system

Abstract

We provide the system. [Solution] A means of collecting employees' closing talks as data, A method for converting collected data into text data and cleansing it, A method for training a generative AI using cleansed data, A means of providing real-time customer support using trained AI models, A means of providing employees with a simulation environment to learn closing techniques, A system that includes this.

Description

Technical Field

[0004] , , ,

[0005] , , ,

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

Background Art

[0002] Patent Document 1 discloses a method for controlling a persona chatbot performed by at least one processor, the method including steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a character of the chatbot, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance as a response to the user utterance.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Currently, maintaining the quality and consistency of closing conversations by employees requires a lot of time and resources. Also, in order for employees to acquire effective closing techniques, actual customer service experience is essential, resulting in a decrease in training efficiency. Furthermore, due to the problem of manpower shortage, there is a problem that it is difficult to provide sufficient customer service at undeveloped stores and sites with high workloads.

Means for Solving the Problems

[0005] The system includes means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generative AI using the cleansed data, means for performing real-time customer service using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This system allows AI to handle customer service on behalf of employees in stores that have not yet been deployed or in locations with high workloads, and also improves the efficiency of training.

[0006] "Closing talk" refers to the final conversation with a customer to finalize the purchase of a product or the contract for a service.

[0007] "Data collection" refers to the process of systematically gathering audio data, text data, and other data for a specific purpose.

[0008] "Converting to text data" is the process of digitizing audio data into textual information.

[0009] "Cleansing" is the process of removing noise and unnecessary information from data, transforming it into a state suitable for analysis and learning.

[0010] "Generative AI" is an artificial intelligence technology that uses machine learning techniques to automatically generate new content and responses.

[0011] "Training" refers to the process of using a large amount of data to train an AI model so that it can perform a specific task.

[0012] "Real-time customer support" means that the AI ​​crew will immediately provide appropriate responses to customer questions and requests.

[0013] A "simulation environment" is a virtual environment that mimics real-world business scenarios and is used by users for training and testing.

[0014] An "unopened store" refers to a store that has not yet started operations or does not have employees permanently stationed there. [Brief explanation of the drawing]

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

Embodiments for Carrying Out the Invention

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

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

[0018] In the following embodiments, a processor with a reference numeral (hereinafter simply referred to as "processor") may be one arithmetic unit or a combination of a plurality of arithmetic units. Also, the processor may be one type of arithmetic unit or a combination of a plurality of types of arithmetic units. Examples of arithmetic units include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), and the like.

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

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

[0021] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0022] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0023] [First Embodiment]

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

[0025] As shown in Figure 1, the 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 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

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

[0028] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and 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 perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

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

[0032] As shown in Figure 2, in the data processing device 12, a specific processing 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" related to the technology of this 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 according to the specific processing program 56 executed on the RAM 30.

[0033] The storage 32 stores the data generation model 58 and the 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 processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

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

[0036] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training.

[0037] Program processing

[0038] Data collection

[0039] The server periodically collects audio and text data of employee closing conversations from each store. This includes actual customer interaction audio conversations from the field and successful examples of existing closing conversations. For example, a recording of a "points card explanation" from store A will be collected.

[0040] Data preprocessing

[0041] The collected data is first converted from audio data to text data by the server. Automatic speech recognition (ASR) technology is used for this process. After that, data cleansing is performed to remove noise and unnecessary information. For example, background noise and irrelevant conversations are removed from data such as "Explanation of loyalty cards."

[0042] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add related contextual information. For example, keywords such as "point card," "deals," and "how to use" are extracted, and their contextual information is organized.

[0043] AI model training

[0044] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses. During the training process, the AI's ability to generate appropriate responses improves.

[0045] Real-time response

[0046] Using a trained AI model, the terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to a server via the terminal. The server uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[0047] OJT tool

[0048] Employees can use the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The terminal receives questions from the user and sends them to the server. The server generates responses using a trained AI model and provides them to the user through the terminal. Through this process, employees can learn closing techniques in a situation close to actual customer interaction. For example, if a new employee asks, "How do I use this loyalty card?", the AI ​​crew will respond, "This loyalty card can be used at all stores, and you can easily accumulate points."

[0049] As described above, this system collects data on employees' closing talks, trains a generating AI, and uses the trained AI model to handle customer interactions in real time. It can also be used as an on-the-job training (OJT) tool to support employees in efficiently learning closing techniques.

[0050] The following describes the processing flow.

[0051] Step 1:

[0052] The server receives audio and text data of employee closing speeches collected from each store. This data is uploaded from store microphones and recording devices.

[0053] Step 2:

[0054] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine operates and generates text information from the audio.

[0055] Step 3:

[0056] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts, and prepares unstructured data. This makes the data suitable for analysis.

[0057] Step 4:

[0058] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, words like "bargain," "points card," and "campaign" might be extracted.

[0059] Step 5:

[0060] The server adds contextual information related to the extracted keywords. Specifically, it describes how the keywords are used and in what scenarios they are effective.

[0061] Step 6:

[0062] The server uses the processed text data to train an AI model. This includes a process of training the model with successful examples from the collected conversation data. Machine learning algorithms are used to enable the AI ​​model to learn patterns in customer interactions.

[0063] Step 7:

[0064] The server deploys the AI ​​model after training is complete. This model is located in the cloud or on an edge server. This enables real-time customer support.

[0065] Step 8:

[0066] When a user interacts with a customer in a store, the terminal receives customer questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[0067] Step 9:

[0068] The terminal receives voice input and sends it to the server as text data. Communication takes place in real time over the internet.

[0069] Step 10:

[0070] The server uses a trained AI model to generate an appropriate response based on the received text data. For example, in response to the question "What are this month's special offers?", it will generate the response "This month's special offer is a newly released smartphone."

[0071] Step 11:

[0072] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[0073] Step 12:

[0074] The device converts received text data into speech and provides a response to the customer. This allows customers to obtain information in a natural conversational flow.

[0075] Step 13:

[0076] When a user uses AI Crew as an OJT tool, the terminal is operated in simulation mode. The user takes on the role of a customer and inputs questions and responses.

[0077] Step 14:

[0078] The terminal sends the user's entered questions to the server. The questions arrive at the server in real time.

[0079] Step 15:

[0080] The server uses a trained AI model to generate responses to input questions. For example, in response to the question, "How do I use this loyalty card?", it generates the response, "This loyalty card can be used at all stores, and you can easily accumulate points."

[0081] Step 16:

[0082] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[0083] (Example 1)

[0084] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0085] Improving the quality of employees' closing pitches and efficiently serving customers is a crucial challenge for many service industries. In particular, in under-established stores or locations with insufficient staff, standardizing closing techniques is difficult, leading to inconsistencies in customer satisfaction. Furthermore, training new employees requires significant time and effort, highlighting the need for effective training methods.

[0086] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0087] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected audio data into text data and cleaning it, means for extracting important keywords and phrases from the cleansed data, means for training a generative AI model using the extracted data, means for providing real-time customer support using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables data collection of closing talks, training of a generative AI model, real-time response, and effective training of employees.

[0088] "Employee closing talk" refers to the final persuasive or explanatory remarks made by an employee when interacting with a customer.

[0089] "Means of collecting data" refers to systems and methods for collecting closing talks in audio or text format from stores and on-site locations.

[0090] "Means of converting collected audio data into text data" refers to the process or technology of converting audio data into text information using speech recognition technology.

[0091] "Methods of cleansing" refer to methods and techniques for removing noise and unnecessary information from converted text data, and organizing and formatting the data.

[0092] "Methods for extracting important keywords and phrases" refers to methods that use natural language processing techniques to extract important words and phrases from text data and add contextual information.

[0093] "Means for training generative AI models" refers to the processes and techniques for training and updating artificial intelligence models using collected and pre-processed data.

[0094] "Methods for providing real-time customer support using trained AI models" refers to systems and methods that use responses generated by generative AI models to immediately address customer questions and requests.

[0095] "Means of providing a simulation environment for employees to learn closing techniques" refers to a training environment in which employees can acquire closing techniques in situations that closely resemble actual customer interactions.

[0096] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI model, and utilizes the trained AI model for customer service and employee training.

[0097] The server periodically collects audio and text data of employee closing speeches from each store. The collected audio data is converted into text data using the Google® Speech-to-Text API. For example, a recording of a "points card explanation" from store A might be used.

[0098] Next, the server cleanses the converted text data, removing noise and unnecessary information. The Python pandas library is used for this process. From the cleansed data, important keywords and phrases are extracted using natural language processing (NLP) techniques. For example, the NLP library spaCy is used to extract words such as "point card," "bargain," and "how to use."

[0099] Subsequently, based on the pre-processed data, the server uses a generative AI model such as GPT-3® to train the model on a large amount of closing talk data. Through this training, the generative AI model learns effective closing scenarios and customer responses. For example, by training the AI ​​with a large amount of data that includes "explanation of loyalty cards," the AI ​​learns to understand the optimal way to explain them.

[0100] Using a trained AI model, the terminal can provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server then uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[0101] Furthermore, the system provides a simulation environment for employee training. Users can take on the role of customers and input questions and responses, and the AI ​​crew will respond in real time. In this simulation mode, if a new employee asks, "How do I use this point card?", the AI ​​crew will respond, "This point card can be used at all stores, and you can easily accumulate points."

[0102] Examples of specific cases and prompt statements

[0103] Specific Example 1: Data Collection and Preprocessing

[0104] Audio data of the closing talk "Explanation of the point card" recorded at store A.

[0105] The Google Speech-to-Text API is used to convert audio data into text data. For example, audio data such as "Explanation of the loyalty card" is converted into text data such as "This is an explanation of the loyalty card."

[0106] For data cleansing, we use the Python pandas library to remove unnecessary background noise and irrelevant conversations.

[0107] Example 2: Real-time response

[0108] A customer asks, "What are the best deals this month?"

[0109] The server uses generative AI models such as GPT-3 to generate the optimal response.

[0110] The device responds in real time with, "This month's special offer is a newly released smartphone."

[0111] Example of a prompt:

[0112] Collect audio data of employee closing speeches, convert it to text data, and then de-noise it. Next, use natural language processing to extract key keywords and train an AI model to create a system for real-time customer support and on-the-job training. Please refer to the following example:

[0113] Audio data from store A: "What are this month's recommended products?"

[0114] Text data conversion: "This month's recommended product is the newly released smartphone."

[0115] Keywords extracted: "Recommended products," "New releases," "Smartphones"

[0116] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer support and effective employee training.

[0117] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0118] Step 1: Data Collection

[0119] The server collects audio and text data of employee closing speeches from each store. It receives the input audio data (e.g., "Explanation of loyalty cards") and saves it as text data. Specifically, it uploads data to the server using protocols such as FTP and HTTP. The input is audio data and text data, and the output is the storage of the collected data.

[0120] Step 2: Convert audio data

[0121] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. The Google Speech-to-Text API is used for this conversion. For example, it converts audio data of an employee saying, "This loyalty card can be used at all stores," into text. The input is audio data, and the output is the converted text data.

[0122] Step 3: Data Cleansing

[0123] The server cleanses the converted text data. Here, noise and unwanted elements, such as background noise and irrelevant conversations, are removed. The Python pandas library is used for this process. Speech utterances like "um" and "uh" are removed from the cleansed data. The input is text data, and the output is cleansed text data.

[0124] Step 4: Keyword Extraction

[0125] The server uses natural language processing (NLP) techniques to extract important keywords and phrases from text data. For example, the NLP library spaCy is used to extract words like "point card," "deals," and "how to use." The input is cleansed text data, and the output consists of important keywords and phrases.

[0126] Step 5: Training the AI ​​model

[0127] The server trains a generative AI model based on pre-processed data. It uses a large amount of closing talk data to train generative AI models such as GPT-3, learning closing scenarios and customer responses. For example, by training the AI ​​with a large amount of conversation data that includes "explanation of loyalty cards," it learns the optimal explanation method. The input is data containing extracted keywords, and the output is the trained generative AI model.

[0128] Step 6: Real-time response

[0129] The terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offers are the newly released smartphones." The input is the customer's question text, and the output is the generated optimal response.

[0130] Step 7: On-the-Job Training Tools

[0131] When the user selects employee training mode, the terminal begins a simulated dialogue with an AI crew. The terminal sends the user's question as text data to the server, which uses a trained AI model to generate and return the optimal response. When a new employee asks, "Are there any special offers?", the AI ​​crew responds, "We're having a double points campaign this month." The input is the user's question, and the output is the AI ​​crew's response.

[0132] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer service and effective employee training.

[0133] (Application Example 1)

[0134] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0135] The problems this invention aims to solve are to provide a means for effectively collecting employees' closing talks and training them in closing techniques, thereby improving the quality of customer service. Furthermore, it aims to improve the efficiency of store operations by enabling employees to provide real-time customer service using smart devices. Additionally, it aims to support employee skill development by providing a simulated training environment.

[0136] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0137] This invention includes a server comprising means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generative AI using the cleansed data, means for providing real-time customer service on smart devices using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables improvement of employees' closing techniques and the quality of customer service. Furthermore, since AI-based customer service can be provided even in stores where it has not yet been implemented, store operations can be made more efficient. In addition, the employee training environment is enhanced, and practical skills can be improved.

[0138] "Employee closing talk" refers to the statements and responses made by employees at the end of a business negotiation with a customer.

[0139] "Means of data collection" refers to systems and methods for recording employees' closing talks as audio or text data and later storing them in a database.

[0140] "Means of converting to text data and cleansing it" refers to the process and technology of converting collected audio data into text using automatic speech recognition (ASR) technology and removing noise and unwanted information from that text.

[0141] "Means of training generative AI" refers to the processes and techniques used to train generative AI models using cleansed data.

[0142] A "trained AI model" refers to an AI model obtained as a result of training a generative AI using collected and pre-processed data.

[0143] A "smart device" refers to a portable device with computer capabilities, such as a smartphone, smart glasses, or head-mounted display.

[0144] "Means of providing real-time customer support" refers to systems and methods that use trained AI models to instantly generate responses to customer inquiries.

[0145] "Means of providing a simulation environment" refers to a training environment in which employees can experience actual work in a simulated manner, and in this context, it includes methods of learning closing techniques through interaction with an AI model.

[0146] "Unopened stores" refer to stores that have not yet opened for business, or stores that cannot immediately be staffed with employees due to reasons such as labor shortages.

[0147] "Means of acting as a proxy for customer service" refers to systems and methods in stores that use AI models to handle customer inquiries on behalf of employees.

[0148] "A means by which an AI crew responds in real time to questions and responses entered by employees acting as customers in a simulation environment" refers to a training method in which employees play the role of customers and ask questions to the AI, to which the AI ​​responds immediately.

[0149] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training. A specific embodiment for carrying out this invention will be described below.

[0150] Hardware to use

[0151] 1. Server: Cloud server or on-premises server

[0152] 2. Smart devices: Smartphones, smart glasses (e.g., Google Glass®), head-mounted displays

[0153] 3. Recording equipment: Microphones worn by employees or portable recorders.

[0154] Software to use

[0155] 1. Automatic Speech Recognition (ASR) Software: Google Cloud Speech-to-Text

[0156] 2. Natural Language Processing (NLP) Libraries: spaCy, BERT

[0157] 3. AI Generative Models: GPT-4(registered trademark), PaLM

[0158] Data collection

[0159] When employees conduct closing speeches, audio data is collected using smart glasses or portable recorders. The collected audio data is periodically uploaded to a server via the smart devices.

[0160] Data preprocessing

[0161] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Then, natural language processing (NLP) technology is used to extract important keywords from the text data and cleanse it of unnecessary noise and irrelevant information.

[0162] AI model training

[0163] The server trains a generative AI model based on pre-processed data. By using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses.

[0164] Real-time response

[0165] When a user receives a question from a customer, that question is sent to a server as text data via their smart device. The server uses a trained AI model to generate the optimal response and sends it back to the user's smart device in real time.

[0166] Specific example

[0167] When a customer asks, "What are this month's recommended products?", the user's smart glasses display the response, "This month's recommended product is the newly released XYZ smartphone. It features a high-performance camera and long battery life."

[0168] OJT tool

[0169] Users can utilize the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The user's questions and responses are sent to the server via a smart device, and the server uses a trained AI model to generate appropriate responses, which are then returned to the user.

[0170] Example of a prompt

[0171] User question: "How do I use this points card?"

[0172] AI Crew's response: "This point card can be used at all stores and you can easily accumulate points."

[0173] In this way, users can learn closing techniques in an environment that closely resembles actual business scenarios.

[0174] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0175] Step 1: Data Collection

[0176] Subject: terminal

[0177] Specific operation: Employees wear smart glasses or portable recorders to record audio data in real time while interacting with customers.

[0178] Input: Audio data of employee closing remarks

[0179] Output: Recorded audio data file

[0180] Data processing / calculations: None

[0181] Step 2: Sending the audio data

[0182] Subject: terminal

[0183] Specific operation: Periodically upload recorded audio data to the server.

[0184] Input: Audio data file

[0185] Output: Audio data sent to the server

[0186] Data processing / calculations: None

[0187] Step 3: Automatic Speech Recognition (ASR)

[0188] Subject: Server

[0189] Specific operation: The server converts the received audio data into text data using automatic speech recognition (ASR) software.

[0190] Input: Audio data

[0191] Output: Text data

[0192] Data processing and calculation: Converting audio data to text data.

[0193] Step 4: Data Cleansing

[0194] Subject: Server

[0195] Specific operation: Performs data cleansing to remove noise and irrelevant information from the converted text data.

[0196] Input: Text data

[0197] Output: Cleansed text data

[0198] Data processing and calculation: Removal of unnecessary information, noise filtering.

[0199] Step 5: Natural Language Processing (NLP)

[0200] Subject: Server

[0201] Specific operation: Natural language processing (NLP) techniques are used on the cleansed text data to extract important keywords and phrases.

[0202] Input: Cleansed text data

[0203] Output: List of keywords and phrases

[0204] Data processing and calculation: Keyword extraction, contextual understanding

[0205] Step 6: AI Model Training

[0206] Subject: Server

[0207] Specific operation: Train a generative AI model based on pre-processed data.

[0208] Input: List of keywords or phrases, text data

[0209] Output: Trained AI model

[0210] Data processing and computation: AI model training, parameter optimization

[0211] Step 7: Real-time response

[0212] Subject: Terminals and servers

[0213] Specific operation: When a user receives a question from a customer, the question is sent to the server as text data via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal.

[0214] Input: Customer questions, text data

[0215] Output: Response text

[0216] Data processing and calculation: Question analysis, response generation

[0217] Step 8: Display the response

[0218] Subject: terminal

[0219] Specific operation: The returned response text is displayed on the screen of smart glasses or other smart devices.

[0220] Input: Response text

[0221] Output: Response displayed on the smart device screen

[0222] Data processing and calculations: Adjustment of display format

[0223] Step 9: Use as an OJT tool

[0224] Subject: Terminals and servers

[0225] Specific operation: The user uses simulation mode to interact with an AI crew for training purposes. The user's questions and responses are sent to the server via the terminal, and the server uses the trained AI model to generate appropriate responses, which are then sent back to the terminal.

[0226] Input: User's question, response

[0227] Output: AI Crew Response Text

[0228] Data processing and calculation: Question analysis, response generation

[0229] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0230] This invention is a system that collects employee closing talks as data and uses that data to train an AI. Furthermore, this system incorporates an emotion engine that recognizes user emotions to provide more effective customer service and employee training.

[0231] Program processing

[0232] Data collection

[0233] The server receives audio and text data of employee closing speeches collected from stores. Data is uploaded from microphones and recording devices installed in the stores. For example, a recording of a "points card explanation" from store A is collected.

[0234] Data preprocessing

[0235] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. Then, it performs data cleansing to remove noise and unnecessary information. For example, background noise and unnecessary dialogue are removed from data such as "points card explanation."

[0236] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add relevant contextual information. For example, keywords such as "loyalty card," "deals," and "campaign" are extracted, and their contextual information is organized.

[0237] AI model training

[0238] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. For example, it is trained to provide optimal responses to specific customer reactions by learning closing talks from art exhibitions.

[0239] Emotional engine integration

[0240] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. Through voice analysis and facial expression analysis, the emotion engine identifies whether the user is happy, confused, or angry.

[0241] Real-time response

[0242] Using trained AI models and an emotion engine, the terminal provides real-time customer support in undeployed stores and understaffed locations. When a user receives a question from a customer, it sends emotion data along with the question to the server. For example, if a customer asks, "What are this month's special offers?", and the emotion engine determines that the customer is excited, that emotion data is also sent to the server.

[0243] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, if the sentiment engine detects customer excitement, it might generate a response such as, "This month's special is the newly released smartphone! There are lots of bonuses too!"

[0244] The generated response is sent back to the terminal, which converts the text data into speech and provides the response to the customer. This allows customers to receive a more personalized response.

[0245] OJT tool

[0246] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. For example, if a new employee asks, "How do I use this point card?", the terminal sends the question and the employee's emotion data to the server.

[0247] The server uses an emotion engine to analyze employees' emotions and provide feedback along with the most appropriate response. For example, if an employee is confused, it might offer specific advice such as, "This loyalty card can be used at all stores and you can easily accumulate points. Let me show you how to register."

[0248] In this way, the system collects employee closing talk data and uses a generative AI and emotion engine for training, providing real-time customer support and efficient employee training.

[0249] The following describes the processing flow.

[0250] Step 1:

[0251] The server receives audio and text data of employee closing speeches from each store. For example, a recording of a "points card explanation" from store A is uploaded to the server.

[0252] Step 2:

[0253] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine then runs and generates text information from the audio.

[0254] Step 3:

[0255] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts to create well-structured data. For example, background noise and unnecessary conversations are removed from the "points card explanation" data.

[0256] Step 4:

[0257] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, it might extract keywords such as "deals," "points card," and "campaign."

[0258] Step 5:

[0259] The server adds contextual information to the extracted keywords, describing how the keywords are used and in what context they are effective. For example, it might add a typical dialogue scenario when describing a "loyalty card."

[0260] Step 6:

[0261] The server uses the processed text data to train the generative AI. Using a large amount of closing talk data, the AI ​​model learns patterns in customer interaction. For example, it learns dialogue patterns for explaining new products.

[0262] Step 7:

[0263] The server deploys the completed AI model and emotion engine, enabling real-time customer support. The AI ​​model and emotion engine reside on either the cloud or edge servers.

[0264] Step 8:

[0265] When a user interacts with a customer in a store, the terminal receives the customer's questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[0266] Step 9:

[0267] The terminal receives voice input and sends it to the server as text data. Information is transferred in real time via the internet.

[0268] Step 10:

[0269] The server uses a trained AI model to generate an appropriate response based on the received text data and sentiment data. For example, if the sentiment engine determines that the customer is excited along with the question, "What are this month's specials?", it will generate a response that takes that sentiment data into account.

[0270] Step 11:

[0271] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[0272] Step 12:

[0273] The device converts received text data into speech and provides a response to the customer. This allows customers to receive a more personalized service.

[0274] Step 13:

[0275] When a user uses AI Crew as an OJT tool, the terminal operates in simulation mode. The user takes on the role of a customer and inputs questions and responses into the terminal.

[0276] Step 14:

[0277] The device sends user-entered questions to the server in real time. Along with the questions, user sentiment data is also transmitted.

[0278] Step 15:

[0279] The server uses an emotion engine to analyze the user's emotions and generate the most appropriate response. For example, if the user asks, "How do I use this loyalty card?" and the server determines that the user is confused, it will respond, "This loyalty card can be used at all stores, and you can easily accumulate points. Let me show you how to use it."

[0280] Step 16:

[0281] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[0282] (Example 2)

[0283] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0284] There is a need for a system that efficiently collects employees' closing talks as data, trains an AI model based on this data to provide real-time customer service, and improves employees' closing techniques. In particular, a system that can analyze users' emotions in real time and provide appropriate responses accordingly contributes to improving customer satisfaction. Also, a system that can effectively handle situations in unopened stores or sites with a shortage of employees is required.

[0285] The specific processing by the specific processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0286] In this invention, the server includes means for collecting employees' closing talks as data, means for converting the collected data into text data using speech recognition technology and performing data cleansing, means for extracting keywords and related information from the cleansed text data using natural language processing technology, means for training a generative AI model using the extracted data, means for analyzing users' emotions using an emotion analysis engine, means for providing real-time customer service using the trained AI model and the emotion analysis engine, and means for providing a simulation environment for employees to learn closing techniques. Thereby, it becomes possible to effectively collect and analyze employees' closing talks and realize real-time response and employee training using an AI model.

[0287] "Employees" refer to people who are employed by a company or organization and engage in work.

[0288] "Closing talk" refers to the conversation or explanation for promoting a conclusion or contract through interaction with customers at the final stage of sales or service provision.

[0289] "Data" refers to a collection of information and numerical values collected and stored for a specific purpose or analysis.

[0290] "Speech recognition technology" is a technology that converts human speech into text data using a computer.

[0291] "Text data" refers to information stored in the form of characters or sentences.

[0292] "Data cleansing" is the process of removing noise and unnecessary information from data to improve its consistency and quality.

[0293] "Natural language processing technology" refers to the technology that enables computers to understand and analyze human natural language.

[0294] A "keyword" is a word or phrase that has particularly important meaning within a document or dialogue.

[0295] A "generative AI model" is an artificial intelligence model generated through machine learning or deep learning based on a large amount of data.

[0296] An "emotion analysis engine" is software or a system that identifies and analyzes a user's emotional state from factors such as voice and facial expressions.

[0297] "Real-time" refers to a state where processing and responses occur instantaneously without delay.

[0298] A "simulation environment" refers to a system or software used to virtually reproduce real-world situations.

[0299] A "customer" is a person or organization that purchases or uses a product or service.

[0300] An "unopened store" refers to a store that has not yet officially started operations.

[0301] This invention relates to a system that collects employees' closing talks, trains a generative AI model using the data, integrates a sentiment analysis engine, and provides real-time customer service and support for employee training.

[0302] System Configuration

[0303] Data Collection

[0304] The server receives voice data of employees' closing talks from microphones and recording devices installed in each store. For example, the voice data of "explanation of the point card" conducted at Store A is the target of collection. In addition, text data can also be received from store logs and message systems.

[0305] Data Preprocessing

[0306] The server converts the received voice data into text data using automatic speech recognition (ASR) technology. For example, Google Cloud Speech-to-Text API is used. After that, data cleaning is performed to remove noise and unnecessary information. For example, background noise and small talk parts are filtered using pandas and NLTK libraries. Next, the server uses natural language processing (NLP) technology to extract important keywords and phrases from the text data and organize the context information. Specific technologies such as spaCy and BERT models are used.

[0307] AI Model Training

[0308] The server trains a generative AI model based on the preprocessed data. AI models used include OpenAI (registered trademark) GPT-3 and Google BERT. For example, the AI model is trained based on scenarios such as "closing talk at an art exhibition" and "closing talk at a mobile shop".

[0309] Integration of Sentiment Engine

[0310] The server uses an emotion analysis engine to collect and analyze emotional data from the user's voice and facial expressions. Software used includes Microsoft® Azure® Emotion API and Affectiva SDK. For example, it can identify emotions such as "confusion" or "excitement" from the user's voice data.

[0311] Real-time response

[0312] The terminal uses a trained AI model and sentiment analysis engine to provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, it sends the question and sentiment data to the server. The server uses the received data to generate the optimal response using a trained AI model, and the generated response is sent back to the terminal. The terminal converts the text data into speech and provides the response to the customer, for example, using Google Cloud Text-to-Speech.

[0313] OJT tool

[0314] Users undergo training using an OJT tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. The server analyzes these questions and employee emotion data, providing feedback along with the most appropriate response. For example, it generates an appropriate response to the question, "How do I use this loyalty card?"

[0315] Specific examples and prompt statements

[0316] For example, the server trains an AI model using closing talk data collected at store A. During this process, an emotion analysis engine is integrated to analyze the user's emotions and generate the optimal response for the customer. In real-time responses, when a customer asks, "Tell me about the new service," the emotion analysis engine senses the customer's interest, and the server generates a response such as, "The new service has the following features..."

[0317] Examples of prompt statements to input into the generative AI model are as follows:

[0318] "Based on training data for closing pitches, generate responses to the customer's question, 'What are the best deals this month?'"

[0319] "Use the emotion engine to provide the best possible response when the user is confused."

[0320] This system effectively collects and analyzes employees' closing pitches, and uses AI models to enable real-time responses and employee training.

[0321] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0322] Step 1:

[0323] Data collection

[0324] The server receives audio data from microphones and recording devices installed in the store. This audio data consists of closing conversations between employees and customers; for example, "audio data of an explanation about the loyalty card at store A" might be input. The server saves this audio data.

[0325] Input: Audio data collected from stores

[0326] Output: Audio data stored on the server

[0327] Specifically, the server periodically runs a script to collect recording data from each store and uploads the data to a central server.

[0328] Step 2:

[0329] Convert to text data

[0330] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, the Google Cloud Speech-to-Text API is used for this conversion. For instance, the text output might read, "Explanation of the benefits of the loyalty card."

[0331] Input: Received audio data

[0332] Output: Text data converted using ASR technology

[0333] Specifically, the server calls the ASR API and executes a script that converts the audio data into text data.

[0334] Step 3:

[0335] Data cleansing

[0336] The server performs data cleansing on the converted text data to remove noise and unnecessary information. For example, it uses libraries such as pandas or NLTK to filter out background noise and casual conversation.

[0337] Input: Text data

[0338] Output: Cleansed text data with noise and unwanted information removed.

[0339] Specifically, the server analyzes the text data and removes unnecessary parts using regular expressions and filtering algorithms.

[0340] Step 4:

[0341] Extraction of keywords and contextual information

[0342] The server applies natural language processing (NLP) techniques to the cleansed text data to extract important keywords and phrases. For example, it uses spaCy or BERT models to extract keywords such as "point card," "deals," and "campaign," and then organizes their contextual information.

[0343] Input: Cleansed text data

[0344] Output: Extracted keywords and contextual information

[0345] In terms of specific operations, the server uses an NLP library to identify specific patterns and keywords from text data and extract relevant contextual data.

[0346] Step 5:

[0347] AI model training

[0348] The server trains a generative AI model based on pre-processed data. Using a large amount of closing talk data, for example, OpenAI GPT-3 or Google BERT, the AI ​​model learns appropriate customer interaction patterns.

[0349] Input: Preprocessed text data

[0350] Output: Trained generative AI model

[0351] In terms of specific operations, the server feeds the dataset into an AI training framework (such as TENSORFLOW® or PyTorch) and performs model training.

[0352] Step 6:

[0353] Emotion analysis

[0354] The server uses an emotion analysis engine to acquire and analyze emotional data from the user's voice and facial expressions. For example, it uses the Microsoft Azure Emotion API or the Affectiva SDK to identify the user's emotional state (joy, confusion, anger, etc.).

[0355] Input: User voice and facial expression data

[0356] Output: Analyzed user sentiment data

[0357] Specifically, the server calls an emotion analysis API to analyze the collected audio data and obtain emotion data.

[0358] Step 7:

[0359] Real-time response

[0360] When the device receives a question from a customer, it sends the question and sentiment data to the server. The server uses this data to generate the optimal response using a trained AI model, and the generated response is sent back to the device. The device then converts the text data into speech using, for example, Google Cloud Text-to-Speech, and provides the response to the customer.

[0361] Input: Customer questions and sentiment data

[0362] Output: Optimal response text and speech-converted response

[0363] In terms of specific operations, the terminal receives user input, provides an interface for sending it to the server, and plays back the response received from the server as audio.

[0364] Step 8:

[0365] Training using OJT tools

[0366] The user conducts employee training using an OJT tool. The user acts as a customer, inputting questions and responses into a terminal. The server analyzes the questions and employee sentiment data to provide optimal responses and feedback.

[0367] Input: Questions, responses, and sentiment data entered by employees.

[0368] Output: Optimal response and feedback

[0369] Specifically, the terminal receives questions and responses entered by employees and sends them to the server. The server uses sentiment analysis and an AI model to generate the optimal response and feedback, and sends it back to the terminal.

[0370] (Application Example 2)

[0371] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0372] Traditional customer service systems lacked sufficient training methods to improve the quality of employee closing pitches, making efficient real-time customer service difficult. Furthermore, limited means of recognizing customer emotions and providing appropriate responses posed a risk of decreased customer satisfaction. Additionally, achieving high-quality customer service was challenging in stores where the system was not yet implemented or in locations with insufficient staff.

[0373] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0374] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generating AI using the cleansed data, means for acquiring emotional data from the user's voice and facial expressions and generating the optimal customer response based on that emotional data, means for providing customer support in real time using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables employee training and high-quality customer support in stores where the system has not yet been deployed.

[0375] "Means for collecting employee closing talks as data" refers to methods and apparatus for recording and collecting closing talks conducted by employees with customers as audio or text data.

[0376] "Means for converting collected data into text data and cleansing it" refers to data cleansing techniques and methods that convert collected audio data into text data using automatic speech recognition (ASR) technology and remove noise and unnecessary information.

[0377] "Methods for training generative AI using cleansed data" refers to training processes and techniques for training generative AI models using pre-processed text data.

[0378] "Methods for providing real-time customer support using trained AI models" refers to systems or methods that use trained AI models to provide immediate and appropriate responses to customer questions and requests.

[0379] "Means of providing a simulation environment for employees to learn closing techniques" refers to training environments and methods that enable employees to improve their closing techniques through customer interaction simulations.

[0380] "Methods for acquiring emotional data from a user's voice and facial expressions and generating the optimal customer response based on that emotional data" refers to technologies and methods that analyze a user's voice and facial expressions to extract emotional data, and then use that data to generate the optimal response using an AI model.

[0381] "Methods for using AI models trained to handle customer service on behalf of stores that do not yet have a presence" refers to systems or methods that use AI models to automatically handle customer service in areas where physical stores do not yet exist or in stores that lack sufficient staff.

[0382] "A method for an AI crew to respond in real time to questions and responses entered by employees acting as customers" refers to a method or system in which employees take on the role of customers in a training environment, input questions and responses, and an AI model immediately responds to those questions.

[0383] This invention is a "smart on-site assistant" system for improving customer service in physical stores, and is an application using smart glasses. This system collects employee closing talk data and combines it with a generative AI model and an emotion engine.

[0384] Data collection

[0385] The server records the closing conversations that store employees have with customers using the microphone on smart glasses and collects the data as audio. The collected data is later converted into text data.

[0386] Data preprocessing

[0387] The server converts the collected audio data into text using automatic speech recognition (ASR) technology and performs data cleansing to remove noise and irrelevant information. This process eliminates background noise and unnecessary conversation, highlighting important closing remarks.

[0388] AI model training

[0389] The server trains the generative AI using cleansed text data. Based on a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. The trained generative AI model is then used to provide optimal customer service.

[0390] Emotional engine integration

[0391] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. The emotion engine identifies whether the user is happy, confused, or angry through voice analysis and facial expression analysis.

[0392] Real-time response

[0393] Using a trained AI model and emotion engine, the device provides real-time customer support. When a user receives a question from a customer, it sends the question and emotion data to the server. For example, if a customer asks, "Are there any special offers for this product?", and the emotion engine identifies that the customer is interested, this emotion data is also sent.

[0394] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, it might generate a response like, "This product has a limited-time discount this month. Plus, you'll earn points!"

[0395] OJT tool

[0396] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. In a simulated environment where employees act as customers, they input questions and responses in real time through smart glasses, and the AI ​​model responds instantly. This process allows employees to efficiently improve their customer service skills for actual store environments.

[0397] Specific example

[0398] In a real store, salesperson A, wearing smart glasses, is interacting with customer B. When customer B asks, "Are there any special offers for this product?", the smart glasses record the statement and send it to a server. The server converts the audio data into text data, an emotion engine identifies customer B's interests, and a generative AI model generates a response based on that information. The response, "This product has a limited-time discount offer this month. Plus, you'll earn points!", is displayed on the smart glasses or communicated to the customer via voice.

[0399] Example of a prompt

[0400] "customer_question: Are there any special offers for this product? emotion: Interested"

[0401] Output: "This month only, the product offers a discount and the opportunity to earn points."

[0402] Thus, this invention provides a specific system and method for improving the quality of customer service in physical stores and supporting employee training.

[0403] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0404] Step 1:

[0405] Input: Voice and facial expression data of employees and customers at physical stores.

[0406] The server records and videos employee-customer conversations and customer facial expressions in real time via the microphone and camera built into the smart glasses.

[0407] Output: Recorded audio data and customer facial image data

[0408] Step 2:

[0409] Input: Recorded audio data

[0410] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Specifically, it uses Google's automatic speech recognition API to convert audio to text, removing noise and unwanted parts during the process.

[0411] Output: Converted text data

[0412] Step 3:

[0413] Input: Recorded audio data and customer facial image data

[0414] The server analyzes audio data and facial imagery, and uses an emotion engine to identify the customer's emotions. Specifically, it uses the DeepFace library to analyze emotions from the customer's facial expressions and also estimates emotions from their voice.

[0415] Output: Identified sentiment data (e.g., interest, joy, confusion, etc.)

[0416] Step 4:

[0417] Input: Converted text data and identified sentiment data

[0418] The server generates the optimal response using a generative AI model. Specifically, it uses OpenAI's GPT-3 to create prompt sentences based on input text and sentiment data, and then generates an appropriate response accordingly.

[0419] Output: Generated response text

[0420] Step 5:

[0421] Input: Generated response text

[0422] The device delivers the generated response text to the customer. The text is displayed on the smart glasses' screen or delivered to the customer as audio via a text-to-speech function.

[0423] Output: Customer response display or voice response

[0424] Step 6:

[0425] Input: Input of employee questions and responses in the OJT simulation environment.

[0426] In employee training mode, users act as customers and input questions and responses into smart glasses.

[0427] Output: Input data in the simulation environment

[0428] Step 7:

[0429] Input: Input data in the simulation environment

[0430] The server uses an emotion engine and a generative AI model to generate optimal responses to employee questions and reactions, and provides the results as feedback.

[0431] Output: Generated response feedback, analysis results

[0432] This system allows for efficient customer service at physical stores while simultaneously conducting employee training.

[0433] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating 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.

[0434] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0435] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0436] [Second Embodiment]

[0437] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0438] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0439] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0440] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0441] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0442] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0443] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0444] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0445] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0446] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0447] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0448] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. 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".

[0449] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training.

[0450] Program processing

[0451] Data collection

[0452] The server periodically collects audio and text data of employee closing conversations from each store. This includes actual customer interaction audio conversations from the field and successful examples of existing closing conversations. For example, a recording of a "points card explanation" from store A will be collected.

[0453] Data preprocessing

[0454] The collected data is first converted from audio data to text data by the server. Automatic speech recognition (ASR) technology is used for this process. After that, data cleansing is performed to remove noise and unnecessary information. For example, background noise and irrelevant conversations are removed from data such as "Explanation of loyalty cards."

[0455] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add related contextual information. For example, keywords such as "point card," "deals," and "how to use" are extracted, and their contextual information is organized.

[0456] AI model training

[0457] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses. During the training process, the AI's ability to generate appropriate responses improves.

[0458] Real-time response

[0459] Using a trained AI model, the terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to a server via the terminal. The server uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[0460] OJT tool

[0461] Employees can use the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The terminal receives questions from the user and sends them to the server. The server generates responses using a trained AI model and provides them to the user through the terminal. Through this process, employees can learn closing techniques in a situation close to actual customer interaction. For example, if a new employee asks, "How do I use this loyalty card?", the AI ​​crew will respond, "This loyalty card can be used at all stores, and you can easily accumulate points."

[0462] As described above, this system collects data on employees' closing talks, trains a generating AI, and uses the trained AI model to handle customer interactions in real time. It can also be used as an on-the-job training (OJT) tool to support employees in efficiently learning closing techniques.

[0463] The following describes the processing flow.

[0464] Step 1:

[0465] The server receives audio and text data of employee closing speeches collected from each store. This data is uploaded from store microphones and recording devices.

[0466] Step 2:

[0467] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine operates and generates text information from the audio.

[0468] Step 3:

[0469] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts, and prepares unstructured data. This makes the data suitable for analysis.

[0470] Step 4:

[0471] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, words like "bargain," "points card," and "campaign" might be extracted.

[0472] Step 5:

[0473] The server adds contextual information related to the extracted keywords. Specifically, it describes how the keywords are used and in what scenarios they are effective.

[0474] Step 6:

[0475] The server uses the processed text data to train an AI model. This includes a process of training the model with successful examples from the collected conversation data. Machine learning algorithms are used to enable the AI ​​model to learn patterns in customer interactions.

[0476] Step 7:

[0477] The server deploys the AI ​​model after training is complete. This model is located in the cloud or on an edge server. This enables real-time customer support.

[0478] Step 8:

[0479] When a user interacts with a customer in a store, the terminal receives customer questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[0480] Step 9:

[0481] The terminal receives voice input and sends it to the server as text data. Communication takes place in real time over the internet.

[0482] Step 10:

[0483] The server uses a trained AI model to generate an appropriate response based on the received text data. For example, in response to the question "What are this month's special offers?", it will generate the response "This month's special offer is a newly released smartphone."

[0484] Step 11:

[0485] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[0486] Step 12:

[0487] The device converts received text data into speech and provides a response to the customer. This allows customers to obtain information in a natural conversational flow.

[0488] Step 13:

[0489] When a user uses AI Crew as an OJT tool, the terminal is operated in simulation mode. The user takes on the role of a customer and inputs questions and responses.

[0490] Step 14:

[0491] The terminal sends the user's entered questions to the server. The questions arrive at the server in real time.

[0492] Step 15:

[0493] The server uses a trained AI model to generate responses to input questions. For example, in response to the question, "How do I use this loyalty card?", it generates the response, "This loyalty card can be used at all stores, and you can easily accumulate points."

[0494] Step 16:

[0495] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[0496] (Example 1)

[0497] Next, we will describe Example 1. 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".

[0498] Improving the quality of employees' closing pitches and efficiently serving customers is a crucial challenge for many service industries. In particular, in under-established stores or locations with insufficient staff, standardizing closing techniques is difficult, leading to inconsistencies in customer satisfaction. Furthermore, training new employees requires significant time and effort, highlighting the need for effective training methods.

[0499] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0500] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected audio data into text data and cleaning it, means for extracting important keywords and phrases from the cleansed data, means for training a generative AI model using the extracted data, means for providing real-time customer support using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables data collection of closing talks, training of a generative AI model, real-time response, and effective training of employees.

[0501] "Employee closing talk" refers to the final persuasive or explanatory remarks made by an employee when interacting with a customer.

[0502] "Means of collecting data" refers to systems and methods for collecting closing talks in audio or text format from stores and on-site locations.

[0503] "Means of converting collected audio data into text data" refers to the process or technology of converting audio data into text information using speech recognition technology.

[0504] "Methods of cleansing" refer to methods and techniques for removing noise and unnecessary information from converted text data, and organizing and formatting the data.

[0505] "Methods for extracting important keywords and phrases" refers to methods that use natural language processing techniques to extract important words and phrases from text data and add contextual information.

[0506] "Means for training generative AI models" refers to the processes and techniques for training and updating artificial intelligence models using collected and pre-processed data.

[0507] "Methods for providing real-time customer support using trained AI models" refers to systems and methods that use responses generated by generative AI models to immediately address customer questions and requests.

[0508] "Means of providing a simulation environment for employees to learn closing techniques" refers to a training environment in which employees can acquire closing techniques in situations that closely resemble actual customer interactions.

[0509] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI model, and utilizes the trained AI model for customer service and employee training.

[0510] The server periodically collects audio and text data of employee closing speeches from each store. The collected audio data is converted into text data using the Google Speech-to-Text API. For example, a recording of a "loyalty card explanation" from store A might be included.

[0511] Next, the server cleanses the converted text data, removing noise and unnecessary information. The Python pandas library is used for this process. From the cleansed data, important keywords and phrases are extracted using natural language processing (NLP) techniques. For example, the NLP library spaCy is used to extract words such as "point card," "bargain," and "how to use."

[0512] Subsequently, based on the pre-processed data, the server uses a generative AI model such as GPT-3 to train the model on a large amount of closing talk data. Through this training, the generative AI model learns effective closing scenarios and customer responses. For example, by training the AI ​​with a large amount of data that includes "explanation of loyalty cards," the AI ​​learns to understand the optimal way to explain them.

[0513] Using a trained AI model, the terminal can provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server then uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[0514] Furthermore, the system provides a simulation environment for employee training. Users can take on the role of customers and input questions and responses, and the AI ​​crew will respond in real time. In this simulation mode, if a new employee asks, "How do I use this point card?", the AI ​​crew will respond, "This point card can be used at all stores, and you can easily accumulate points."

[0515] Examples of specific cases and prompt statements

[0516] Specific Example 1: Data Collection and Preprocessing

[0517] Audio data of the closing talk "Explanation of the point card" recorded at store A.

[0518] The Google Speech-to-Text API is used to convert audio data into text data. For example, audio data such as "Explanation of the loyalty card" is converted into text data such as "This is an explanation of the loyalty card."

[0519] For data cleansing, we use the Python pandas library to remove unnecessary background noise and irrelevant conversations.

[0520] Example 2: Real-time response

[0521] A customer asks, "What are the best deals this month?"

[0522] The server uses generative AI models such as GPT-3 to generate the optimal response.

[0523] The device responds in real time with, "This month's special offer is a newly released smartphone."

[0524] Example of a prompt:

[0525] Collect audio data of employee closing speeches, convert it to text data, and then de-noise it. Next, use natural language processing to extract key keywords and train an AI model to create a system for real-time customer support and on-the-job training. Please refer to the following example:

[0526] Audio data from store A: "What are this month's recommended products?"

[0527] Text data conversion: "This month's recommended product is the newly released smartphone."

[0528] Keywords extracted: "Recommended products," "New releases," "Smartphones"

[0529] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer support and effective employee training.

[0530] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0531] Step 1: Data Collection

[0532] The server collects audio and text data of employee closing speeches from each store. It receives the input audio data (e.g., "Explanation of loyalty cards") and saves it as text data. Specifically, it uploads data to the server using protocols such as FTP and HTTP. The input is audio data and text data, and the output is the storage of the collected data.

[0533] Step 2: Convert audio data

[0534] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. The Google Speech-to-Text API is used for this conversion. For example, it converts audio data of an employee saying, "This loyalty card can be used at all stores," into text. The input is audio data, and the output is the converted text data.

[0535] Step 3: Data Cleansing

[0536] The server cleanses the converted text data. Here, noise and unwanted elements, such as background noise and irrelevant conversations, are removed. The Python pandas library is used for this process. Speech utterances like "um" and "uh" are removed from the cleansed data. The input is text data, and the output is cleansed text data.

[0537] Step 4: Keyword Extraction

[0538] The server uses natural language processing (NLP) techniques to extract important keywords and phrases from text data. For example, the NLP library spaCy is used to extract words like "point card," "deals," and "how to use." The input is cleansed text data, and the output consists of important keywords and phrases.

[0539] Step 5: Training the AI ​​model

[0540] The server trains a generative AI model based on pre-processed data. It uses a large amount of closing talk data to train generative AI models such as GPT-3, learning closing scenarios and customer responses. For example, by training the AI ​​with a large amount of conversation data that includes "explanation of loyalty cards," it learns the optimal explanation method. The input is data containing extracted keywords, and the output is the trained generative AI model.

[0541] Step 6: Real-time response

[0542] The terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offers are the newly released smartphones." The input is the customer's question text, and the output is the generated optimal response.

[0543] Step 7: On-the-Job Training Tools

[0544] When the user selects employee training mode, the terminal begins a simulated dialogue with an AI crew. The terminal sends the user's question as text data to the server, which uses a trained AI model to generate and return the optimal response. When a new employee asks, "Are there any special offers?", the AI ​​crew responds, "We're having a double points campaign this month." The input is the user's question, and the output is the AI ​​crew's response.

[0545] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer service and effective employee training.

[0546] (Application Example 1)

[0547] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0548] The problems this invention aims to solve are to provide a means for effectively collecting employees' closing talks and training them in closing techniques, thereby improving the quality of customer service. Furthermore, it aims to improve the efficiency of store operations by enabling employees to provide real-time customer service using smart devices. Additionally, it aims to support employee skill development by providing a simulated training environment.

[0549] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0550] This invention includes a server comprising means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generative AI using the cleansed data, means for providing real-time customer service on smart devices using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables improvement of employees' closing techniques and the quality of customer service. Furthermore, since AI-based customer service can be provided even in stores where it has not yet been implemented, store operations can be made more efficient. In addition, the employee training environment is enhanced, and practical skills can be improved.

[0551] "Employee closing talk" refers to the statements and responses made by employees at the end of a business negotiation with a customer.

[0552] "Means of data collection" refers to systems and methods for recording employees' closing talks as audio or text data and later storing them in a database.

[0553] "Means of converting to text data and cleansing it" refers to the process and technology of converting collected audio data into text using automatic speech recognition (ASR) technology and removing noise and unwanted information from that text.

[0554] "Means of training generative AI" refers to the processes and techniques used to train generative AI models using cleansed data.

[0555] A "trained AI model" refers to an AI model obtained as a result of training a generative AI using collected and pre-processed data.

[0556] A "smart device" refers to a portable device with computer capabilities, such as a smartphone, smart glasses, or head-mounted display.

[0557] "Means of providing real-time customer support" refers to systems and methods that use trained AI models to instantly generate responses to customer inquiries.

[0558] "Means of providing a simulation environment" refers to a training environment in which employees can experience actual work in a simulated manner, and in this context, it includes methods of learning closing techniques through interaction with an AI model.

[0559] "Unopened stores" refer to stores that have not yet opened for business, or stores that cannot immediately be staffed with employees due to reasons such as labor shortages.

[0560] "Means of acting as a proxy for customer service" refers to systems and methods in stores that use AI models to handle customer inquiries on behalf of employees.

[0561] "A means by which an AI crew responds in real time to questions and responses entered by employees acting as customers in a simulation environment" refers to a training method in which employees play the role of customers and ask questions to the AI, to which the AI ​​responds immediately.

[0562] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training. A specific embodiment for carrying out this invention will be described below.

[0563] Hardware to use

[0564] 1. Server: Cloud server or on-premises server

[0565] 2. Smart devices: Smartphones, smart glasses (e.g., Google Glass), head-mounted displays

[0566] 3. Recording equipment: Microphones worn by employees or portable recorders.

[0567] Software to use

[0568] 1. Automatic Speech Recognition (ASR) Software: Google Cloud Speech-to-Text

[0569] 2. Natural Language Processing (NLP) Libraries: spaCy, BERT

[0570] 3. AI Generative Models: GPT-4, PaLM

[0571] Data collection

[0572] When employees conduct closing speeches, audio data is collected using smart glasses or portable recorders. The collected audio data is periodically uploaded to a server via the smart devices.

[0573] Data preprocessing

[0574] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Then, natural language processing (NLP) technology is used to extract important keywords from the text data and cleanse it of unnecessary noise and irrelevant information.

[0575] AI model training

[0576] The server trains a generative AI model based on pre-processed data. By using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses.

[0577] Real-time response

[0578] When a user receives a question from a customer, that question is sent to a server as text data via their smart device. The server uses a trained AI model to generate the optimal response and sends it back to the user's smart device in real time.

[0579] Specific example

[0580] When a customer asks, "What are this month's recommended products?", the user's smart glasses display the response, "This month's recommended product is the newly released XYZ smartphone. It features a high-performance camera and long battery life."

[0581] OJT tool

[0582] Users can utilize the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The user's questions and responses are sent to the server via a smart device, and the server uses a trained AI model to generate appropriate responses, which are then returned to the user.

[0583] Example of a prompt

[0584] User question: "How do I use this points card?"

[0585] AI Crew's response: "This point card can be used at all stores and you can easily accumulate points."

[0586] In this way, users can learn closing techniques in an environment that closely resembles actual business scenarios.

[0587] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0588] Step 1: Data Collection

[0589] Subject: terminal

[0590] Specific operation: Employees wear smart glasses or portable recorders to record audio data in real time while interacting with customers.

[0591] Input: Audio data of employee closing remarks

[0592] Output: Recorded audio data file

[0593] Data processing / calculations: None

[0594] Step 2: Sending the audio data

[0595] Subject: terminal

[0596] Specific operation: Periodically upload recorded audio data to the server.

[0597] Input: Audio data file

[0598] Output: Audio data sent to the server

[0599] Data processing / calculations: None

[0600] Step 3: Automatic Speech Recognition (ASR)

[0601] Subject: Server

[0602] Specific operation: The server converts the received audio data into text data using automatic speech recognition (ASR) software.

[0603] Input: Audio data

[0604] Output: Text data

[0605] Data processing and calculation: Converting audio data to text data.

[0606] Step 4: Data Cleansing

[0607] Subject: Server

[0608] Specific operation: Performs data cleansing to remove noise and irrelevant information from the converted text data.

[0609] Input: Text data

[0610] Output: Cleansed text data

[0611] Data processing and calculation: Removal of unnecessary information, noise filtering.

[0612] Step 5: Natural Language Processing (NLP)

[0613] Subject: Server

[0614] Specific operation: Natural language processing (NLP) techniques are used on the cleansed text data to extract important keywords and phrases.

[0615] Input: Cleansed text data

[0616] Output: List of keywords and phrases

[0617] Data processing and calculation: Keyword extraction, contextual understanding

[0618] Step 6: AI Model Training

[0619] Subject: Server

[0620] Specific operation: Train a generative AI model based on pre-processed data.

[0621] Input: List of keywords or phrases, text data

[0622] Output: Trained AI model

[0623] Data processing and computation: AI model training, parameter optimization

[0624] Step 7: Real-time response

[0625] Subject: Terminals and servers

[0626] Specific operation: When a user receives a question from a customer, the question is sent to the server as text data via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal.

[0627] Input: Customer questions, text data

[0628] Output: Response text

[0629] Data processing and calculation: Question analysis, response generation

[0630] Step 8: Display the response

[0631] Subject: terminal

[0632] Specific operation: The returned response text is displayed on the screen of smart glasses or other smart devices.

[0633] Input: Response text

[0634] Output: Response displayed on the smart device screen

[0635] Data processing and calculations: Adjustment of display format

[0636] Step 9: Use as an OJT tool

[0637] Subject: Terminals and servers

[0638] Specific operation: The user uses simulation mode to interact with an AI crew for training purposes. The user's questions and responses are sent to the server via the terminal, and the server uses the trained AI model to generate appropriate responses, which are then sent back to the terminal.

[0639] Input: User's question, response

[0640] Output: AI Crew Response Text

[0641] Data processing and calculation: Question analysis, response generation

[0642] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0643] This invention is a system that collects employee closing talks as data and uses that data to train an AI. Furthermore, this system incorporates an emotion engine that recognizes user emotions to provide more effective customer service and employee training.

[0644] Program processing

[0645] Data collection

[0646] The server receives audio and text data of employee closing speeches collected from stores. Data is uploaded from microphones and recording devices installed in the stores. For example, a recording of a "points card explanation" from store A is collected.

[0647] Data preprocessing

[0648] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. Then, it performs data cleansing to remove noise and unnecessary information. For example, background noise and unnecessary dialogue are removed from data such as "points card explanation."

[0649] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add relevant contextual information. For example, keywords such as "loyalty card," "deals," and "campaign" are extracted, and their contextual information is organized.

[0650] AI model training

[0651] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. For example, it is trained to provide optimal responses to specific customer reactions by learning closing talks from art exhibitions.

[0652] Emotional engine integration

[0653] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. Through voice analysis and facial expression analysis, the emotion engine identifies whether the user is happy, confused, or angry.

[0654] Real-time response

[0655] Using trained AI models and an emotion engine, the terminal provides real-time customer support in undeployed stores and understaffed locations. When a user receives a question from a customer, it sends emotion data along with the question to the server. For example, if a customer asks, "What are this month's special offers?", and the emotion engine determines that the customer is excited, that emotion data is also sent to the server.

[0656] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, if the sentiment engine detects customer excitement, it might generate a response such as, "This month's special is the newly released smartphone! There are lots of bonuses too!"

[0657] The generated response is sent back to the terminal, which converts the text data into speech and provides the response to the customer. This allows customers to receive a more personalized response.

[0658] OJT tool

[0659] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. For example, if a new employee asks, "How do I use this point card?", the terminal sends the question and the employee's emotion data to the server.

[0660] The server uses an emotion engine to analyze employees' emotions and provide feedback along with the most appropriate response. For example, if an employee is confused, it might offer specific advice such as, "This loyalty card can be used at all stores and you can easily accumulate points. Let me show you how to register."

[0661] In this way, the system collects employee closing talk data and uses a generative AI and emotion engine for training, providing real-time customer support and efficient employee training.

[0662] The following describes the processing flow.

[0663] Step 1:

[0664] The server receives audio and text data of employee closing speeches from each store. For example, a recording of a "points card explanation" from store A is uploaded to the server.

[0665] Step 2:

[0666] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine then runs and generates text information from the audio.

[0667] Step 3:

[0668] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts to create well-structured data. For example, background noise and unnecessary conversations are removed from the "points card explanation" data.

[0669] Step 4:

[0670] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, it might extract keywords such as "deals," "points card," and "campaign."

[0671] Step 5:

[0672] The server adds contextual information to the extracted keywords, describing how the keywords are used and in what context they are effective. For example, it might add a typical dialogue scenario when describing a "loyalty card."

[0673] Step 6:

[0674] The server uses the processed text data to train the generative AI. Using a large amount of closing talk data, the AI ​​model learns patterns in customer interaction. For example, it learns dialogue patterns for explaining new products.

[0675] Step 7:

[0676] The server deploys the completed AI model and emotion engine, enabling real-time customer support. The AI ​​model and emotion engine reside on either the cloud or edge servers.

[0677] Step 8:

[0678] When a user interacts with a customer in a store, the terminal receives the customer's questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[0679] Step 9:

[0680] The terminal receives voice input and sends it to the server as text data. Information is transferred in real time via the internet.

[0681] Step 10:

[0682] The server uses a trained AI model to generate an appropriate response based on the received text data and sentiment data. For example, if the sentiment engine determines that the customer is excited along with the question, "What are this month's specials?", it will generate a response that takes that sentiment data into account.

[0683] Step 11:

[0684] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[0685] Step 12:

[0686] The device converts received text data into speech and provides a response to the customer. This allows customers to receive a more personalized service.

[0687] Step 13:

[0688] When a user uses AI Crew as an OJT tool, the terminal operates in simulation mode. The user takes on the role of a customer and inputs questions and responses into the terminal.

[0689] Step 14:

[0690] The device sends user-entered questions to the server in real time. Along with the questions, user sentiment data is also transmitted.

[0691] Step 15:

[0692] The server uses an emotion engine to analyze the user's emotions and generate the most appropriate response. For example, if the user asks, "How do I use this loyalty card?" and the server determines that the user is confused, it will respond, "This loyalty card can be used at all stores, and you can easily accumulate points. Let me show you how to use it."

[0693] Step 16:

[0694] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[0695] (Example 2)

[0696] Next, we will describe Example 2. 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".

[0697] There is a need for a system that efficiently collects data on employees' closing pitches, uses this data to train AI models for real-time customer service, and improves employees' closing skills. In particular, a system that can analyze user emotions in real time and provide appropriate responses will contribute to improving customer satisfaction. Furthermore, a system that can effectively operate in stores where the system is not yet implemented or in locations with insufficient staff is also necessary.

[0698] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0699] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data using speech recognition technology and performing data cleansing, means for extracting keywords and related information from the cleansed text data using natural language processing technology, means for training a generative AI model using the extracted data, means for analyzing user emotions using an emotion analysis engine, means for providing real-time customer support using the trained AI model and emotion analysis engine, and means for providing a simulation environment for employees to learn closing techniques. This makes it possible to effectively collect and analyze employee closing talks and to realize real-time support and employee training using an AI model.

[0700] An "employee" is a person who is employed by a company or organization and engages in work.

[0701] "Closing talk" refers to the dialogue and explanations used to encourage customers to close a deal or sign a contract during the final stages of sales or service provision.

[0702] "Data" refers to a collection of information or numerical data that is collected and stored for a specific purpose or analysis.

[0703] "Speech recognition technology" is a technology that converts human speech into text data using a computer.

[0704] "Text data" refers to information stored in the form of characters or sentences.

[0705] "Data cleansing" is the process of removing noise and unnecessary information from data to improve its consistency and quality.

[0706] "Natural language processing technology" refers to the technology that enables computers to understand and analyze human natural language.

[0707] A "keyword" is a word or phrase that has particularly important meaning within a document or dialogue.

[0708] A "generative AI model" is an artificial intelligence model generated through machine learning or deep learning based on a large amount of data.

[0709] An "emotion analysis engine" is software or a system that identifies and analyzes a user's emotional state from factors such as voice and facial expressions.

[0710] "Real-time" refers to a state where processing and responses occur instantaneously without delay.

[0711] A "simulation environment" refers to a system or software used to virtually reproduce real-world situations.

[0712] A "customer" is a person or organization that purchases or uses a product or service.

[0713] An "unopened store" refers to a store that has not yet officially started operations.

[0714] This invention relates to a system that collects employee closing talks, uses that data to train a generative AI model, and further integrates an emotion analysis engine to provide real-time customer service and support employee training.

[0715] System Configuration

[0716] Data collection

[0717] The server receives audio data of employees' closing speeches from microphones and recording devices installed in each store. For example, audio data of a "points card explanation" given at store A will be collected. Text data can also be received from store logs and messaging systems.

[0718] Data preprocessing

[0719] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, it uses the Google Cloud Speech-to-Text API. After that, it performs data cleansing to remove noise and unnecessary information. For example, it filters out background noise and casual conversation using libraries such as pandas or NLTK. Next, the server uses natural language processing (NLP) technology to extract important keywords and phrases from the text data and organize contextual information. Specific technologies used include spaCy and the BERT model.

[0720] AI model training

[0721] The server trains a generative AI model based on pre-processed data. AI models used include OpenAI GPT-3 and Google BERT. For example, the AI ​​model is trained based on scenarios such as "closing speech at an art exhibition" or "closing speech at a mobile phone shop."

[0722] Emotional engine integration

[0723] The server uses an emotion analysis engine to collect and analyze emotional data from the user's voice and facial expressions. Software used includes the Microsoft Azure Emotion API and the Affectiva SDK. For example, it can identify emotions such as "confusion" or "excitement" from the user's voice data.

[0724] Real-time response

[0725] The terminal uses a trained AI model and sentiment analysis engine to provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, it sends the question and sentiment data to the server. The server uses the received data to generate the optimal response using a trained AI model, and the generated response is sent back to the terminal. The terminal converts the text data into speech and provides the response to the customer, for example, using Google Cloud Text-to-Speech.

[0726] OJT tool

[0727] Users undergo training using an OJT tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. The server analyzes these questions and employee emotion data, providing feedback along with the most appropriate response. For example, it generates an appropriate response to the question, "How do I use this loyalty card?"

[0728] Specific examples and prompt statements

[0729] For example, the server trains an AI model using closing talk data collected at store A. During this process, an emotion analysis engine is integrated to analyze the user's emotions and generate the optimal response for the customer. In real-time responses, when a customer asks, "Tell me about the new service," the emotion analysis engine senses the customer's interest, and the server generates a response such as, "The new service has the following features..."

[0730] Examples of prompt statements to input into the generative AI model are as follows:

[0731] "Based on training data for closing pitches, generate responses to the customer's question, 'What are the best deals this month?'"

[0732] "Use the emotion engine to provide the best possible response when the user is confused."

[0733] This system effectively collects and analyzes employees' closing pitches, and uses AI models to enable real-time responses and employee training.

[0734] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0735] Step 1:

[0736] Data collection

[0737] The server receives audio data from microphones and recording devices installed in the store. This audio data consists of closing conversations between employees and customers; for example, "audio data of an explanation about the loyalty card at store A" might be input. The server saves this audio data.

[0738] Input: Audio data collected from stores

[0739] Output: Audio data stored on the server

[0740] Specifically, the server periodically runs a script to collect recording data from each store and uploads the data to a central server.

[0741] Step 2:

[0742] Convert to text data

[0743] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, the Google Cloud Speech-to-Text API is used for this conversion. For instance, the text output might read, "Explanation of the benefits of the loyalty card."

[0744] Input: Received audio data

[0745] Output: Text data converted using ASR technology

[0746] Specifically, the server calls the ASR API and executes a script that converts the audio data into text data.

[0747] Step 3:

[0748] Data cleansing

[0749] The server performs data cleansing on the converted text data to remove noise and unnecessary information. For example, it uses libraries such as pandas or NLTK to filter out background noise and casual conversation.

[0750] Input: Text data

[0751] Output: Cleansed text data with noise and unwanted information removed.

[0752] Specifically, the server analyzes the text data and removes unnecessary parts using regular expressions and filtering algorithms.

[0753] Step 4:

[0754] Extraction of keywords and contextual information

[0755] The server applies natural language processing (NLP) techniques to the cleansed text data to extract important keywords and phrases. For example, it uses spaCy or BERT models to extract keywords such as "point card," "deals," and "campaign," and then organizes their contextual information.

[0756] Input: Cleansed text data

[0757] Output: Extracted keywords and contextual information

[0758] In terms of specific operations, the server uses an NLP library to identify specific patterns and keywords from text data and extract relevant contextual data.

[0759] Step 5:

[0760] AI model training

[0761] The server trains a generative AI model based on pre-processed data. Using a large amount of closing talk data, for example, OpenAI GPT-3 or Google BERT, the AI ​​model learns appropriate customer interaction patterns.

[0762] Input: Preprocessed text data

[0763] Output: Trained generative AI model

[0764] In terms of specific operations, the server feeds the dataset into an AI training framework (such as TensorFlow or PyTorch) and performs model training.

[0765] Step 6:

[0766] Emotion analysis

[0767] The server uses an emotion analysis engine to acquire and analyze emotional data from the user's voice and facial expressions. For example, it uses the Microsoft Azure Emotion API or the Affectiva SDK to identify the user's emotional state (joy, confusion, anger, etc.).

[0768] Input: User voice and facial expression data

[0769] Output: Analyzed user sentiment data

[0770] Specifically, the server calls an emotion analysis API to analyze the collected audio data and obtain emotion data.

[0771] Step 7:

[0772] Real-time response

[0773] When the device receives a question from a customer, it sends the question and sentiment data to the server. The server uses this data to generate the optimal response using a trained AI model, and the generated response is sent back to the device. The device then converts the text data into speech using, for example, Google Cloud Text-to-Speech, and provides the response to the customer.

[0774] Input: Customer questions and sentiment data

[0775] Output: Optimal response text and speech-converted response

[0776] In terms of specific operations, the terminal receives user input, provides an interface for sending it to the server, and plays back the response received from the server as audio.

[0777] Step 8:

[0778] Training using OJT tools

[0779] The user conducts employee training using an OJT tool. The user acts as a customer, inputting questions and responses into a terminal. The server analyzes the questions and employee sentiment data to provide optimal responses and feedback.

[0780] Input: Questions, responses, and sentiment data entered by employees.

[0781] Output: Optimal response and feedback

[0782] Specifically, the terminal receives questions and responses entered by employees and sends them to the server. The server uses sentiment analysis and an AI model to generate the optimal response and feedback, and sends it back to the terminal.

[0783] (Application Example 2)

[0784] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0785] Traditional customer service systems lacked sufficient training methods to improve the quality of employee closing pitches, making efficient real-time customer service difficult. Furthermore, limited means of recognizing customer emotions and providing appropriate responses posed a risk of decreased customer satisfaction. Additionally, achieving high-quality customer service was challenging in stores where the system was not yet implemented or in locations with insufficient staff.

[0786] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0787] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generating AI using the cleansed data, means for acquiring emotional data from the user's voice and facial expressions and generating the optimal customer response based on that emotional data, means for providing customer support in real time using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables employee training and high-quality customer support in stores where the system has not yet been deployed.

[0788] "Means for collecting employee closing talks as data" refers to methods and apparatus for recording and collecting closing talks conducted by employees with customers as audio or text data.

[0789] "Means for converting collected data into text data and cleansing it" refers to data cleansing techniques and methods that convert collected audio data into text data using automatic speech recognition (ASR) technology and remove noise and unnecessary information.

[0790] "Methods for training generative AI using cleansed data" refers to training processes and techniques for training generative AI models using pre-processed text data.

[0791] "Methods for providing real-time customer support using trained AI models" refers to systems or methods that use trained AI models to provide immediate and appropriate responses to customer questions and requests.

[0792] "Means of providing a simulation environment for employees to learn closing techniques" refers to training environments and methods that enable employees to improve their closing techniques through customer interaction simulations.

[0793] "Methods for acquiring emotional data from a user's voice and facial expressions and generating the optimal customer response based on that emotional data" refers to technologies and methods that analyze a user's voice and facial expressions to extract emotional data, and then use that data to generate the optimal response using an AI model.

[0794] "Methods for using AI models trained to handle customer service on behalf of stores that do not yet have a presence" refers to systems or methods that use AI models to automatically handle customer service in areas where physical stores do not yet exist or in stores that lack sufficient staff.

[0795] "A method for an AI crew to respond in real time to questions and responses entered by employees acting as customers" refers to a method or system in which employees take on the role of customers in a training environment, input questions and responses, and an AI model immediately responds to those questions.

[0796] This invention is a "smart on-site assistant" system for improving customer service in physical stores, and is an application using smart glasses. This system collects employee closing talk data and combines it with a generative AI model and an emotion engine.

[0797] Data collection

[0798] The server records the closing conversations that store employees have with customers using the microphone on smart glasses and collects the data as audio. The collected data is later converted into text data.

[0799] Data preprocessing

[0800] The server converts the collected audio data into text using automatic speech recognition (ASR) technology and performs data cleansing to remove noise and irrelevant information. This process eliminates background noise and unnecessary conversation, highlighting important closing remarks.

[0801] AI model training

[0802] The server trains the generative AI using cleansed text data. Based on a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. The trained generative AI model is then used to provide optimal customer service.

[0803] Emotional engine integration

[0804] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. The emotion engine identifies whether the user is happy, confused, or angry through voice analysis and facial expression analysis.

[0805] Real-time response

[0806] Using a trained AI model and emotion engine, the device provides real-time customer support. When a user receives a question from a customer, it sends the question and emotion data to the server. For example, if a customer asks, "Are there any special offers for this product?", and the emotion engine identifies that the customer is interested, this emotion data is also sent.

[0807] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, it might generate a response like, "This product has a limited-time discount this month. Plus, you'll earn points!"

[0808] OJT tool

[0809] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. In a simulated environment where employees act as customers, they input questions and responses in real time through smart glasses, and the AI ​​model responds instantly. This process allows employees to efficiently improve their customer service skills for actual store environments.

[0810] Specific example

[0811] In a real store, salesperson A, wearing smart glasses, is interacting with customer B. When customer B asks, "Are there any special offers for this product?", the smart glasses record the statement and send it to a server. The server converts the audio data into text data, an emotion engine identifies customer B's interests, and a generative AI model generates a response based on that information. The response, "This product has a limited-time discount offer this month. Plus, you'll earn points!", is displayed on the smart glasses or communicated to the customer via voice.

[0812] Example of a prompt

[0813] "customer_question: Are there any special offers for this product? emotion: Interested"

[0814] Output: "This month only, the product offers a discount and the opportunity to earn points."

[0815] Thus, this invention provides a specific system and method for improving the quality of customer service in physical stores and supporting employee training.

[0816] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0817] Step 1:

[0818] Input: Voice and facial expression data of employees and customers at physical stores.

[0819] The server records and videos employee-customer conversations and customer facial expressions in real time via the microphone and camera built into the smart glasses.

[0820] Output: Recorded audio data and customer facial image data

[0821] Step 2:

[0822] Input: Recorded audio data

[0823] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Specifically, it uses Google's automatic speech recognition API to convert audio to text, removing noise and unwanted parts during the process.

[0824] Output: Converted text data

[0825] Step 3:

[0826] Input: Recorded audio data and customer facial image data

[0827] The server analyzes audio data and facial imagery, and uses an emotion engine to identify the customer's emotions. Specifically, it uses the DeepFace library to analyze emotions from the customer's facial expressions and also estimates emotions from their voice.

[0828] Output: Identified sentiment data (e.g., interest, joy, confusion, etc.)

[0829] Step 4:

[0830] Input: Converted text data and identified sentiment data

[0831] The server generates the optimal response using a generative AI model. Specifically, it uses OpenAI's GPT-3 to create prompt sentences based on input text and sentiment data, and then generates an appropriate response accordingly.

[0832] Output: Generated response text

[0833] Step 5:

[0834] Input: Generated response text

[0835] The device delivers the generated response text to the customer. The text is displayed on the smart glasses' screen or delivered to the customer as audio via a text-to-speech function.

[0836] Output: Customer response display or voice response

[0837] Step 6:

[0838] Input: Input of employee questions and responses in the OJT simulation environment.

[0839] In employee training mode, users act as customers and input questions and responses into smart glasses.

[0840] Output: Input data in the simulation environment

[0841] Step 7:

[0842] Input: Input data in the simulation environment

[0843] The server uses an emotion engine and a generative AI model to generate optimal responses to employee questions and reactions, and provides the results as feedback.

[0844] Output: Generated response feedback, analysis results

[0845] This system allows for efficient customer service at physical stores while simultaneously conducting employee training.

[0846] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0847] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0848] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0849] [Third Embodiment]

[0850] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0851] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0852] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0853] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0854] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0855] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0856] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0857] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0858] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0859] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0860] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0861] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0862] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training.

[0863] Program processing

[0864] Data collection

[0865] The server periodically collects audio and text data of employee closing conversations from each store. This includes actual customer interaction audio conversations from the field and successful examples of existing closing conversations. For example, a recording of a "points card explanation" from store A will be collected.

[0866] Data preprocessing

[0867] The collected data is first converted from audio data to text data by the server. Automatic speech recognition (ASR) technology is used for this process. After that, data cleansing is performed to remove noise and unnecessary information. For example, background noise and irrelevant conversations are removed from data such as "Explanation of loyalty cards."

[0868] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add related contextual information. For example, keywords such as "point card," "deals," and "how to use" are extracted, and their contextual information is organized.

[0869] AI model training

[0870] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses. During the training process, the AI's ability to generate appropriate responses improves.

[0871] Real-time response

[0872] Using a trained AI model, the terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to a server via the terminal. The server uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[0873] OJT tool

[0874] Employees can use the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The terminal receives questions from the user and sends them to the server. The server generates responses using a trained AI model and provides them to the user through the terminal. Through this process, employees can learn closing techniques in a situation close to actual customer interaction. For example, if a new employee asks, "How do I use this loyalty card?", the AI ​​crew will respond, "This loyalty card can be used at all stores, and you can easily accumulate points."

[0875] As described above, this system collects data on employees' closing talks, trains a generating AI, and uses the trained AI model to handle customer interactions in real time. It can also be used as an on-the-job training (OJT) tool to support employees in efficiently learning closing techniques.

[0876] The following describes the processing flow.

[0877] Step 1:

[0878] The server receives audio and text data of employee closing speeches collected from each store. This data is uploaded from store microphones and recording devices.

[0879] Step 2:

[0880] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine operates and generates text information from the audio.

[0881] Step 3:

[0882] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts, and prepares unstructured data. This makes the data suitable for analysis.

[0883] Step 4:

[0884] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, words like "bargain," "points card," and "campaign" might be extracted.

[0885] Step 5:

[0886] The server adds contextual information related to the extracted keywords. Specifically, it describes how the keywords are used and in what scenarios they are effective.

[0887] Step 6:

[0888] The server uses the processed text data to train an AI model. This includes a process of training the model with successful examples from the collected conversation data. Machine learning algorithms are used to enable the AI ​​model to learn patterns in customer interactions.

[0889] Step 7:

[0890] The server deploys the AI ​​model after training is complete. This model is located in the cloud or on an edge server. This enables real-time customer support.

[0891] Step 8:

[0892] When a user interacts with a customer in a store, the terminal receives customer questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[0893] Step 9:

[0894] The terminal receives voice input and sends it to the server as text data. Communication takes place in real time over the internet.

[0895] Step 10:

[0896] The server uses a trained AI model to generate an appropriate response based on the received text data. For example, in response to the question "What are this month's special offers?", it will generate the response "This month's special offer is a newly released smartphone."

[0897] Step 11:

[0898] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[0899] Step 12:

[0900] The device converts received text data into speech and provides a response to the customer. This allows customers to obtain information in a natural conversational flow.

[0901] Step 13:

[0902] When a user uses AI Crew as an OJT tool, the terminal is operated in simulation mode. The user takes on the role of a customer and inputs questions and responses.

[0903] Step 14:

[0904] The terminal sends the user's entered questions to the server. The questions arrive at the server in real time.

[0905] Step 15:

[0906] The server uses a trained AI model to generate responses to input questions. For example, in response to the question, "How do I use this loyalty card?", it generates the response, "This loyalty card can be used at all stores, and you can easily accumulate points."

[0907] Step 16:

[0908] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[0909] (Example 1)

[0910] Next, we will describe Example 1. 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."

[0911] Improving the quality of employees' closing pitches and efficiently serving customers is a crucial challenge for many service industries. In particular, in under-established stores or locations with insufficient staff, standardizing closing techniques is difficult, leading to inconsistencies in customer satisfaction. Furthermore, training new employees requires significant time and effort, highlighting the need for effective training methods.

[0912] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0913] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected audio data into text data and cleaning it, means for extracting important keywords and phrases from the cleansed data, means for training a generative AI model using the extracted data, means for providing real-time customer support using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables data collection of closing talks, training of a generative AI model, real-time response, and effective training of employees.

[0914] "Employee closing talk" refers to the final persuasive or explanatory remarks made by an employee when interacting with a customer.

[0915] "Means of collecting data" refers to systems and methods for collecting closing talks in audio or text format from stores and on-site locations.

[0916] "Means of converting collected audio data into text data" refers to the process or technology of converting audio data into text information using speech recognition technology.

[0917] "Methods of cleansing" refer to methods and techniques for removing noise and unnecessary information from converted text data, and organizing and formatting the data.

[0918] "Methods for extracting important keywords and phrases" refers to methods that use natural language processing techniques to extract important words and phrases from text data and add contextual information.

[0919] "Means for training generative AI models" refers to the processes and techniques for training and updating artificial intelligence models using collected and pre-processed data.

[0920] "Methods for providing real-time customer support using trained AI models" refers to systems and methods that use responses generated by generative AI models to immediately address customer questions and requests.

[0921] "Means of providing a simulation environment for employees to learn closing techniques" refers to a training environment in which employees can acquire closing techniques in situations that closely resemble actual customer interactions.

[0922] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI model, and utilizes the trained AI model for customer service and employee training.

[0923] The server periodically collects audio and text data of employee closing speeches from each store. The collected audio data is converted into text data using the Google Speech-to-Text API. For example, a recording of a "loyalty card explanation" from store A might be included.

[0924] Next, the server cleanses the converted text data, removing noise and unnecessary information. The Python pandas library is used for this process. From the cleansed data, important keywords and phrases are extracted using natural language processing (NLP) techniques. For example, the NLP library spaCy is used to extract words such as "point card," "bargain," and "how to use."

[0925] Subsequently, based on the pre-processed data, the server uses a generative AI model such as GPT-3 to train the model on a large amount of closing talk data. Through this training, the generative AI model learns effective closing scenarios and customer responses. For example, by training the AI ​​with a large amount of data that includes "explanation of loyalty cards," the AI ​​learns to understand the optimal way to explain them.

[0926] Using a trained AI model, the terminal can provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server then uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[0927] Furthermore, the system provides a simulation environment for employee training. Users can take on the role of customers and input questions and responses, and the AI ​​crew will respond in real time. In this simulation mode, if a new employee asks, "How do I use this point card?", the AI ​​crew will respond, "This point card can be used at all stores, and you can easily accumulate points."

[0928] Examples of specific cases and prompt statements

[0929] Specific Example 1: Data Collection and Preprocessing

[0930] Audio data of the closing talk "Explanation of the point card" recorded at store A.

[0931] The Google Speech-to-Text API is used to convert audio data into text data. For example, audio data such as "Explanation of the loyalty card" is converted into text data such as "This is an explanation of the loyalty card."

[0932] For data cleansing, we use the Python pandas library to remove unnecessary background noise and irrelevant conversations.

[0933] Example 2: Real-time response

[0934] A customer asks, "What are the best deals this month?"

[0935] The server uses generative AI models such as GPT-3 to generate the optimal response.

[0936] The device responds in real time with, "This month's special offer is a newly released smartphone."

[0937] Example of a prompt:

[0938] Collect audio data of employee closing speeches, convert it to text data, and then de-noise it. Next, use natural language processing to extract key keywords and train an AI model to create a system for real-time customer support and on-the-job training. Please refer to the following example:

[0939] Audio data from store A: "What are this month's recommended products?"

[0940] Text data conversion: "This month's recommended product is the newly released smartphone."

[0941] Keywords extracted: "Recommended products," "New releases," "Smartphones"

[0942] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer support and effective employee training.

[0943] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0944] Step 1: Data Collection

[0945] The server collects audio and text data of employee closing speeches from each store. It receives the input audio data (e.g., "Explanation of loyalty cards") and saves it as text data. Specifically, it uploads data to the server using protocols such as FTP and HTTP. The input is audio data and text data, and the output is the storage of the collected data.

[0946] Step 2: Convert audio data

[0947] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. The Google Speech-to-Text API is used for this conversion. For example, it converts audio data of an employee saying, "This loyalty card can be used at all stores," into text. The input is audio data, and the output is the converted text data.

[0948] Step 3: Data Cleansing

[0949] The server cleanses the converted text data. Here, noise and unwanted elements, such as background noise and irrelevant conversations, are removed. The Python pandas library is used for this process. Speech utterances like "um" and "uh" are removed from the cleansed data. The input is text data, and the output is cleansed text data.

[0950] Step 4: Keyword Extraction

[0951] The server uses natural language processing (NLP) techniques to extract important keywords and phrases from text data. For example, the NLP library spaCy is used to extract words like "point card," "deals," and "how to use." The input is cleansed text data, and the output consists of important keywords and phrases.

[0952] Step 5: Training the AI ​​model

[0953] The server trains a generative AI model based on pre-processed data. It uses a large amount of closing talk data to train generative AI models such as GPT-3, learning closing scenarios and customer responses. For example, by training the AI ​​with a large amount of conversation data that includes "explanation of loyalty cards," it learns the optimal explanation method. The input is data containing extracted keywords, and the output is the trained generative AI model.

[0954] Step 6: Real-time response

[0955] The terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offers are the newly released smartphones." The input is the customer's question text, and the output is the generated optimal response.

[0956] Step 7: On-the-Job Training Tools

[0957] When the user selects employee training mode, the terminal begins a simulated dialogue with an AI crew. The terminal sends the user's question as text data to the server, which uses a trained AI model to generate and return the optimal response. When a new employee asks, "Are there any special offers?", the AI ​​crew responds, "We're having a double points campaign this month." The input is the user's question, and the output is the AI ​​crew's response.

[0958] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer service and effective employee training.

[0959] (Application Example 1)

[0960] Next, we will explain Application Example 1. In the following explanation, 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."

[0961] The problems this invention aims to solve are to provide a means for effectively collecting employees' closing talks and training them in closing techniques, thereby improving the quality of customer service. Furthermore, it aims to improve the efficiency of store operations by enabling employees to provide real-time customer service using smart devices. Additionally, it aims to support employee skill development by providing a simulated training environment.

[0962] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0963] This invention includes a server comprising means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generative AI using the cleansed data, means for providing real-time customer service on smart devices using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables improvement of employees' closing techniques and the quality of customer service. Furthermore, since AI-based customer service can be provided even in stores where it has not yet been implemented, store operations can be made more efficient. In addition, the employee training environment is enhanced, and practical skills can be improved.

[0964] "Employee closing talk" refers to the statements and responses made by employees at the end of a business negotiation with a customer.

[0965] "Means of data collection" refers to systems and methods for recording employees' closing talks as audio or text data and later storing them in a database.

[0966] "Means of converting to text data and cleansing it" refers to the process and technology of converting collected audio data into text using automatic speech recognition (ASR) technology and removing noise and unwanted information from that text.

[0967] "Means of training generative AI" refers to the processes and techniques used to train generative AI models using cleansed data.

[0968] A "trained AI model" refers to an AI model obtained as a result of training a generative AI using collected and pre-processed data.

[0969] A "smart device" refers to a portable device with computer capabilities, such as a smartphone, smart glasses, or head-mounted display.

[0970] "Means of providing real-time customer support" refers to systems and methods that use trained AI models to instantly generate responses to customer inquiries.

[0971] "Means of providing a simulation environment" refers to a training environment in which employees can experience actual work in a simulated manner, and in this context, it includes methods of learning closing techniques through interaction with an AI model.

[0972] "Unopened stores" refer to stores that have not yet opened for business, or stores that cannot immediately be staffed with employees due to reasons such as labor shortages.

[0973] "Means of acting as a proxy for customer service" refers to systems and methods in stores that use AI models to handle customer inquiries on behalf of employees.

[0974] "A means by which an AI crew responds in real time to questions and responses entered by employees acting as customers in a simulation environment" refers to a training method in which employees play the role of customers and ask questions to the AI, to which the AI ​​responds immediately.

[0975] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training. A specific embodiment for carrying out this invention will be described below.

[0976] Hardware to use

[0977] 1. Server: Cloud server or on-premises server

[0978] 2. Smart devices: Smartphones, smart glasses (e.g., Google Glass), head-mounted displays

[0979] 3. Recording equipment: Microphones worn by employees or portable recorders.

[0980] Software to use

[0981] 1. Automatic Speech Recognition (ASR) Software: Google Cloud Speech-to-Text

[0982] 2. Natural Language Processing (NLP) Libraries: spaCy, BERT

[0983] 3. AI Generative Models: GPT-4, PaLM

[0984] Data collection

[0985] When employees conduct closing speeches, audio data is collected using smart glasses or portable recorders. The collected audio data is periodically uploaded to a server via the smart devices.

[0986] Data preprocessing

[0987] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Then, natural language processing (NLP) technology is used to extract important keywords from the text data and cleanse it of unnecessary noise and irrelevant information.

[0988] AI model training

[0989] The server trains a generative AI model based on pre-processed data. By using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses.

[0990] Real-time response

[0991] When a user receives a question from a customer, that question is sent to a server as text data via their smart device. The server uses a trained AI model to generate the optimal response and sends it back to the user's smart device in real time.

[0992] Specific example

[0993] When a customer asks, "What are this month's recommended products?", the user's smart glasses display the response, "This month's recommended product is the newly released XYZ smartphone. It features a high-performance camera and long battery life."

[0994] OJT tool

[0995] Users can utilize the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The user's questions and responses are sent to the server via a smart device, and the server uses a trained AI model to generate appropriate responses, which are then returned to the user.

[0996] Example of a prompt

[0997] User question: "How do I use this points card?"

[0998] AI Crew's response: "This point card can be used at all stores and you can easily accumulate points."

[0999] In this way, users can learn closing techniques in an environment that closely resembles actual business scenarios.

[1000] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1001] Step 1: Data Collection

[1002] Subject: terminal

[1003] Specific operation: Employees wear smart glasses or portable recorders to record audio data in real time while interacting with customers.

[1004] Input: Audio data of employee closing remarks

[1005] Output: Recorded audio data file

[1006] Data processing / calculations: None

[1007] Step 2: Sending the audio data

[1008] Subject: terminal

[1009] Specific operation: Periodically upload recorded audio data to the server.

[1010] Input: Audio data file

[1011] Output: Audio data sent to the server

[1012] Data processing / calculations: None

[1013] Step 3: Automatic Speech Recognition (ASR)

[1014] Subject: Server

[1015] Specific operation: The server converts the received audio data into text data using automatic speech recognition (ASR) software.

[1016] Input: Audio data

[1017] Output: Text data

[1018] Data processing and calculation: Converting audio data to text data.

[1019] Step 4: Data Cleansing

[1020] Subject: Server

[1021] Specific operation: Performs data cleansing to remove noise and irrelevant information from the converted text data.

[1022] Input: Text data

[1023] Output: Cleansed text data

[1024] Data processing and calculation: Removal of unnecessary information, noise filtering.

[1025] Step 5: Natural Language Processing (NLP)

[1026] Subject: Server

[1027] Specific operation: Natural language processing (NLP) techniques are used on the cleansed text data to extract important keywords and phrases.

[1028] Input: Cleansed text data

[1029] Output: List of keywords and phrases

[1030] Data processing and calculation: Keyword extraction, contextual understanding

[1031] Step 6: AI Model Training

[1032] Subject: Server

[1033] Specific operation: Train a generative AI model based on pre-processed data.

[1034] Input: List of keywords or phrases, text data

[1035] Output: Trained AI model

[1036] Data processing and computation: AI model training, parameter optimization

[1037] Step 7: Real-time response

[1038] Subject: Terminals and servers

[1039] Specific operation: When a user receives a question from a customer, the question is sent to the server as text data via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal.

[1040] Input: Customer questions, text data

[1041] Output: Response text

[1042] Data processing and calculation: Question analysis, response generation

[1043] Step 8: Display the response

[1044] Subject: terminal

[1045] Specific operation: The returned response text is displayed on the screen of smart glasses or other smart devices.

[1046] Input: Response text

[1047] Output: Response displayed on the smart device screen

[1048] Data processing and calculations: Adjustment of display format

[1049] Step 9: Use as an OJT tool

[1050] Subject: Terminals and servers

[1051] Specific operation: The user uses simulation mode to interact with an AI crew for training purposes. The user's questions and responses are sent to the server via the terminal, and the server uses the trained AI model to generate appropriate responses, which are then sent back to the terminal.

[1052] Input: User's question, response

[1053] Output: AI Crew Response Text

[1054] Data processing and calculation: Question analysis, response generation

[1055] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1056] This invention is a system that collects employee closing talks as data and uses that data to train an AI. Furthermore, this system incorporates an emotion engine that recognizes user emotions to provide more effective customer service and employee training.

[1057] Program processing

[1058] Data collection

[1059] The server receives audio and text data of employee closing speeches collected from stores. Data is uploaded from microphones and recording devices installed in the stores. For example, a recording of a "points card explanation" from store A is collected.

[1060] Data preprocessing

[1061] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. Then, it performs data cleansing to remove noise and unnecessary information. For example, background noise and unnecessary dialogue are removed from data such as "points card explanation."

[1062] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add relevant contextual information. For example, keywords such as "loyalty card," "deals," and "campaign" are extracted, and their contextual information is organized.

[1063] AI model training

[1064] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. For example, it is trained to provide optimal responses to specific customer reactions by learning closing talks from art exhibitions.

[1065] Emotional engine integration

[1066] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. Through voice analysis and facial expression analysis, the emotion engine identifies whether the user is happy, confused, or angry.

[1067] Real-time response

[1068] Using trained AI models and an emotion engine, the terminal provides real-time customer support in undeployed stores and understaffed locations. When a user receives a question from a customer, it sends emotion data along with the question to the server. For example, if a customer asks, "What are this month's special offers?", and the emotion engine determines that the customer is excited, that emotion data is also sent to the server.

[1069] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, if the sentiment engine detects customer excitement, it might generate a response such as, "This month's special is the newly released smartphone! There are lots of bonuses too!"

[1070] The generated response is sent back to the terminal, which converts the text data into speech and provides the response to the customer. This allows customers to receive a more personalized response.

[1071] OJT tool

[1072] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. For example, if a new employee asks, "How do I use this point card?", the terminal sends the question and the employee's emotion data to the server.

[1073] The server uses an emotion engine to analyze employees' emotions and provide feedback along with the most appropriate response. For example, if an employee is confused, it might offer specific advice such as, "This loyalty card can be used at all stores and you can easily accumulate points. Let me show you how to register."

[1074] In this way, the system collects employee closing talk data and uses a generative AI and emotion engine for training, providing real-time customer support and efficient employee training.

[1075] The following describes the processing flow.

[1076] Step 1:

[1077] The server receives audio and text data of employee closing speeches from each store. For example, a recording of a "points card explanation" from store A is uploaded to the server.

[1078] Step 2:

[1079] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine then runs and generates text information from the audio.

[1080] Step 3:

[1081] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts to create well-structured data. For example, background noise and unnecessary conversations are removed from the "points card explanation" data.

[1082] Step 4:

[1083] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, it might extract keywords such as "deals," "points card," and "campaign."

[1084] Step 5:

[1085] The server adds contextual information to the extracted keywords, describing how the keywords are used and in what context they are effective. For example, it might add a typical dialogue scenario when describing a "loyalty card."

[1086] Step 6:

[1087] The server uses the processed text data to train the generative AI. Using a large amount of closing talk data, the AI ​​model learns patterns in customer interaction. For example, it learns dialogue patterns for explaining new products.

[1088] Step 7:

[1089] The server deploys the completed AI model and emotion engine, enabling real-time customer support. The AI ​​model and emotion engine reside on either the cloud or edge servers.

[1090] Step 8:

[1091] When a user interacts with a customer in a store, the terminal receives the customer's questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[1092] Step 9:

[1093] The terminal receives voice input and sends it to the server as text data. Information is transferred in real time via the internet.

[1094] Step 10:

[1095] The server uses a trained AI model to generate an appropriate response based on the received text data and sentiment data. For example, if the sentiment engine determines that the customer is excited along with the question, "What are this month's specials?", it will generate a response that takes that sentiment data into account.

[1096] Step 11:

[1097] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[1098] Step 12:

[1099] The device converts received text data into speech and provides a response to the customer. This allows customers to receive a more personalized service.

[1100] Step 13:

[1101] When a user uses AI Crew as an OJT tool, the terminal operates in simulation mode. The user takes on the role of a customer and inputs questions and responses into the terminal.

[1102] Step 14:

[1103] The device sends user-entered questions to the server in real time. Along with the questions, user sentiment data is also transmitted.

[1104] Step 15:

[1105] The server uses an emotion engine to analyze the user's emotions and generate the most appropriate response. For example, if the user asks, "How do I use this loyalty card?" and the server determines that the user is confused, it will respond, "This loyalty card can be used at all stores, and you can easily accumulate points. Let me show you how to use it."

[1106] Step 16:

[1107] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[1108] (Example 2)

[1109] Next, we will describe Example 2. 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."

[1110] There is a need for a system that efficiently collects data on employees' closing pitches, uses this data to train AI models for real-time customer service, and improves employees' closing skills. In particular, a system that can analyze user emotions in real time and provide appropriate responses will contribute to improving customer satisfaction. Furthermore, a system that can effectively operate in stores where the system is not yet implemented or in locations with insufficient staff is also necessary.

[1111] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[1112] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data using speech recognition technology and performing data cleansing, means for extracting keywords and related information from the cleansed text data using natural language processing technology, means for training a generative AI model using the extracted data, means for analyzing user emotions using an emotion analysis engine, means for providing real-time customer support using the trained AI model and emotion analysis engine, and means for providing a simulation environment for employees to learn closing techniques. This makes it possible to effectively collect and analyze employee closing talks and to realize real-time support and employee training using an AI model.

[1113] An "employee" is a person who is employed by a company or organization and engages in work.

[1114] "Closing talk" refers to the dialogue and explanations used to encourage customers to close a deal or sign a contract during the final stages of sales or service provision.

[1115] "Data" refers to a collection of information or numerical data that is collected and stored for a specific purpose or analysis.

[1116] "Speech recognition technology" is a technology that converts human speech into text data using a computer.

[1117] "Text data" refers to information stored in the form of characters or sentences.

[1118] "Data cleansing" is the process of removing noise and unnecessary information from data to improve its consistency and quality.

[1119] "Natural language processing technology" refers to the technology that enables computers to understand and analyze human natural language.

[1120] A "keyword" is a word or phrase that has particularly important meaning within a document or dialogue.

[1121] A "generative AI model" is an artificial intelligence model generated through machine learning or deep learning based on a large amount of data.

[1122] An "emotion analysis engine" is software or a system that identifies and analyzes a user's emotional state from factors such as voice and facial expressions.

[1123] "Real-time" refers to a state where processing and responses occur instantaneously without delay.

[1124] A "simulation environment" refers to a system or software used to virtually reproduce real-world situations.

[1125] A "customer" is a person or organization that purchases or uses a product or service.

[1126] An "unopened store" refers to a store that has not yet officially started operations.

[1127] This invention relates to a system that collects employee closing talks, uses that data to train a generative AI model, and further integrates an emotion analysis engine to provide real-time customer service and support employee training.

[1128] System Configuration

[1129] Data collection

[1130] The server receives audio data of employees' closing speeches from microphones and recording devices installed in each store. For example, audio data of a "points card explanation" given at store A will be collected. Text data can also be received from store logs and messaging systems.

[1131] Data preprocessing

[1132] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, it uses the Google Cloud Speech-to-Text API. After that, it performs data cleansing to remove noise and unnecessary information. For example, it filters out background noise and casual conversation using libraries such as pandas or NLTK. Next, the server uses natural language processing (NLP) technology to extract important keywords and phrases from the text data and organize contextual information. Specific technologies used include spaCy and the BERT model.

[1133] AI model training

[1134] The server trains a generative AI model based on pre-processed data. AI models used include OpenAI GPT-3 and Google BERT. For example, the AI ​​model is trained based on scenarios such as "closing speech at an art exhibition" or "closing speech at a mobile phone shop."

[1135] Emotional engine integration

[1136] The server uses an emotion analysis engine to collect and analyze emotional data from the user's voice and facial expressions. Software used includes the Microsoft Azure Emotion API and the Affectiva SDK. For example, it can identify emotions such as "confusion" or "excitement" from the user's voice data.

[1137] Real-time response

[1138] The terminal uses a trained AI model and sentiment analysis engine to provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, it sends the question and sentiment data to the server. The server uses the received data to generate the optimal response using a trained AI model, and the generated response is sent back to the terminal. The terminal converts the text data into speech and provides the response to the customer, for example, using Google Cloud Text-to-Speech.

[1139] OJT tool

[1140] Users undergo training using an OJT tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. The server analyzes these questions and employee emotion data, providing feedback along with the most appropriate response. For example, it generates an appropriate response to the question, "How do I use this loyalty card?"

[1141] Specific examples and prompt statements

[1142] For example, the server trains an AI model using closing talk data collected at store A. During this process, an emotion analysis engine is integrated to analyze the user's emotions and generate the optimal response for the customer. In real-time responses, when a customer asks, "Tell me about the new service," the emotion analysis engine senses the customer's interest, and the server generates a response such as, "The new service has the following features..."

[1143] Examples of prompt statements to input into the generative AI model are as follows:

[1144] "Based on training data for closing pitches, generate responses to the customer's question, 'What are the best deals this month?'"

[1145] "Use the emotion engine to provide the best possible response when the user is confused."

[1146] This system effectively collects and analyzes employees' closing pitches, and uses AI models to enable real-time responses and employee training.

[1147] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1148] Step 1:

[1149] Data collection

[1150] The server receives audio data from microphones and recording devices installed in the store. This audio data consists of closing conversations between employees and customers; for example, "audio data of an explanation about the loyalty card at store A" might be input. The server saves this audio data.

[1151] Input: Audio data collected from stores

[1152] Output: Audio data stored on the server

[1153] Specifically, the server periodically runs a script to collect recording data from each store and uploads the data to a central server.

[1154] Step 2:

[1155] Convert to text data

[1156] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, the Google Cloud Speech-to-Text API is used for this conversion. For instance, the text output might read, "Explanation of the benefits of the loyalty card."

[1157] Input: Received audio data

[1158] Output: Text data converted using ASR technology

[1159] Specifically, the server calls the ASR API and executes a script that converts the audio data into text data.

[1160] Step 3:

[1161] Data cleansing

[1162] The server performs data cleansing on the converted text data to remove noise and unnecessary information. For example, it uses libraries such as pandas or NLTK to filter out background noise and casual conversation.

[1163] Input: Text data

[1164] Output: Cleansed text data with noise and unwanted information removed.

[1165] Specifically, the server analyzes the text data and removes unnecessary parts using regular expressions and filtering algorithms.

[1166] Step 4:

[1167] Extraction of keywords and contextual information

[1168] The server applies natural language processing (NLP) techniques to the cleansed text data to extract important keywords and phrases. For example, it uses spaCy or BERT models to extract keywords such as "point card," "deals," and "campaign," and then organizes their contextual information.

[1169] Input: Cleansed text data

[1170] Output: Extracted keywords and contextual information

[1171] In terms of specific operations, the server uses an NLP library to identify specific patterns and keywords from text data and extract relevant contextual data.

[1172] Step 5:

[1173] AI model training

[1174] The server trains a generative AI model based on pre-processed data. Using a large amount of closing talk data, for example, OpenAI GPT-3 or Google BERT, the AI ​​model learns appropriate customer interaction patterns.

[1175] Input: Preprocessed text data

[1176] Output: Trained generative AI model

[1177] In terms of specific operations, the server feeds the dataset into an AI training framework (such as TensorFlow or PyTorch) and performs model training.

[1178] Step 6:

[1179] Emotion analysis

[1180] The server uses an emotion analysis engine to acquire and analyze emotional data from the user's voice and facial expressions. For example, it uses the Microsoft Azure Emotion API or the Affectiva SDK to identify the user's emotional state (joy, confusion, anger, etc.).

[1181] Input: User voice and facial expression data

[1182] Output: Analyzed user sentiment data

[1183] Specifically, the server calls an emotion analysis API to analyze the collected audio data and obtain emotion data.

[1184] Step 7:

[1185] Real-time response

[1186] When the device receives a question from a customer, it sends the question and sentiment data to the server. The server uses this data to generate the optimal response using a trained AI model, and the generated response is sent back to the device. The device then converts the text data into speech using, for example, Google Cloud Text-to-Speech, and provides the response to the customer.

[1187] Input: Customer questions and sentiment data

[1188] Output: Optimal response text and speech-converted response

[1189] In terms of specific operations, the terminal receives user input, provides an interface for sending it to the server, and plays back the response received from the server as audio.

[1190] Step 8:

[1191] Training using OJT tools

[1192] The user conducts employee training using an OJT tool. The user acts as a customer, inputting questions and responses into a terminal. The server analyzes the questions and employee sentiment data to provide optimal responses and feedback.

[1193] Input: Questions, responses, and sentiment data entered by employees.

[1194] Output: Optimal response and feedback

[1195] Specifically, the terminal receives questions and responses entered by employees and sends them to the server. The server uses sentiment analysis and an AI model to generate the optimal response and feedback, and sends it back to the terminal.

[1196] (Application Example 2)

[1197] Next, we will explain application example 2. In the following explanation, 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."

[1198] Traditional customer service systems lacked sufficient training methods to improve the quality of employee closing pitches, making efficient real-time customer service difficult. Furthermore, limited means of recognizing customer emotions and providing appropriate responses posed a risk of decreased customer satisfaction. Additionally, achieving high-quality customer service was challenging in stores where the system was not yet implemented or in locations with insufficient staff.

[1199] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[1200] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generating AI using the cleansed data, means for acquiring emotional data from the user's voice and facial expressions and generating the optimal customer response based on that emotional data, means for providing customer support in real time using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables employee training and high-quality customer support in stores where the system has not yet been deployed.

[1201] "Means for collecting employee closing talks as data" refers to methods and apparatus for recording and collecting closing talks conducted by employees with customers as audio or text data.

[1202] "Means for converting collected data into text data and cleansing it" refers to data cleansing techniques and methods that convert collected audio data into text data using automatic speech recognition (ASR) technology and remove noise and unnecessary information.

[1203] "Methods for training generative AI using cleansed data" refers to training processes and techniques for training generative AI models using pre-processed text data.

[1204] "Methods for providing real-time customer support using trained AI models" refers to systems or methods that use trained AI models to provide immediate and appropriate responses to customer questions and requests.

[1205] "Means of providing a simulation environment for employees to learn closing techniques" refers to training environments and methods that enable employees to improve their closing techniques through customer interaction simulations.

[1206] "Methods for acquiring emotional data from a user's voice and facial expressions and generating the optimal customer response based on that emotional data" refers to technologies and methods that analyze a user's voice and facial expressions to extract emotional data, and then use that data to generate the optimal response using an AI model.

[1207] "Methods for using AI models trained to handle customer service on behalf of stores that do not yet have a presence" refers to systems or methods that use AI models to automatically handle customer service in areas where physical stores do not yet exist or in stores that lack sufficient staff.

[1208] "A method for an AI crew to respond in real time to questions and responses entered by employees acting as customers" refers to a method or system in which employees take on the role of customers in a training environment, input questions and responses, and an AI model immediately responds to those questions.

[1209] This invention is a "smart on-site assistant" system for improving customer service in physical stores, and is an application using smart glasses. This system collects employee closing talk data and combines it with a generative AI model and an emotion engine.

[1210] Data collection

[1211] The server records the closing conversations that store employees have with customers using the microphone on smart glasses and collects the data as audio. The collected data is later converted into text data.

[1212] Data preprocessing

[1213] The server converts the collected audio data into text using automatic speech recognition (ASR) technology and performs data cleansing to remove noise and irrelevant information. This process eliminates background noise and unnecessary conversation, highlighting important closing remarks.

[1214] AI model training

[1215] The server trains the generative AI using cleansed text data. Based on a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. The trained generative AI model is then used to provide optimal customer service.

[1216] Emotional engine integration

[1217] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. The emotion engine identifies whether the user is happy, confused, or angry through voice analysis and facial expression analysis.

[1218] Real-time response

[1219] Using a trained AI model and emotion engine, the device provides real-time customer support. When a user receives a question from a customer, it sends the question and emotion data to the server. For example, if a customer asks, "Are there any special offers for this product?", and the emotion engine identifies that the customer is interested, this emotion data is also sent.

[1220] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, it might generate a response like, "This product has a limited-time discount this month. Plus, you'll earn points!"

[1221] OJT tool

[1222] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. In a simulated environment where employees act as customers, they input questions and responses in real time through smart glasses, and the AI ​​model responds instantly. This process allows employees to efficiently improve their customer service skills for actual store environments.

[1223] Specific example

[1224] In a real store, salesperson A, wearing smart glasses, is interacting with customer B. When customer B asks, "Are there any special offers for this product?", the smart glasses record the statement and send it to a server. The server converts the audio data into text data, an emotion engine identifies customer B's interests, and a generative AI model generates a response based on that information. The response, "This product has a limited-time discount offer this month. Plus, you'll earn points!", is displayed on the smart glasses or communicated to the customer via voice.

[1225] Example of a prompt

[1226] "customer_question: Are there any special offers for this product? emotion: Interested"

[1227] Output: "This month only, the product offers a discount and the opportunity to earn points."

[1228] Thus, this invention provides a specific system and method for improving the quality of customer service in physical stores and supporting employee training.

[1229] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1230] Step 1:

[1231] Input: Voice and facial expression data of employees and customers at physical stores.

[1232] The server records and videos employee-customer conversations and customer facial expressions in real time via the microphone and camera built into the smart glasses.

[1233] Output: Recorded audio data and customer facial image data

[1234] Step 2:

[1235] Input: Recorded audio data

[1236] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Specifically, it uses Google's automatic speech recognition API to convert audio to text, removing noise and unwanted parts during the process.

[1237] Output: Converted text data

[1238] Step 3:

[1239] Input: Recorded audio data and customer facial image data

[1240] The server analyzes audio data and facial imagery, and uses an emotion engine to identify the customer's emotions. Specifically, it uses the DeepFace library to analyze emotions from the customer's facial expressions and also estimates emotions from their voice.

[1241] Output: Identified sentiment data (e.g., interest, joy, confusion, etc.)

[1242] Step 4:

[1243] Input: Converted text data and identified sentiment data

[1244] The server generates the optimal response using a generative AI model. Specifically, it uses OpenAI's GPT-3 to create prompt sentences based on input text and sentiment data, and then generates an appropriate response accordingly.

[1245] Output: Generated response text

[1246] Step 5:

[1247] Input: Generated response text

[1248] The device delivers the generated response text to the customer. The text is displayed on the smart glasses' screen or delivered to the customer as audio via a text-to-speech function.

[1249] Output: Customer response display or voice response

[1250] Step 6:

[1251] Input: Input of employee questions and responses in the OJT simulation environment.

[1252] In employee training mode, users act as customers and input questions and responses into smart glasses.

[1253] Output: Input data in the simulation environment

[1254] Step 7:

[1255] Input: Input data in the simulation environment

[1256] The server uses an emotion engine and a generative AI model to generate optimal responses to employee questions and reactions, and provides the results as feedback.

[1257] Output: Generated response feedback, analysis results

[1258] This system allows for efficient customer service at physical stores while simultaneously conducting employee training.

[1259] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1260] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1261] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[1262] [Fourth Embodiment]

[1263] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[1264] As shown in Figure 7, the 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.

[1265] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1266] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[1267] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[1268] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[1269] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[1270] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive 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 robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[1271] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[1272] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[1273] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1274] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[1275] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1276] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training.

[1277] Program processing

[1278] Data collection

[1279] The server periodically collects audio and text data of employee closing conversations from each store. This includes actual customer interaction audio conversations from the field and successful examples of existing closing conversations. For example, a recording of a "points card explanation" from store A will be collected.

[1280] Data preprocessing

[1281] The collected data is first converted from audio data to text data by the server. Automatic speech recognition (ASR) technology is used for this process. After that, data cleansing is performed to remove noise and unnecessary information. For example, background noise and irrelevant conversations are removed from data such as "Explanation of loyalty cards."

[1282] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add related contextual information. For example, keywords such as "point card," "deals," and "how to use" are extracted, and their contextual information is organized.

[1283] AI model training

[1284] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses. During the training process, the AI's ability to generate appropriate responses improves.

[1285] Real-time response

[1286] Using a trained AI model, the terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to a server via the terminal. The server uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[1287] OJT tool

[1288] Employees can use the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The terminal receives questions from the user and sends them to the server. The server generates responses using a trained AI model and provides them to the user through the terminal. Through this process, employees can learn closing techniques in a situation close to actual customer interaction. For example, if a new employee asks, "How do I use this loyalty card?", the AI ​​crew will respond, "This loyalty card can be used at all stores, and you can easily accumulate points."

[1289] As described above, this system collects data on employees' closing talks, trains a generating AI, and uses the trained AI model to handle customer interactions in real time. It can also be used as an on-the-job training (OJT) tool to support employees in efficiently learning closing techniques.

[1290] The following describes the processing flow.

[1291] Step 1:

[1292] The server receives audio and text data of employee closing speeches collected from each store. This data is uploaded from store microphones and recording devices.

[1293] Step 2:

[1294] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine operates and generates text information from the audio.

[1295] Step 3:

[1296] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts, and prepares unstructured data. This makes the data suitable for analysis.

[1297] Step 4:

[1298] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, words like "bargain," "points card," and "campaign" might be extracted.

[1299] Step 5:

[1300] The server adds contextual information related to the extracted keywords. Specifically, it describes how the keywords are used and in what scenarios they are effective.

[1301] Step 6:

[1302] The server uses the processed text data to train an AI model. This includes a process of training the model with successful examples from the collected conversation data. Machine learning algorithms are used to enable the AI ​​model to learn patterns in customer interactions.

[1303] Step 7:

[1304] The server deploys the AI ​​model after training is complete. This model is located in the cloud or on an edge server. This enables real-time customer support.

[1305] Step 8:

[1306] When a user interacts with a customer in a store, the terminal receives customer questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[1307] Step 9:

[1308] The terminal receives voice input and sends it to the server as text data. Communication takes place in real time over the internet.

[1309] Step 10:

[1310] The server uses a trained AI model to generate an appropriate response based on the received text data. For example, in response to the question "What are this month's special offers?", it will generate the response "This month's special offer is a newly released smartphone."

[1311] Step 11:

[1312] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[1313] Step 12:

[1314] The device converts received text data into speech and provides a response to the customer. This allows customers to obtain information in a natural conversational flow.

[1315] Step 13:

[1316] When a user uses AI Crew as an OJT tool, the terminal is operated in simulation mode. The user takes on the role of a customer and inputs questions and responses.

[1317] Step 14:

[1318] The terminal sends the user's entered questions to the server. The questions arrive at the server in real time.

[1319] Step 15:

[1320] The server uses a trained AI model to generate responses to input questions. For example, in response to the question, "How do I use this loyalty card?", it generates the response, "This loyalty card can be used at all stores, and you can easily accumulate points."

[1321] Step 16:

[1322] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[1323] (Example 1)

[1324] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1325] Improving the quality of employees' closing pitches and efficiently serving customers is a crucial challenge for many service industries. In particular, in under-established stores or locations with insufficient staff, standardizing closing techniques is difficult, leading to inconsistencies in customer satisfaction. Furthermore, training new employees requires significant time and effort, highlighting the need for effective training methods.

[1326] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[1327] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected audio data into text data and cleaning it, means for extracting important keywords and phrases from the cleansed data, means for training a generative AI model using the extracted data, means for providing real-time customer support using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables data collection of closing talks, training of a generative AI model, real-time response, and effective training of employees.

[1328] "Employee closing talk" refers to the final persuasive or explanatory remarks made by an employee when interacting with a customer.

[1329] "Means of collecting data" refers to systems and methods for collecting closing talks in audio or text format from stores and on-site locations.

[1330] "Means of converting collected audio data into text data" refers to the process or technology of converting audio data into text information using speech recognition technology.

[1331] "Methods of cleansing" refer to methods and techniques for removing noise and unnecessary information from converted text data, and organizing and formatting the data.

[1332] "Methods for extracting important keywords and phrases" refers to methods that use natural language processing techniques to extract important words and phrases from text data and add contextual information.

[1333] "Means for training generative AI models" refers to the processes and techniques for training and updating artificial intelligence models using collected and pre-processed data.

[1334] "Methods for providing real-time customer support using trained AI models" refers to systems and methods that use responses generated by generative AI models to immediately address customer questions and requests.

[1335] "Means of providing a simulation environment for employees to learn closing techniques" refers to a training environment in which employees can acquire closing techniques in situations that closely resemble actual customer interactions.

[1336] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI model, and utilizes the trained AI model for customer service and employee training.

[1337] The server periodically collects audio and text data of employee closing speeches from each store. The collected audio data is converted into text data using the Google Speech-to-Text API. For example, a recording of a "loyalty card explanation" from store A might be included.

[1338] Next, the server cleanses the converted text data, removing noise and unnecessary information. The Python pandas library is used for this process. From the cleansed data, important keywords and phrases are extracted using natural language processing (NLP) techniques. For example, the NLP library spaCy is used to extract words such as "point card," "bargain," and "how to use."

[1339] Subsequently, based on the pre-processed data, the server uses a generative AI model such as GPT-3 to train the model on a large amount of closing talk data. Through this training, the generative AI model learns effective closing scenarios and customer responses. For example, by training the AI ​​with a large amount of data that includes "explanation of loyalty cards," the AI ​​learns to understand the optimal way to explain them.

[1340] Using a trained AI model, the terminal can provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server then uses the trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offer is the newly released smartphone."

[1341] Furthermore, the system provides a simulation environment for employee training. Users can take on the role of customers and input questions and responses, and the AI ​​crew will respond in real time. In this simulation mode, if a new employee asks, "How do I use this point card?", the AI ​​crew will respond, "This point card can be used at all stores, and you can easily accumulate points."

[1342] Examples of specific cases and prompt statements

[1343] Specific Example 1: Data Collection and Preprocessing

[1344] Audio data of the closing talk "Explanation of the point card" recorded at store A.

[1345] The Google Speech-to-Text API is used to convert audio data into text data. For example, audio data such as "Explanation of the loyalty card" is converted into text data such as "This is an explanation of the loyalty card."

[1346] For data cleansing, we use the Python pandas library to remove unnecessary background noise and irrelevant conversations.

[1347] Example 2: Real-time response

[1348] A customer asks, "What are the best deals this month?"

[1349] The server uses generative AI models such as GPT-3 to generate the optimal response.

[1350] The device responds in real time with, "This month's special offer is a newly released smartphone."

[1351] Example of a prompt:

[1352] Collect audio data of employee closing speeches, convert it to text data, and then de-noise it. Next, use natural language processing to extract key keywords and train an AI model to create a system for real-time customer support and on-the-job training. Please refer to the following example:

[1353] Audio data from store A: "What are this month's recommended products?"

[1354] Text data conversion: "This month's recommended product is the newly released smartphone."

[1355] Keywords extracted: "Recommended products," "New releases," "Smartphones"

[1356] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer support and effective employee training.

[1357] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1358] Step 1: Data Collection

[1359] The server collects audio and text data of employee closing speeches from each store. It receives the input audio data (e.g., "Explanation of loyalty cards") and saves it as text data. Specifically, it uploads data to the server using protocols such as FTP and HTTP. The input is audio data and text data, and the output is the storage of the collected data.

[1360] Step 2: Convert audio data

[1361] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. The Google Speech-to-Text API is used for this conversion. For example, it converts audio data of an employee saying, "This loyalty card can be used at all stores," into text. The input is audio data, and the output is the converted text data.

[1362] Step 3: Data Cleansing

[1363] The server cleanses the converted text data. Here, noise and unwanted elements, such as background noise and irrelevant conversations, are removed. The Python pandas library is used for this process. Speech utterances like "um" and "uh" are removed from the cleansed data. The input is text data, and the output is cleansed text data.

[1364] Step 4: Keyword Extraction

[1365] The server uses natural language processing (NLP) techniques to extract important keywords and phrases from text data. For example, the NLP library spaCy is used to extract words like "point card," "deals," and "how to use." The input is cleansed text data, and the output consists of important keywords and phrases.

[1366] Step 5: Training the AI ​​model

[1367] The server trains a generative AI model based on pre-processed data. It uses a large amount of closing talk data to train generative AI models such as GPT-3, learning closing scenarios and customer responses. For example, by training the AI ​​with a large amount of conversation data that includes "explanation of loyalty cards," it learns the optimal explanation method. The input is data containing extracted keywords, and the output is the trained generative AI model.

[1368] Step 6: Real-time response

[1369] The terminal provides real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, the question is sent as text data to the server via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal. For example, if a customer asks, "What are this month's special offers?", the terminal will respond in real time with, "This month's special offers are the newly released smartphones." The input is the customer's question text, and the output is the generated optimal response.

[1370] Step 7: On-the-Job Training Tools

[1371] When the user selects employee training mode, the terminal begins a simulated dialogue with an AI crew. The terminal sends the user's question as text data to the server, which uses a trained AI model to generate and return the optimal response. When a new employee asks, "Are there any special offers?", the AI ​​crew responds, "We're having a double points campaign this month." The input is the user's question, and the output is the AI ​​crew's response.

[1372] In this way, the system collects employee closing talks as data, trains a generative AI model, and uses the trained AI model to enable real-time customer service and effective employee training.

[1373] (Application Example 1)

[1374] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1375] The problems this invention aims to solve are to provide a means for effectively collecting employees' closing talks and training them in closing techniques, thereby improving the quality of customer service. Furthermore, it aims to improve the efficiency of store operations by enabling employees to provide real-time customer service using smart devices. Additionally, it aims to support employee skill development by providing a simulated training environment.

[1376] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[1377] This invention includes a server comprising means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generative AI using the cleansed data, means for providing real-time customer service on smart devices using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables improvement of employees' closing techniques and the quality of customer service. Furthermore, since AI-based customer service can be provided even in stores where it has not yet been implemented, store operations can be made more efficient. In addition, the employee training environment is enhanced, and practical skills can be improved.

[1378] "Employee closing talk" refers to the statements and responses made by employees at the end of a business negotiation with a customer.

[1379] "Means of data collection" refers to systems and methods for recording employees' closing talks as audio or text data and later storing them in a database.

[1380] "Means of converting to text data and cleansing it" refers to the process and technology of converting collected audio data into text using automatic speech recognition (ASR) technology and removing noise and unwanted information from that text.

[1381] "Means of training generative AI" refers to the processes and techniques used to train generative AI models using cleansed data.

[1382] A "trained AI model" refers to an AI model obtained as a result of training a generative AI using collected and pre-processed data.

[1383] A "smart device" refers to a portable device with computer capabilities, such as a smartphone, smart glasses, or head-mounted display.

[1384] "Means of providing real-time customer support" refers to systems and methods that use trained AI models to instantly generate responses to customer inquiries.

[1385] "Means of providing a simulation environment" refers to a training environment in which employees can experience actual work in a simulated manner, and in this context, it includes methods of learning closing techniques through interaction with an AI model.

[1386] "Unopened stores" refer to stores that have not yet opened for business, or stores that cannot immediately be staffed with employees due to reasons such as labor shortages.

[1387] "Means of acting as a proxy for customer service" refers to systems and methods in stores that use AI models to handle customer inquiries on behalf of employees.

[1388] "A means by which an AI crew responds in real time to questions and responses entered by employees acting as customers in a simulation environment" refers to a training method in which employees play the role of customers and ask questions to the AI, to which the AI ​​responds immediately.

[1389] This invention is a system that collects employee closing talks as data, uses that data to train a generative AI, and utilizes the trained AI model for customer service and employee training. A specific embodiment for carrying out this invention will be described below.

[1390] Hardware to use

[1391] 1. Server: Cloud server or on-premises server

[1392] 2. Smart devices: Smartphones, smart glasses (e.g., Google Glass), head-mounted displays

[1393] 3. Recording equipment: Microphones worn by employees or portable recorders.

[1394] Software to use

[1395] 1. Automatic Speech Recognition (ASR) Software: Google Cloud Speech-to-Text

[1396] 2. Natural Language Processing (NLP) Libraries: spaCy, BERT

[1397] 3. AI Generative Models: GPT-4, PaLM

[1398] Data collection

[1399] When employees conduct closing speeches, audio data is collected using smart glasses or portable recorders. The collected audio data is periodically uploaded to a server via the smart devices.

[1400] Data preprocessing

[1401] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Then, natural language processing (NLP) technology is used to extract important keywords from the text data and cleanse it of unnecessary noise and irrelevant information.

[1402] AI model training

[1403] The server trains a generative AI model based on pre-processed data. By using a large amount of closing talk data, the AI ​​model learns closing scenarios and customer responses.

[1404] Real-time response

[1405] When a user receives a question from a customer, that question is sent to a server as text data via their smart device. The server uses a trained AI model to generate the optimal response and sends it back to the user's smart device in real time.

[1406] Specific example

[1407] When a customer asks, "What are this month's recommended products?", the user's smart glasses display the response, "This month's recommended product is the newly released XYZ smartphone. It features a high-performance camera and long battery life."

[1408] OJT tool

[1409] Users can utilize the AI ​​crew as an on-the-job training (OJT) tool. In this simulation mode, the user acts as a customer and interacts with the AI ​​crew. The user's questions and responses are sent to the server via a smart device, and the server uses a trained AI model to generate appropriate responses, which are then returned to the user.

[1410] Example of a prompt

[1411] User question: "How do I use this points card?"

[1412] AI Crew's response: "This point card can be used at all stores and you can easily accumulate points."

[1413] In this way, users can learn closing techniques in an environment that closely resembles actual business scenarios.

[1414] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1415] Step 1: Data Collection

[1416] Subject: terminal

[1417] Specific operation: Employees wear smart glasses or portable recorders to record audio data in real time while interacting with customers.

[1418] Input: Audio data of employee closing remarks

[1419] Output: Recorded audio data file

[1420] Data processing / calculations: None

[1421] Step 2: Sending the audio data

[1422] Subject: terminal

[1423] Specific operation: Periodically upload recorded audio data to the server.

[1424] Input: Audio data file

[1425] Output: Audio data sent to the server

[1426] Data processing / calculations: None

[1427] Step 3: Automatic Speech Recognition (ASR)

[1428] Subject: Server

[1429] Specific operation: The server converts the received audio data into text data using automatic speech recognition (ASR) software.

[1430] Input: Audio data

[1431] Output: Text data

[1432] Data processing and calculation: Converting audio data to text data.

[1433] Step 4: Data Cleansing

[1434] Subject: Server

[1435] Specific operation: Performs data cleansing to remove noise and irrelevant information from the converted text data.

[1436] Input: Text data

[1437] Output: Cleansed text data

[1438] Data processing and calculation: Removal of unnecessary information, noise filtering.

[1439] Step 5: Natural Language Processing (NLP)

[1440] Subject: Server

[1441] Specific operation: Natural language processing (NLP) techniques are used on the cleansed text data to extract important keywords and phrases.

[1442] Input: Cleansed text data

[1443] Output: List of keywords and phrases

[1444] Data processing and calculation: Keyword extraction, contextual understanding

[1445] Step 6: AI Model Training

[1446] Subject: Server

[1447] Specific operation: Train a generative AI model based on pre-processed data.

[1448] Input: List of keywords or phrases, text data

[1449] Output: Trained AI model

[1450] Data processing and computation: AI model training, parameter optimization

[1451] Step 7: Real-time response

[1452] Subject: Terminals and servers

[1453] Specific operation: When a user receives a question from a customer, the question is sent to the server as text data via the terminal. The server uses a trained AI model to generate the optimal response and sends it back to the terminal.

[1454] Input: Customer questions, text data

[1455] Output: Response text

[1456] Data processing and calculation: Question analysis, response generation

[1457] Step 8: Display the response

[1458] Subject: terminal

[1459] Specific operation: The returned response text is displayed on the screen of smart glasses or other smart devices.

[1460] Input: Response text

[1461] Output: Response displayed on the smart device screen

[1462] Data processing and calculations: Adjustment of display format

[1463] Step 9: Use as an OJT tool

[1464] Subject: Terminals and servers

[1465] Specific operation: The user uses simulation mode to interact with an AI crew for training purposes. The user's questions and responses are sent to the server via the terminal, and the server uses the trained AI model to generate appropriate responses, which are then sent back to the terminal.

[1466] Input: User's question, response

[1467] Output: AI Crew Response Text

[1468] Data processing and calculation: Question analysis, response generation

[1469] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1470] This invention is a system that collects employee closing talks as data and uses that data to train an AI. Furthermore, this system incorporates an emotion engine that recognizes user emotions to provide more effective customer service and employee training.

[1471] Program processing

[1472] Data collection

[1473] The server receives audio and text data of employee closing speeches collected from stores. Data is uploaded from microphones and recording devices installed in the stores. For example, a recording of a "points card explanation" from store A is collected.

[1474] Data preprocessing

[1475] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. Then, it performs data cleansing to remove noise and unnecessary information. For example, background noise and unnecessary dialogue are removed from data such as "points card explanation."

[1476] Next, the server uses natural language processing (NLP) techniques to extract important keywords and phrases from the text data and add relevant contextual information. For example, keywords such as "loyalty card," "deals," and "campaign" are extracted, and their contextual information is organized.

[1477] AI model training

[1478] Based on the pre-processed data, the server trains the generative AI. Using a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. For example, it is trained to provide optimal responses to specific customer reactions by learning closing talks from art exhibitions.

[1479] Emotional engine integration

[1480] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. Through voice analysis and facial expression analysis, the emotion engine identifies whether the user is happy, confused, or angry.

[1481] Real-time response

[1482] Using trained AI models and an emotion engine, the terminal provides real-time customer support in undeployed stores and understaffed locations. When a user receives a question from a customer, it sends emotion data along with the question to the server. For example, if a customer asks, "What are this month's special offers?", and the emotion engine determines that the customer is excited, that emotion data is also sent to the server.

[1483] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, if the sentiment engine detects customer excitement, it might generate a response such as, "This month's special is the newly released smartphone! There are lots of bonuses too!"

[1484] The generated response is sent back to the terminal, which converts the text data into speech and provides the response to the customer. This allows customers to receive a more personalized response.

[1485] OJT tool

[1486] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. For example, if a new employee asks, "How do I use this point card?", the terminal sends the question and the employee's emotion data to the server.

[1487] The server uses an emotion engine to analyze employees' emotions and provide feedback along with the most appropriate response. For example, if an employee is confused, it might offer specific advice such as, "This loyalty card can be used at all stores and you can easily accumulate points. Let me show you how to register."

[1488] In this way, the system collects employee closing talk data and uses a generative AI and emotion engine for training, providing real-time customer support and efficient employee training.

[1489] The following describes the processing flow.

[1490] Step 1:

[1491] The server receives audio and text data of employee closing speeches from each store. For example, a recording of a "points card explanation" from store A is uploaded to the server.

[1492] Step 2:

[1493] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. The speech recognition engine then runs and generates text information from the audio.

[1494] Step 3:

[1495] The server performs a cleansing process on the text data. Specifically, it removes noise and unnecessary parts to create well-structured data. For example, background noise and unnecessary conversations are removed from the "points card explanation" data.

[1496] Step 4:

[1497] The server uses natural language processing (NLP) techniques to extract keywords and important phrases from the cleansed text data. For example, it might extract keywords such as "deals," "points card," and "campaign."

[1498] Step 5:

[1499] The server adds contextual information to the extracted keywords, describing how the keywords are used and in what context they are effective. For example, it might add a typical dialogue scenario when describing a "loyalty card."

[1500] Step 6:

[1501] The server uses the processed text data to train the generative AI. Using a large amount of closing talk data, the AI ​​model learns patterns in customer interaction. For example, it learns dialogue patterns for explaining new products.

[1502] Step 7:

[1503] The server deploys the completed AI model and emotion engine, enabling real-time customer support. The AI ​​model and emotion engine reside on either the cloud or edge servers.

[1504] Step 8:

[1505] When a user interacts with a customer in a store, the terminal receives the customer's questions as voice input. For example, a tablet or smartphone records the customer's statements through a voice recognition module.

[1506] Step 9:

[1507] The terminal receives voice input and sends it to the server as text data. Information is transferred in real time via the internet.

[1508] Step 10:

[1509] The server uses a trained AI model to generate an appropriate response based on the received text data and sentiment data. For example, if the sentiment engine determines that the customer is excited along with the question, "What are this month's specials?", it will generate a response that takes that sentiment data into account.

[1510] Step 11:

[1511] The server sends the generated response back to the terminal. It is transmitted in real time as text data.

[1512] Step 12:

[1513] The device converts received text data into speech and provides a response to the customer. This allows customers to receive a more personalized service.

[1514] Step 13:

[1515] When a user uses AI Crew as an OJT tool, the terminal operates in simulation mode. The user takes on the role of a customer and inputs questions and responses into the terminal.

[1516] Step 14:

[1517] The device sends user-entered questions to the server in real time. Along with the questions, user sentiment data is also transmitted.

[1518] Step 15:

[1519] The server uses an emotion engine to analyze the user's emotions and generate the most appropriate response. For example, if the user asks, "How do I use this loyalty card?" and the server determines that the user is confused, it will respond, "This loyalty card can be used at all stores, and you can easily accumulate points. Let me show you how to use it."

[1520] Step 16:

[1521] The device converts the generated responses into audio and provides them to the user. This allows the user to receive realistic training from a customer's perspective.

[1522] (Example 2)

[1523] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1524] There is a need for a system that efficiently collects data on employees' closing pitches, uses this data to train AI models for real-time customer service, and improves employees' closing skills. In particular, a system that can analyze user emotions in real time and provide appropriate responses will contribute to improving customer satisfaction. Furthermore, a system that can effectively operate in stores where the system is not yet implemented or in locations with insufficient staff is also necessary.

[1525] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[1526] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data using speech recognition technology and performing data cleansing, means for extracting keywords and related information from the cleansed text data using natural language processing technology, means for training a generative AI model using the extracted data, means for analyzing user emotions using an emotion analysis engine, means for providing real-time customer support using the trained AI model and emotion analysis engine, and means for providing a simulation environment for employees to learn closing techniques. This makes it possible to effectively collect and analyze employee closing talks and to realize real-time support and employee training using an AI model.

[1527] An "employee" is a person who is employed by a company or organization and engages in work.

[1528] "Closing talk" refers to the dialogue and explanations used to encourage customers to close a deal or sign a contract during the final stages of sales or service provision.

[1529] "Data" refers to a collection of information or numerical data that is collected and stored for a specific purpose or analysis.

[1530] "Speech recognition technology" is a technology that converts human speech into text data using a computer.

[1531] "Text data" refers to information stored in the form of characters or sentences.

[1532] "Data cleansing" is the process of removing noise and unnecessary information from data to improve its consistency and quality.

[1533] "Natural language processing technology" refers to the technology that enables computers to understand and analyze human natural language.

[1534] A "keyword" is a word or phrase that has particularly important meaning within a document or dialogue.

[1535] A "generative AI model" is an artificial intelligence model generated through machine learning or deep learning based on a large amount of data.

[1536] An "emotion analysis engine" is software or a system that identifies and analyzes a user's emotional state from factors such as voice and facial expressions.

[1537] "Real-time" refers to a state where processing and responses occur instantaneously without delay.

[1538] A "simulation environment" refers to a system or software used to virtually reproduce real-world situations.

[1539] A "customer" is a person or organization that purchases or uses a product or service.

[1540] An "unopened store" refers to a store that has not yet officially started operations.

[1541] This invention relates to a system that collects employee closing talks, uses that data to train a generative AI model, and further integrates an emotion analysis engine to provide real-time customer service and support employee training.

[1542] System Configuration

[1543] Data collection

[1544] The server receives audio data of employees' closing speeches from microphones and recording devices installed in each store. For example, audio data of a "points card explanation" given at store A will be collected. Text data can also be received from store logs and messaging systems.

[1545] Data preprocessing

[1546] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, it uses the Google Cloud Speech-to-Text API. After that, it performs data cleansing to remove noise and unnecessary information. For example, it filters out background noise and casual conversation using libraries such as pandas or NLTK. Next, the server uses natural language processing (NLP) technology to extract important keywords and phrases from the text data and organize contextual information. Specific technologies used include spaCy and the BERT model.

[1547] AI model training

[1548] The server trains a generative AI model based on pre-processed data. AI models used include OpenAI GPT-3 and Google BERT. For example, the AI ​​model is trained based on scenarios such as "closing speech at an art exhibition" or "closing speech at a mobile phone shop."

[1549] Emotional engine integration

[1550] The server uses an emotion analysis engine to collect and analyze emotional data from the user's voice and facial expressions. Software used includes the Microsoft Azure Emotion API and the Affectiva SDK. For example, it can identify emotions such as "confusion" or "excitement" from the user's voice data.

[1551] Real-time response

[1552] The terminal uses a trained AI model and sentiment analysis engine to provide real-time customer support in undeployed stores or locations with insufficient staff. When a user receives a question from a customer, it sends the question and sentiment data to the server. The server uses the received data to generate the optimal response using a trained AI model, and the generated response is sent back to the terminal. The terminal converts the text data into speech and provides the response to the customer, for example, using Google Cloud Text-to-Speech.

[1553] OJT tool

[1554] Users undergo training using an OJT tool equipped with an AI crew and an emotion engine. Users act as customers, inputting questions and responses into a terminal. The server analyzes these questions and employee emotion data, providing feedback along with the most appropriate response. For example, it generates an appropriate response to the question, "How do I use this loyalty card?"

[1555] Specific examples and prompt statements

[1556] For example, the server trains an AI model using closing talk data collected at store A. During this process, an emotion analysis engine is integrated to analyze the user's emotions and generate the optimal response for the customer. In real-time responses, when a customer asks, "Tell me about the new service," the emotion analysis engine senses the customer's interest, and the server generates a response such as, "The new service has the following features..."

[1557] Examples of prompt statements to input into the generative AI model are as follows:

[1558] "Based on training data for closing pitches, generate responses to the customer's question, 'What are the best deals this month?'"

[1559] "Use the emotion engine to provide the best possible response when the user is confused."

[1560] This system effectively collects and analyzes employees' closing pitches, and uses AI models to enable real-time responses and employee training.

[1561] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1562] Step 1:

[1563] Data collection

[1564] The server receives audio data from microphones and recording devices installed in the store. This audio data consists of closing conversations between employees and customers; for example, "audio data of an explanation about the loyalty card at store A" might be input. The server saves this audio data.

[1565] Input: Audio data collected from stores

[1566] Output: Audio data stored on the server

[1567] Specifically, the server periodically runs a script to collect recording data from each store and uploads the data to a central server.

[1568] Step 2:

[1569] Convert to text data

[1570] The server converts the received audio data into text data using automatic speech recognition (ASR) technology. For example, the Google Cloud Speech-to-Text API is used for this conversion. For instance, the text output might read, "Explanation of the benefits of the loyalty card."

[1571] Input: Received audio data

[1572] Output: Text data converted using ASR technology

[1573] Specifically, the server calls the ASR API and executes a script that converts the audio data into text data.

[1574] Step 3:

[1575] Data cleansing

[1576] The server performs data cleansing on the converted text data to remove noise and unnecessary information. For example, it uses libraries such as pandas or NLTK to filter out background noise and casual conversation.

[1577] Input: Text data

[1578] Output: Cleansed text data with noise and unwanted information removed.

[1579] Specifically, the server analyzes the text data and removes unnecessary parts using regular expressions and filtering algorithms.

[1580] Step 4:

[1581] Extraction of keywords and contextual information

[1582] The server applies natural language processing (NLP) techniques to the cleansed text data to extract important keywords and phrases. For example, it uses spaCy or BERT models to extract keywords such as "point card," "deals," and "campaign," and then organizes their contextual information.

[1583] Input: Cleansed text data

[1584] Output: Extracted keywords and contextual information

[1585] In terms of specific operations, the server uses an NLP library to identify specific patterns and keywords from text data and extract relevant contextual data.

[1586] Step 5:

[1587] AI model training

[1588] The server trains a generative AI model based on pre-processed data. Using a large amount of closing talk data, for example, OpenAI GPT-3 or Google BERT, the AI ​​model learns appropriate customer interaction patterns.

[1589] Input: Preprocessed text data

[1590] Output: Trained generative AI model

[1591] In terms of specific operations, the server feeds the dataset into an AI training framework (such as TensorFlow or PyTorch) and performs model training.

[1592] Step 6:

[1593] Emotion analysis

[1594] The server uses an emotion analysis engine to acquire and analyze emotional data from the user's voice and facial expressions. For example, it uses the Microsoft Azure Emotion API or the Affectiva SDK to identify the user's emotional state (joy, confusion, anger, etc.).

[1595] Input: User voice and facial expression data

[1596] Output: Analyzed user sentiment data

[1597] Specifically, the server calls an emotion analysis API to analyze the collected audio data and obtain emotion data.

[1598] Step 7:

[1599] Real-time response

[1600] When the device receives a question from a customer, it sends the question and sentiment data to the server. The server uses this data to generate the optimal response using a trained AI model, and the generated response is sent back to the device. The device then converts the text data into speech using, for example, Google Cloud Text-to-Speech, and provides the response to the customer.

[1601] Input: Customer questions and sentiment data

[1602] Output: Optimal response text and speech-converted response

[1603] In terms of specific operations, the terminal receives user input, provides an interface for sending it to the server, and plays back the response received from the server as audio.

[1604] Step 8:

[1605] Training using OJT tools

[1606] The user conducts employee training using an OJT tool. The user acts as a customer, inputting questions and responses into a terminal. The server analyzes the questions and employee sentiment data to provide optimal responses and feedback.

[1607] Input: Questions, responses, and sentiment data entered by employees.

[1608] Output: Optimal response and feedback

[1609] Specifically, the terminal receives questions and responses entered by employees and sends them to the server. The server uses sentiment analysis and an AI model to generate the optimal response and feedback, and sends it back to the terminal.

[1610] (Application Example 2)

[1611] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1612] Traditional customer service systems lacked sufficient training methods to improve the quality of employee closing pitches, making efficient real-time customer service difficult. Furthermore, limited means of recognizing customer emotions and providing appropriate responses posed a risk of decreased customer satisfaction. Additionally, achieving high-quality customer service was challenging in stores where the system was not yet implemented or in locations with insufficient staff.

[1613] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[1614] In this invention, the server includes means for collecting employee closing talks as data, means for converting the collected data into text data and cleaning it, means for training a generating AI using the cleansed data, means for acquiring emotional data from the user's voice and facial expressions and generating the optimal customer response based on that emotional data, means for providing customer support in real time using the trained AI model, and means for providing a simulation environment for employees to learn closing techniques. This enables employee training and high-quality customer support in stores where the system has not yet been deployed.

[1615] "Means for collecting employee closing talks as data" refers to methods and apparatus for recording and collecting closing talks conducted by employees with customers as audio or text data.

[1616] "Means for converting collected data into text data and cleansing it" refers to data cleansing techniques and methods that convert collected audio data into text data using automatic speech recognition (ASR) technology and remove noise and unnecessary information.

[1617] "Methods for training generative AI using cleansed data" refers to training processes and techniques for training generative AI models using pre-processed text data.

[1618] "Methods for providing real-time customer support using trained AI models" refers to systems or methods that use trained AI models to provide immediate and appropriate responses to customer questions and requests.

[1619] "Means of providing a simulation environment for employees to learn closing techniques" refers to training environments and methods that enable employees to improve their closing techniques through customer interaction simulations.

[1620] "Methods for acquiring emotional data from a user's voice and facial expressions and generating the optimal customer response based on that emotional data" refers to technologies and methods that analyze a user's voice and facial expressions to extract emotional data, and then use that data to generate the optimal response using an AI model.

[1621] "Methods for using AI models trained to handle customer service on behalf of stores that do not yet have a presence" refers to systems or methods that use AI models to automatically handle customer service in areas where physical stores do not yet exist or in stores that lack sufficient staff.

[1622] "A method for an AI crew to respond in real time to questions and responses entered by employees acting as customers" refers to a method or system in which employees take on the role of customers in a training environment, input questions and responses, and an AI model immediately responds to those questions.

[1623] This invention is a "smart on-site assistant" system for improving customer service in physical stores, and is an application using smart glasses. This system collects employee closing talk data and combines it with a generative AI model and an emotion engine.

[1624] Data collection

[1625] The server records the closing conversations that store employees have with customers using the microphone on smart glasses and collects the data as audio. The collected data is later converted into text data.

[1626] Data preprocessing

[1627] The server converts the collected audio data into text using automatic speech recognition (ASR) technology and performs data cleansing to remove noise and irrelevant information. This process eliminates background noise and unnecessary conversation, highlighting important closing remarks.

[1628] AI model training

[1629] The server trains the generative AI using cleansed text data. Based on a large amount of closing talk data, the AI ​​model learns appropriate customer interaction patterns. The trained generative AI model is then used to provide optimal customer service.

[1630] Emotional engine integration

[1631] The server uses an emotion engine to acquire emotional data from the user's voice and facial expressions. The emotion engine identifies whether the user is happy, confused, or angry through voice analysis and facial expression analysis.

[1632] Real-time response

[1633] Using a trained AI model and emotion engine, the device provides real-time customer support. When a user receives a question from a customer, it sends the question and emotion data to the server. For example, if a customer asks, "Are there any special offers for this product?", and the emotion engine identifies that the customer is interested, this emotion data is also sent.

[1634] The server uses text and sentiment data to generate the optimal response using a trained AI model. For example, it might generate a response like, "This product has a limited-time discount this month. Plus, you'll earn points!"

[1635] OJT tool

[1636] Employees undergo training using an on-the-job training (OJT) tool equipped with an AI crew and an emotion engine. In a simulated environment where employees act as customers, they input questions and responses in real time through smart glasses, and the AI ​​model responds instantly. This process allows employees to efficiently improve their customer service skills for actual store environments.

[1637] Specific example

[1638] In a real store, salesperson A, wearing smart glasses, is interacting with customer B. When customer B asks, "Are there any special offers for this product?", the smart glasses record the statement and send it to a server. The server converts the audio data into text data, an emotion engine identifies customer B's interests, and a generative AI model generates a response based on that information. The response, "This product has a limited-time discount offer this month. Plus, you'll earn points!", is displayed on the smart glasses or communicated to the customer via voice.

[1639] Example of a prompt

[1640] "customer_question: Are there any special offers for this product? emotion: Interested"

[1641] Output: "This month only, the product offers a discount and the opportunity to earn points."

[1642] Thus, this invention provides a specific system and method for improving the quality of customer service in physical stores and supporting employee training.

[1643] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1644] Step 1:

[1645] Input: Voice and facial expression data of employees and customers at physical stores.

[1646] The server records and videos employee-customer conversations and customer facial expressions in real time via the microphone and camera built into the smart glasses.

[1647] Output: Recorded audio data and customer facial image data

[1648] Step 2:

[1649] Input: Recorded audio data

[1650] The server converts the collected audio data into text data using automatic speech recognition (ASR) technology. Specifically, it uses Google's automatic speech recognition API to convert audio to text, removing noise and unwanted parts during the process.

[1651] Output: Converted text data

[1652] Step 3:

[1653] Input: Recorded audio data and customer facial image data

[1654] The server analyzes audio data and facial imagery, and uses an emotion engine to identify the customer's emotions. Specifically, it uses the DeepFace library to analyze emotions from the customer's facial expressions and also estimates emotions from their voice.

[1655] Output: Identified sentiment data (e.g., interest, joy, confusion, etc.)

[1656] Step 4:

[1657] Input: Converted text data and identified sentiment data

[1658] The server generates the optimal response using a generative AI model. Specifically, it uses OpenAI's GPT-3 to create prompt sentences based on input text and sentiment data, and then generates an appropriate response accordingly.

[1659] Output: Generated response text

[1660] Step 5:

[1661] Input: Generated response text

[1662] The device delivers the generated response text to the customer. The text is displayed on the smart glasses' screen or delivered to the customer as audio via a text-to-speech function.

[1663] Output: Customer response display or voice response

[1664] Step 6:

[1665] Input: Input of employee questions and responses in the OJT simulation environment.

[1666] In employee training mode, users act as customers and input questions and responses into smart glasses.

[1667] Output: Input data in the simulation environment

[1668] Step 7:

[1669] Input: Input data in the simulation environment

[1670] The server uses an emotion engine and a generative AI model to generate optimal responses to employee questions and reactions, and provides the results as feedback.

[1671] Output: Generated response feedback, analysis results

[1672] This system allows for efficient customer service at physical stores while simultaneously conducting employee training.

[1673] 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 controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1674] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1675] 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 this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[1676] Furthermore, the emotion identification model 59, acting 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 a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1677] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[1678] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[1679] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[1680] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[1681] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is 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 the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[1682] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[1683] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[1684] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[1685] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[1686] 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.

[1687] Furthermore, it is not necessary to store the entirety 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 the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[1688] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[1689] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of 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). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[1690] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[1691] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[1692] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[1693] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[1694] The following is further disclosed regarding the embodiments described above.

[1695] (Claim 1)

[1696] A means of collecting employee closing talks as data,

[1697] A method for converting collected data into text data and cleansing it,

[1698] A method for training a generative AI using cleansed data,

[1699] A means of providing real-time customer support using trained AI models,

[1700] A means of providing employees with a simulation environment to learn closing techniques,

[1701] A system that includes this.

[1702] (Claim 2)

[1703] The system according to claim 1, comprising means of using an AI model trained to perform customer service on behalf of customers in stores that have not yet been opened.

[1704] (Claim 3)

[1705] The system according to claim 1, which includes a means for an AI crew to respond in real time to questions and responses entered by an employee acting as a customer, as a simulation environment for employee training.

[1706] "Example 1"

[1707] (Claim 1)

[1708] A means of collecting employee closing talks as data,

[1709] A method for converting collected audio data into text data and then cleansing it,

[1710] A method for extracting important keywords and phrases from cleansed data,

[1711] A method for training a generative AI model using extracted data,

[1712] A means of providing real-time customer support using trained AI models,

[1713] A means of providing employees with a simulation environment to learn closing techniques,

[1714] A system that includes this.

[1715] (Claim 2)

[1716] The system according to claim 1, comprising means of using an AI model trained to perform customer service on behalf of customers in stores that have not yet been opened.

[1717] (Claim 3)

[1718] The system according to claim 1, comprising means for a generating AI model to respond in real time to questions and responses entered by employees acting as customers, as a simulation environment for employee training.

[1719] "Application Example 1"

[1720] (Claim 1)

[1721] A means of collecting employee closing talks as data,

[1722] A method for converting collected data into text data and cleansing it,

[1723] A method for training a generative AI using cleansed data,

[1724] A means of providing real-time customer support on smart devices using trained AI models,

[1725] A means of providing employees with a simulation environment to learn closing techniques,

[1726] A system that includes this.

[1727] (Claim 2)

[1728] The system according to claim 1, comprising means of using an AI model trained to perform customer service on behalf of customers in stores that have not yet been opened.

[1729] (Claim 3)

[1730] The system according to claim 1, comprising means for an AI crew to respond in real time to questions and responses entered by an employee acting as a customer in a simulation environment for employee training.

[1731] "Example 2 of combining an emotion engine"

[1732] (Claim 1)

[1733] A means of collecting employee closing talks as data,

[1734] A method for converting collected data into text data using speech recognition technology and performing data cleansing,

[1735] A method for extracting keywords and related information from cleansed text data using natural language processing techniques,

[1736] A method for training a generative AI model using extracted data,

[1737] A means of analyzing a user's emotions using an emotion analysis engine,

[1738] A means of providing real-time customer support using trained AI models and sentiment analysis engines,

[1739] A means of providing employees with a simulation environment to learn closing techniques,

[1740] A system that includes this.

[1741] (Claim 2)

[1742] The system according to claim 1, comprising means of utilizing an AI model and sentiment analysis engine trained to perform customer service on behalf of customers in stores that have not yet been opened.

[1743] (Claim 3)

[1744] The system according to claim 1, comprising means for an AI model and an emotion analysis engine to respond in real time to questions and responses entered by an employee acting as a customer, as a simulation environment for employee training.

[1745] "Application example 2 when combining with an emotional engine"

[1746] (Claim 1)

[1747] A means of collecting employee closing talks as data,

[1748] A method for converting collected data into text data and cleansing it,

[1749] A method for training a generative AI using cleansed data,

[1750] A means of providing real-time customer support using trained AI models,

[1751] A means of providing employees with a simulation environment to learn closing techniques,

[1752] A means of acquiring emotional data from the user's voice and facial expressions, and generating the optimal customer response based on that emotional data,

[1753] A system that includes this.

[1754] (Claim 2)

[1755] The system according to claim 1, comprising means of using an AI model trained to perform customer service on behalf of customers in stores that have not yet been opened.

[1756] (Claim 3)

[1757] The system according to claim 1, which includes a means for an AI crew to respond in real time to questions and responses entered by an employee acting as a customer, as a simulation environment for employee training. [Explanation of Symbols]

[1758] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting employee closing talks as data, A method for converting collected data into text data and cleansing it, A method for training a generative AI using cleansed data, A means of providing real-time customer support using trained AI models, A means of providing employees with a simulation environment to learn closing techniques, A system that includes this.

2. The system according to claim 1, which includes means of using an AI model trained to perform customer service on behalf of customers in stores that have not yet been opened.

3. The system according to claim 1, which includes a means for an AI crew to respond in real time to questions and responses entered by an employee acting as a customer, as a simulation environment for employee training.

Images