods for informing users of the optimal solutions to anticipated questions or problems.

[0368] "Means for recognizing emotional states" refers to technology that identifies the emotions (e.g., joy, anger, sadness, surprise, etc.) present in a user's mind from their text and behavioral data.

[0369] "Customized response" refers to services and support that are optimally tailored to each individual user's situation and emotional state.

[0370] The present invention relates to a user support system that can be used in an autonomous vehicle. Specific embodiments of the system will be described below.

[0371] First, the server collects past inquiry data. This data includes questions and trouble reports submitted by users in the past, and records the inquiry content, date and time of occurrence, solution method, and user attribute information. The collected data is then normalized to remove duplicate and incomplete data.

[0372] The server then analyzes this data and builds a predictive model. Specifically, it extracts query features (keywords, time of occurrence, user attributes, etc.) from the data and trains the model using machine learning algorithms (e.g., random forest, support vector machine, deep learning, etc.). This predictive model is used to predict potential user questions.

[0373] The server also monitors user behavior data in real time, recording clicks, browsing time, and the sequence of operations as the user interacts with the in-car infotainment system. Based on this information, the server predicts the type of questions or concerns the user may have. It also uses an emotion recognition engine to identify the user's emotional state. For example, if the user shows signs of frustration, the system can provide a particularly prompt and courteous response.

[0374] For predicted questions, the server generates appropriate solutions and notifies the user. Solutions include FAQs, manuals, video guides, etc. This allows users to quickly find a way to solve their problems.

[0375] For example, if a user in a car says, "My car won't start," the emotion recognition engine recognizes the user's frustration. The server uses a predictive model to anticipate possible questions and solutions, and responds promptly and courteously. The notification system sends a message saying, "Please try restarting the engine. If the problem persists, please contact our support center."

[0376] An example of a prompt using a generative AI model is "When a user expresses frustration by asking if their car won't start, what should I do?" Using this prompt, the required solution can be quickly generated.

[0377] As described above, the present invention provides a system that can quickly and flexibly respond to questions and problems that users of autonomous vehicles may encounter. Furthermore, by providing customized responses based on the user's emotional state, user satisfaction can be significantly improved.

[0378] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0379] Step 1: Data collection

[0380] The server collects past inquiry data from the customer support system. This data includes inquiry content, the date and time of occurrence, the solution, and user attribute information. As input, this information is extracted from the inquiry database and temporarily stored inside the server. The output is inquiry data formatted in a format that can be used in subsequent processing steps.

[0381] Step 2: Data Preprocessing

[0382] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. It receives the temporarily stored query data as input and uses data cleansing techniques. Specifically, it executes database queries to remove duplicate rows, handle missing values, and normalize the data into a consistent text format. The output is the clean query data.

[0383] Step 3: Text analysis

[0384] The server performs text analysis on the preprocessed data using natural language processing techniques. The input is the cleaned query data, and the output is the extraction of topics and keywords for each query. Specific operations include text tokenization, morphological analysis, and application of topic models.

[0385] Step 4: Building a predictive model

[0386] The server builds a predictive model based on the features obtained from text analysis. The input is the analyzed feature data. The server uses a machine learning algorithm (random forest or deep learning) to train the model. The output is a model that can predict potential questions from users. Specific operations include training and cross-validation on a dataset.

[0387] Step 5: Emotion Recognition

[0388] The server performs real-time emotion recognition based on the user's behavioral data and text input. The input is the text and behavioral data entered by the user into the device. The server uses an emotion recognition engine to analyze the user's emotional state (joy, anger, sadness, surprise, etc.). The output is the result of the emotional state. The specific operation is to execute a text emotion analysis algorithm.

[0389] Step 6: Anticipate potential questions

[0390] The server combines the emotion recognition results with a prediction model to predict the user's potential questions. The input is the user's current emotional state and past behavioral data. The output is a list of likely questions. Specific operations include running a label prediction algorithm.

[0391] Step 7: Generate a suitable solution

[0392] The server generates appropriate solutions for predicted questions. The input is a list of predicted questions, and the output is solutions such as FAQs, manuals, video guides, etc. Specific operations include searching and composing relevant solutions from a database.

[0393] Step 8: Notify users

[0394] The server customizes the generated solution according to the user's emotional state and notifies the terminal at an appropriate time. The input is the solution and the user's emotional state. The output is a customized notification message. Specific operations include message tone adjustment and notification scheduling.

[0395] Through these processing steps, the server can provide quick and appropriate solutions to potential user questions within the autonomous vehicle, realizing emotionally responsive and customized support.

[0396] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0397] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0398] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0399] [Second embodiment]

[0400] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0401] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0402] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0403] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0404] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0405] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0406] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0407] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0408] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0409] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0410] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0411] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0412] The system of the present invention aims to collect and analyze past inquiry data, predict potential questions from users, and provide appropriate solutions.

[0413] Data collection

[0414] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[0415] Data analysis

[0416] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[0417] Model Building

[0418] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[0419] Predictions and Recommendations

[0420] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and suggest solutions. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts that the user might ask, "I want to check the delivery status of my order."

[0421] notification

[0422] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions, and the device notifies the user of the relevant information. By allowing users to receive information in real time, inquiries can be reduced. For example, if a user wants to know the "current delivery status" three days after placing an order, the server will immediately provide that information, and the device will notify the user as a pop-up message.

[0423] Specific examples

[0424] Example 1: Online shopping site

[0425] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. On the third day after an order is placed, the terminal notifies the user that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0426] Example 2: IT Support Help Desk

[0427] The server collects and organizes inquiry data about "VPN connection problems" at the IT support help desk. Anticipating that new employees will frequently ask questions about VPN connections at the beginning of the week, the server prepares a "VPN connection guide" in advance. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[0428] In this way, the system of the present invention can improve the efficiency of customer support and also significantly improve user convenience.

[0429] The processing flow will be explained below.

[0430] Step 1:

[0431] The server connects to the customer support system and collects past inquiry data, including inquiry content, occurrence time, resolution method, and user attribute information.

[0432] Step 2:

[0433] The server removes duplicates and incomplete data from the collected data, and also normalizes the data, for example, correcting spelling errors and using the same case.

[0434] Step 3:

[0435] The server performs text analysis on past inquiry data, automatically classifying subjects (topics), and statistically analyzing the frequency and occurrence patterns of each topic.

[0436] Step 4:

[0437] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[0438] Step 5:

[0439] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[0440] Step 6:

[0441] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[0442] Step 7:

[0443] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time.

[0444] Step 8:

[0445] The server uses a trained model to predict potential questions based on the data it monitors, for example, if a user spends a lot of time on a particular page, it predicts questions related to that page.

[0446] Step 9:

[0447] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc.

[0448] Step 10:

[0449] The server converts the generated solution into a message format and prepares it for notification to the user.

[0450] Step 11:

[0451] The device will notify users of anticipated questions and solutions in real time. For example, if a user is viewing the "Order History" page, a pop-up window will display the current delivery status.

[0452] In this way, users can obtain the information they need in a timely manner, reducing the time and effort required to make inquiries and reducing support costs.

[0453] Example 1

[0454] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0455] Conventional customer support systems have difficulty presenting appropriate solutions to user inquiries in real time. Furthermore, the analysis of inquiry data and the construction of predictive models are inefficient, making it difficult to make accurate predictions. This results in poor user convenience and a lack of improvement in the efficiency of support operations.

[0456] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0457] In this invention, the server includes means for connecting to a customer support system and collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing the text data, means for performing text analysis on the past inquiry data to automatically classify themes and extract features, means for building a predictive model using a machine learning algorithm based on the features, means for monitoring user behavior logs in real time and predicting questions, and means for generating appropriate solutions to the predicted questions and notifying the user. This makes it possible to predict potential user questions with high accuracy and provide appropriate solutions in real time.

[0458] A "customer support system" is a system for responding to user inquiries and assisting in resolving problems.

[0459] "Inquiry Data" means data containing information relating to a question or problem that a user submits to our customer support system.

[0460] "Duplicate data" refers to data in which the same inquiry or information is recorded multiple times.

[0461] "Incomplete data" refers to data where necessary information is missing or recorded incompletely.

[0462] "Text data normalization" is a process that removes special characters and unnecessary spaces in order to standardize the format of text data and improve the quality of the data.

[0463] "Text analysis" is the process of analyzing the content of text data using natural language processing technology and extracting themes and keywords.

[0464] "Automatic subject categorization" is the process of automatically categorizing each piece of data into a specific category or topic based on the content of the inquiry data.

[0465] "Features" refer to important attributes or keywords that machine learning algorithms use to learn from data.

[0466] A "predictive model" is a model built using machine learning algorithms to forecast future trends or outcomes based on past data.

[0467] "Machine learning algorithms" refer to mathematical models and methods for learning patterns from data and making predictions or classifications.

[0468] An "action log" is data that records the history of operations and actions performed by a user on a system.

[0469] "Real-time monitoring" is the process by which a system tracks user behavior in real time and analyzes it immediately.

[0470] "Question prediction" is the process of inferring the questions or problems a user may have in the future based on their behavior and past data.

[0471] "Solution generation" is the process of automatically creating appropriate answers or solutions to anticipated questions.

[0472] "Notification" is the process of presenting solutions to anticipated questions to the user.

[0473] The system of the present invention is primarily intended to improve the efficiency of customer support and enhance user convenience. Each component of the present invention and a specific implementation method thereof will be described below.

[0474] Data collection and preprocessing

[0475] The server connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. When collecting data, the data is obtained from the database via an API connection. The collected data is preprocessed using programming languages ​​such as Python and R. Specifically, libraries such as Pandas and Numpy are used to remove duplicate and incomplete data and normalize text data.

[0476] Text analysis and feature extraction

[0477] The server uses natural language processing libraries (such as NLTK or spaCy) to perform text analysis of the collected inquiry data. It automatically classifies each inquiry by topic and extracts features, such as keywords in the inquiry, the time of occurrence, and user attributes.

[0478] Building a machine learning model

[0479] The server selects an appropriate machine learning algorithm based on the extracted features and builds a predictive model. Examples of algorithms used include random forests, support vector machines, and deep learning. The model is trained using libraries such as Scikit-learn, TensorFlow, and PyTorch. Cross-validation is performed to evaluate the model's performance, and the model is periodically retrained as necessary.

[0480] Monitoring user behavior logs and predicting questions

[0481] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time. Apache Kafka and RabbitMQ are used to capture real-time data. Potential questions that users may have are predicted based on the collected behavior logs.

[0482] Solution generation and notification

[0483] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions. This allows the device to notify the user of the relevant information, reducing the number of inquiries. For example, if a user wants to know the "current delivery status" three days after placing an order, the device will display a pop-up message saying, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0484] Specific examples

[0485] online shopping site

[0486] The server collects and organizes inquiry data about the "delivery status of an order" from the online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. For example, if a user wants to know the "current delivery status" three days after placing an order, the terminal will notify them that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0487] IT Support Help Desk

[0488] The server collects inquiry data from the IT support help desk about the problem of "VPN connection not working" and predicts that new employees will often ask this question at the beginning of the week. It prepares a "VPN connection guide" in advance, and when a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[0489] As described above, by effectively combining each component, the system of the present invention is able to predict potential questions from users with high accuracy and provide appropriate solutions in real time.

[0490] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0491] Step 1:

[0492] The server connects to the customer support system and collects past inquiry data. It uses an API to access the database as input, obtaining information such as the inquiry content, time of occurrence, solution, and user attribute information. The output is a collection of raw inquiry data. Specifically, it sends an HTTP request to obtain the raw data and saves the obtained data in JSON format.

[0493] Step 2:

[0494] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. The input is the raw query data obtained in step 1. The output is the preprocessed, clean data. Specifically, it uses the Pandas library to remove duplicate rows and rows containing missing values. It also removes special characters and excess whitespace to normalize the text.

[0495] Step 3:

[0496] The server performs text analysis on past inquiry data, automatically classifying topics and extracting features. The input is the clean text data obtained in step 2. The output is the analyzed topics and their corresponding features. Specifically, it uses a natural language processing library (e.g., NLTK or spaCy) to tokenize the text and apply a topic model (e.g., LDA) to perform topic classification.

[0497] Step 4:

[0498] The server selects an appropriate machine learning algorithm based on the extracted features and constructs a predictive model. The input is the features obtained in step 3. The output is the constructed predictive model. Specifically, it uses libraries such as Scikit-learn, TensorFlow, and PyTorch to train the model using a machine learning algorithm (e.g., random forest or support vector machine).

[0499] Step 5:

[0500] The server evaluates the performance of the predictive model using cross-validation and performs optimization. The input is the predictive model constructed in step 4 and the evaluation data. The output is the performance score of the evaluated model and the optimized model. Specifically, it applies the cross-validation method, calculates evaluation indicators such as the precision and recall of the model, and performs parameter tuning.

[0501] Step 6:

[0502] The server monitors user behavior logs in real time and predicts potential questions. The input is the user behavior log data. The output is the predicted potential questions. Specifically, it uses Apache Kafka or RabbitMQ to capture real-time data and applies a predictive model to predict questions.

[0503] Step 7:

[0504] The server generates an appropriate solution for the predicted question and sends a notification to the device. The input is the question predicted in step 6 and the solution database. The output is the solution provided to the user. Specifically, the server refers to FAQs, manuals, and video guides, selects an appropriate solution, and sends an automatically generated notification message to the device.

[0505] Step 8:

[0506] The terminal notifies the user of the solution received from the server. The input is the notification message from the server. The output is the solution displayed to the user. As a specific operation, the solution is presented to the user as a pop-up message or an alert.

[0507] Through these steps, the system is able to predict users' potential questions with high accuracy and provide appropriate solutions in real time.

[0508] (Application example 1)

[0509] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0510] When shopping online, users often make inquiries seeking specific information, which increases the burden on customer support and reduces user satisfaction. To solve this problem, it is necessary to quickly provide appropriate information before users make inquiries. Furthermore, there is a need for a system that can predict users' potential questions and notify them of appropriate solutions in real time.

[0511] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0512] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, and means for monitoring user behavior data in real time, thereby making it possible to predict potential user questions using the prediction model, predict inquiries in real time based on user behavior logs, and provide appropriate information via push notifications.

[0513] "Past inquiry data" refers to data such as questions and complaints received from users in the past, the content of the inquiry, the time of occurrence, the solution, and user attribute information.

[0514] A "predictive model" is a machine learning algorithm built to predict potential user questions based on patterns and features derived from collected and analyzed past inquiry data.

[0515] "User behavior data" refers to behavioral logs such as the actions a user takes on a website or application, click patterns, input content, and viewing time.

[0516] "Push notifications" are a feature that allows applications or systems to automatically send information or messages to a user's device in real time.

[0517] "Duplicate data" refers to data in which the same content is recorded multiple times, and is data that can cause a decrease in the accuracy of analysis.

[0518] "Incomplete data" is data that lacks necessary information or has insufficient content, and is a factor that hinders accurate analysis and prediction.

[0519] "Retraining" is the process of repeatedly learning an existing predictive model using new data in order to improve its accuracy.

[0520] "Providing relevant information through push notifications" means sending solutions and related information to a user's device in real time in response to questions that are inferred based on the user's behavior.

[0521] This invention is a system that collects and analyzes past inquiry data, predicts potential user questions, and provides appropriate solutions. This system is realized based on server, terminal, and user behavior data.

[0522] Data Collection and Cleansing

[0523] The server first connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data. This process can be performed using data processing tools such as Python's pandas and nltk libraries.

[0524] Data analysis and feature extraction

[0525] The server then performs text analysis on the collected inquiry data and automatically classifies the subject matter (topics). It then statistically analyzes the frequency and occurrence patterns of each topic and extracts inquiry features (e.g., keywords, occurrence time, user attributes, etc.). Machine learning libraries such as sklearn can be used for this process.

[0526] Building and retraining the model

[0527] The server selects a machine learning algorithm (e.g., random forest or support vector machine) based on these features and builds a predictive model. The model's performance is evaluated using cross-validation, and accuracy is improved by periodic retraining.

[0528] Real-time monitoring and user prediction

[0529] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and provide solutions. For example, if a user spends a long time viewing the "Order Delivery Status" page, the server predicts the question, "I would like to check the current delivery status."

[0530] Notification of Solution

[0531] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The device provides this information to the user via push notifications. These notifications are displayed on the user's smartphone or in a web application. Push notifications can be delivered using notification services such as Firebase Cloud Messaging (FCM) or Apple Push Notification Service (APNs).

[0532] Specific examples

[0533] For example, in an application on an online shopping site, the server collects and analyzes past inquiry data about delivery status and predicts that there will be many inquiries about "delivery status" a few days after an order is placed. When a user tries to check their order history, the server sends a notification such as, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0534] Examples of prompt statements

[0535] "I have a request. I would like to train an AI model that predicts that when a user searches for "I want to know the delivery status of my order," the user has a question about delivery. Based on the past inquiry data below, please extract features based on the inquiry content, inquiry time, user attributes, and resolution status, and build a model using a random forest classifier."

[0536] In this way, the system of the present invention can effectively reduce the burden on customer support and improve user convenience.

[0537] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0538] Step 1:

[0539] The server connects to the customer support system and collects past inquiry data. This data includes the inquiry content, time of occurrence, solution method, and user attribute information. The server stores this data in a database, which is used for subsequent data analysis.

[0540] Input: Inquiry data from the customer support system

[0541] Output: Query data stored in a database

[0542] Step 2:

[0543] The server removes duplicate and incomplete data from the collected data and normalizes the text data. Specifically, it uses a data cleaning algorithm to remove unnecessary whitespace and special characters and maintain data consistency, resulting in a clean dataset.

[0544] Input: Collected inquiry data

[0545] Output: A clean dataset

[0546] Step 3:

[0547] The server performs text analysis on the clean data to automatically classify the subject matter. Next, it statistically analyzes the frequency and occurrence patterns of each topic to extract query features. This process uses natural language processing libraries to extract keywords and perform topic modeling.

[0548] Input: A clean dataset

[0549] Output: Feature-extracted data

[0550] Step 4:

[0551] The server selects a machine learning algorithm based on the feature values ​​and builds a predictive model. The model's performance is evaluated through cross-validation, and regular retraining is performed to improve accuracy. This results in a model that can predict potential user questions with high accuracy.

[0552] Input: Feature-extracted data

[0553] Output: Highly accurate predictive model

[0554] Step 5:

[0555] When a user uses a website, the server monitors the user's activity log in real time. Specifically, the pages the user accesses, their click patterns, and the content of their input are collected and analyzed as logs. Based on this data, the user's current behavioral patterns can be understood.

[0556] Input: Real-time user activity log

[0557] Output: Parsed behavior log

[0558] Step 6:

[0559] The server uses real-time behavior logs to run a predictive model to predict potential questions users may have. For example, if a user spends a long time viewing a particular page, questions related to that page are predicted. This results in predicted questions.

[0560] Input: Parsed behavior log

[0561] Output: Prediction results for the question

[0562] Step 7:

[0563] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The generated information is sent to the user's device as a push notification. This notification provides the user with the information they need to quickly solve their problem.

[0564] Input: Predicted result of question

[0565] Output: Push notification to user device

[0566] Step 8:

[0567] Users can check the push notification sent to their device and obtain the appropriate solution or information, which allows them to quickly resolve the problem without having to contact the company.

[0568] Input: Push notification

[0569] Output: The solution or information provided to the user

[0570] In this way, through a series of steps, a system is realized that predicts potential questions from users and provides appropriate information in real time.

[0571] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0572] The system of the present invention aims to collect and analyze past inquiry data, predict users' potential questions, and provide appropriate solutions, as well as combine an emotion engine to recognize users' emotions and provide more appropriate and customized support.

[0573] Data collection

[0574] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[0575] Data analysis

[0576] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[0577] Model Building

[0578] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[0579] emotion recognition

[0580] The server combines an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, thereby identifying the user's current emotion (e.g., joy, anger, sadness, surprise, etc.).

[0581] Predictions and Recommendations

[0582] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts the question, "I'd like to check the delivery status of my order." If the user's emotions indicate irritability, the server provides a particularly fast and courteous response.

[0583] notification

[0584] The server generates appropriate solutions for predicted questions. Solutions can include FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state and provides information at the appropriate time. For example, if the user is dissatisfied, the server will send a notification using particularly polite language and with a prompt response.

[0585] Specific examples

[0586] Example 1: Online shopping site

[0587] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0588] Example 2: IT Support Help Desk

[0589] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[0590] In this way, the system of the present invention can improve the efficiency of customer support and significantly improve user convenience. Furthermore, by combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[0591] The processing flow will be explained below.

[0592] Step 1:

[0593] The server connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[0594] Step 2:

[0595] The server removes duplicates and incomplete data from the collected data and normalizes the text data, for example, by standardizing uppercase and lowercase letters and correcting spelling errors.

[0596] Step 3:

[0597] The server performs text analysis on past inquiry data and automatically classifies subjects (topics), allowing for statistical analysis of the frequency and occurrence patterns of each topic.

[0598] Step 4:

[0599] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[0600] Step 5:

[0601] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[0602] Step 6:

[0603] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[0604] Step 7:

[0605] The server uses an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, including joy, anger, sadness, surprise, etc.

[0606] Step 8:

[0607] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time viewing a particular page, it predicts questions related to that page.

[0608] Step 9:

[0609] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state.

[0610] Step 10:

[0611] The server converts the generated solution into a message format and prepares it for notification to the user, including particularly polite language and prompt responses depending on the user's emotional state.

[0612] Step 11:

[0613] The device will notify the user of anticipated questions and solutions in real time. For example, if the user is viewing the "Order History" page, the device will notify the user, "The current delivery status is 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0614] In this way, users can obtain the information they need in a timely manner, saving them the trouble of making inquiries and reducing support response costs.By combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[0615] Example 2

[0616] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0617] Modern customer support systems are required to improve the speed and appropriateness of responses to user inquiries, but conventional systems lack the ability to take into account the user's emotional state. Therefore, to improve user satisfaction, a system that recognizes the user's emotions in real time and provides more customized responses is needed.

[0618] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0619] In this invention, the server includes means for collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing text data, means for analyzing the past inquiry data to classify topics and extract features, means for building a predictive model using a machine learning algorithm based on the extracted features, means for evaluating the performance of the predictive model and optimizing its accuracy, means for monitoring user behavior data in real time and recognizing the user's emotional state using an emotion engine, means for predicting potential user questions using the predictive model, and means for generating appropriate solutions to the predicted questions, customizing notification content based on the user's emotional state, and providing information at an appropriate time, thereby enabling appropriate and prompt customized support according to the user's emotional state.

[0620] "Past inquiry data" refers to historical information such as the content of inquiries made by the user, the time of occurrence, the solution, and user attribute information.

[0621] "Duplicate data" refers to data in which the same content is recorded multiple times.

[0622] "Incomplete data" refers to data that is missing necessary information.

[0623] "Text data normalization" refers to the process of converting text data into a unified format to make it easier to analyze.

[0624] "Topic" refers to the themes or categories used to classify the content of text data.

[0625] "Features" refer to important elements or parameters that machine learning algorithms use to analyze data.

[0626] A "machine learning algorithm" refers to a computational method for learning patterns and rules from data and making predictions and classifications.

[0627] "Predictive model" refers to a statistical or machine learning-based model built to forecast future events based on historical data.

[0628] "Cross-validation" refers to a method of dividing data and performing cross-validation to evaluate the performance of a model.

[0629] An "emotion engine" refers to algorithms or software that automatically recognize a user's emotional state from their input and behavioral data.

[0630] "Potential questions" refer to questions that are anticipated before a user explicitly asks them.

[0631] "Solutions" refer to answers or solutions to potential user questions or problems.

[0632] "Customizing notification content" refers to changing the content and presentation of the information to be notified according to the user's situation and emotions.

[0633] The system of the present invention aims to collect and analyze past inquiry data, predict potential user questions, and provide appropriate solutions. Furthermore, it uses an emotion engine to recognize user emotions and provide customized support.

[0634] Hardware and software configuration

[0635] The server is used to implement the following functions:

[0636] Data collection and cleansing: We use a dedicated script to connect to your customer support system and collect inquiry data.

[0637] Natural Language Processing (NLP) engine: Used to analyze the query content and classify the subject (topic).

[0638] Machine learning model: Based on the extracted features, a predictive model is constructed using an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.).

[0639] Emotion recognition engine: Recognizes the user's emotional state in real time based on user behavioral data and input information.

[0640] The terminal is used to implement the following functions:

[0641] Notifications: Generate appropriate solutions to predicted questions and provide customized notification content based on the user's emotional state.

[0642] System operation flow

[0643] 1. Data Collection

[0644] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[0645] Duplicate and incomplete data is removed from the collected data, and the text data is normalized.

[0646] 2. Data analysis

[0647] The server performs text analysis on past inquiry data to classify subjects (topics), and then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry.

[0648] 3. Model Building

[0649] The server selects and trains an appropriate machine learning algorithm based on the extracted features. It evaluates the model's performance using cross-validation and optimizes its accuracy. It then periodically retrains the model to improve its accuracy.

[0650] 4. Emotion recognition

[0651] The server combines an emotion engine to recognize the user's emotional state in real time from the user's behavioral data and input information.

[0652] 5. Predictions and Recommendations

[0653] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing the "Order History" page on a website, it predicts the question, "I want to check the delivery status of my order."

[0654] 6. Notification

[0655] The server generates appropriate solutions for predicted questions, such as FAQs, manuals, video guides, etc. It customizes notifications based on the user's emotional state and provides information at the right time.

[0656] Specific examples

[0657] Example 1: Online shopping site

[0658] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0659] Example 2: IT Support Help Desk

[0660] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[0661] Examples of prompt statements

[0662] Here are some example prompts to input to a generative AI model:

[0663] "How can I implement an algorithm that predicts potential questions users might have based on past inquiry data?"

[0664] "Describe the design of a system that recognizes a user's emotions and responds appropriately."

[0665] This system will significantly improve the efficiency of customer support, enhance user convenience, and provide more accurate and courteous support that reflects the user's emotional state.

[0666] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0667] System program processing flow

[0668] Step 1: Data collection

[0669] Input: Connection information for customer support system

[0670] Output: Collected historical inquiry data

[0671] What happens:

[0672] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, time of occurrence, resolution method, and user attribute information. Data collection is generally performed using APIs or database queries.

[0673] Step 2: Cleanse the data

[0674] Input: Collected inquiry data

[0675] Output: Cleansed data

[0676] What happens:

[0677] The server analyzes the collected data and removes duplicates and incomplete data. Specifically, it performs the following processes:

[0678] If there are multiple identical queries in the database, they are merged into one.

[0679] Detect rows with missing data or incomplete information and filter them out or complete them.

[0680] Step 3: Normalize the text data

[0681] Input: Cleansed data

[0682] Output: Normalized text data

[0683] What happens:

[0684] The server normalizes the collected text data, which includes lowercasing all text, removing special characters and extra spaces, and standardizing the text according to the user's language and region.

[0685] Step 4: Text Analysis

[0686] Input: normalized text data

[0687] Output: Categorized topic data

[0688] What happens:

[0689] The server uses a natural language processing (NLP) engine to classify the topics of the inquiry data by extracting keywords from the text data and categorizing them into categories (e.g., order status, returns, technical support, etc.).

[0690] Step 5: Statistical analysis

[0691] Input: Categorized topic data

[0692] Output: Thematic frequency and pattern analysis results

[0693] What happens:

[0694] The server statistically analyzes the frequency and occurrence patterns of each topic. For example, by analyzing whether a particular topic occurs frequently during a particular time period, it can identify problem occurrence patterns.

[0695] Step 6: Feature extraction

[0696] Input: Categorized topic data and analysis results

[0697] Output: Extracted features

[0698] What happens:

[0699] The server extracts features (keywords, occurrence time, user attributes, etc.) for each topic, allowing for a detailed understanding of the characteristics of each topic, which can be used in the next machine learning process.

[0700] Step 7: Select and train a machine learning model

[0701] Input: Extracted features

[0702] Output: A trained machine learning model

[0703] What happens:

[0704] The server selects an appropriate machine learning algorithm based on the extracted features and trains the model, using algorithms such as random forest, SVM, and deep learning to build a predictive model based on the query content.

[0705] Step 8: Evaluate and optimize the model

[0706] Input: A trained machine learning model

[0707] Output: Evaluated and optimized model

[0708] What happens:

[0709] The server evaluates the model's performance using cross-validation and adjusts hyperparameters as needed, for example, by performing 5-fold cross-validation and optimizing for accuracy above 90%.

[0710] Step 9: Emotion Recognition

[0711] Input: User behavior data and input information

[0712] Output: Classified emotion data

[0713] What happens:

[0714] The server uses an emotion engine to recognize the user's emotional state in real time based on their behavioral data and input information. For example, if a user sends messages in rapid succession, it can detect irritation.

[0715] Step 10: Anticipate potential questions

[0716] Input: User behavior log and emotional state

[0717] Output: Predicted potential questions

[0718] What happens:

[0719] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing their "order history," it predicts that they will "check the delivery status."

[0720] Step 11: Generate a solution

[0721] Input: predicted potential questions

[0722] Output: The generated solution

[0723] What happens:

[0724] The server generates appropriate solutions for predicted questions, providing specific solutions based on FAQs, manuals, video guides, etc.

[0725] Step 12: Notification

[0726] Input: Generated solution and user's emotional state

[0727] Output: Customized notification content

[0728] What happens:

[0729] The server customizes the notification content based on the user's emotional state and provides information at the appropriate time, such as "Your order has now been shipped and is expected to arrive tomorrow. If you have any questions, please feel free to contact us."

[0730] The above are the specific processing steps of the system program.

[0731] (Application example 2)

[0732] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0733] There is a need for a support system that can quickly and accurately respond to any problems or questions that users of autonomous vehicles may encounter while using the vehicle. In particular, it is necessary to improve user satisfaction by recognizing the user's emotional state and responding flexibly accordingly.

[0734] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0735] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, means for monitoring user behavior data in real time, means for predicting potential questions from users using the prediction model, means for notifying users of appropriate solutions to the predicted questions, and means for recognizing the user's emotional state and customizing a response.

[0736] This will enable us to respond quickly and flexibly to the various questions and problems that users of autonomous vehicles may encounter.Furthermore, by providing customized responses based on the user's emotional state, we can significantly improve user satisfaction.

[0737] "Past inquiry data" refers to records of questions and trouble reports submitted by users in the past, including the content of the inquiry, the date and time of the occurrence, the solution, and user attribute information.

[0738] "Analysis" is the act of extracting meaning and trends from collected data using statistical or computational techniques.

[0739] A "predictive model" is a collection of mathematical and statistical methods for predicting future conditions or outcomes based on past data.

[0740] "User behavior data" refers to data that records the actions and movements of users when using a system or device, including clicks, viewing time, and operation sequences.

[0741] "Real-time monitoring" refers to the system's ability to instantly collect and analyze user operations and situations.

[0742] "Means for predicting potential questions" refers to a technology that uses a predictive model to predict in advance what questions a user will ask in the future.

[0743] "Means for notifying appropriate solutions" refers to methods for informing users of the optimal solutions to anticipated questions or problems.

[0744] "Means for recognizing emotional states" refers to technology that identifies the emotions (e.g., joy, anger, sadness, surprise, etc.) present in a user's mind from their text and behavioral data.

[0745] "Customized response" refers to services and support that are optimally tailored to each individual user's situation and emotional state.

[0746] The present invention relates to a user support system that can be used in an autonomous vehicle. Specific embodiments of the system will be described below.

[0747] First, the server collects past inquiry data. This data includes questions and trouble reports submitted by users in the past, and records the inquiry content, date and time of occurrence, solution method, and user attribute information. The collected data is then normalized to remove duplicate and incomplete data.

[0748] The server then analyzes this data and builds a predictive model. Specifically, it extracts query features (keywords, time of occurrence, user attributes, etc.) from the data and trains the model using machine learning algorithms (e.g., random forest, support vector machine, deep learning, etc.). This predictive model is used to predict potential user questions.

[0749] The server also monitors user behavior data in real time, recording clicks, browsing time, and the sequence of operations as the user interacts with the in-car infotainment system. Based on this information, the server predicts the type of questions or concerns the user may have. It also uses an emotion recognition engine to identify the user's emotional state. For example, if the user shows signs of frustration, the system can provide a particularly prompt and courteous response.

[0750] For predicted questions, the server generates appropriate solutions and notifies the user. Solutions include FAQs, manuals, video guides, etc. This allows users to quickly find a way to solve their problems.

[0751] For example, if a user in a car says, "My car won't start," the emotion recognition engine recognizes the user's frustration. The server uses a predictive model to anticipate possible questions and solutions, and responds promptly and courteously. The notification system sends a message saying, "Please try restarting the engine. If the problem persists, please contact our support center."

[0752] An example of a prompt using a generative AI model is "When a user expresses frustration by asking if their car won't start, what should I do?" Using this prompt, the required solution can be quickly generated.

[0753] As described above, the present invention provides a system that can quickly and flexibly respond to questions and problems that users of autonomous vehicles may encounter. Furthermore, by providing customized responses based on the user's emotional state, user satisfaction can be significantly improved.

[0754] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0755] Step 1: Data collection

[0756] The server collects past inquiry data from the customer support system. This data includes inquiry content, the date and time of occurrence, the solution, and user attribute information. As input, this information is extracted from the inquiry database and temporarily stored inside the server. The output is inquiry data formatted in a format that can be used in subsequent processing steps.

[0757] Step 2: Data Preprocessing

[0758] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. It receives the temporarily stored query data as input and uses data cleansing techniques. Specifically, it executes database queries to remove duplicate rows, handle missing values, and normalize the data into a consistent text format. The output is the clean query data.

[0759] Step 3: Text analysis

[0760] The server performs text analysis on the preprocessed data using natural language processing techniques. The input is the cleaned query data, and the output is the extraction of topics and keywords for each query. Specific operations include text tokenization, morphological analysis, and application of topic models.

[0761] Step 4: Building a predictive model

[0762] The server builds a predictive model based on the features obtained from text analysis. The input is the analyzed feature data. The server uses a machine learning algorithm (random forest or deep learning) to train the model. The output is a model that can predict potential questions from users. Specific operations include training and cross-validation on a dataset.

[0763] Step 5: Emotion Recognition

[0764] The server performs real-time emotion recognition based on the user's behavioral data and text input. The input is the text and behavioral data entered by the user into the device. The server uses an emotion recognition engine to analyze the user's emotional state (joy, anger, sadness, surprise, etc.). The output is the result of the emotional state. The specific operation is to execute a text emotion analysis algorithm.

[0765] Step 6: Anticipate potential questions

[0766] The server combines the emotion recognition results with a prediction model to predict the user's potential questions. The input is the user's current emotional state and past behavioral data. The output is a list of likely questions. Specific operations include running a label prediction algorithm.

[0767] Step 7: Generate a suitable solution

[0768] The server generates appropriate solutions for predicted questions. The input is a list of predicted questions, and the output is solutions such as FAQs, manuals, video guides, etc. Specific operations include searching and composing relevant solutions from a database.

[0769] Step 8: Notify users

[0770] The server customizes the generated solution according to the user's emotional state and notifies the terminal at an appropriate time. The input is the solution and the user's emotional state. The output is a customized notification message. Specific operations include message tone adjustment and notification scheduling.

[0771] Through these processing steps, the server can provide quick and appropriate solutions to potential user questions within the autonomous vehicle, realizing emotionally responsive and customized support.

[0772] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0773] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0774] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0775] [Third embodiment]

[0776] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0777] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0778] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0779] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0780] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0781] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0782] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0783] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0784] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0785] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0786] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0787] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0788] The system of the present invention aims to collect and analyze past inquiry data, predict potential questions from users, and provide appropriate solutions.

[0789] Data collection

[0790] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[0791] Data analysis

[0792] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[0793] Model Building

[0794] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[0795] Predictions and Recommendations

[0796] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and suggest solutions. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts that the user might ask, "I want to check the delivery status of my order."

[0797] notification

[0798] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions, and the device notifies the user of the relevant information. By allowing users to receive information in real time, inquiries can be reduced. For example, if a user wants to know the "current delivery status" three days after placing an order, the server will immediately provide that information, and the device will notify the user as a pop-up message.

[0799] Specific examples

[0800] Example 1: Online shopping site

[0801] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. On the third day after an order is placed, the terminal notifies the user that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0802] Example 2: IT Support Help Desk

[0803] The server collects and organizes inquiry data about "VPN connection problems" at the IT support help desk. Anticipating that new employees will frequently ask questions about VPN connections at the beginning of the week, the server prepares a "VPN connection guide" in advance. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[0804] In this way, the system of the present invention can improve the efficiency of customer support and also significantly improve user convenience.

[0805] The processing flow will be explained below.

[0806] Step 1:

[0807] The server connects to the customer support system and collects past inquiry data, including inquiry content, occurrence time, resolution method, and user attribute information.

[0808] Step 2:

[0809] The server removes duplicates and incomplete data from the collected data, and also normalizes the data, for example, correcting spelling errors and using the same case.

[0810] Step 3:

[0811] The server performs text analysis on past inquiry data, automatically classifying subjects (topics), and statistically analyzing the frequency and occurrence patterns of each topic.

[0812] Step 4:

[0813] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[0814] Step 5:

[0815] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[0816] Step 6:

[0817] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[0818] Step 7:

[0819] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time.

[0820] Step 8:

[0821] The server uses a trained model to predict potential questions based on the data it monitors, for example, if a user spends a lot of time on a particular page, it predicts questions related to that page.

[0822] Step 9:

[0823] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc.

[0824] Step 10:

[0825] The server converts the generated solution into a message format and prepares it for notification to the user.

[0826] Step 11:

[0827] The device will notify users of anticipated questions and solutions in real time. For example, if a user is viewing the "Order History" page, a pop-up window will display the current delivery status.

[0828] In this way, users can obtain the information they need in a timely manner, reducing the time and effort required to make inquiries and reducing support costs.

[0829] Example 1

[0830] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0831] Conventional customer support systems have difficulty presenting appropriate solutions to user inquiries in real time. Furthermore, the analysis of inquiry data and the construction of predictive models are inefficient, making it difficult to make accurate predictions. This results in poor user convenience and a lack of improvement in the efficiency of support operations.

[0832] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0833] In this invention, the server includes means for connecting to a customer support system and collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing the text data, means for performing text analysis on the past inquiry data to automatically classify themes and extract features, means for building a predictive model using a machine learning algorithm based on the features, means for monitoring user behavior logs in real time and predicting questions, and means for generating appropriate solutions to the predicted questions and notifying the user. This makes it possible to predict potential user questions with high accuracy and provide appropriate solutions in real time.

[0834] A "customer support system" is a system for responding to user inquiries and assisting in resolving problems.

[0835] "Inquiry Data" means data containing information relating to a question or problem that a user submits to our customer support system.

[0836] "Duplicate data" refers to data in which the same inquiry or information is recorded multiple times.

[0837] "Incomplete data" refers to data where necessary information is missing or recorded incompletely.

[0838] "Text data normalization" is a process that removes special characters and unnecessary spaces in order to standardize the format of text data and improve the quality of the data.

[0839] "Text analysis" is the process of analyzing the content of text data using natural language processing technology and extracting themes and keywords.

[0840] "Automatic subject categorization" is the process of automatically categorizing each piece of data into a specific category or topic based on the content of the inquiry data.

[0841] "Features" refer to important attributes or keywords that machine learning algorithms use to learn from data.

[0842] A "predictive model" is a model built using machine learning algorithms to forecast future trends or outcomes based on past data.

[0843] "Machine learning algorithms" refer to mathematical models and methods for learning patterns from data and making predictions or classifications.

[0844] An "action log" is data that records the history of operations and actions performed by a user on a system.

[0845] "Real-time monitoring" is the process by which a system tracks user behavior in real time and analyzes it immediately.

[0846] "Question prediction" is the process of inferring the questions or problems a user may have in the future based on their behavior and past data.

[0847] "Solution generation" is the process of automatically creating appropriate answers or solutions to anticipated questions.

[0848] "Notification" is the process of presenting solutions to anticipated questions to the user.

[0849] The system of the present invention is primarily intended to improve the efficiency of customer support and enhance user convenience. Each component of the present invention and a specific implementation method thereof will be described below.

[0850] Data collection and preprocessing

[0851] The server connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. When collecting data, the data is obtained from the database via an API connection. The collected data is preprocessed using programming languages ​​such as Python and R. Specifically, libraries such as Pandas and Numpy are used to remove duplicate and incomplete data and normalize text data.

[0852] Text analysis and feature extraction

[0853] The server uses natural language processing libraries (such as NLTK or spaCy) to perform text analysis of the collected inquiry data. It automatically classifies each inquiry by topic and extracts features, such as keywords in the inquiry, the time of occurrence, and user attributes.

[0854] Building a machine learning model

[0855] The server selects an appropriate machine learning algorithm based on the extracted features and builds a predictive model. Examples of algorithms used include random forests, support vector machines, and deep learning. The model is trained using libraries such as Scikit-learn, TensorFlow, and PyTorch. Cross-validation is performed to evaluate the model's performance, and the model is periodically retrained as necessary.

[0856] Monitoring user behavior logs and predicting questions

[0857] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time. Apache Kafka and RabbitMQ are used to capture real-time data. Potential questions that users may have are predicted based on the collected behavior logs.

[0858] Solution generation and notification

[0859] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions. This allows the device to notify the user of the relevant information, reducing the number of inquiries. For example, if a user wants to know the "current delivery status" three days after placing an order, the device will display a pop-up message saying, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0860] Specific examples

[0861] online shopping site

[0862] The server collects and organizes inquiry data about the "delivery status of an order" from the online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. For example, if a user wants to know the "current delivery status" three days after placing an order, the terminal will notify them that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0863] IT Support Help Desk

[0864] The server collects inquiry data from the IT support help desk about the problem of "VPN connection not working" and predicts that new employees will often ask this question at the beginning of the week. It prepares a "VPN connection guide" in advance, and when a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[0865] As described above, by effectively combining each component, the system of the present invention is able to predict potential questions from users with high accuracy and provide appropriate solutions in real time.

[0866] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0867] Step 1:

[0868] The server connects to the customer support system and collects past inquiry data. It uses an API to access the database as input, obtaining information such as the inquiry content, time of occurrence, solution, and user attribute information. The output is a collection of raw inquiry data. Specifically, it sends an HTTP request to obtain the raw data and saves the obtained data in JSON format.

[0869] Step 2:

[0870] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. The input is the raw query data obtained in step 1. The output is the preprocessed, clean data. Specifically, it uses the Pandas library to remove duplicate rows and rows containing missing values. It also removes special characters and excess whitespace to normalize the text.

[0871] Step 3:

[0872] The server performs text analysis on past inquiry data, automatically classifying topics and extracting features. The input is the clean text data obtained in step 2. The output is the analyzed topics and their corresponding features. Specifically, it uses a natural language processing library (e.g., NLTK or spaCy) to tokenize the text and apply a topic model (e.g., LDA) to perform topic classification.

[0873] Step 4:

[0874] The server selects an appropriate machine learning algorithm based on the extracted features and constructs a predictive model. The input is the features obtained in step 3. The output is the constructed predictive model. Specifically, it uses libraries such as Scikit-learn, TensorFlow, and PyTorch to train the model using a machine learning algorithm (e.g., random forest or support vector machine).

[0875] Step 5:

[0876] The server evaluates the performance of the predictive model using cross-validation and performs optimization. The input is the predictive model constructed in step 4 and the evaluation data. The output is the performance score of the evaluated model and the optimized model. Specifically, it applies the cross-validation method, calculates evaluation indicators such as the precision and recall of the model, and performs parameter tuning.

[0877] Step 6:

[0878] The server monitors user behavior logs in real time and predicts potential questions. The input is the user behavior log data. The output is the predicted potential questions. Specifically, it uses Apache Kafka or RabbitMQ to capture real-time data and applies a predictive model to predict questions.

[0879] Step 7:

[0880] The server generates an appropriate solution for the predicted question and sends a notification to the device. The input is the question predicted in step 6 and the solution database. The output is the solution provided to the user. Specifically, the server refers to FAQs, manuals, and video guides, selects an appropriate solution, and sends an automatically generated notification message to the device.

[0881] Step 8:

[0882] The terminal notifies the user of the solution received from the server. The input is the notification message from the server. The output is the solution displayed to the user. As a specific operation, the solution is presented to the user as a pop-up message or an alert.

[0883] Through these steps, the system is able to predict users' potential questions with high accuracy and provide appropriate solutions in real time.

[0884] (Application example 1)

[0885] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0886] When shopping online, users often make inquiries seeking specific information, which increases the burden on customer support and reduces user satisfaction. To solve this problem, it is necessary to quickly provide appropriate information before users make inquiries. Furthermore, there is a need for a system that can predict users' potential questions and notify them of appropriate solutions in real time.

[0887] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0888] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, and means for monitoring user behavior data in real time, thereby making it possible to predict potential user questions using the prediction model, predict inquiries in real time based on user behavior logs, and provide appropriate information via push notifications.

[0889] "Past inquiry data" refers to data such as questions and complaints received from users in the past, the content of the inquiry, the time of occurrence, the solution, and user attribute information.

[0890] A "predictive model" is a machine learning algorithm built to predict potential user questions based on patterns and features derived from collected and analyzed past inquiry data.

[0891] "User behavior data" refers to behavioral logs such as the actions a user takes on a website or application, click patterns, input content, and viewing time.

[0892] "Push notifications" are a feature that allows applications or systems to automatically send information or messages to a user's device in real time.

[0893] "Duplicate data" refers to data in which the same content is recorded multiple times, and is data that can cause a decrease in the accuracy of analysis.

[0894] "Incomplete data" is data that lacks necessary information or has insufficient content, and is a factor that hinders accurate analysis and prediction.

[0895] "Retraining" is the process of repeatedly learning an existing predictive model using new data in order to improve its accuracy.

[0896] "Providing relevant information through push notifications" means sending solutions and related information to a user's device in real time in response to questions that are inferred based on the user's behavior.

[0897] This invention is a system that collects and analyzes past inquiry data, predicts potential user questions, and provides appropriate solutions. This system is realized based on server, terminal, and user behavior data.

[0898] Data Collection and Cleansing

[0899] The server first connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data. This process can be performed using data processing tools such as Python's pandas and nltk libraries.

[0900] Data analysis and feature extraction

[0901] The server then performs text analysis on the collected inquiry data and automatically classifies the subject matter (topics). It then statistically analyzes the frequency and occurrence patterns of each topic and extracts inquiry features (e.g., keywords, occurrence time, user attributes, etc.). Machine learning libraries such as sklearn can be used for this process.

[0902] Building and retraining the model

[0903] The server selects a machine learning algorithm (e.g., random forest or support vector machine) based on these features and builds a predictive model. The model's performance is evaluated using cross-validation, and accuracy is improved by periodic retraining.

[0904] Real-time monitoring and user prediction

[0905] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and provide solutions. For example, if a user spends a long time viewing the "Order Delivery Status" page, the server predicts the question, "I would like to check the current delivery status."

[0906] Notification of Solution

[0907] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The device provides this information to the user via push notifications. These notifications are displayed on the user's smartphone or in a web application. Push notifications can be delivered using notification services such as Firebase Cloud Messaging (FCM) or Apple Push Notification Service (APNs).

[0908] Specific examples

[0909] For example, in an application on an online shopping site, the server collects and analyzes past inquiry data about delivery status and predicts that there will be many inquiries about "delivery status" a few days after an order is placed. When a user tries to check their order history, the server sends a notification such as, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[0910] Examples of prompt statements

[0911] "I have a request. I would like to train an AI model that predicts that when a user searches for "I want to know the delivery status of my order," the user has a question about delivery. Based on the past inquiry data below, please extract features based on the inquiry content, inquiry time, user attributes, and resolution status, and build a model using a random forest classifier."

[0912] In this way, the system of the present invention can effectively reduce the burden on customer support and improve user convenience.

[0913] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0914] Step 1:

[0915] The server connects to the customer support system and collects past inquiry data. This data includes the inquiry content, time of occurrence, solution method, and user attribute information. The server stores this data in a database, which is used for subsequent data analysis.

[0916] Input: Inquiry data from the customer support system

[0917] Output: Query data stored in a database

[0918] Step 2:

[0919] The server removes duplicate and incomplete data from the collected data and normalizes the text data. Specifically, it uses a data cleaning algorithm to remove unnecessary whitespace and special characters and maintain data consistency, resulting in a clean dataset.

[0920] Input: Collected inquiry data

[0921] Output: A clean dataset

[0922] Step 3:

[0923] The server performs text analysis on the clean data to automatically classify the subject matter. Next, it statistically analyzes the frequency and occurrence patterns of each topic to extract query features. This process uses natural language processing libraries to extract keywords and perform topic modeling.

[0924] Input: A clean dataset

[0925] Output: Feature-extracted data

[0926] Step 4:

[0927] The server selects a machine learning algorithm based on the feature values ​​and builds a predictive model. The model's performance is evaluated through cross-validation, and regular retraining is performed to improve accuracy. This results in a model that can predict potential user questions with high accuracy.

[0928] Input: Feature-extracted data

[0929] Output: Highly accurate predictive model

[0930] Step 5:

[0931] When a user uses a website, the server monitors the user's activity log in real time. Specifically, the pages the user accesses, their click patterns, and the content of their input are collected and analyzed as logs. Based on this data, the user's current behavioral patterns can be understood.

[0932] Input: Real-time user activity log

[0933] Output: Parsed behavior log

[0934] Step 6:

[0935] The server uses real-time behavior logs to run a predictive model to predict potential questions users may have. For example, if a user spends a long time viewing a particular page, questions related to that page are predicted. This results in predicted questions.

[0936] Input: Parsed behavior log

[0937] Output: Prediction results for the question

[0938] Step 7:

[0939] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The generated information is sent to the user's device as a push notification. This notification provides the user with the information they need to quickly solve their problem.

[0940] Input: Predicted result of question

[0941] Output: Push notification to user device

[0942] Step 8:

[0943] Users can check the push notification sent to their device and obtain the appropriate solution or information, which allows them to quickly resolve the problem without having to contact the company.

[0944] Input: Push notification

[0945] Output: The solution or information provided to the user

[0946] In this way, through a series of steps, a system is realized that predicts potential questions from users and provides appropriate information in real time.

[0947] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0948] The system of the present invention aims to collect and analyze past inquiry data, predict users' potential questions, and provide appropriate solutions, as well as combine an emotion engine to recognize users' emotions and provide more appropriate and customized support.

[0949] Data collection

[0950] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[0951] Data analysis

[0952] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[0953] Model Building

[0954] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[0955] emotion recognition

[0956] The server combines an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, thereby identifying the user's current emotion (e.g., joy, anger, sadness, surprise, etc.).

[0957] Predictions and Recommendations

[0958] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts the question, "I'd like to check the delivery status of my order." If the user's emotions indicate irritability, the server provides a particularly fast and courteous response.

[0959] notification

[0960] The server generates appropriate solutions for predicted questions. Solutions can include FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state and provides information at the appropriate time. For example, if the user is dissatisfied, the server will send a notification using particularly polite language and with a prompt response.

[0961] Specific examples

[0962] Example 1: Online shopping site

[0963] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0964] Example 2: IT Support Help Desk

[0965] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[0966] In this way, the system of the present invention can improve the efficiency of customer support and significantly improve user convenience. Furthermore, by combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[0967] The processing flow will be explained below.

[0968] Step 1:

[0969] The server connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[0970] Step 2:

[0971] The server removes duplicates and incomplete data from the collected data and normalizes the text data, for example, by standardizing uppercase and lowercase letters and correcting spelling errors.

[0972] Step 3:

[0973] The server performs text analysis on past inquiry data and automatically classifies subjects (topics), allowing for statistical analysis of the frequency and occurrence patterns of each topic.

[0974] Step 4:

[0975] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[0976] Step 5:

[0977] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[0978] Step 6:

[0979] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[0980] Step 7:

[0981] The server uses an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, including joy, anger, sadness, surprise, etc.

[0982] Step 8:

[0983] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time viewing a particular page, it predicts questions related to that page.

[0984] Step 9:

[0985] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state.

[0986] Step 10:

[0987] The server converts the generated solution into a message format and prepares it for notification to the user, including particularly polite language and prompt responses depending on the user's emotional state.

[0988] Step 11:

[0989] The device will notify the user of anticipated questions and solutions in real time. For example, if the user is viewing the "Order History" page, the device will notify the user, "The current delivery status is 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[0990] In this way, users can obtain the information they need in a timely manner, saving them the trouble of making inquiries and reducing support response costs.By combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[0991] Example 2

[0992] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0993] Modern customer support systems are required to improve the speed and appropriateness of responses to user inquiries, but conventional systems lack the ability to take into account the user's emotional state. Therefore, to improve user satisfaction, a system that recognizes the user's emotions in real time and provides more customized responses is needed.

[0994] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0995] In this invention, the server includes means for collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing text data, means for analyzing the past inquiry data to classify topics and extract features, means for building a predictive model using a machine learning algorithm based on the extracted features, means for evaluating the performance of the predictive model and optimizing its accuracy, means for monitoring user behavior data in real time and recognizing the user's emotional state using an emotion engine, means for predicting potential user questions using the predictive model, and means for generating appropriate solutions to the predicted questions, customizing notification content based on the user's emotional state, and providing information at an appropriate time, thereby enabling appropriate and prompt customized support according to the user's emotional state.

[0996] "Past inquiry data" refers to historical information such as the content of inquiries made by the user, the time of occurrence, the solution, and user attribute information.

[0997] "Duplicate data" refers to data in which the same content is recorded multiple times.

[0998] "Incomplete data" refers to data that is missing necessary information.

[0999] "Text data normalization" refers to the process of converting text data into a unified format to make it easier to analyze.

[1000] "Topic" refers to the themes or categories used to classify the content of text data.

[1001] "Features" refer to important elements or parameters that machine learning algorithms use to analyze data.

[1002] A "machine learning algorithm" refers to a computational method for learning patterns and rules from data and making predictions and classifications.

[1003] "Predictive model" refers to a statistical or machine learning-based model built to forecast future events based on historical data.

[1004] "Cross-validation" refers to a method of dividing data and performing cross-validation to evaluate the performance of a model.

[1005] An "emotion engine" refers to algorithms or software that automatically recognize a user's emotional state from their input and behavioral data.

[1006] "Potential questions" refer to questions that are anticipated before a user explicitly asks them.

[1007] "Solutions" refer to answers or solutions to potential user questions or problems.

[1008] "Customizing notification content" refers to changing the content and presentation of the information to be notified according to the user's situation and emotions.

[1009] The system of the present invention aims to collect and analyze past inquiry data, predict potential user questions, and provide appropriate solutions. Furthermore, it uses an emotion engine to recognize user emotions and provide customized support.

[1010] Hardware and software configuration

[1011] The server is used to implement the following functions:

[1012] Data collection and cleansing: We use a dedicated script to connect to your customer support system and collect inquiry data.

[1013] Natural Language Processing (NLP) engine: Used to analyze the query content and classify the subject (topic).

[1014] Machine learning model: Based on the extracted features, a predictive model is constructed using an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.).

[1015] Emotion recognition engine: Recognizes the user's emotional state in real time based on user behavioral data and input information.

[1016] The terminal is used to implement the following functions:

[1017] Notifications: Generate appropriate solutions to predicted questions and provide customized notification content based on the user's emotional state.

[1018] System operation flow

[1019] 1. Data Collection

[1020] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[1021] Duplicate and incomplete data is removed from the collected data, and the text data is normalized.

[1022] 2. Data analysis

[1023] The server performs text analysis on past inquiry data to classify subjects (topics), and then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry.

[1024] 3. Model Building

[1025] The server selects and trains an appropriate machine learning algorithm based on the extracted features. It evaluates the model's performance using cross-validation and optimizes its accuracy. It then periodically retrains the model to improve its accuracy.

[1026] 4. Emotion recognition

[1027] The server combines an emotion engine to recognize the user's emotional state in real time from the user's behavioral data and input information.

[1028] 5. Predictions and Recommendations

[1029] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing the "Order History" page on a website, it predicts the question, "I want to check the delivery status of my order."

[1030] 6. Notification

[1031] The server generates appropriate solutions for predicted questions, such as FAQs, manuals, video guides, etc. It customizes notifications based on the user's emotional state and provides information at the right time.

[1032] Specific examples

[1033] Example 1: Online shopping site

[1034] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[1035] Example 2: IT Support Help Desk

[1036] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[1037] Examples of prompt statements

[1038] Here are some example prompts to input to a generative AI model:

[1039] "How can I implement an algorithm that predicts potential questions users might have based on past inquiry data?"

[1040] "Describe the design of a system that recognizes a user's emotions and responds appropriately."

[1041] This system will significantly improve the efficiency of customer support, enhance user convenience, and provide more accurate and courteous support that reflects the user's emotional state.

[1042] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1043] System program processing flow

[1044] Step 1: Data collection

[1045] Input: Connection information for customer support system

[1046] Output: Collected historical inquiry data

[1047] What happens:

[1048] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, time of occurrence, resolution method, and user attribute information. Data collection is generally performed using APIs or database queries.

[1049] Step 2: Cleanse the data

[1050] Input: Collected inquiry data

[1051] Output: Cleansed data

[1052] What happens:

[1053] The server analyzes the collected data and removes duplicates and incomplete data. Specifically, it performs the following processes:

[1054] If there are multiple identical queries in the database, they are merged into one.

[1055] Detect rows with missing data or incomplete information and filter them out or complete them.

[1056] Step 3: Normalize the text data

[1057] Input: Cleansed data

[1058] Output: Normalized text data

[1059] What happens:

[1060] The server normalizes the collected text data, which includes lowercasing all text, removing special characters and extra spaces, and standardizing the text according to the user's language and region.

[1061] Step 4: Text Analysis

[1062] Input: normalized text data

[1063] Output: Categorized topic data

[1064] What happens:

[1065] The server uses a natural language processing (NLP) engine to classify the topics of the inquiry data by extracting keywords from the text data and categorizing them into categories (e.g., order status, returns, technical support, etc.).

[1066] Step 5: Statistical analysis

[1067] Input: Categorized topic data

[1068] Output: Thematic frequency and pattern analysis results

[1069] What happens:

[1070] The server statistically analyzes the frequency and occurrence patterns of each topic. For example, by analyzing whether a particular topic occurs frequently during a particular time period, it can identify problem occurrence patterns.

[1071] Step 6: Feature extraction

[1072] Input: Categorized topic data and analysis results

[1073] Output: Extracted features

[1074] What happens:

[1075] The server extracts features (keywords, occurrence time, user attributes, etc.) for each topic, allowing for a detailed understanding of the characteristics of each topic, which can be used in the next machine learning process.

[1076] Step 7: Select and train a machine learning model

[1077] Input: Extracted features

[1078] Output: A trained machine learning model

[1079] What happens:

[1080] The server selects an appropriate machine learning algorithm based on the extracted features and trains the model, using algorithms such as random forest, SVM, and deep learning to build a predictive model based on the query content.

[1081] Step 8: Evaluate and optimize the model

[1082] Input: A trained machine learning model

[1083] Output: Evaluated and optimized model

[1084] What happens:

[1085] The server evaluates the model's performance using cross-validation and adjusts hyperparameters as needed, for example, by performing 5-fold cross-validation and optimizing for accuracy above 90%.

[1086] Step 9: Emotion Recognition

[1087] Input: User behavior data and input information

[1088] Output: Classified emotion data

[1089] What happens:

[1090] The server uses an emotion engine to recognize the user's emotional state in real time based on their behavioral data and input information. For example, if a user sends messages in rapid succession, it can detect irritation.

[1091] Step 10: Anticipate potential questions

[1092] Input: User behavior log and emotional state

[1093] Output: Predicted potential questions

[1094] What happens:

[1095] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing their "order history," it predicts that they will "check the delivery status."

[1096] Step 11: Generate a solution

[1097] Input: predicted potential questions

[1098] Output: The generated solution

[1099] What happens:

[1100] The server generates appropriate solutions for predicted questions, providing specific solutions based on FAQs, manuals, video guides, etc.

[1101] Step 12: Notification

[1102] Input: Generated solution and user's emotional state

[1103] Output: Customized notification content

[1104] What happens:

[1105] The server customizes the notification content based on the user's emotional state and provides information at the appropriate time, such as "Your order has now been shipped and is expected to arrive tomorrow. If you have any questions, please feel free to contact us."

[1106] The above are the specific processing steps of the system program.

[1107] (Application example 2)

[1108] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1109] There is a need for a support system that can quickly and accurately respond to any problems or questions that users of autonomous vehicles may encounter while using the vehicle. In particular, it is necessary to improve user satisfaction by recognizing the user's emotional state and responding flexibly accordingly.

[1110] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1111] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, means for monitoring user behavior data in real time, means for predicting potential questions from users using the prediction model, means for notifying users of appropriate solutions to the predicted questions, and means for recognizing the user's emotional state and customizing a response.

[1112] This will enable us to respond quickly and flexibly to the various questions and problems that users of autonomous vehicles may encounter.Furthermore, by providing customized responses based on the user's emotional state, we can significantly improve user satisfaction.

[1113] "Past inquiry data" refers to records of questions and trouble reports submitted by users in the past, including the content of the inquiry, the date and time of the occurrence, the solution, and user attribute information.

[1114] "Analysis" is the act of extracting meaning and trends from collected data using statistical or computational techniques.

[1115] A "predictive model" is a collection of mathematical and statistical methods for predicting future conditions or outcomes based on past data.

[1116] "User behavior data" refers to data that records the actions and movements of users when using a system or device, including clicks, viewing time, and operation sequences.

[1117] "Real-time monitoring" refers to the system's ability to instantly collect and analyze user operations and situations.

[1118] "Means for predicting potential questions" refers to a technology that uses a predictive model to predict in advance what questions a user will ask in the future.

[1119] "Means for notifying appropriate solutions" refers to methods for informing users of the optimal solutions to anticipated questions or problems.

[1120] "Means for recognizing emotional states" refers to technology that identifies the emotions (e.g., joy, anger, sadness, surprise, etc.) present in a user's mind from their text and behavioral data.

[1121] "Customized response" refers to services and support that are optimally tailored to each individual user's situation and emotional state.

[1122] The present invention relates to a user support system that can be used in an autonomous vehicle. Specific embodiments of the system will be described below.

[1123] First, the server collects past inquiry data. This data includes questions and trouble reports submitted by users in the past, and records the inquiry content, date and time of occurrence, solution method, and user attribute information. The collected data is then normalized to remove duplicate and incomplete data.

[1124] The server then analyzes this data and builds a predictive model. Specifically, it extracts query features (keywords, time of occurrence, user attributes, etc.) from the data and trains the model using machine learning algorithms (e.g., random forest, support vector machine, deep learning, etc.). This predictive model is used to predict potential user questions.

[1125] The server also monitors user behavior data in real time, recording clicks, browsing time, and the sequence of operations as the user interacts with the in-car infotainment system. Based on this information, the server predicts the type of questions or concerns the user may have. It also uses an emotion recognition engine to identify the user's emotional state. For example, if the user shows signs of frustration, the system can provide a particularly prompt and courteous response.

[1126] For predicted questions, the server generates appropriate solutions and notifies the user. Solutions include FAQs, manuals, video guides, etc. This allows users to quickly find a way to solve their problems.

[1127] For example, if a user in a car says, "My car won't start," the emotion recognition engine recognizes the user's frustration. The server uses a predictive model to anticipate possible questions and solutions, and responds promptly and courteously. The notification system sends a message saying, "Please try restarting the engine. If the problem persists, please contact our support center."

[1128] An example of a prompt using a generative AI model is "When a user expresses frustration by asking if their car won't start, what should I do?" Using this prompt, the required solution can be quickly generated.

[1129] As described above, the present invention provides a system that can quickly and flexibly respond to questions and problems that users of autonomous vehicles may encounter. Furthermore, by providing customized responses based on the user's emotional state, user satisfaction can be significantly improved.

[1130] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1131] Step 1: Data collection

[1132] The server collects past inquiry data from the customer support system. This data includes inquiry content, the date and time of occurrence, the solution, and user attribute information. As input, this information is extracted from the inquiry database and temporarily stored inside the server. The output is inquiry data formatted in a format that can be used in subsequent processing steps.

[1133] Step 2: Data Preprocessing

[1134] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. It receives the temporarily stored query data as input and uses data cleansing techniques. Specifically, it executes database queries to remove duplicate rows, handle missing values, and normalize the data into a consistent text format. The output is the clean query data.

[1135] Step 3: Text analysis

[1136] The server performs text analysis on the preprocessed data using natural language processing techniques. The input is the cleaned query data, and the output is the extraction of topics and keywords for each query. Specific operations include text tokenization, morphological analysis, and application of topic models.

[1137] Step 4: Building a predictive model

[1138] The server builds a predictive model based on the features obtained from text analysis. The input is the analyzed feature data. The server uses a machine learning algorithm (random forest or deep learning) to train the model. The output is a model that can predict potential questions from users. Specific operations include training and cross-validation on a dataset.

[1139] Step 5: Emotion Recognition

[1140] The server performs real-time emotion recognition based on the user's behavioral data and text input. The input is the text and behavioral data entered by the user into the device. The server uses an emotion recognition engine to analyze the user's emotional state (joy, anger, sadness, surprise, etc.). The output is the result of the emotional state. The specific operation is to execute a text emotion analysis algorithm.

[1141] Step 6: Anticipate potential questions

[1142] The server combines the emotion recognition results with a prediction model to predict the user's potential questions. The input is the user's current emotional state and past behavioral data. The output is a list of likely questions. Specific operations include running a label prediction algorithm.

[1143] Step 7: Generate a suitable solution

[1144] The server generates appropriate solutions for predicted questions. The input is a list of predicted questions, and the output is solutions such as FAQs, manuals, video guides, etc. Specific operations include searching and composing relevant solutions from a database.

[1145] Step 8: Notify users

[1146] The server customizes the generated solution according to the user's emotional state and notifies the terminal at an appropriate time. The input is the solution and the user's emotional state. The output is a customized notification message. Specific operations include message tone adjustment and notification scheduling.

[1147] Through these processing steps, the server can provide quick and appropriate solutions to potential user questions within the autonomous vehicle, realizing emotionally responsive and customized support.

[1148] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1149] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1150] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1151] [Fourth embodiment]

[1152] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1153] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1154] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1155] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1156] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1157] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1158] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1159] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1160] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1161] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1162] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1163] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1164] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1165] The system of the present invention aims to collect and analyze past inquiry data, predict potential questions from users, and provide appropriate solutions.

[1166] Data collection

[1167] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[1168] Data analysis

[1169] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[1170] Model Building

[1171] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[1172] Predictions and Recommendations

[1173] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and suggest solutions. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts that the user might ask, "I want to check the delivery status of my order."

[1174] notification

[1175] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions, and the device notifies the user of the relevant information. By allowing users to receive information in real time, inquiries can be reduced. For example, if a user wants to know the "current delivery status" three days after placing an order, the server will immediately provide that information, and the device will notify the user as a pop-up message.

[1176] Specific examples

[1177] Example 1: Online shopping site

[1178] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. On the third day after an order is placed, the terminal notifies the user that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[1179] Example 2: IT Support Help Desk

[1180] The server collects and organizes inquiry data about "VPN connection problems" at the IT support help desk. Anticipating that new employees will frequently ask questions about VPN connections at the beginning of the week, the server prepares a "VPN connection guide" in advance. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[1181] In this way, the system of the present invention can improve the efficiency of customer support and also significantly improve user convenience.

[1182] The processing flow will be explained below.

[1183] Step 1:

[1184] The server connects to the customer support system and collects past inquiry data, including inquiry content, occurrence time, resolution method, and user attribute information.

[1185] Step 2:

[1186] The server removes duplicates and incomplete data from the collected data, and also normalizes the data, for example, correcting spelling errors and using the same case.

[1187] Step 3:

[1188] The server performs text analysis on past inquiry data, automatically classifying subjects (topics), and statistically analyzing the frequency and occurrence patterns of each topic.

[1189] Step 4:

[1190] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[1191] Step 5:

[1192] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[1193] Step 6:

[1194] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[1195] Step 7:

[1196] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time.

[1197] Step 8:

[1198] The server uses a trained model to predict potential questions based on the data it monitors, for example, if a user spends a lot of time on a particular page, it predicts questions related to that page.

[1199] Step 9:

[1200] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc.

[1201] Step 10:

[1202] The server converts the generated solution into a message format and prepares it for notification to the user.

[1203] Step 11:

[1204] The device will notify users of anticipated questions and solutions in real time. For example, if a user is viewing the "Order History" page, a pop-up window will display the current delivery status.

[1205] In this way, users can obtain the information they need in a timely manner, reducing the time and effort required to make inquiries and reducing support costs.

[1206] Example 1

[1207] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1208] Conventional customer support systems have difficulty presenting appropriate solutions to user inquiries in real time. Furthermore, the analysis of inquiry data and the construction of predictive models are inefficient, making it difficult to make accurate predictions. This results in poor user convenience and a lack of improvement in the efficiency of support operations.

[1209] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1210] In this invention, the server includes means for connecting to a customer support system and collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing the text data, means for performing text analysis on the past inquiry data to automatically classify themes and extract features, means for building a predictive model using a machine learning algorithm based on the features, means for monitoring user behavior logs in real time and predicting questions, and means for generating appropriate solutions to the predicted questions and notifying the user. This makes it possible to predict potential user questions with high accuracy and provide appropriate solutions in real time.

[1211] A "customer support system" is a system for responding to user inquiries and assisting in resolving problems.

[1212] "Inquiry Data" means data containing information relating to a question or problem that a user submits to our customer support system.

[1213] "Duplicate data" refers to data in which the same inquiry or information is recorded multiple times.

[1214] "Incomplete data" refers to data where necessary information is missing or recorded incompletely.

[1215] "Text data normalization" is a process that removes special characters and unnecessary spaces in order to standardize the format of text data and improve the quality of the data.

[1216] "Text analysis" is the process of analyzing the content of text data using natural language processing technology and extracting themes and keywords.

[1217] "Automatic subject categorization" is the process of automatically categorizing each piece of data into a specific category or topic based on the content of the inquiry data.

[1218] "Features" refer to important attributes or keywords that machine learning algorithms use to learn from data.

[1219] A "predictive model" is a model built using machine learning algorithms to forecast future trends or outcomes based on past data.

[1220] "Machine learning algorithms" refer to mathematical models and methods for learning patterns from data and making predictions or classifications.

[1221] An "action log" is data that records the history of operations and actions performed by a user on a system.

[1222] "Real-time monitoring" is the process by which a system tracks user behavior in real time and analyzes it immediately.

[1223] "Question prediction" is the process of inferring the questions or problems a user may have in the future based on their behavior and past data.

[1224] "Solution generation" is the process of automatically creating appropriate answers or solutions to anticipated questions.

[1225] "Notification" is the process of presenting solutions to anticipated questions to the user.

[1226] The system of the present invention is primarily intended to improve the efficiency of customer support and enhance user convenience. Each component of the present invention and a specific implementation method thereof will be described below.

[1227] Data collection and preprocessing

[1228] The server connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. When collecting data, the data is obtained from the database via an API connection. The collected data is preprocessed using programming languages ​​such as Python and R. Specifically, libraries such as Pandas and Numpy are used to remove duplicate and incomplete data and normalize text data.

[1229] Text analysis and feature extraction

[1230] The server uses natural language processing libraries (such as NLTK or spaCy) to perform text analysis of the collected inquiry data. It automatically classifies each inquiry by topic and extracts features, such as keywords in the inquiry, the time of occurrence, and user attributes.

[1231] Building a machine learning model

[1232] The server selects an appropriate machine learning algorithm based on the extracted features and builds a predictive model. Examples of algorithms used include random forests, support vector machines, and deep learning. The model is trained using libraries such as Scikit-learn, TensorFlow, and PyTorch. Cross-validation is performed to evaluate the model's performance, and the model is periodically retrained as necessary.

[1233] Monitoring user behavior logs and predicting questions

[1234] The server monitors user behavior logs (website operations, click patterns, input content, etc.) in real time. Apache Kafka and RabbitMQ are used to capture real-time data. Potential questions that users may have are predicted based on the collected behavior logs.

[1235] Solution generation and notification

[1236] The server generates appropriate solutions (FAQs, manuals, video guides, etc.) for predicted questions. This allows the device to notify the user of the relevant information, reducing the number of inquiries. For example, if a user wants to know the "current delivery status" three days after placing an order, the device will display a pop-up message saying, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[1237] Specific examples

[1238] online shopping site

[1239] The server collects and organizes inquiry data about the "delivery status of an order" from the online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. For example, if a user wants to know the "current delivery status" three days after placing an order, the terminal will notify them that "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[1240] IT Support Help Desk

[1241] The server collects inquiry data from the IT support help desk about the problem of "VPN connection not working" and predicts that new employees will often ask this question at the beginning of the week. It prepares a "VPN connection guide" in advance, and when a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide."

[1242] As described above, by effectively combining each component, the system of the present invention is able to predict potential questions from users with high accuracy and provide appropriate solutions in real time.

[1243] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1244] Step 1:

[1245] The server connects to the customer support system and collects past inquiry data. It uses an API to access the database as input, obtaining information such as the inquiry content, time of occurrence, solution, and user attribute information. The output is a collection of raw inquiry data. Specifically, it sends an HTTP request to obtain the raw data and saves the obtained data in JSON format.

[1246] Step 2:

[1247] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. The input is the raw query data obtained in step 1. The output is the preprocessed, clean data. Specifically, it uses the Pandas library to remove duplicate rows and rows containing missing values. It also removes special characters and excess whitespace to normalize the text.

[1248] Step 3:

[1249] The server performs text analysis on past inquiry data, automatically classifying topics and extracting features. The input is the clean text data obtained in step 2. The output is the analyzed topics and their corresponding features. Specifically, it uses a natural language processing library (e.g., NLTK or spaCy) to tokenize the text and apply a topic model (e.g., LDA) to perform topic classification.

[1250] Step 4:

[1251] The server selects an appropriate machine learning algorithm based on the extracted features and constructs a predictive model. The input is the features obtained in step 3. The output is the constructed predictive model. Specifically, it uses libraries such as Scikit-learn, TensorFlow, and PyTorch to train the model using a machine learning algorithm (e.g., random forest or support vector machine).

[1252] Step 5:

[1253] The server evaluates the performance of the predictive model using cross-validation and performs optimization. The input is the predictive model constructed in step 4 and the evaluation data. The output is the performance score of the evaluated model and the optimized model. Specifically, it applies the cross-validation method, calculates evaluation indicators such as the precision and recall of the model, and performs parameter tuning.

[1254] Step 6:

[1255] The server monitors user behavior logs in real time and predicts potential questions. The input is the user behavior log data. The output is the predicted potential questions. Specifically, it uses Apache Kafka or RabbitMQ to capture real-time data and applies a predictive model to predict questions.

[1256] Step 7:

[1257] The server generates an appropriate solution for the predicted question and sends a notification to the device. The input is the question predicted in step 6 and the solution database. The output is the solution provided to the user. Specifically, the server refers to FAQs, manuals, and video guides, selects an appropriate solution, and sends an automatically generated notification message to the device.

[1258] Step 8:

[1259] The terminal notifies the user of the solution received from the server. The input is the notification message from the server. The output is the solution displayed to the user. As a specific operation, the solution is presented to the user as a pop-up message or an alert.

[1260] Through these steps, the system is able to predict users' potential questions with high accuracy and provide appropriate solutions in real time.

[1261] (Application example 1)

[1262] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1263] When shopping online, users often make inquiries seeking specific information, which increases the burden on customer support and reduces user satisfaction. To solve this problem, it is necessary to quickly provide appropriate information before users make inquiries. Furthermore, there is a need for a system that can predict users' potential questions and notify them of appropriate solutions in real time.

[1264] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1265] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, and means for monitoring user behavior data in real time, thereby making it possible to predict potential user questions using the prediction model, predict inquiries in real time based on user behavior logs, and provide appropriate information via push notifications.

[1266] "Past inquiry data" refers to data such as questions and complaints received from users in the past, the content of the inquiry, the time of occurrence, the solution, and user attribute information.

[1267] A "predictive model" is a machine learning algorithm built to predict potential user questions based on patterns and features derived from collected and analyzed past inquiry data.

[1268] "User behavior data" refers to behavioral logs such as the actions a user takes on a website or application, click patterns, input content, and viewing time.

[1269] "Push notifications" are a feature that allows applications or systems to automatically send information or messages to a user's device in real time.

[1270] "Duplicate data" refers to data in which the same content is recorded multiple times, and is data that can cause a decrease in the accuracy of analysis.

[1271] "Incomplete data" is data that lacks necessary information or has insufficient content, and is a factor that hinders accurate analysis and prediction.

[1272] "Retraining" is the process of repeatedly learning an existing predictive model using new data in order to improve its accuracy.

[1273] "Providing relevant information through push notifications" means sending solutions and related information to a user's device in real time in response to questions that are inferred based on the user's behavior.

[1274] This invention is a system that collects and analyzes past inquiry data, predicts potential user questions, and provides appropriate solutions. This system is realized based on server, terminal, and user behavior data.

[1275] Data Collection and Cleansing

[1276] The server first connects to the customer support system and automatically collects past inquiry data. This data includes the inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data. This process can be performed using data processing tools such as Python's pandas and nltk libraries.

[1277] Data analysis and feature extraction

[1278] The server then performs text analysis on the collected inquiry data and automatically classifies the subject matter (topics). It then statistically analyzes the frequency and occurrence patterns of each topic and extracts inquiry features (e.g., keywords, occurrence time, user attributes, etc.). Machine learning libraries such as sklearn can be used for this process.

[1279] Building and retraining the model

[1280] The server selects a machine learning algorithm (e.g., random forest or support vector machine) based on these features and builds a predictive model. The model's performance is evaluated using cross-validation, and accuracy is improved by periodic retraining.

[1281] Real-time monitoring and user prediction

[1282] The server monitors user behavior logs (website operations, click patterns, input contents, etc.) in real time. This allows it to predict what questions the user may have and provide solutions. For example, if a user spends a long time viewing the "Order Delivery Status" page, the server predicts the question, "I would like to check the current delivery status."

[1283] Notification of Solution

[1284] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The device provides this information to the user via push notifications. These notifications are displayed on the user's smartphone or in a web application. Push notifications can be delivered using notification services such as Firebase Cloud Messaging (FCM) or Apple Push Notification Service (APNs).

[1285] Specific examples

[1286] For example, in an application on an online shopping site, the server collects and analyzes past inquiry data about delivery status and predicts that there will be many inquiries about "delivery status" a few days after an order is placed. When a user tries to check their order history, the server sends a notification such as, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'."

[1287] Examples of prompt statements

[1288] "I have a request. I would like to train an AI model that predicts that when a user searches for "I want to know the delivery status of my order," the user has a question about delivery. Based on the past inquiry data below, please extract features based on the inquiry content, inquiry time, user attributes, and resolution status, and build a model using a random forest classifier."

[1289] In this way, the system of the present invention can effectively reduce the burden on customer support and improve user convenience.

[1290] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1291] Step 1:

[1292] The server connects to the customer support system and collects past inquiry data. This data includes the inquiry content, time of occurrence, solution method, and user attribute information. The server stores this data in a database, which is used for subsequent data analysis.

[1293] Input: Inquiry data from the customer support system

[1294] Output: Query data stored in a database

[1295] Step 2:

[1296] The server removes duplicate and incomplete data from the collected data and normalizes the text data. Specifically, it uses a data cleaning algorithm to remove unnecessary whitespace and special characters and maintain data consistency, resulting in a clean dataset.

[1297] Input: Collected inquiry data

[1298] Output: A clean dataset

[1299] Step 3:

[1300] The server performs text analysis on the clean data to automatically classify the subject matter. Next, it statistically analyzes the frequency and occurrence patterns of each topic to extract query features. This process uses natural language processing libraries to extract keywords and perform topic modeling.

[1301] Input: A clean dataset

[1302] Output: Feature-extracted data

[1303] Step 4:

[1304] The server selects a machine learning algorithm based on the feature values ​​and builds a predictive model. The model's performance is evaluated through cross-validation, and regular retraining is performed to improve accuracy. This results in a model that can predict potential user questions with high accuracy.

[1305] Input: Feature-extracted data

[1306] Output: Highly accurate predictive model

[1307] Step 5:

[1308] When a user uses a website, the server monitors the user's activity log in real time. Specifically, the pages the user accesses, their click patterns, and the content of their input are collected and analyzed as logs. Based on this data, the user's current behavioral patterns can be understood.

[1309] Input: Real-time user activity log

[1310] Output: Parsed behavior log

[1311] Step 6:

[1312] The server uses real-time behavior logs to run a predictive model to predict potential questions users may have. For example, if a user spends a long time viewing a particular page, questions related to that page are predicted. This results in predicted questions.

[1313] Input: Parsed behavior log

[1314] Output: Prediction results for the question

[1315] Step 7:

[1316] The server generates optimal solutions (FAQs, manuals, video guides, etc.) for predicted questions. The generated information is sent to the user's device as a push notification. This notification provides the user with the information they need to quickly solve their problem.

[1317] Input: Predicted result of question

[1318] Output: Push notification to user device

[1319] Step 8:

[1320] Users can check the push notification sent to their device and obtain the appropriate solution or information, which allows them to quickly resolve the problem without having to contact the company.

[1321] Input: Push notification

[1322] Output: The solution or information provided to the user

[1323] In this way, through a series of steps, a system is realized that predicts potential questions from users and provides appropriate information in real time.

[1324] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1325] The system of the present invention aims to collect and analyze past inquiry data, predict users' potential questions, and provide appropriate solutions, as well as combine an emotion engine to recognize users' emotions and provide more appropriate and customized support.

[1326] Data collection

[1327] The server first connects to the customer support system and automatically collects past inquiry data, including inquiry content, time of occurrence, resolution method, and user attribute information. The server then removes duplicate and incomplete data from the collected data and normalizes the text data.

[1328] Data analysis

[1329] The server then performs text analysis on past inquiry data to automatically classify topics. It then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry. These features include keywords in the inquiry, the time of occurrence, and user attributes.

[1330] Model Building

[1331] Based on these features, the server selects and trains an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.). The server evaluates the model's performance using cross-validation, optimizes its accuracy, and then periodically retrains it to improve its accuracy.

[1332] emotion recognition

[1333] The server combines an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, thereby identifying the user's current emotion (e.g., joy, anger, sadness, surprise, etc.).

[1334] Predictions and Recommendations

[1335] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time browsing the "Order History" page on a website, the server predicts the question, "I'd like to check the delivery status of my order." If the user's emotions indicate irritability, the server provides a particularly fast and courteous response.

[1336] notification

[1337] The server generates appropriate solutions for predicted questions. Solutions can include FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state and provides information at the appropriate time. For example, if the user is dissatisfied, the server will send a notification using particularly polite language and with a prompt response.

[1338] Specific examples

[1339] Example 1: Online shopping site

[1340] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[1341] Example 2: IT Support Help Desk

[1342] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[1343] In this way, the system of the present invention can improve the efficiency of customer support and significantly improve user convenience. Furthermore, by combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[1344] The processing flow will be explained below.

[1345] Step 1:

[1346] The server connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[1347] Step 2:

[1348] The server removes duplicates and incomplete data from the collected data and normalizes the text data, for example, by standardizing uppercase and lowercase letters and correcting spelling errors.

[1349] Step 3:

[1350] The server performs text analysis on past inquiry data and automatically classifies subjects (topics), allowing for statistical analysis of the frequency and occurrence patterns of each topic.

[1351] Step 4:

[1352] The server uses a machine learning algorithm to extract features from the inquiry data, including inquiry keywords, time of occurrence, and user attributes.

[1353] Step 5:

[1354] Based on the extracted features, the server selects an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.) and trains the model.

[1355] Step 6:

[1356] The server uses cross-validation to evaluate the model's performance (prediction accuracy, recall, etc.) and performs parameter adjustments to optimize the model's accuracy.

[1357] Step 7:

[1358] The server uses an emotion engine to recognize the user's emotional state in real time based on the user's behavioral data and input information, including joy, anger, sadness, surprise, etc.

[1359] Step 8:

[1360] The server monitors the user's behavioral log and emotional state in real time, and uses a predictive model to predict potential questions the user may have. For example, if a user spends a long time viewing a particular page, it predicts questions related to that page.

[1361] Step 9:

[1362] The server generates appropriate solutions for predicted questions, including FAQs, manuals, video guides, etc. The server customizes notification content based on the user's emotional state.

[1363] Step 10:

[1364] The server converts the generated solution into a message format and prepares it for notification to the user, including particularly polite language and prompt responses depending on the user's emotional state.

[1365] Step 11:

[1366] The device will notify the user of anticipated questions and solutions in real time. For example, if the user is viewing the "Order History" page, the device will notify the user, "The current delivery status is 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[1367] In this way, users can obtain the information they need in a timely manner, saving them the trouble of making inquiries and reducing support response costs.By combining it with an emotion recognition engine, it is possible to provide appropriate support according to the user's emotional state, further improving user satisfaction.

[1368] Example 2

[1369] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1370] Modern customer support systems are required to improve the speed and appropriateness of responses to user inquiries, but conventional systems lack the ability to take into account the user's emotional state. Therefore, to improve user satisfaction, a system that recognizes the user's emotions in real time and provides more customized responses is needed.

[1371] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1372] In this invention, the server includes means for collecting past inquiry data, means for deleting duplicate data and incomplete data from the past inquiry data and normalizing text data, means for analyzing the past inquiry data to classify topics and extract features, means for building a predictive model using a machine learning algorithm based on the extracted features, means for evaluating the performance of the predictive model and optimizing its accuracy, means for monitoring user behavior data in real time and recognizing the user's emotional state using an emotion engine, means for predicting potential user questions using the predictive model, and means for generating appropriate solutions to the predicted questions, customizing notification content based on the user's emotional state, and providing information at an appropriate time, thereby enabling appropriate and prompt customized support according to the user's emotional state.

[1373] "Past inquiry data" refers to historical information such as the content of inquiries made by the user, the time of occurrence, the solution, and user attribute information.

[1374] "Duplicate data" refers to data in which the same content is recorded multiple times.

[1375] "Incomplete data" refers to data that is missing necessary information.

[1376] "Text data normalization" refers to the process of converting text data into a unified format to make it easier to analyze.

[1377] "Topic" refers to the themes or categories used to classify the content of text data.

[1378] "Features" refer to important elements or parameters that machine learning algorithms use to analyze data.

[1379] A "machine learning algorithm" refers to a computational method for learning patterns and rules from data and making predictions and classifications.

[1380] "Predictive model" refers to a statistical or machine learning-based model built to forecast future events based on historical data.

[1381] "Cross-validation" refers to a method of dividing data and performing cross-validation to evaluate the performance of a model.

[1382] An "emotion engine" refers to algorithms or software that automatically recognize a user's emotional state from their input and behavioral data.

[1383] "Potential questions" refer to questions that are anticipated before a user explicitly asks them.

[1384] "Solutions" refer to answers or solutions to potential user questions or problems.

[1385] "Customizing notification content" refers to changing the content and presentation of the information to be notified according to the user's situation and emotions.

[1386] The system of the present invention aims to collect and analyze past inquiry data, predict potential user questions, and provide appropriate solutions. Furthermore, it uses an emotion engine to recognize user emotions and provide customized support.

[1387] Hardware and software configuration

[1388] The server is used to implement the following functions:

[1389] Data collection and cleansing: We use a dedicated script to connect to your customer support system and collect inquiry data.

[1390] Natural Language Processing (NLP) engine: Used to analyze the query content and classify the subject (topic).

[1391] Machine learning model: Based on the extracted features, a predictive model is constructed using an appropriate machine learning algorithm (e.g., random forest, support vector machine, deep learning, etc.).

[1392] Emotion recognition engine: Recognizes the user's emotional state in real time based on user behavioral data and input information.

[1393] The terminal is used to implement the following functions:

[1394] Notifications: Generate appropriate solutions to predicted questions and provide customized notification content based on the user's emotional state.

[1395] System operation flow

[1396] 1. Data Collection

[1397] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, the time of occurrence, the solution, and user attribute information.

[1398] Duplicate and incomplete data is removed from the collected data, and the text data is normalized.

[1399] 2. Data analysis

[1400] The server performs text analysis on past inquiry data to classify subjects (topics), and then statistically analyzes the frequency and occurrence patterns of each topic to extract features of the inquiry.

[1401] 3. Model Building

[1402] The server selects and trains an appropriate machine learning algorithm based on the extracted features. It evaluates the model's performance using cross-validation and optimizes its accuracy. It then periodically retrains the model to improve its accuracy.

[1403] 4. Emotion recognition

[1404] The server combines an emotion engine to recognize the user's emotional state in real time from the user's behavioral data and input information.

[1405] 5. Predictions and Recommendations

[1406] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing the "Order History" page on a website, it predicts the question, "I want to check the delivery status of my order."

[1407] 6. Notification

[1408] The server generates appropriate solutions for predicted questions, such as FAQs, manuals, video guides, etc. It customizes notifications based on the user's emotional state and provides information at the right time.

[1409] Specific examples

[1410] Example 1: Online shopping site

[1411] The server collects and organizes inquiry data about the "delivery status of an order" from an online shopping site. Based on this, it predicts that many questions about the "delivery status" will occur three days after an order is placed, and prepares "information about the delivery status" in advance. Furthermore, it uses an emotion engine to recognize whether the user is dissatisfied with a delay in their order. On the third day after an order is placed, the terminal notifies the user, "The delivery status of your order is currently 'Shipped'. The expected arrival date is 'Tomorrow'. If you have any questions, please feel free to contact us."

[1412] Example 2: IT Support Help Desk

[1413] The server collects and organizes inquiry data about "VPN connection problems" from the IT support help desk. It predicts that new employees will frequently ask about VPN connections at the beginning of the week, and prepares a "VPN connection guide" in advance. It also uses an emotion engine to recognize whether users are likely to become frustrated with the connection. When a new employee logs in at the beginning of the week, the device automatically sends the "VPN connection procedure guide," complete with particularly polite instructions.

[1414] Examples of prompt statements

[1415] Here are some example prompts to input to a generative AI model:

[1416] "How can I implement an algorithm that predicts potential questions users might have based on past inquiry data?"

[1417] "Describe the design of a system that recognizes a user's emotions and responds appropriately."

[1418] This system will significantly improve the efficiency of customer support, enhance user convenience, and provide more accurate and courteous support that reflects the user's emotional state.

[1419] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1420] System program processing flow

[1421] Step 1: Data collection

[1422] Input: Connection information for customer support system

[1423] Output: Collected historical inquiry data

[1424] What happens:

[1425] The server periodically connects to the customer support system and automatically collects past inquiry data, including the inquiry content, time of occurrence, resolution method, and user attribute information. Data collection is generally performed using APIs or database queries.

[1426] Step 2: Cleanse the data

[1427] Input: Collected inquiry data

[1428] Output: Cleansed data

[1429] What happens:

[1430] The server analyzes the collected data and removes duplicates and incomplete data. Specifically, it performs the following processes:

[1431] If there are multiple identical queries in the database, they are merged into one.

[1432] Detect rows with missing data or incomplete information and filter them out or complete them.

[1433] Step 3: Normalize the text data

[1434] Input: Cleansed data

[1435] Output: Normalized text data

[1436] What happens:

[1437] The server normalizes the collected text data, which includes lowercasing all text, removing special characters and extra spaces, and standardizing the text according to the user's language and region.

[1438] Step 4: Text Analysis

[1439] Input: normalized text data

[1440] Output: Categorized topic data

[1441] What happens:

[1442] The server uses a natural language processing (NLP) engine to classify the topics of the inquiry data by extracting keywords from the text data and categorizing them into categories (e.g., order status, returns, technical support, etc.).

[1443] Step 5: Statistical analysis

[1444] Input: Categorized topic data

[1445] Output: Thematic frequency and pattern analysis results

[1446] What happens:

[1447] The server statistically analyzes the frequency and occurrence patterns of each topic. For example, by analyzing whether a particular topic occurs frequently during a particular time period, it can identify problem occurrence patterns.

[1448] Step 6: Feature extraction

[1449] Input: Categorized topic data and analysis results

[1450] Output: Extracted features

[1451] What happens:

[1452] The server extracts features (keywords, occurrence time, user attributes, etc.) for each topic, allowing for a detailed understanding of the characteristics of each topic, which can be used in the next machine learning process.

[1453] Step 7: Select and train a machine learning model

[1454] Input: Extracted features

[1455] Output: A trained machine learning model

[1456] What happens:

[1457] The server selects an appropriate machine learning algorithm based on the extracted features and trains the model, using algorithms such as random forest, SVM, and deep learning to build a predictive model based on the query content.

[1458] Step 8: Evaluate and optimize the model

[1459] Input: A trained machine learning model

[1460] Output: Evaluated and optimized model

[1461] What happens:

[1462] The server evaluates the model's performance using cross-validation and adjusts hyperparameters as needed, for example, by performing 5-fold cross-validation and optimizing for accuracy above 90%.

[1463] Step 9: Emotion Recognition

[1464] Input: User behavior data and input information

[1465] Output: Classified emotion data

[1466] What happens:

[1467] The server uses an emotion engine to recognize the user's emotional state in real time based on their behavioral data and input information. For example, if a user sends messages in rapid succession, it can detect irritation.

[1468] Step 10: Anticipate potential questions

[1469] Input: User behavior log and emotional state

[1470] Output: Predicted potential questions

[1471] What happens:

[1472] The server uses a predictive model to predict potential questions based on the user's behavioral log and emotional state. For example, if a user spends a long time browsing their "order history," it predicts that they will "check the delivery status."

[1473] Step 11: Generate a solution

[1474] Input: predicted potential questions

[1475] Output: The generated solution

[1476] What happens:

[1477] The server generates appropriate solutions for predicted questions, providing specific solutions based on FAQs, manuals, video guides, etc.

[1478] Step 12: Notification

[1479] Input: Generated solution and user's emotional state

[1480] Output: Customized notification content

[1481] What happens:

[1482] The server customizes the notification content based on the user's emotional state and provides information at the appropriate time, such as "Your order has now been shipped and is expected to arrive tomorrow. If you have any questions, please feel free to contact us."

[1483] The above are the specific processing steps of the system program.

[1484] (Application example 2)

[1485] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1486] There is a need for a support system that can quickly and accurately respond to any problems or questions that users of autonomous vehicles may encounter while using the vehicle. In particular, it is necessary to improve user satisfaction by recognizing the user's emotional state and responding flexibly accordingly.

[1487] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1488] In this invention, the server includes means for collecting past inquiry data, means for analyzing the past inquiry data to build a prediction model, means for monitoring user behavior data in real time, means for predicting potential questions from users using the prediction model, means for notifying users of appropriate solutions to the predicted questions, and means for recognizing the user's emotional state and customizing a response.

[1489] This will enable us to respond quickly and flexibly to the various questions and problems that users of autonomous vehicles may encounter.Furthermore, by providing customized responses based on the user's emotional state, we can significantly improve user satisfaction.

[1490] "Past inquiry data" refers to records of questions and trouble reports submitted by users in the past, including the content of the inquiry, the date and time of the occurrence, the solution, and user attribute information.

[1491] "Analysis" is the act of extracting meaning and trends from collected data using statistical or computational techniques.

[1492] A "predictive model" is a collection of mathematical and statistical methods for predicting future conditions or outcomes based on past data.

[1493] "User behavior data" refers to data that records the actions and movements of users when using a system or device, including clicks, viewing time, and operation sequences.

[1494] "Real-time monitoring" refers to the system's ability to instantly collect and analyze user operations and situations.

[1495] "Means for predicting potential questions" refers to a technology that uses a predictive model to predict in advance what questions a user will ask in the future.

[1496] "Means for notifying appropriate solutions" refers to methods for informing users of the optimal solutions to anticipated questions or problems.

[1497] "Means for recognizing emotional states" refers to technology that identifies the emotions (e.g., joy, anger, sadness, surprise, etc.) present in a user's mind from their text and behavioral data.

[1498] "Customized response" refers to services and support that are optimally tailored to each individual user's situation and emotional state.

[1499] The present invention relates to a user support system that can be used in an autonomous vehicle. Specific embodiments of the system will be described below.

[1500] First, the server collects past inquiry data. This data includes questions and trouble reports submitted by users in the past, and records the inquiry content, date and time of occurrence, solution method, and user attribute information. The collected data is then normalized to remove duplicate and incomplete data.

[1501] The server then analyzes this data and builds a predictive model. Specifically, it extracts query features (keywords, time of occurrence, user attributes, etc.) from the data and trains the model using machine learning algorithms (e.g., random forest, support vector machine, deep learning, etc.). This predictive model is used to predict potential user questions.

[1502] The server also monitors user behavior data in real time, recording clicks, browsing time, and the sequence of operations as the user interacts with the in-car infotainment system. Based on this information, the server predicts the type of questions or concerns the user may have. It also uses an emotion recognition engine to identify the user's emotional state. For example, if the user shows signs of frustration, the system can provide a particularly prompt and courteous response.

[1503] For predicted questions, the server generates appropriate solutions and notifies the user. Solutions include FAQs, manuals, video guides, etc. This allows users to quickly find a way to solve their problems.

[1504] For example, if a user in a car says, "My car won't start," the emotion recognition engine recognizes the user's frustration. The server uses a predictive model to anticipate possible questions and solutions, and responds promptly and courteously. The notification system sends a message saying, "Please try restarting the engine. If the problem persists, please contact our support center."

[1505] An example of a prompt using a generative AI model is "When a user expresses frustration by asking if their car won't start, what should I do?" Using this prompt, the required solution can be quickly generated.

[1506] As described above, the present invention provides a system that can quickly and flexibly respond to questions and problems that users of autonomous vehicles may encounter. Furthermore, by providing customized responses based on the user's emotional state, user satisfaction can be significantly improved.

[1507] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1508] Step 1: Data collection

[1509] The server collects past inquiry data from the customer support system. This data includes inquiry content, the date and time of occurrence, the solution, and user attribute information. As input, this information is extracted from the inquiry database and temporarily stored inside the server. The output is inquiry data formatted in a format that can be used in subsequent processing steps.

[1510] Step 2: Data Preprocessing

[1511] The server removes duplicate and incomplete data from the collected query data and normalizes the text data. It receives the temporarily stored query data as input and uses data cleansing techniques. Specifically, it executes database queries to remove duplicate rows, handle missing values, and normalize the data into a consistent text format. The output is the clean query data.

[1512] Step 3: Text analysis

[1513] The server performs text analysis on the preprocessed data using natural language processing techniques. The input is the cleaned query data, and the output is the extraction of topics and keywords for each query. Specific operations include text tokenization, morphological analysis, and application of topic models.

[1514] Step 4: Building a predictive model

[1515] The server builds a predictive model based on the features obtained from text analysis. The input is the analyzed feature data. The server uses a machine learning algorithm (random forest or deep learning) to train the model. The output is a model that can predict potential questions from users. Specific operations include training and cross-validation on a dataset.

[1516] Step 5: Emotion Recognition

[1517] The server performs real-time emotion recognition based on the user's behavioral data and text input. The input is the text and behavioral data entered by the user into the device. The server uses an emotion recognition engine to analyze the user's emotional state (joy, anger, sadness, surprise, etc.). The output is the result of the emotional state. The specific operation is to execute a text emotion analysis algorithm.

[1518] Step 6: Anticipate potential questions

[1519] The server combines the emotion recognition results with a prediction model to predict the user's potential questions. The input is the user's current emotional state and past behavioral data. The output is a list of likely questions. Specific operations include running a label prediction algorithm.

[1520] Step 7: Generate a suitable solution

[1521] The server generates appropriate solutions for predicted questions. The input is a list of predicted questions, and the output is solutions such as FAQs, manuals, video guides, etc. Specific operations include searching and composing relevant solutions from a database.

[1522] Step 8: Notify users

[1523] The server customizes the generated solution according to the user's emotional state and notifies the terminal at an appropriate time. The input is the solution and the user's emotional state. The output is a customized notification message. Specific operations include message tone adjustment and notification scheduling.

[1524] Through these processing steps, the server can provide quick and appropriate solutions to potential user questions within the autonomous vehicle, realizing emotionally responsive and customized support.

[1525] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1526] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1527] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1528] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1529] FIG. 9 illustrates an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and behaviors arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1530] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1531] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1532] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1533] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1534] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1535] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1536] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1537] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1538] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1539] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1540] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1541] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1542] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1543] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1544] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1545] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1546] The following is further disclosed regarding the above embodiment.

[1547] (Claim 1)

[1548] a means of collecting historical inquiry data;

[1549] means for analyzing the historical inquiry data to build a predictive model;

[1550] a means for monitoring user behavior data in real time;

[1551] means for predicting potential questions of a user using the predictive model;

[1552] a means for informing the user of appropriate solutions to predicted questions;

[1553] A system including:

[1554] (Claim 2)

[1555] 2. The system according to claim 1, further comprising means for sorting out duplicate data and incomplete data of past inquiries.

[1556] (Claim 3)

[1557] The system of claim 1 , further comprising means for periodically retraining the predictive model to improve its accuracy.

[1558] "Example 1"

[1559] (Claim 1)

[1560] a means for connecting to a customer support system and collecting past inquiry data;

[1561] means for deleting duplicate data and incomplete data from the past inquiry data and normalizing the text data;

[1562] means for performing text analysis on the past inquiry data, automatically classifying the subject matter, and extracting features;

[1563] A means for constructing a prediction model using a machine learning algorithm based on the feature amount;

[1564] A means of monitoring user behavior logs in real time and predicting questions,

[1565] A means for generating appropriate solutions to predicted questions and notifying the user;

[1566] A system including:

[1567] (Claim 2)

[1568] 10. The system of claim 1, further comprising means for using cross-validation to evaluate the performance of the predictive model and optimizing accuracy.

[1569] (Claim 3)

[1570] The system of claim 1 , further comprising means for periodically retraining the predictive model to improve its accuracy.

[1571] "Application Example 1"

[1572] (Claim 1)

[1573] a means of collecting historical inquiry data;

[1574] means for analyzing the historical inquiry data to build a predictive model;

[1575] a means for monitoring user behavior data in real time;

[1576] means for predicting potential questions of a user using the predictive model;

[1577] a means for informing the user of appropriate solutions to predicted questions;

[1578] A means to predict inquiries in real time based on user behavior logs and provide appropriate information via push notifications,

[1579] A system including:

[1580] (Claim 2)

[1581] 2. The system according to claim 1, further comprising means for sorting out duplicate data and incomplete data of past inquiries.

[1582] (Claim 3)

[1583] The system of claim 1 , further comprising means for periodically retraining the predictive model to improve its accuracy.

[1584] "Example 2: Combining Emotion Engines"

[1585] (Claim 1)

[1586] a means of collecting historical inquiry data;

[1587] means for deleting duplicate data and incomplete data from the past inquiry data and normalizing the text data;

[1588] means for analyzing the past inquiry data, classifying the subject matter, and extracting feature quantities;

[1589] A means for constructing a predictive model using a machine learning algorithm based on the extracted features;

[1590] means for evaluating the performance of the predictive model and optimizing its accuracy;

[1591] means for monitoring user behavior data in real time and recognizing the user's emotional state using an emotion engine;

[1592] means for predicting potential questions of a user using the predictive model;

[1593] A means for generating appropriate solutions to predicted questions, customizing notifications based on the user's emotional state, and providing information at the right time;

[1594] A system including:

[1595] (Claim 2)

[1596] 2. The system according to claim 1, further comprising means for sorting out duplicate data and incomplete data of past inquiries.

[1597] (Claim 3)

[1598] The system of claim 1 , further comprising means for periodically retraining the predictive model to improve its accuracy.

[1599] "Application example 2 when combining emotion engines"

[1600] (Claim 1)

[1601] a means of collecting historical inquiry data;

[1602] means for analyzing the historical inquiry data to build a predictive model;

[1603] a means for monitoring user behavior data in real time;

[1604] means for predicting potential questions of a user using the predictive model;

[1605] a means for informing the user of appropriate solutions to predicted questions;

[1606] a means for recognizing a user's emotional state and customizing a response;

[1607] A system including:

[1608] (Claim 2)

[1609] 2. The system according to claim 1, further comprising means for sorting out duplicate data and incomplete data of past inquiries.

[1610] (Claim 3)

[1611] The system of claim 1 , further comprising means for periodically retraining the predictive model to improve its accuracy. [Explanation of symbols]

[1612] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means of collecting historical inquiry data; means for analyzing the historical inquiry data to build a predictive model; a means for monitoring user behavior data in real time; means for predicting potential questions of a user using the predictive model; a means for informing the user of appropriate solutions to predicted questions; A system including:

2. 2. The system according to claim 1, further comprising means for sorting out duplicate data and incomplete data of past inquiries.

3. The system of claim 1 , further comprising means for periodically retraining the predictive model to improve its accuracy.

Images