System

Abstract

A system is provided.SOLUTION: A system, comprising: means for initializing and loading a generative AI model; means for receiving a conversational request from a user; means for analyzing the received conversational request; means for generating a response based on the analysis; means for transmitting the generated response to the user; and means for storing the conversational history in a database for use in continuous learning.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] In the nursing care industry, there is a need to solve the problem of caregivers being overwhelmed with miscellaneous tasks and not having enough time to have heartfelt conversations with users, as well as the problem of users feeling lonely and lacking time to talk to someone. This will eliminate the lack of communication within nursing care facilities and provide an environment that provides psychological support to users. [Means for solving the problem]

[0005] The present invention is a system that includes a means for initializing and loading a generative AI model, a means for receiving a conversation request from a user, a natural language understanding means for analyzing the received conversation request, a natural language generation means for generating a response based on the analysis result, a means for sending the generated response to the user, and a means for storing the conversation history in a database and using it for continuous learning. It also includes a means for displaying a user interface, receiving messages from the user, and displaying responses via a terminal that communicates with a server. This configuration reduces the burden on caregivers and enables continuous and personalized conversations to be provided to users.

[0006] A "generative AI model" is an artificial intelligence model that generates natural language based on user input.

[0007] The "initialization and loading means" is a means for configuring the generative AI model and loading the necessary data into memory in order to make it usable.

[0008] A "conversation request" refers to a communication content such as a message or question that a user sends to the system.

[0009] The "receiving means" is a means by which the system receives a conversation request from a user through a terminal or a user interface.

[0010] "Natural language understanding means" refers to means for analyzing the content, sentiment, and context of a received conversation request.

[0011] A "natural language generation means" is a means for generating an appropriate response based on the analyzed information.

[0012] The "transmission means" is a means for transmitting the generated response to the user through a user interface or terminal.

[0013] "Conversation history" is a record of past conversations between a user and the system.

[0014] A "database" is a system for storing conversation histories and other related information and making them accessible as needed.

[0015] "Continuous learning means" refers to the learning process used to improve the performance of the generative AI model based on stored conversation history and data.

[0016] A "user interface" is a means for providing a screen or input area for a user to interact with a system.

[0017] "Terminal" refers to a device through which a user accesses and interacts with the system.

[0018] A "server" is a central processing unit that executes key components such as generative AI models, analytical means, and generation means, and provides services. [Brief explanation of the drawings]

[0019] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0020] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0021] First, the terms used in the following description will be explained.

[0022] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0023] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0024] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0025] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0026] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0027] [First embodiment]

[0028] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0029] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0030] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0031] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0032] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0033] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0034] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0035] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0036] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0037] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0038] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0039] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0040] The present invention is a service system dedicated to conversation using a generative AI model, and specific embodiments are described below.

[0041] Server Action:

[0042] 1. Initializing and loading the model

[0043] The server initializes the generative AI model at system startup and loads the necessary data and settings into memory, making the AI ​​ready to converse with the user.

[0044] Example: A server loads a pre-trained generative AI model (e.g., GPT-4).

[0045] 2. Request acceptance

[0046] The server receives a conversation request sent from the terminal, which includes a message or question entered by the user.

[0047] Example: The server receives user input saying "The weather is nice today."

[0048] 3. Natural language understanding

[0049] The server passes the received message to a natural language understanding module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[0050] Example: The server reads positive sentiment from the message "The weather is nice."

[0051] 4. Generating the Response

[0052] The server uses a natural language generation module to generate appropriate responses based on the analyzed information. The responses are formulated in natural language, enabling smooth conversation with the user.

[0053] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[0054] 5. Sending the Response

[0055] The server generates a response and sends it back to the terminal, where it is displayed to the user and the dialogue proceeds.

[0056] Example: The server generates a response and sends it to the device.

[0057] 6. Conversation history storage and continuous learning

[0058] The server stores the conversation history in a database and uses it to continuously train the AI ​​model, allowing it to learn about the user's characteristics and preferences over time.

[0059] Example: The server stores the conversation history in a database to help with the next interaction.

[0060] Terminal handling:

[0061] 1. Displaying the user interface

[0062] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0063] Example: A tablet device displays a chat app.

[0064] 2. Getting and Sending User Input

[0065] The terminal receives input from the user (such as a message or question) and sends the content to the server.

[0066] Example: A user types "What day is it today?" and the device sends the message to the server.

[0067] 3. View the response

[0068] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0069] Example: The device displays the response from the server: "Today is World Environment Day."

[0070] User Action:

[0071] 1. Start a conversation

[0072] Users initiate conversations through the device interface and can start conversations at any time.

[0073] Example: A user types "Good morning" into a terminal.

[0074] 2. Enter your message

[0075] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[0076] Example: A user types "What's in the news today?" and presses the send button.

[0077] 3. Check the response and enter the next message

[0078] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[0079] Example: User continues with "Thank you, is there anything else?"

[0080] As described above, by combining a generative AI model with multiple processing means, the present invention enables natural and continuous conversation, thereby reducing the burden on caregivers and eliminating feelings of loneliness among users.

[0081] The processing flow will be explained below.

[0082] Server processing steps:

[0083] Step 1:

[0084] The server initializes the generative AI model and loads the necessary data and settings into memory, so the AI ​​is ready to converse with the user.

[0085] Step 2:

[0086] The server receives a conversation request from the device, which includes a message or question typed by the user.

[0087] Step 3:

[0088] The server passes the received message to a natural language understanding (NLU) module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[0089] Step 4:

[0090] Based on the analysis results, the server uses a natural language generation (NLG) module to generate an appropriate response, which is formulated in natural language.

[0091] Step 5:

[0092] The server then sends the generated response to the terminal, allowing the conversation to proceed.

[0093] Step 6:

[0094] The server stores the conversation history in a database and uses it to continuously train the generative AI model, allowing the model to learn about the user's characteristics and preferences over time.

[0095] Terminal processing steps:

[0096] Step 1:

[0097] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0098] Step 2:

[0099] The device receives input from the user (such as a message or question) and sends it to the server, which then forwards the user's message to the server.

[0100] Step 3:

[0101] The terminal receives the response sent by the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0102] User steps:

[0103] Step 1:

[0104] Users initiate conversations through the device interface, and conversations can begin at any time.

[0105] Step 2:

[0106] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[0107] Step 3:

[0108] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[0109] Example 1

[0110] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0111] Existing conversational systems using generative AI models have difficulty accurately understanding the user's intentions and emotions and continuously learning to achieve more natural dialogue. Furthermore, they lack a means to effectively utilize dialogue history and smoothly advance ongoing dialogue with the user. As a result, users sometimes experience unnatural and fragmented dialogue, resulting in reduced satisfaction with the system.

[0112] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0113] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, natural language generation means for generating a response based on the analysis result, means for sending the generated response to the user, means for saving the conversation history in a database and using it for continuous learning, means for loading necessary data and settings of the generative AI model into memory when the system starts up, means for analyzing the emotion and context of a message from the user to understand the user's intention, and means for generating a response written in natural language. This makes it possible to accurately understand the user's intention and emotion, realize natural conversations, and improve the quality of conversations with the user through continuous learning.

[0114] A "generative AI model" is a model that can generate text using artificial intelligence, and is used for response generation and text completion in natural language processing.

[0115] "Initialization and loading means" refers to the methods or processes for initializing a generative AI model at system startup and loading the necessary data and settings from storage into memory.

[0116] "Means for receiving conversation requests" refers to the method or protocol by which the system receives text messages or questions entered by the user.

[0117] "Natural language understanding" refers to technologies and algorithms that analyze the content, sentiment, and context of received text messages to understand the user's intent.

[0118] "Natural language generation means" refers to technologies and algorithms that generate appropriate responses based on information obtained through natural language understanding.

[0119] "Means for sending a response to a user" refers to a data transmission method or protocol for sending a generated response back to a user to continue the dialogue.

[0120] "Means for storing conversation history in a database" refers to technologies and methods for persistently storing the content of conversations with users and using them to improve the quality of future interactions.

[0121] "Means used for continuous learning" refers to the learning process of retraining the generative AI model based on stored conversation history to improve response accuracy and the naturalness of the dialogue.

[0122] "Means for loading the necessary data and settings for the generative AI model into memory when the system starts up" refers to a method for loading the files and setting information necessary for the generative AI model into memory space when the system starts up.

[0123] "Means for understanding a user's intent by analyzing the emotion and context of messages from the user" refers to technologies and algorithms for analyzing the emotion and context from the user's input text and understanding their intent.

[0124] "Means for generating responses formulated in natural language" refers to techniques or processes for generating responses to a user in a natural language expression.

[0125] This invention relates to a conversation system that uses a generative AI model, and specifically has a configuration in which three elements - a server, a terminal, and a user - function in cooperation with each other.

[0126] 1. Server Processing

[0127] The server initializes the generative AI model (e.g., GPT-4, a model specialized for natural language processing) at system startup and loads the necessary data and settings into memory. The server listens to and receives conversation requests sent from the device, and when a message is received, it is passed to a natural language understanding (NLU) module for analysis of content, sentiment, and context.

[0128] Based on the analyzed information, the server uses a natural language generation (NLG) module to generate an appropriate response. The response is then sent back to the device via the server. Furthermore, the conversation history is stored in a database and used for continuous learning. This allows the generative AI model to improve its accuracy over time, enabling it to provide answers that are tailored to the user's characteristics and preferences.

[0129] Specific examples

[0130] When the server starts up, it loads the GPT-4 model and reads the configuration file into memory.

[0131] The server receives a message from the user saying, "The weather is nice today," and the NLU module analyzes the positive emotion as "the weather is nice."

[0132] The server generates a response saying, "Today is indeed a perfect day to go out," and sends it to the terminal.

[0133] The server stores the conversation history in a database for future use.

[0134] 2. Terminal Processing

[0135] The terminal displays an interface such as a chat window that allows the user to interact with the system. The user can enter messages through the interface. The terminal obtains input from the user and sends the content to the server. The terminal receives responses sent from the server and displays them in the user interface to provide feedback to the user.

[0136] Specific examples

[0137] The tablet device launches a chat app and displays a chat window.

[0138] The user types "What day is it today?" and the device sends the message to the server.

[0139] The terminal displays the response "Today is World Environment Day" received from the server and prompts the user to confirm it.

[0140] 3. User Operation

[0141] Users can freely start a conversation through the interface on their device. A conversation begins by entering a message or question in the chat window and sending it. By checking the response sent back from the server and entering the next message based on that, a natural, continuous conversation is realized.

[0142] Specific examples

[0143] A user opens a chat app and types "Good morning."

[0144] Check the server's response, "Good morning! How can I help you today?", then type, "What's the news today?"

[0145] Look at the generated response, "Today's news is..." and continue with more detailed questions.

[0146] Prompt Sentence Examples

[0147] Below are some specific examples of prompt sentences to input to the generative AI model.

[0148] "Tell me the news today."

[0149] "How's the weather?"

[0150] "What are some popular movies these days?"

[0151] "Tell me your ideas for our next vacation."

[0152] "Please recommend some books."

[0153] This system, which uses a generative AI model (e.g., GPT-4) to enable natural, continuous conversations with users, was developed with the aim of reducing the burden on caregivers and eliminating feelings of loneliness among users.

[0154] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0155] Server processing steps:

[0156] Step 1:

[0157] Initializing and Loading the Model

[0158] Input: System Start Trigger

[0159] Processing: The server initializes the generative AI model (e.g., GPT-4), which involves loading the model file and configuration file from disk into memory.

[0160] Output: An initialized generative AI model

[0161] Specific operation: The server "loads the generative AI model file from disk." For example, "checks the model parameter settings."

[0162] Step 2:

[0163] Request reception

[0164] Input: conversation request from device

[0165] Processing: The server listens on a port for requests and receives conversational requests via HTTP or WebSocket protocols.

[0166] Output: Received user message

[0167] Specific action: The server "listens for requests on port 8080." For example, "receives a message from the user saying 'The weather is nice today.'"

[0168] Step 3:

[0169] natural language understanding

[0170] Input: Received user message

[0171] Processing: The server passes the message to a Natural Language Understanding (NLU) module for content, sentiment, and context analysis, which helps understand the user's intent and sentiment.

[0172] Output: Analysis results (intent, sentiment, context)

[0173] Specific operation: The server sends the user's message to the NLU engine. For example, it determines the positive sentiment from the message 'The weather is good'.

[0174] Step 4:

[0175] Generating a response

[0176] Input: Analysis result of NLU module

[0177] Processing: The server uses a natural language generation (NLG) module to generate an appropriate response based on the parsed information.

[0178] Output: The generated response message

[0179] Specific action: The server generates a response using an NLG engine. For example, it generates a response such as, "Today is indeed a perfect day to go out."

[0180] Step 5:

[0181] Sending a Response

[0182] Input: The generated response message

[0183] Processing: The server generates a response and sends it to the terminal as an HTTP response.

[0184] Output: Response message sent to the terminal

[0185] Specific action: The server "returns the generated response as an HTTP response."

[0186] Step 6:

[0187] Conversation history storage and continuous learning

[0188] Input: Conversation history (user messages and server responses)

[0189] Processing: The server stores the conversation history in a database and uses it to continuously train the generative AI model.

[0190] Output: Saved conversation history, updated generative AI model

[0191] Specific operation: The server "records the conversation history in a database." For example, "saves the content of the conversation with the user."

[0192] Terminal processing steps:

[0193] Step 1:

[0194] User Interface Display

[0195] Input: User-initiated interface launch

[0196] Processing: The terminal displays an interface (such as a chat window) that allows the user to interact with the system.

[0197] Output: Chat window displayed

[0198] Specific action: The device "launches a chat app and displays the screen." For example, "displays a chat input field."

[0199] Step 2:

[0200] Getting and Sending User Input

[0201] Input: Message input from the user

[0202] Processing: The terminal detects input from the user and sends a message to the server.

[0203] Output: User message sent

[0204] Specific action: The device "detects input from the user." For example, "sends the message 'Hello' to the server."

[0205] Step 3:

[0206] Viewing the response

[0207] Input: Response message from the server

[0208] Processing: The terminal receives the response sent by the server and displays it on the user interface.

[0209] Output: The displayed response message

[0210] Specific action: The device "receives the response from the server and displays it on the screen." For example, "displays the response 'What a great day today' in the chat window."

[0211] User steps:

[0212] Step 1:

[0213] Start a conversation

[0214] Input: Chat interface launched by user action

[0215] Process: The user initiates a conversation through the interface on the device.

[0216] Output: User's first message

[0217] Specific action: The user "opens a chat app." For example, "types 'Good morning.'"

[0218] Step 2:

[0219] Enter your message

[0220] Input: Message entered into the user interface

[0221] Process: The user enters a message or question through the interface and presses the send button.

[0222] Output: Message sent to the server

[0223] Specific action: The user "types a question into the chat window." For example, "types 'What's the weather like today?' and presses the send button."

[0224] Step 3:

[0225] Check the response and enter the next message

[0226] Input: Response from the server displayed on the terminal

[0227] Action: The user checks the response from the server and enters the next message based on it.

[0228] Output: The following message

[0229] Specific action: The user "reads the server's response." For example, "after seeing the response 'It's sunny today,' he / she re-enters the question."

[0230] (Application example 1)

[0231] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0232] Physical stores require systems that allow customers to quickly and efficiently obtain product information and ask questions. However, conventional methods require in-store staff to respond to each customer individually, which can lead to delayed responses or insufficient explanations. In addition, it can be difficult for customers to instantly obtain the information they want, which can lead to a decrease in customer satisfaction. To solve these issues, it is necessary to develop a system that uses AI technology to quickly and accurately respond to customers' questions.

[0233] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0234] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, and means for customers to obtain product information and ask questions using a user interface. This allows customers to instantly obtain the product information they need, reducing the burden on in-store staff and improving customer satisfaction.

[0235] A "generative AI model" is an artificial intelligence model that generates natural responses based on sentences entered by the user.

[0236] "Means for initializing and loading" refers to the function of loading the generated AI model and necessary data and settings into memory when the system starts up.

[0237] The "means for receiving conversation requests from users" is a function for sending messages and questions entered by users to the server.

[0238] "Natural language understanding means" is a function that analyzes received user messages and understands their content, emotions, and context.

[0239] The "natural language generation means" is a function that generates an appropriate response in natural language based on the analyzed information.

[0240] The "means for sending to the user" is a function for sending the generated response back to the user's terminal.

[0241] "Means for saving conversation history in a database and using it for continuous learning" is a function that saves conversation history with the user and uses it to improve the accuracy of the model.

[0242] A "user interface" is a function that displays screens and widgets that allow users to interact with the system and accepts user input.

[0243] "Means for customers to obtain product information and ask questions" refers to functions that allow customers to obtain product information and ask questions via smartphones or smart glasses.

[0244] In this invention, we build a system that uses generative AI models to realize smart concierge services in physical stores.

[0245] Server Processing

[0246] The server has the means to initialize and load the generative AI model, loading the model and necessary data into memory. This step prepares the server for a conversation with the user.

[0247] A conversation request from a user is received by the server using a receiving means. The received request includes a message or question, which the server analyzes using a natural language understanding means to understand the content, emotion, and context. Then, based on the analysis results, a natural language generation means generates an appropriate response, which is then sent to the user using a transmission means. This response is displayed on the user's device.

[0248] Additionally, by storing conversation history in a database and providing a means for continuous learning, the server can learn about the user's characteristics and preferences over time and generate more appropriate responses.

[0249] Terminal handling

[0250] The terminal has a means for displaying a user interface, allowing the user to input messages. The terminal receives input from the user and sends it to the server. The response sent from the server is again displayed to the user through the terminal's interface.

[0251] This user interface is used by customers to obtain product information and ask questions, allowing customers to instantly obtain the information they need and reducing the burden on store staff.

[0252] Specific Examples

[0253] The server uses a server machine equipped with a high-performance GPU, and the terminals can be smartphones or smart glasses. The software library used is the transformers library from Hugging Face.

[0254] For example, if a customer uses their smartphone to type, "Good morning. What products do you recommend?", the generative AI model will analyze the information and respond with, "Good morning. Today's recommendation is our newly released smartwatch. Would you like to hear more about it?" In this way, customers can easily obtain information, making in-store customer service more efficient.

[0255] Prompt Sentence Examples

[0256] Here is an example of a real prompt:

[0257] "Good morning. What products do you recommend?"

[0258] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0259] Step 1:

[0260] The server initializes the generative AI model at system startup, loading the necessary data and configuration into memory. At this stage, the AI ​​model is ready to converse with the user. Specifically, a server machine equipped with a powerful GPU loads a pre-trained generative AI model (e.g., GPT-4) using Hugging Face's transformers library. The input is the model and associated configuration data, and the output is the model in its ready state.

[0261] Step 2:

[0262] The user initiates a dialogue with the system using a user interface on a smartphone or smart glasses. Specifically, the user types "Good morning. What products do you recommend?" into the interface. The device receives this message in real time and sends it to the server. The input is the user's message, and the output is a transmission request to the server.

[0263] Step 3:

[0264] The server receives a conversation request from the user. Specifically, the server receives a message sent by the user and passes the content to the natural language understanding module for analysis. The input is the message sent by the user, and the output is data to the understanding module.

[0265] Step 4:

[0266] The server analyzes the received message using natural language understanding. Specifically, the server analyzes the message content, sentiment, and context using a natural language understanding module (e.g., BERT) to understand the user's intent. The input is the user's message data, and the output is the analyzed information.

[0267] Step 5:

[0268] The server generates an appropriate response using natural language generation means based on the analysis results. Specifically, the server uses a natural language generation module (e.g., GPT-4) to generate a natural response from the analyzed information. The input is the analysis results, and the output is the generated response sentence.

[0269] Step 6:

[0270] The server sends the generated response to the user. Specifically, the server sends the generated response to the terminal, which displays it on the user interface. The input is the generated response sentence, and the output is a message displayed on the user's terminal.

[0271] Step 7:

[0272] The terminal displays the response received from the server on the user interface. Specifically, the terminal displays the message received from the server on the interface and notifies the user. The input is the response message from the server, and the output is the display on the user interface.

[0273] Step 8:

[0274] The server stores the conversation history in a database and uses it for continuous learning. Specifically, the server records all conversation history in a database and uses it as learning data to improve the accuracy of future conversations. The input is the conversation history data, and the output is an updated database and a trained model.

[0275] This process will enable a smart concierge service to be realized in physical stores, allowing customers to smoothly obtain product information and have questions answered using their smartphones or smart glasses.

[0276] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0277] The present invention combines an emotion engine with a system that uses generative AI models to resolve conversational shortcomings in nursing homes, thereby achieving more natural and emotionally sensitive dialogue. The following describes specific embodiments.

[0278] Server Action:

[0279] 1. Initializing and loading the model

[0280] At system startup, the server initializes the generative AI model and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[0281] Example: A server loads a pre-trained generative AI model (e.g., GPT-4) and an emotion engine.

[0282] 2. Request acceptance

[0283] The server receives a conversation request from the terminal, which includes a message or question entered by the user.

[0284] Example: The server receives user input saying "The weather is nice today."

[0285] 3. Natural Language Understanding and Sentiment Analysis

[0286] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, and also analyzes the user's emotions using an emotion engine, thereby accurately understanding the user's intentions and emotions.

[0287] Example: The server reads positive sentiment from the message "The weather is nice."

[0288] 4. Generating the Response

[0289] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analyzed information and the results of sentiment analysis. The response is formulated in natural language and takes into account the user's sentiment.

[0290] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[0291] 5. Sending the Response

[0292] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[0293] Example: The server generates a response and sends it to the device.

[0294] 6. Conversation history storage and continuous learning

[0295] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn user characteristics, preferences, and sentiment trends over time.

[0296] Example: The server stores the conversation history in a database to help with the next interaction.

[0297] Terminal handling:

[0298] 1. Displaying the user interface

[0299] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0300] Example: A tablet device displays a chat app.

[0301] 2. Getting and Sending User Input

[0302] The terminal receives input from the user (such as a message or question) and sends the content to the server. This operation transfers the user's message to the server.

[0303] Example: A user types "What day is it today?" and the device sends the message to the server.

[0304] 3. View the response

[0305] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0306] Example: The device displays the response from the server: "Today is World Environment Day."

[0307] User Action:

[0308] 1. Start a conversation

[0309] Users initiate conversations through the device interface and can start conversations at any time.

[0310] Example: A user types "Good morning" into a terminal.

[0311] 2. Enter your message

[0312] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[0313] Example: A user types "What's in the news today?" and presses the send button.

[0314] 3. Check the response and enter the next message

[0315] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[0316] Example: User continues with "Thank you, is there anything else?"

[0317] As described above, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural, emotion-sensitive conversations, which can simultaneously reduce the burden on caregivers and provide emotional support to users.

[0318] The processing flow will be explained below.

[0319] Server processing steps:

[0320] Step 1:

[0321] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, so that the AI ​​and emotion engine are ready to converse with the user.

[0322] Step 2:

[0323] The server receives the user's conversation request sent from the device, which includes a message or question typed by the user.

[0324] Step 3:

[0325] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, thereby understanding the user's intent and determining how to respond.

[0326] Step 4:

[0327] At the same time, the server uses an emotion engine to analyze the emotions from the user's message, which clarifies whether the user has positive, negative, or neutral emotions.

[0328] Step 5:

[0329] The server combines the analysis results of the NLU module and the emotion engine and passes them to the natural language generation (NLG) module, which provides information for generating emotion-sensitive responses.

[0330] Step 6:

[0331] The server uses the NLG module to generate an appropriate response based on the integrated information, which is formulated in natural language and is emotionally relevant to the user.

[0332] Step 7:

[0333] The server generates a response and sends it to the device, allowing the conversation to continue.

[0334] Step 8:

[0335] The server stores the conversation history and sentiment analysis results in a database, allowing the generative AI model and sentiment engine to continuously learn and be more effective in the next conversation.

[0336] Terminal processing steps:

[0337] Step 1:

[0338] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0339] Step 2:

[0340] The device receives messages and questions entered by the user and sends them to the server. This operation transfers the user's messages to the server.

[0341] Step 3:

[0342] The terminal receives the response from the server and displays it on the user interface, allowing the user to see the system's response and enter the next message.

[0343] User steps:

[0344] Step 1:

[0345] The user initiates the conversation through the terminal interface, entering the initial message and beginning a dialogue with the system.

[0346] Step 2:

[0347] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[0348] Step 3:

[0349] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[0350] Through the above processing steps, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural and emotion-sensitive conversations. This system can simultaneously reduce the burden on caregivers and provide emotional support to users.

[0351] Example 2

[0352] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0353] Conventional communication systems in nursing care facilities have difficulty in communicating with users while taking their emotions into full consideration, which makes it difficult to increase the psychological satisfaction of users. Furthermore, there are few systems that support continuous learning by effectively utilizing conversation history, so long-term improvement cannot be expected.

[0354] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for initializing and loading the generative AI model and emotion engine, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, means for analyzing emotions based on the analysis results, natural language generation means for generating a response based on the analysis results and the emotion analysis results, means for sending the generated response to the user, and means for saving the conversation history and the emotion analysis results in a database and using them for continuous learning. This enables natural conversation that takes user emotions into consideration, and continuous learning can be expected to improve the quality of the conversation.

[0355] A "generative AI model" is an artificial intelligence model that generates responses in natural language based on input from the user.

[0356] An "emotion engine" is a means of analyzing emotions from user input data and generating a response according to those emotions.

[0357] "Initialization" refers to the operation of setting a program or system into an operational state.

[0358] "Load" refers to the operation of reading a program or data into memory.

[0359] A "conversation request" refers to a message or question that a user inputs to the system.

[0360] "Natural language understanding" is a technology that analyzes the content and context of received conversation requests and understands their meaning.

[0361] "Natural language generation" is a technology that generates responses in natural language based on analyzed information.

[0362] "Send" refers to the act of transferring generated data or responses to another device or terminal.

[0363] "Conversation history" refers to a record of all interactions between a user and a system.

[0364] "Continuous learning" is a learning process that improves the system's performance based on past conversation history and new data.

[0365] "User interface" refers to the display screen and input devices that allow a user to interact with a system.

[0366] "Terminal" refers to a device for communicating with the system and displaying responses.

[0367] MODE FOR CARRYING OUT THE INVENTION

[0368] This invention is a system that combines a generative AI model and an emotion engine to resolve the lack of conversation in nursing care facilities and realize natural, emotion-sensitive dialogue.

[0369] Server Processing

[0370] When the system starts up, the server initializes the generative AI model and emotion engine and loads the necessary data and settings into memory. Specifically, it loads the parameters of a pre-trained generative AI model such as GPT-4 and the emotion engine that performs emotion analysis into memory. This prepares the server to respond immediately to conversations with users.

[0371] The server receives requests from users. These requests include messages or questions entered by the user, which are sent from the device to the server. The received messages are parsed by a natural language understanding (NLU) module, which tokenizes the messages and analyzes their content and context. At the same time, an emotion engine calculates an emotion value (e.g., positive, negative, neutral) based on the message tokens.

[0372] Based on the analysis results, the server generates a response using a natural language generation (NLG) module. The generative AI model uses the user's message and sentiment value as prompts to predict the next conversation flow and generate a response. The generated response is formatted in natural language and sent to the user.

[0373] Conversation history and sentiment analysis results are stored in a database and used as input for continuous learning. This process allows the generative AI model and emotion engine to improve performance with each use, resulting in more appropriate and natural responses.

[0374] Terminal handling

[0375] The terminal displays an interface through which the user can interact with the system, including a chat window and a text entry field where the user can type a message. After the user types a message, the terminal sends it to the server.

[0376] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to enter the next message based on the displayed response and progress the conversation.

[0377] User operations

[0378] A user can initiate a conversation at any time through the device's interface. For example, the user can type "Good morning" into the device and press the send button. The user can then use the interface to type and send questions or messages. For example, the user can type and send "Please tell me the news today."

[0379] The response generated by the server is displayed on the terminal, and after checking it, the user can input the next message, enabling continuous dialogue. For example, the user can input a message such as "Thank you, is there anything else?"

[0380] Specific examples

[0381] Specific examples are shown below.

[0382] When a user types "Are there any events today?" into their device, the server receives the message and analyzes it. Based on the analysis results and the emotion value, the generative AI model generates a response such as "There will be music activities this afternoon." This response is sent to the device and displayed to the user.

[0383] An example prompt is:

[0384] "Is there any event today?"

[0385] "Tell me about what happened yesterday."

[0386] "Please tell me the latest news."

[0387] As described above, this system, which includes a generative AI model, an emotion engine, and a natural language understanding and generation module, is designed to facilitate conversations in nursing homes. By conducting conversations in a natural and emotionally sensitive manner, it can improve the psychological satisfaction of users and reduce the burden on caregivers.

[0388] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0389] Step 1: Initialize and load the model

[0390] When the system starts up, the server initializes the parameters of the generative AI model (e.g., GPT-4) and emotion engine and loads them into memory. Specifically, it calls the initialize_model() function. The input of this step is the trained data for the generative AI model and emotion engine, and the output is the initialized model and engine deployed in memory.

[0391] Step 2: Request acceptance

[0392] The server receives a conversation request from the terminal. It uses the receive_request(connection) method to receive a data packet containing a message or question entered by the user. The input is the message from the user, and the output is the received conversation request.

[0393] Step 3: Natural Language Understanding and Sentiment Analysis

[0394] The server passes the received message to the Natural Language Understanding (NLU) module for content and context analysis. It also performs sentiment analysis using the emotion engine. Specifically, nlu_module.analyze(text) performs tokenization and syntactic analysis, and emotion_engine.analyze(text) calculates the sentiment value. The input to this step is the received conversation request, and the output is the analysis result and sentiment value.

[0395] Step 4: Generate a response

[0396] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analysis result and emotion value. It generates a response by calling the nlg_module.generate_response(context, emotion) method. The input is the analysis result and emotion value, and the output is the generated response sentence.

[0397] Step 5: Sending a Response

[0398] The server sends the generated response to the terminal using the send_response(response, client) method. The input is the generated response text, and the output is the response sent to the terminal.

[0399] Step 6: Save conversation history and continue learning

[0400] The server saves the conversation history and sentiment analysis results in a database for continuous learning. Specifically, it uses the save_to_database(user_id, conversation_data) method. The input of this step is the conversation history and sentiment analysis results, and the output is the data saved in the database.

[0401] Step 7: Displaying the User Interface

[0402] The terminal displays an interface for the user to interact with the system. Specifically, the display_interface() function is called to display the chat window. The input is the interface design loaded when the system started, and the output is the chat screen displayed to the user.

[0403] Step 8: Getting and Sending User Input

[0404] The terminal gets input from the user and sends the contents to the server. get_user_input() receives user input and send_to_server(input_text) sends it to the server. The input is the text entered by the user and the output is the message sent to the server.

[0405] Step 9: View the response

[0406] The terminal receives the response sent from the server and displays it on the user interface. The response is received with receive_from_server() and updated with update_interface(response). The input is the response from the server, and the output is the response displayed to the user.

[0407] Through these steps, natural, emotion-sensitive dialogue is realized using a generative AI model and emotion engine. Appropriate processing is performed based on the input data at each step, and the resulting output data is passed on to the next step. The user can engage in natural dialogue through their device, and the system can generate responses that are in tune with the user's emotions.

[0408] (Application example 2)

[0409] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0410] Customer service is a major issue in modern brick-and-mortar stores. In particular, when customers ask about products, store clerks may lack sufficient knowledge. While a response that takes into consideration the customer's feelings is required, not all store clerks have the skills to do so. This can lead to a decline in customer satisfaction. The present invention aims to solve these issues and realize natural, emotionally sensitive conversations in brick-and-mortar stores.

[0411] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0412] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, means for extracting information about products and explaining the products in a natural conversational format, means for generating responses according to emotions and taking customer emotions into consideration, and means for analyzing customer reactions and questions in real time. This enables improvement in the quality of customer service in physical stores and improvement in customer satisfaction.

[0413] A "generative AI model" refers to an artificial intelligence technology that is trained and learned in advance using large amounts of data to understand human language and generate responses.

[0414] "Means to take emotions into consideration" refers to the function of analyzing emotions from the user's input message and generating an appropriate response based on those emotions.

[0415] "Natural language understanding means" refers to a function for analyzing input messages from users and understanding their content and intent.

[0416] "Natural language generator" refers to a function that generates responses in natural language based on analyzed user input.

[0417] "User interface" refers to the collection of screens and input devices that allow a user to interact with a system.

[0418] A "conversation request" refers to a question or message that a user inputs to the system.

[0419] "Means for receiving" refers to a function for taking in conversation requests input from the outside.

[0420] "Means for generating a response" refers to the function of generating an appropriate reply based on a received conversation request.

[0421] "Means for sending" refers to a function for sending the generated response back to the user.

[0422] "Conversation history" refers to data that stores the contents of past conversations exchanged between a user and the system.

[0423] A "database" refers to a digital storage device that systematically stores information such as conversation history.

[0424] "Means for continuous learning" refers to a function that allows the AI ​​model to further learn based on saved conversation history, improving the accuracy and quality of responses.

[0425] "Product Information" refers to detailed information about a specific product, such as its characteristics, usage, and price.

[0426] "Means of analysis" refers to the function of analyzing input data in various ways and understanding its meaning and intent.

[0427] The present invention relates to a system that uses a generative AI model to improve the efficiency of customer service in brick-and-mortar stores and provide natural conversations that take emotions into consideration. Specific embodiments are described below.

[0428] Server Processing

[0429] The server performs the following process.

[0430] 1. Initializing and loading the model

[0431] At system startup, the server initializes the generative AI model (e.g., GPT-4) and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[0432] 2. Receiving a conversation request

[0433] The server receives a conversation request from the terminal, which includes a message or question entered by the customer.

[0434] 3. Natural Language Understanding and Sentiment Analysis

[0435] The server passes the received message to a natural language understanding module to analyze the content and context, and also analyzes the customer's emotions using an emotion engine, thereby accurately understanding the customer's intentions and emotions.

[0436] 4. Generating the Response

[0437] The server uses a natural language generation module to generate an appropriate response based on the analyzed information and the sentiment analysis results. The generated response is written in natural language and takes into consideration the customer's sentiment.

[0438] 5. Sending the Response

[0439] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[0440] 6. Conversation history storage and continuous learning

[0441] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn customer characteristics, preferences, and sentiment trends over time.

[0442] Terminal handling

[0443] The terminal performs the following process.

[0444] 1. Displaying the user interface

[0445] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the customer to type messages.

[0446] 2. Capture and submit customer input

[0447] The terminal receives input from the customer (such as a message or question) and sends the content to the server. This operation transfers the customer's message to the server.

[0448] 3. View the response

[0449] The terminal receives the response sent from the server and displays it on the user interface, allowing the customer to check the system's response and enter the next message.

[0450] Hardware and software used

[0451] The system is implemented using the following hardware and software:

[0452] Hardware: Smartphones, smart glasses, head-mounted displays

[0453] Software: Flask (web server), OpenAI GPT-4 (generative AI model), TextBlob (sentiment analysis)

[0454] Specific examples

[0455] 1. A customer asks through smart glasses, "Tell me about this shirt."

[0456] The server generates a response such as "This shirt has a modern design and is highly breathable" and sends it to the device, which displays the response on the smart glasses.

[0457] 2. A customer types, "I'm tired from work today."

[0458] The server generates a response that takes into account the user's emotions, such as "Thank you for your hard work. Would you like something to help you relax?", and sends it to the device. The device then displays this response on its smartphone.

[0459] Prompt Sentence Examples

[0460] Customer: Tell me about this shirt.

[0461] clerk:

[0462] Customer: I'm tired from work today.

[0463] clerk:

[0464] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0465] Step 1:

[0466] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, which prepares the generative AI model and emotion engine to generate customer responses. The input is data from the pre-trained AI model, and the output is the system configuration information loaded into memory.

[0467] Step 2:

[0468] A user inputs a conversation request through a terminal. Specifically, the user inputs a question or message into the interface (e.g., a chat window) of a smartphone or smart glasses. The input is the user's text message, and the output is the transfer of the request from the terminal to the server.

[0469] Step 3:

[0470] The server passes the conversation request received from the device to a natural language understanding module, which analyzes the content and context. At the same time, it uses an emotion engine to analyze the user's emotions. The input is the user's text message, and the output is the analysis results of the message's content, context, and emotions. Natural language processing and emotion analysis technologies are used for the analysis.

[0471] Step 4:

[0472] The server uses a natural language generation module based on the analysis results to generate an appropriate response. A generative AI model (e.g., GPT-4) is used to form a natural-language response based on the user's intent and sentiment. The input is the analysis results of the message's content, context, and sentiment, and the output is the generated text response.

[0473] Step 5:

[0474] The server sends the generated response to the terminal, which allows the interaction to reach the user and proceed to the next step. The input is the generated text response, and the output is the display of the response in the user interface.

[0475] Step 6:

[0476] The terminal displays the response received from the server on the user interface. The user checks the displayed response and prepares to enter the next message. The input is the text response sent from the server, and the output is the text displayed on the screen of the smartphone or smart glasses.

[0477] Step 7:

[0478] Through the user interface, the user inputs the next message and sends it back to the server. This cycle continues, creating a continuous dialogue: the input is the user's new message, and the output is again a request to be parsed and a response generated.

[0479] Step 8:

[0480] The server stores the conversation history and emotion analysis results in a database and uses them for continuous learning of the generative AI model and emotion engine. This allows the system to learn customer characteristics, preferences, and emotional trends over time and use them for the next interaction. The input is the conversation history and emotion data to be saved, and the output is the information accumulated in the database.

[0481] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0482] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0483] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0484] [Second embodiment]

[0485] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0486] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0487] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0488] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0489] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0490] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0491] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0492] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0493] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0494] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0495] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0496] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0497] The present invention is a service system dedicated to conversation using a generative AI model, and specific embodiments are described below.

[0498] Server Action:

[0499] 1. Initializing and loading the model

[0500] The server initializes the generative AI model at system startup and loads the necessary data and settings into memory, making the AI ​​ready to converse with the user.

[0501] Example: A server loads a pre-trained generative AI model (e.g., GPT-4).

[0502] 2. Request acceptance

[0503] The server receives a conversation request sent from the terminal, which includes a message or question entered by the user.

[0504] Example: The server receives user input saying "The weather is nice today."

[0505] 3. Natural language understanding

[0506] The server passes the received message to a natural language understanding module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[0507] Example: The server reads positive sentiment from the message "The weather is nice."

[0508] 4. Generating the Response

[0509] The server uses a natural language generation module to generate appropriate responses based on the analyzed information. The responses are formulated in natural language, enabling smooth conversation with the user.

[0510] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[0511] 5. Sending the Response

[0512] The server generates a response and sends it back to the terminal, where it is displayed to the user and the dialogue proceeds.

[0513] Example: The server generates a response and sends it to the device.

[0514] 6. Conversation history storage and continuous learning

[0515] The server stores the conversation history in a database and uses it to continuously train the AI ​​model, allowing it to learn about the user's characteristics and preferences over time.

[0516] Example: The server stores the conversation history in a database to help with the next interaction.

[0517] Terminal handling:

[0518] 1. Displaying the user interface

[0519] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0520] Example: A tablet device displays a chat app.

[0521] 2. Getting and Sending User Input

[0522] The terminal receives input from the user (such as a message or question) and sends the content to the server.

[0523] Example: A user types "What day is it today?" and the device sends the message to the server.

[0524] 3. View the response

[0525] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0526] Example: The device displays the response from the server: "Today is World Environment Day."

[0527] User Action:

[0528] 1. Start a conversation

[0529] Users initiate conversations through the device interface and can start conversations at any time.

[0530] Example: A user types "Good morning" into a terminal.

[0531] 2. Enter your message

[0532] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[0533] Example: A user types "What's in the news today?" and presses the send button.

[0534] 3. Check the response and enter the next message

[0535] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[0536] Example: User continues with "Thank you, is there anything else?"

[0537] As described above, by combining a generative AI model with multiple processing means, the present invention enables natural and continuous conversation, thereby reducing the burden on caregivers and eliminating feelings of loneliness among users.

[0538] The processing flow will be explained below.

[0539] Server processing steps:

[0540] Step 1:

[0541] The server initializes the generative AI model and loads the necessary data and settings into memory, so the AI ​​is ready to converse with the user.

[0542] Step 2:

[0543] The server receives a conversation request from the device, which includes a message or question typed by the user.

[0544] Step 3:

[0545] The server passes the received message to a natural language understanding (NLU) module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[0546] Step 4:

[0547] Based on the analysis results, the server uses a natural language generation (NLG) module to generate an appropriate response, which is formulated in natural language.

[0548] Step 5:

[0549] The server then sends the generated response to the terminal, allowing the conversation to proceed.

[0550] Step 6:

[0551] The server stores the conversation history in a database and uses it to continuously train the generative AI model, allowing the model to learn about the user's characteristics and preferences over time.

[0552] Terminal processing steps:

[0553] Step 1:

[0554] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0555] Step 2:

[0556] The device receives input from the user (such as a message or question) and sends it to the server, which then forwards the user's message to the server.

[0557] Step 3:

[0558] The terminal receives the response sent by the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0559] User steps:

[0560] Step 1:

[0561] Users initiate conversations through the device interface, and conversations can begin at any time.

[0562] Step 2:

[0563] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[0564] Step 3:

[0565] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[0566] Example 1

[0567] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0568] Existing conversational systems using generative AI models have difficulty accurately understanding the user's intentions and emotions and continuously learning to achieve more natural dialogue. Furthermore, they lack a means to effectively utilize dialogue history and smoothly advance ongoing dialogue with the user. As a result, users sometimes experience unnatural and fragmented dialogue, resulting in reduced satisfaction with the system.

[0569] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0570] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, natural language generation means for generating a response based on the analysis result, means for sending the generated response to the user, means for saving the conversation history in a database and using it for continuous learning, means for loading necessary data and settings of the generative AI model into memory when the system starts up, means for analyzing the emotion and context of a message from the user to understand the user's intention, and means for generating a response written in natural language. This makes it possible to accurately understand the user's intention and emotion, realize natural conversations, and improve the quality of conversations with the user through continuous learning.

[0571] A "generative AI model" is a model that can generate text using artificial intelligence, and is used for response generation and text completion in natural language processing.

[0572] "Initialization and loading means" refers to the methods or processes for initializing a generative AI model at system startup and loading the necessary data and settings from storage into memory.

[0573] "Means for receiving conversation requests" refers to the method or protocol by which the system receives text messages or questions entered by the user.

[0574] "Natural language understanding" refers to technologies and algorithms that analyze the content, sentiment, and context of received text messages to understand the user's intent.

[0575] "Natural language generation means" refers to technologies and algorithms that generate appropriate responses based on information obtained through natural language understanding.

[0576] "Means for sending a response to a user" refers to a data transmission method or protocol for sending a generated response back to a user to continue the dialogue.

[0577] "Means for storing conversation history in a database" refers to technologies and methods for persistently storing the content of conversations with users and using them to improve the quality of future interactions.

[0578] "Means used for continuous learning" refers to the learning process of retraining the generative AI model based on stored conversation history to improve response accuracy and the naturalness of the dialogue.

[0579] "Means for loading the necessary data and settings for the generative AI model into memory when the system starts up" refers to a method for loading the files and setting information necessary for the generative AI model into memory space when the system starts up.

[0580] "Means for understanding a user's intent by analyzing the emotion and context of messages from the user" refers to technologies and algorithms for analyzing the emotion and context from the user's input text and understanding their intent.

[0581] "Means for generating responses formulated in natural language" refers to techniques or processes for generating responses to a user in a natural language expression.

[0582] This invention relates to a conversation system that uses a generative AI model, and specifically has a configuration in which three elements - a server, a terminal, and a user - function in cooperation with each other.

[0583] 1. Server Processing

[0584] The server initializes the generative AI model (e.g., GPT-4, a model specialized for natural language processing) at system startup and loads the necessary data and settings into memory. The server listens to and receives conversation requests sent from the device, and when a message is received, it is passed to a natural language understanding (NLU) module for analysis of content, sentiment, and context.

[0585] Based on the analyzed information, the server uses a natural language generation (NLG) module to generate an appropriate response. The response is then sent back to the device via the server. Furthermore, the conversation history is stored in a database and used for continuous learning. This allows the generative AI model to improve its accuracy over time, enabling it to provide answers that are tailored to the user's characteristics and preferences.

[0586] Specific examples

[0587] When the server starts up, it loads the GPT-4 model and reads the configuration file into memory.

[0588] The server receives a message from the user saying, "The weather is nice today," and the NLU module analyzes the positive emotion as "the weather is nice."

[0589] The server generates a response saying, "Today is indeed a perfect day to go out," and sends it to the terminal.

[0590] The server stores the conversation history in a database for future use.

[0591] 2. Terminal Processing

[0592] The terminal displays an interface such as a chat window that allows the user to interact with the system. The user can enter messages through the interface. The terminal obtains input from the user and sends the content to the server. The terminal receives responses sent from the server and displays them in the user interface to provide feedback to the user.

[0593] Specific examples

[0594] The tablet device launches a chat app and displays a chat window.

[0595] The user types "What day is it today?" and the device sends the message to the server.

[0596] The terminal displays the response "Today is World Environment Day" received from the server and prompts the user to confirm it.

[0597] 3. User Operation

[0598] Users can freely start a conversation through the interface on their device. A conversation begins by entering a message or question in the chat window and sending it. By checking the response sent back from the server and entering the next message based on that, a natural, continuous conversation is realized.

[0599] Specific examples

[0600] A user opens a chat app and types "Good morning."

[0601] Check the server's response, "Good morning! How can I help you today?", then type, "What's the news today?"

[0602] Look at the generated response, "Today's news is..." and continue with more detailed questions.

[0603] Prompt Sentence Examples

[0604] Below are some specific examples of prompt sentences to input to the generative AI model.

[0605] "Tell me the news today."

[0606] "How's the weather?"

[0607] "What are some popular movies these days?"

[0608] "Tell me your ideas for our next vacation."

[0609] "Please recommend some books."

[0610] This system, which uses a generative AI model (e.g., GPT-4) to enable natural, continuous conversations with users, was developed with the aim of reducing the burden on caregivers and eliminating feelings of loneliness among users.

[0611] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0612] Server processing steps:

[0613] Step 1:

[0614] Initializing and Loading the Model

[0615] Input: System Start Trigger

[0616] Processing: The server initializes the generative AI model (e.g., GPT-4), which involves loading the model file and configuration file from disk into memory.

[0617] Output: An initialized generative AI model

[0618] Specific operation: The server "loads the generative AI model file from disk." For example, "checks the model parameter settings."

[0619] Step 2:

[0620] Request reception

[0621] Input: conversation request from device

[0622] Processing: The server listens on a port for requests and receives conversational requests via HTTP or WebSocket protocols.

[0623] Output: Received user message

[0624] Specific action: The server "listens for requests on port 8080." For example, "receives a message from the user saying 'The weather is nice today.'"

[0625] Step 3:

[0626] natural language understanding

[0627] Input: Received user message

[0628] Processing: The server passes the message to a Natural Language Understanding (NLU) module for content, sentiment, and context analysis, which helps understand the user's intent and sentiment.

[0629] Output: Analysis results (intent, sentiment, context)

[0630] Specific operation: The server sends the user's message to the NLU engine. For example, it determines the positive sentiment from the message 'The weather is good'.

[0631] Step 4:

[0632] Generating a response

[0633] Input: Analysis result of NLU module

[0634] Processing: The server uses a natural language generation (NLG) module to generate an appropriate response based on the parsed information.

[0635] Output: The generated response message

[0636] Specific action: The server generates a response using an NLG engine. For example, it generates a response such as, "Today is indeed a perfect day to go out."

[0637] Step 5:

[0638] Sending a Response

[0639] Input: The generated response message

[0640] Processing: The server generates a response and sends it to the terminal as an HTTP response.

[0641] Output: Response message sent to the terminal

[0642] Specific action: The server "returns the generated response as an HTTP response."

[0643] Step 6:

[0644] Conversation history storage and continuous learning

[0645] Input: Conversation history (user messages and server responses)

[0646] Processing: The server stores the conversation history in a database and uses it to continuously train the generative AI model.

[0647] Output: Saved conversation history, updated generative AI model

[0648] Specific operation: The server "records the conversation history in a database." For example, "saves the content of the conversation with the user."

[0649] Terminal processing steps:

[0650] Step 1:

[0651] User Interface Display

[0652] Input: User-initiated interface launch

[0653] Processing: The terminal displays an interface (such as a chat window) that allows the user to interact with the system.

[0654] Output: Chat window displayed

[0655] Specific action: The device "launches a chat app and displays the screen." For example, "displays a chat input field."

[0656] Step 2:

[0657] Getting and Sending User Input

[0658] Input: Message input from the user

[0659] Processing: The terminal detects input from the user and sends a message to the server.

[0660] Output: User message sent

[0661] Specific action: The device "detects input from the user." For example, "sends the message 'Hello' to the server."

[0662] Step 3:

[0663] Viewing the response

[0664] Input: Response message from the server

[0665] Processing: The terminal receives the response sent by the server and displays it on the user interface.

[0666] Output: The displayed response message

[0667] Specific action: The device "receives the response from the server and displays it on the screen." For example, "displays the response 'What a great day today' in the chat window."

[0668] User steps:

[0669] Step 1:

[0670] Start a conversation

[0671] Input: Chat interface launched by user action

[0672] Process: The user initiates a conversation through the interface on the device.

[0673] Output: User's first message

[0674] Specific action: The user "opens a chat app." For example, "types 'Good morning.'"

[0675] Step 2:

[0676] Enter your message

[0677] Input: Message entered into the user interface

[0678] Process: The user enters a message or question through the interface and presses the send button.

[0679] Output: Message sent to the server

[0680] Specific action: The user "types a question into the chat window." For example, "types 'What's the weather like today?' and presses the send button."

[0681] Step 3:

[0682] Check the response and enter the next message

[0683] Input: Response from the server displayed on the terminal

[0684] Action: The user checks the response from the server and enters the next message based on it.

[0685] Output: The following message

[0686] Specific action: The user "reads the server's response." For example, "after seeing the response 'It's sunny today,' he / she re-enters the question."

[0687] (Application example 1)

[0688] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0689] Physical stores require systems that allow customers to quickly and efficiently obtain product information and ask questions. However, conventional methods require in-store staff to respond to each customer individually, which can lead to delayed responses or insufficient explanations. In addition, it can be difficult for customers to instantly obtain the information they want, which can lead to a decrease in customer satisfaction. To solve these issues, it is necessary to develop a system that uses AI technology to quickly and accurately respond to customers' questions.

[0690] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0691] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, and means for customers to obtain product information and ask questions using a user interface. This allows customers to instantly obtain the product information they need, reducing the burden on in-store staff and improving customer satisfaction.

[0692] A "generative AI model" is an artificial intelligence model that generates natural responses based on sentences entered by the user.

[0693] "Means for initializing and loading" refers to the function of loading the generated AI model and necessary data and settings into memory when the system starts up.

[0694] The "means for receiving conversation requests from users" is a function for sending messages and questions entered by users to the server.

[0695] "Natural language understanding means" is a function that analyzes received user messages and understands their content, emotions, and context.

[0696] The "natural language generation means" is a function that generates an appropriate response in natural language based on the analyzed information.

[0697] The "means for sending to the user" is a function for sending the generated response back to the user's terminal.

[0698] "Means for saving conversation history in a database and using it for continuous learning" is a function that saves conversation history with the user and uses it to improve the accuracy of the model.

[0699] A "user interface" is a function that displays screens and widgets that allow users to interact with the system and accepts user input.

[0700] "Means for customers to obtain product information and ask questions" refers to functions that allow customers to obtain product information and ask questions via smartphones or smart glasses.

[0701] In this invention, we build a system that uses generative AI models to realize smart concierge services in physical stores.

[0702] Server Processing

[0703] The server has the means to initialize and load the generative AI model, loading the model and necessary data into memory. This step prepares the server for a conversation with the user.

[0704] A conversation request from a user is received by the server using a receiving means. The received request includes a message or question, which the server analyzes using a natural language understanding means to understand the content, emotion, and context. Then, based on the analysis results, a natural language generation means generates an appropriate response, which is then sent to the user using a transmission means. This response is displayed on the user's device.

[0705] Additionally, by storing conversation history in a database and providing a means for continuous learning, the server can learn about the user's characteristics and preferences over time and generate more appropriate responses.

[0706] Terminal handling

[0707] The terminal has a means for displaying a user interface, allowing the user to input messages. The terminal receives input from the user and sends it to the server. The response sent from the server is again displayed to the user through the terminal's interface.

[0708] This user interface is used by customers to obtain product information and ask questions, allowing customers to instantly obtain the information they need and reducing the burden on store staff.

[0709] Specific Examples

[0710] The server uses a server machine equipped with a high-performance GPU, and the terminals can be smartphones or smart glasses. The software library used is the transformers library from Hugging Face.

[0711] For example, if a customer uses their smartphone to type, "Good morning. What products do you recommend?", the generative AI model will analyze the information and respond with, "Good morning. Today's recommendation is our newly released smartwatch. Would you like to hear more about it?" In this way, customers can easily obtain information, making in-store customer service more efficient.

[0712] Prompt Sentence Examples

[0713] Here is an example of a real prompt:

[0714] "Good morning. What products do you recommend?"

[0715] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0716] Step 1:

[0717] The server initializes the generative AI model at system startup, loading the necessary data and configuration into memory. At this stage, the AI ​​model is ready to converse with the user. Specifically, a server machine equipped with a powerful GPU loads a pre-trained generative AI model (e.g., GPT-4) using Hugging Face's transformers library. The input is the model and associated configuration data, and the output is the model in its ready state.

[0718] Step 2:

[0719] The user initiates a dialogue with the system using a user interface on a smartphone or smart glasses. Specifically, the user types "Good morning. What products do you recommend?" into the interface. The device receives this message in real time and sends it to the server. The input is the user's message, and the output is a transmission request to the server.

[0720] Step 3:

[0721] The server receives a conversation request from the user. Specifically, the server receives a message sent by the user and passes the content to the natural language understanding module for analysis. The input is the message sent by the user, and the output is data to the understanding module.

[0722] Step 4:

[0723] The server analyzes the received message using natural language understanding. Specifically, the server analyzes the message content, sentiment, and context using a natural language understanding module (e.g., BERT) to understand the user's intent. The input is the user's message data, and the output is the analyzed information.

[0724] Step 5:

[0725] The server generates an appropriate response using natural language generation means based on the analysis results. Specifically, the server uses a natural language generation module (e.g., GPT-4) to generate a natural response from the analyzed information. The input is the analysis results, and the output is the generated response sentence.

[0726] Step 6:

[0727] The server sends the generated response to the user. Specifically, the server sends the generated response to the terminal, which displays it on the user interface. The input is the generated response sentence, and the output is a message displayed on the user's terminal.

[0728] Step 7:

[0729] The terminal displays the response received from the server on the user interface. Specifically, the terminal displays the message received from the server on the interface and notifies the user. The input is the response message from the server, and the output is the display on the user interface.

[0730] Step 8:

[0731] The server stores the conversation history in a database and uses it for continuous learning. Specifically, the server records all conversation history in a database and uses it as learning data to improve the accuracy of future conversations. The input is the conversation history data, and the output is an updated database and a trained model.

[0732] This process will enable a smart concierge service to be realized in physical stores, allowing customers to smoothly obtain product information and have questions answered using their smartphones or smart glasses.

[0733] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0734] The present invention combines an emotion engine with a system that uses generative AI models to resolve conversational shortcomings in nursing homes, thereby achieving more natural and emotionally sensitive dialogue. The following describes specific embodiments.

[0735] Server Action:

[0736] 1. Initializing and loading the model

[0737] At system startup, the server initializes the generative AI model and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[0738] Example: A server loads a pre-trained generative AI model (e.g., GPT-4) and an emotion engine.

[0739] 2. Request acceptance

[0740] The server receives a conversation request from the terminal, which includes a message or question entered by the user.

[0741] Example: The server receives user input saying "The weather is nice today."

[0742] 3. Natural Language Understanding and Sentiment Analysis

[0743] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, and also analyzes the user's emotions using an emotion engine, thereby accurately understanding the user's intentions and emotions.

[0744] Example: The server reads positive sentiment from the message "The weather is nice."

[0745] 4. Generating the Response

[0746] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analyzed information and the results of sentiment analysis. The response is formulated in natural language and takes into account the user's sentiment.

[0747] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[0748] 5. Sending the Response

[0749] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[0750] Example: The server generates a response and sends it to the device.

[0751] 6. Conversation history storage and continuous learning

[0752] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn user characteristics, preferences, and sentiment trends over time.

[0753] Example: The server stores the conversation history in a database to help with the next interaction.

[0754] Terminal handling:

[0755] 1. Displaying the user interface

[0756] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0757] Example: A tablet device displays a chat app.

[0758] 2. Getting and Sending User Input

[0759] The terminal receives input from the user (such as a message or question) and sends the content to the server. This operation transfers the user's message to the server.

[0760] Example: A user types "What day is it today?" and the device sends the message to the server.

[0761] 3. View the response

[0762] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0763] Example: The device displays the response from the server: "Today is World Environment Day."

[0764] User Action:

[0765] 1. Start a conversation

[0766] Users initiate conversations through the device interface and can start conversations at any time.

[0767] Example: A user types "Good morning" into a terminal.

[0768] 2. Enter your message

[0769] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[0770] Example: A user types "What's in the news today?" and presses the send button.

[0771] 3. Check the response and enter the next message

[0772] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[0773] Example: User continues with "Thank you, is there anything else?"

[0774] As described above, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural, emotion-sensitive conversations, which can simultaneously reduce the burden on caregivers and provide emotional support to users.

[0775] The processing flow will be explained below.

[0776] Server processing steps:

[0777] Step 1:

[0778] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, so that the AI ​​and emotion engine are ready to converse with the user.

[0779] Step 2:

[0780] The server receives the user's conversation request sent from the device, which includes a message or question typed by the user.

[0781] Step 3:

[0782] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, thereby understanding the user's intent and determining how to respond.

[0783] Step 4:

[0784] At the same time, the server uses an emotion engine to analyze the emotions from the user's message, which clarifies whether the user has positive, negative, or neutral emotions.

[0785] Step 5:

[0786] The server combines the analysis results of the NLU module and the emotion engine and passes them to the natural language generation (NLG) module, which provides information for generating emotion-sensitive responses.

[0787] Step 6:

[0788] The server uses the NLG module to generate an appropriate response based on the integrated information, which is formulated in natural language and is emotionally relevant to the user.

[0789] Step 7:

[0790] The server generates a response and sends it to the device, allowing the conversation to continue.

[0791] Step 8:

[0792] The server stores the conversation history and sentiment analysis results in a database, allowing the generative AI model and sentiment engine to continuously learn and be more effective in the next conversation.

[0793] Terminal processing steps:

[0794] Step 1:

[0795] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0796] Step 2:

[0797] The device receives messages and questions entered by the user and sends them to the server. This operation transfers the user's messages to the server.

[0798] Step 3:

[0799] The terminal receives the response from the server and displays it on the user interface, allowing the user to see the system's response and enter the next message.

[0800] User steps:

[0801] Step 1:

[0802] The user initiates the conversation through the terminal interface, entering the initial message and beginning a dialogue with the system.

[0803] Step 2:

[0804] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[0805] Step 3:

[0806] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[0807] Through the above processing steps, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural and emotion-sensitive conversations. This system can simultaneously reduce the burden on caregivers and provide emotional support to users.

[0808] Example 2

[0809] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0810] Conventional communication systems in nursing care facilities have difficulty in communicating with users while taking their emotions into full consideration, which makes it difficult to increase the psychological satisfaction of users. Furthermore, there are few systems that support continuous learning by effectively utilizing conversation history, so long-term improvement cannot be expected.

[0811] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for initializing and loading the generative AI model and emotion engine, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, means for analyzing emotions based on the analysis results, natural language generation means for generating a response based on the analysis results and the emotion analysis results, means for sending the generated response to the user, and means for saving the conversation history and the emotion analysis results in a database and using them for continuous learning. This enables natural conversation that takes user emotions into consideration, and continuous learning can be expected to improve the quality of the conversation.

[0812] A "generative AI model" is an artificial intelligence model that generates responses in natural language based on input from the user.

[0813] An "emotion engine" is a means of analyzing emotions from user input data and generating a response according to those emotions.

[0814] "Initialization" refers to the operation of setting a program or system into an operational state.

[0815] "Load" refers to the operation of reading a program or data into memory.

[0816] A "conversation request" refers to a message or question that a user inputs to the system.

[0817] "Natural language understanding" is a technology that analyzes the content and context of received conversation requests and understands their meaning.

[0818] "Natural language generation" is a technology that generates responses in natural language based on analyzed information.

[0819] "Send" refers to the act of transferring generated data or responses to another device or terminal.

[0820] "Conversation history" refers to a record of all interactions between a user and a system.

[0821] "Continuous learning" is a learning process that improves the system's performance based on past conversation history and new data.

[0822] "User interface" refers to the display screen and input devices that allow a user to interact with a system.

[0823] "Terminal" refers to a device for communicating with the system and displaying responses.

[0824] MODE FOR CARRYING OUT THE INVENTION

[0825] This invention is a system that combines a generative AI model and an emotion engine to resolve the lack of conversation in nursing care facilities and realize natural, emotion-sensitive dialogue.

[0826] Server Processing

[0827] When the system starts up, the server initializes the generative AI model and emotion engine and loads the necessary data and settings into memory. Specifically, it loads the parameters of a pre-trained generative AI model such as GPT-4 and the emotion engine that performs emotion analysis into memory. This prepares the server to respond immediately to conversations with users.

[0828] The server receives requests from users. These requests include messages or questions entered by the user, which are sent from the device to the server. The received messages are parsed by a natural language understanding (NLU) module, which tokenizes the messages and analyzes their content and context. At the same time, an emotion engine calculates an emotion value (e.g., positive, negative, neutral) based on the message tokens.

[0829] Based on the analysis results, the server generates a response using a natural language generation (NLG) module. The generative AI model uses the user's message and sentiment value as prompts to predict the next conversation flow and generate a response. The generated response is formatted in natural language and sent to the user.

[0830] Conversation history and sentiment analysis results are stored in a database and used as input for continuous learning. This process allows the generative AI model and emotion engine to improve performance with each use, resulting in more appropriate and natural responses.

[0831] Terminal handling

[0832] The terminal displays an interface through which the user can interact with the system, including a chat window and a text entry field where the user can type a message. After the user types a message, the terminal sends it to the server.

[0833] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to enter the next message based on the displayed response and progress the conversation.

[0834] User operations

[0835] A user can initiate a conversation at any time through the device's interface. For example, the user can type "Good morning" into the device and press the send button. The user can then use the interface to type and send questions or messages. For example, the user can type and send "Please tell me the news today."

[0836] The response generated by the server is displayed on the terminal, and after checking it, the user can input the next message, enabling continuous dialogue. For example, the user can input a message such as "Thank you, is there anything else?"

[0837] Specific examples

[0838] Specific examples are shown below.

[0839] When a user types "Are there any events today?" into their device, the server receives the message and analyzes it. Based on the analysis results and the emotion value, the generative AI model generates a response such as "There will be music activities this afternoon." This response is sent to the device and displayed to the user.

[0840] An example prompt is:

[0841] "Is there any event today?"

[0842] "Tell me about what happened yesterday."

[0843] "Please tell me the latest news."

[0844] As described above, this system, which includes a generative AI model, an emotion engine, and a natural language understanding and generation module, is designed to facilitate conversations in nursing homes. By conducting conversations in a natural and emotionally sensitive manner, it can improve the psychological satisfaction of users and reduce the burden on caregivers.

[0845] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0846] Step 1: Initialize and load the model

[0847] When the system starts up, the server initializes the parameters of the generative AI model (e.g., GPT-4) and emotion engine and loads them into memory. Specifically, it calls the initialize_model() function. The input of this step is the trained data for the generative AI model and emotion engine, and the output is the initialized model and engine deployed in memory.

[0848] Step 2: Request acceptance

[0849] The server receives a conversation request from the terminal. It uses the receive_request(connection) method to receive a data packet containing a message or question entered by the user. The input is the message from the user, and the output is the received conversation request.

[0850] Step 3: Natural Language Understanding and Sentiment Analysis

[0851] The server passes the received message to the Natural Language Understanding (NLU) module for content and context analysis. It also performs sentiment analysis using the emotion engine. Specifically, nlu_module.analyze(text) performs tokenization and syntactic analysis, and emotion_engine.analyze(text) calculates the sentiment value. The input to this step is the received conversation request, and the output is the analysis result and sentiment value.

[0852] Step 4: Generate a response

[0853] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analysis result and emotion value. It generates a response by calling the nlg_module.generate_response(context, emotion) method. The input is the analysis result and emotion value, and the output is the generated response sentence.

[0854] Step 5: Sending a Response

[0855] The server sends the generated response to the terminal using the send_response(response, client) method. The input is the generated response text, and the output is the response sent to the terminal.

[0856] Step 6: Save conversation history and continue learning

[0857] The server saves the conversation history and sentiment analysis results in a database for continuous learning. Specifically, it uses the save_to_database(user_id, conversation_data) method. The input of this step is the conversation history and sentiment analysis results, and the output is the data saved in the database.

[0858] Step 7: Displaying the User Interface

[0859] The terminal displays an interface for the user to interact with the system. Specifically, the display_interface() function is called to display the chat window. The input is the interface design loaded when the system started, and the output is the chat screen displayed to the user.

[0860] Step 8: Getting and Sending User Input

[0861] The terminal gets input from the user and sends the contents to the server. get_user_input() receives user input and send_to_server(input_text) sends it to the server. The input is the text entered by the user and the output is the message sent to the server.

[0862] Step 9: View the response

[0863] The terminal receives the response sent from the server and displays it on the user interface. The response is received with receive_from_server() and updated with update_interface(response). The input is the response from the server, and the output is the response displayed to the user.

[0864] Through these steps, natural, emotion-sensitive dialogue is realized using a generative AI model and emotion engine. Appropriate processing is performed based on the input data at each step, and the resulting output data is passed on to the next step. The user can engage in natural dialogue through their device, and the system can generate responses that are in tune with the user's emotions.

[0865] (Application example 2)

[0866] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0867] Customer service is a major issue in modern brick-and-mortar stores. In particular, when customers ask about products, store clerks may lack sufficient knowledge. While a response that takes into consideration the customer's feelings is required, not all store clerks have the skills to do so. This can lead to a decline in customer satisfaction. The present invention aims to solve these issues and realize natural, emotionally sensitive conversations in brick-and-mortar stores.

[0868] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0869] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, means for extracting information about products and explaining the products in a natural conversational format, means for generating responses according to emotions and taking customer emotions into consideration, and means for analyzing customer reactions and questions in real time. This enables improvement in the quality of customer service in physical stores and improvement in customer satisfaction.

[0870] A "generative AI model" refers to an artificial intelligence technology that is trained and learned in advance using large amounts of data to understand human language and generate responses.

[0871] "Means to take emotions into consideration" refers to the function of analyzing emotions from the user's input message and generating an appropriate response based on those emotions.

[0872] "Natural language understanding means" refers to a function for analyzing input messages from users and understanding their content and intent.

[0873] "Natural language generator" refers to a function that generates responses in natural language based on analyzed user input.

[0874] "User interface" refers to the collection of screens and input devices that allow a user to interact with a system.

[0875] A "conversation request" refers to a question or message that a user inputs to the system.

[0876] "Means for receiving" refers to a function for taking in conversation requests input from the outside.

[0877] "Means for generating a response" refers to the function of generating an appropriate reply based on a received conversation request.

[0878] "Means for sending" refers to a function for sending the generated response back to the user.

[0879] "Conversation history" refers to data that stores the contents of past conversations exchanged between a user and the system.

[0880] A "database" refers to a digital storage device that systematically stores information such as conversation history.

[0881] "Means for continuous learning" refers to a function that allows the AI ​​model to further learn based on saved conversation history, improving the accuracy and quality of responses.

[0882] "Product Information" refers to detailed information about a specific product, such as its characteristics, usage, and price.

[0883] "Means of analysis" refers to the function of analyzing input data in various ways and understanding its meaning and intent.

[0884] The present invention relates to a system that uses a generative AI model to improve the efficiency of customer service in brick-and-mortar stores and provide natural conversations that take emotions into consideration. Specific embodiments are described below.

[0885] Server Processing

[0886] The server performs the following process.

[0887] 1. Initializing and loading the model

[0888] At system startup, the server initializes the generative AI model (e.g., GPT-4) and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[0889] 2. Receiving a conversation request

[0890] The server receives a conversation request from the terminal, which includes a message or question entered by the customer.

[0891] 3. Natural Language Understanding and Sentiment Analysis

[0892] The server passes the received message to a natural language understanding module to analyze the content and context, and also analyzes the customer's emotions using an emotion engine, thereby accurately understanding the customer's intentions and emotions.

[0893] 4. Generating the Response

[0894] The server uses a natural language generation module to generate an appropriate response based on the analyzed information and the sentiment analysis results. The generated response is written in natural language and takes into consideration the customer's sentiment.

[0895] 5. Sending the Response

[0896] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[0897] 6. Conversation history storage and continuous learning

[0898] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn customer characteristics, preferences, and sentiment trends over time.

[0899] Terminal handling

[0900] The terminal performs the following process.

[0901] 1. Displaying the user interface

[0902] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the customer to type messages.

[0903] 2. Capture and submit customer input

[0904] The terminal receives input from the customer (such as a message or question) and sends the content to the server. This operation transfers the customer's message to the server.

[0905] 3. View the response

[0906] The terminal receives the response sent from the server and displays it on the user interface, allowing the customer to check the system's response and enter the next message.

[0907] Hardware and software used

[0908] The system is implemented using the following hardware and software:

[0909] Hardware: Smartphones, smart glasses, head-mounted displays

[0910] Software: Flask (web server), OpenAI GPT-4 (generative AI model), TextBlob (sentiment analysis)

[0911] Specific examples

[0912] 1. A customer asks through smart glasses, "Tell me about this shirt."

[0913] The server generates a response such as "This shirt has a modern design and is highly breathable" and sends it to the device, which displays the response on the smart glasses.

[0914] 2. A customer types, "I'm tired from work today."

[0915] The server generates a response that takes into account the user's emotions, such as "Thank you for your hard work. Would you like something to help you relax?", and sends it to the device. The device then displays this response on its smartphone.

[0916] Prompt Sentence Examples

[0917] Customer: Tell me about this shirt.

[0918] clerk:

[0919] Customer: I'm tired from work today.

[0920] clerk:

[0921] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0922] Step 1:

[0923] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, which prepares the generative AI model and emotion engine to generate customer responses. The input is data from the pre-trained AI model, and the output is the system configuration information loaded into memory.

[0924] Step 2:

[0925] A user inputs a conversation request through a terminal. Specifically, the user inputs a question or message into the interface (e.g., a chat window) of a smartphone or smart glasses. The input is the user's text message, and the output is the transfer of the request from the terminal to the server.

[0926] Step 3:

[0927] The server passes the conversation request received from the device to a natural language understanding module, which analyzes the content and context. At the same time, it uses an emotion engine to analyze the user's emotions. The input is the user's text message, and the output is the analysis results of the message's content, context, and emotions. Natural language processing and emotion analysis technologies are used for the analysis.

[0928] Step 4:

[0929] The server uses a natural language generation module based on the analysis results to generate an appropriate response. A generative AI model (e.g., GPT-4) is used to form a natural-language response based on the user's intent and sentiment. The input is the analysis results of the message's content, context, and sentiment, and the output is the generated text response.

[0930] Step 5:

[0931] The server sends the generated response to the terminal, which allows the interaction to reach the user and proceed to the next step. The input is the generated text response, and the output is the display of the response in the user interface.

[0932] Step 6:

[0933] The terminal displays the response received from the server on the user interface. The user checks the displayed response and prepares to enter the next message. The input is the text response sent from the server, and the output is the text displayed on the screen of the smartphone or smart glasses.

[0934] Step 7:

[0935] Through the user interface, the user inputs the next message and sends it back to the server. This cycle continues, creating a continuous dialogue: the input is the user's new message, and the output is again a request to be parsed and a response generated.

[0936] Step 8:

[0937] The server stores the conversation history and emotion analysis results in a database and uses them for continuous learning of the generative AI model and emotion engine. This allows the system to learn customer characteristics, preferences, and emotional trends over time and use them for the next interaction. The input is the conversation history and emotion data to be saved, and the output is the information accumulated in the database.

[0938] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0939] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0940] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0941] [Third embodiment]

[0942] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0943] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0944] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0945] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0946] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0947] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0948] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0949] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0950] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0951] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0952] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0953] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0954] The present invention is a service system dedicated to conversation using a generative AI model, and specific embodiments are described below.

[0955] Server Action:

[0956] 1. Initializing and loading the model

[0957] The server initializes the generative AI model at system startup and loads the necessary data and settings into memory, making the AI ​​ready to converse with the user.

[0958] Example: A server loads a pre-trained generative AI model (e.g., GPT-4).

[0959] 2. Request acceptance

[0960] The server receives a conversation request sent from the terminal, which includes a message or question entered by the user.

[0961] Example: The server receives user input saying "The weather is nice today."

[0962] 3. Natural language understanding

[0963] The server passes the received message to a natural language understanding module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[0964] Example: The server reads positive sentiment from the message "The weather is nice."

[0965] 4. Generating the Response

[0966] The server uses a natural language generation module to generate appropriate responses based on the analyzed information. The responses are formulated in natural language, enabling smooth conversation with the user.

[0967] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[0968] 5. Sending the Response

[0969] The server generates a response and sends it back to the terminal, where it is displayed to the user and the dialogue proceeds.

[0970] Example: The server generates a response and sends it to the device.

[0971] 6. Conversation history storage and continuous learning

[0972] The server stores the conversation history in a database and uses it to continuously train the AI ​​model, allowing it to learn about the user's characteristics and preferences over time.

[0973] Example: The server stores the conversation history in a database to help with the next interaction.

[0974] Terminal handling:

[0975] 1. Displaying the user interface

[0976] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[0977] Example: A tablet device displays a chat app.

[0978] 2. Getting and Sending User Input

[0979] The terminal receives input from the user (such as a message or question) and sends the content to the server.

[0980] Example: A user types "What day is it today?" and the device sends the message to the server.

[0981] 3. View the response

[0982] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[0983] Example: The device displays the response from the server: "Today is World Environment Day."

[0984] User Action:

[0985] 1. Start a conversation

[0986] Users initiate conversations through the device interface and can start conversations at any time.

[0987] Example: A user types "Good morning" into a terminal.

[0988] 2. Enter your message

[0989] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[0990] Example: A user types "What's in the news today?" and presses the send button.

[0991] 3. Check the response and enter the next message

[0992] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[0993] Example: User continues with "Thank you, is there anything else?"

[0994] As described above, by combining a generative AI model with multiple processing means, the present invention enables natural and continuous conversation, thereby reducing the burden on caregivers and eliminating feelings of loneliness among users.

[0995] The processing flow will be explained below.

[0996] Server processing steps:

[0997] Step 1:

[0998] The server initializes the generative AI model and loads the necessary data and settings into memory, so the AI ​​is ready to converse with the user.

[0999] Step 2:

[1000] The server receives a conversation request from the device, which includes a message or question typed by the user.

[1001] Step 3:

[1002] The server passes the received message to a natural language understanding (NLU) module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[1003] Step 4:

[1004] Based on the analysis results, the server uses a natural language generation (NLG) module to generate an appropriate response, which is formulated in natural language.

[1005] Step 5:

[1006] The server then sends the generated response to the terminal, allowing the conversation to proceed.

[1007] Step 6:

[1008] The server stores the conversation history in a database and uses it to continuously train the generative AI model, allowing the model to learn about the user's characteristics and preferences over time.

[1009] Terminal processing steps:

[1010] Step 1:

[1011] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1012] Step 2:

[1013] The device receives input from the user (such as a message or question) and sends it to the server, which then forwards the user's message to the server.

[1014] Step 3:

[1015] The terminal receives the response sent by the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[1016] User steps:

[1017] Step 1:

[1018] Users initiate conversations through the device interface, and conversations can begin at any time.

[1019] Step 2:

[1020] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[1021] Step 3:

[1022] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[1023] Example 1

[1024] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1025] Existing conversational systems using generative AI models have difficulty accurately understanding the user's intentions and emotions and continuously learning to achieve more natural dialogue. Furthermore, they lack a means to effectively utilize dialogue history and smoothly advance ongoing dialogue with the user. As a result, users sometimes experience unnatural and fragmented dialogue, resulting in reduced satisfaction with the system.

[1026] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1027] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, natural language generation means for generating a response based on the analysis result, means for sending the generated response to the user, means for saving the conversation history in a database and using it for continuous learning, means for loading necessary data and settings of the generative AI model into memory when the system starts up, means for analyzing the emotion and context of a message from the user to understand the user's intention, and means for generating a response written in natural language. This makes it possible to accurately understand the user's intention and emotion, realize natural conversations, and improve the quality of conversations with the user through continuous learning.

[1028] A "generative AI model" is a model that can generate text using artificial intelligence, and is used for response generation and text completion in natural language processing.

[1029] "Initialization and loading means" refers to the methods or processes for initializing a generative AI model at system startup and loading the necessary data and settings from storage into memory.

[1030] "Means for receiving conversation requests" refers to the method or protocol by which the system receives text messages or questions entered by the user.

[1031] "Natural language understanding" refers to technologies and algorithms that analyze the content, sentiment, and context of received text messages to understand the user's intent.

[1032] "Natural language generation means" refers to technologies and algorithms that generate appropriate responses based on information obtained through natural language understanding.

[1033] "Means for sending a response to a user" refers to a data transmission method or protocol for sending a generated response back to a user to continue the dialogue.

[1034] "Means for storing conversation history in a database" refers to technologies and methods for persistently storing the content of conversations with users and using them to improve the quality of future interactions.

[1035] "Means used for continuous learning" refers to the learning process of retraining the generative AI model based on stored conversation history to improve response accuracy and the naturalness of the dialogue.

[1036] "Means for loading the necessary data and settings for the generative AI model into memory when the system starts up" refers to a method for loading the files and setting information necessary for the generative AI model into memory space when the system starts up.

[1037] "Means for understanding a user's intent by analyzing the emotion and context of messages from the user" refers to technologies and algorithms for analyzing the emotion and context from the user's input text and understanding their intent.

[1038] "Means for generating responses formulated in natural language" refers to techniques or processes for generating responses to a user in a natural language expression.

[1039] This invention relates to a conversation system that uses a generative AI model, and specifically has a configuration in which three elements - a server, a terminal, and a user - function in cooperation with each other.

[1040] 1. Server Processing

[1041] The server initializes the generative AI model (e.g., GPT-4, a model specialized for natural language processing) at system startup and loads the necessary data and settings into memory. The server listens to and receives conversation requests sent from the device, and when a message is received, it is passed to a natural language understanding (NLU) module for analysis of content, sentiment, and context.

[1042] Based on the analyzed information, the server uses a natural language generation (NLG) module to generate an appropriate response. The response is then sent back to the device via the server. Furthermore, the conversation history is stored in a database and used for continuous learning. This allows the generative AI model to improve its accuracy over time, enabling it to provide answers that are tailored to the user's characteristics and preferences.

[1043] Specific examples

[1044] When the server starts up, it loads the GPT-4 model and reads the configuration file into memory.

[1045] The server receives a message from the user saying, "The weather is nice today," and the NLU module analyzes the positive emotion as "the weather is nice."

[1046] The server generates a response saying, "Today is indeed a perfect day to go out," and sends it to the terminal.

[1047] The server stores the conversation history in a database for future use.

[1048] 2. Terminal Processing

[1049] The terminal displays an interface such as a chat window that allows the user to interact with the system. The user can enter messages through the interface. The terminal obtains input from the user and sends the content to the server. The terminal receives responses sent from the server and displays them in the user interface to provide feedback to the user.

[1050] Specific examples

[1051] The tablet device launches a chat app and displays a chat window.

[1052] The user types "What day is it today?" and the device sends the message to the server.

[1053] The terminal displays the response "Today is World Environment Day" received from the server and prompts the user to confirm it.

[1054] 3. User Operation

[1055] Users can freely start a conversation through the interface on their device. A conversation begins by entering a message or question in the chat window and sending it. By checking the response sent back from the server and entering the next message based on that, a natural, continuous conversation is realized.

[1056] Specific examples

[1057] A user opens a chat app and types "Good morning."

[1058] Check the server's response, "Good morning! How can I help you today?", then type, "What's the news today?"

[1059] Look at the generated response, "Today's news is..." and continue with more detailed questions.

[1060] Prompt Sentence Examples

[1061] Below are some specific examples of prompt sentences to input to the generative AI model.

[1062] "Tell me the news today."

[1063] "How's the weather?"

[1064] "What are some popular movies these days?"

[1065] "Tell me your ideas for our next vacation."

[1066] "Please recommend some books."

[1067] This system, which uses a generative AI model (e.g., GPT-4) to enable natural, continuous conversations with users, was developed with the aim of reducing the burden on caregivers and eliminating feelings of loneliness among users.

[1068] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1069] Server processing steps:

[1070] Step 1:

[1071] Initializing and Loading the Model

[1072] Input: System Start Trigger

[1073] Processing: The server initializes the generative AI model (e.g., GPT-4), which involves loading the model file and configuration file from disk into memory.

[1074] Output: An initialized generative AI model

[1075] Specific operation: The server "loads the generative AI model file from disk." For example, "checks the model parameter settings."

[1076] Step 2:

[1077] Request reception

[1078] Input: conversation request from device

[1079] Processing: The server listens on a port for requests and receives conversational requests via HTTP or WebSocket protocols.

[1080] Output: Received user message

[1081] Specific action: The server "listens for requests on port 8080." For example, "receives a message from the user saying 'The weather is nice today.'"

[1082] Step 3:

[1083] natural language understanding

[1084] Input: Received user message

[1085] Processing: The server passes the message to a Natural Language Understanding (NLU) module for content, sentiment, and context analysis, which helps understand the user's intent and sentiment.

[1086] Output: Analysis results (intent, sentiment, context)

[1087] Specific operation: The server sends the user's message to the NLU engine. For example, it determines the positive sentiment from the message 'The weather is good'.

[1088] Step 4:

[1089] Generating a response

[1090] Input: Analysis result of NLU module

[1091] Processing: The server uses a natural language generation (NLG) module to generate an appropriate response based on the parsed information.

[1092] Output: The generated response message

[1093] Specific action: The server generates a response using an NLG engine. For example, it generates a response such as, "Today is indeed a perfect day to go out."

[1094] Step 5:

[1095] Sending a Response

[1096] Input: The generated response message

[1097] Processing: The server generates a response and sends it to the terminal as an HTTP response.

[1098] Output: Response message sent to the terminal

[1099] Specific action: The server "returns the generated response as an HTTP response."

[1100] Step 6:

[1101] Conversation history storage and continuous learning

[1102] Input: Conversation history (user messages and server responses)

[1103] Processing: The server stores the conversation history in a database and uses it to continuously train the generative AI model.

[1104] Output: Saved conversation history, updated generative AI model

[1105] Specific operation: The server "records the conversation history in a database." For example, "saves the content of the conversation with the user."

[1106] Terminal processing steps:

[1107] Step 1:

[1108] User Interface Display

[1109] Input: User-initiated interface launch

[1110] Processing: The terminal displays an interface (such as a chat window) that allows the user to interact with the system.

[1111] Output: Chat window displayed

[1112] Specific action: The device "launches a chat app and displays the screen." For example, "displays a chat input field."

[1113] Step 2:

[1114] Getting and Sending User Input

[1115] Input: Message input from the user

[1116] Processing: The terminal detects input from the user and sends a message to the server.

[1117] Output: User message sent

[1118] Specific action: The device "detects input from the user." For example, "sends the message 'Hello' to the server."

[1119] Step 3:

[1120] Viewing the response

[1121] Input: Response message from the server

[1122] Processing: The terminal receives the response sent by the server and displays it on the user interface.

[1123] Output: The displayed response message

[1124] Specific action: The device "receives the response from the server and displays it on the screen." For example, "displays the response 'What a great day today' in the chat window."

[1125] User steps:

[1126] Step 1:

[1127] Start a conversation

[1128] Input: Chat interface launched by user action

[1129] Process: The user initiates a conversation through the interface on the device.

[1130] Output: User's first message

[1131] Specific action: The user "opens a chat app." For example, "types 'Good morning.'"

[1132] Step 2:

[1133] Enter your message

[1134] Input: Message entered into the user interface

[1135] Process: The user enters a message or question through the interface and presses the send button.

[1136] Output: Message sent to the server

[1137] Specific action: The user "types a question into the chat window." For example, "types 'What's the weather like today?' and presses the send button."

[1138] Step 3:

[1139] Check the response and enter the next message

[1140] Input: Response from the server displayed on the terminal

[1141] Action: The user checks the response from the server and enters the next message based on it.

[1142] Output: The following message

[1143] Specific action: The user "reads the server's response." For example, "after seeing the response 'It's sunny today,' he / she re-enters the question."

[1144] (Application example 1)

[1145] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1146] Physical stores require systems that allow customers to quickly and efficiently obtain product information and ask questions. However, conventional methods require in-store staff to respond to each customer individually, which can lead to delayed responses or insufficient explanations. In addition, it can be difficult for customers to instantly obtain the information they want, which can lead to a decrease in customer satisfaction. To solve these issues, it is necessary to develop a system that uses AI technology to quickly and accurately respond to customers' questions.

[1147] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1148] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, and means for customers to obtain product information and ask questions using a user interface. This allows customers to instantly obtain the product information they need, reducing the burden on in-store staff and improving customer satisfaction.

[1149] A "generative AI model" is an artificial intelligence model that generates natural responses based on sentences entered by the user.

[1150] "Means for initializing and loading" refers to the function of loading the generated AI model and necessary data and settings into memory when the system starts up.

[1151] The "means for receiving conversation requests from users" is a function for sending messages and questions entered by users to the server.

[1152] "Natural language understanding means" is a function that analyzes received user messages and understands their content, emotions, and context.

[1153] The "natural language generation means" is a function that generates an appropriate response in natural language based on the analyzed information.

[1154] The "means for sending to the user" is a function for sending the generated response back to the user's terminal.

[1155] "Means for saving conversation history in a database and using it for continuous learning" is a function that saves conversation history with the user and uses it to improve the accuracy of the model.

[1156] A "user interface" is a function that displays screens and widgets that allow users to interact with the system and accepts user input.

[1157] "Means for customers to obtain product information and ask questions" refers to functions that allow customers to obtain product information and ask questions via smartphones or smart glasses.

[1158] In this invention, we build a system that uses generative AI models to realize smart concierge services in physical stores.

[1159] Server Processing

[1160] The server has the means to initialize and load the generative AI model, loading the model and necessary data into memory. This step prepares the server for a conversation with the user.

[1161] A conversation request from a user is received by the server using a receiving means. The received request includes a message or question, which the server analyzes using a natural language understanding means to understand the content, emotion, and context. Then, based on the analysis results, a natural language generation means generates an appropriate response, which is then sent to the user using a transmission means. This response is displayed on the user's device.

[1162] Additionally, by storing conversation history in a database and providing a means for continuous learning, the server can learn about the user's characteristics and preferences over time and generate more appropriate responses.

[1163] Terminal handling

[1164] The terminal has a means for displaying a user interface, allowing the user to input messages. The terminal receives input from the user and sends it to the server. The response sent from the server is again displayed to the user through the terminal's interface.

[1165] This user interface is used by customers to obtain product information and ask questions, allowing customers to instantly obtain the information they need and reducing the burden on store staff.

[1166] Specific Examples

[1167] The server uses a server machine equipped with a high-performance GPU, and the terminals can be smartphones or smart glasses. The software library used is the transformers library from Hugging Face.

[1168] For example, if a customer uses their smartphone to type, "Good morning. What products do you recommend?", the generative AI model will analyze the information and respond with, "Good morning. Today's recommendation is our newly released smartwatch. Would you like to hear more about it?" In this way, customers can easily obtain information, making in-store customer service more efficient.

[1169] Prompt Sentence Examples

[1170] Here is an example of a real prompt:

[1171] "Good morning. What products do you recommend?"

[1172] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1173] Step 1:

[1174] The server initializes the generative AI model at system startup, loading the necessary data and configuration into memory. At this stage, the AI ​​model is ready to converse with the user. Specifically, a server machine equipped with a powerful GPU loads a pre-trained generative AI model (e.g., GPT-4) using Hugging Face's transformers library. The input is the model and associated configuration data, and the output is the model in its ready state.

[1175] Step 2:

[1176] The user initiates a dialogue with the system using a user interface on a smartphone or smart glasses. Specifically, the user types "Good morning. What products do you recommend?" into the interface. The device receives this message in real time and sends it to the server. The input is the user's message, and the output is a transmission request to the server.

[1177] Step 3:

[1178] The server receives a conversation request from the user. Specifically, the server receives a message sent by the user and passes the content to the natural language understanding module for analysis. The input is the message sent by the user, and the output is data to the understanding module.

[1179] Step 4:

[1180] The server analyzes the received message using natural language understanding. Specifically, the server analyzes the message content, sentiment, and context using a natural language understanding module (e.g., BERT) to understand the user's intent. The input is the user's message data, and the output is the analyzed information.

[1181] Step 5:

[1182] The server generates an appropriate response using natural language generation means based on the analysis results. Specifically, the server uses a natural language generation module (e.g., GPT-4) to generate a natural response from the analyzed information. The input is the analysis results, and the output is the generated response sentence.

[1183] Step 6:

[1184] The server sends the generated response to the user. Specifically, the server sends the generated response to the terminal, which displays it on the user interface. The input is the generated response sentence, and the output is a message displayed on the user's terminal.

[1185] Step 7:

[1186] The terminal displays the response received from the server on the user interface. Specifically, the terminal displays the message received from the server on the interface and notifies the user. The input is the response message from the server, and the output is the display on the user interface.

[1187] Step 8:

[1188] The server stores the conversation history in a database and uses it for continuous learning. Specifically, the server records all conversation history in a database and uses it as learning data to improve the accuracy of future conversations. The input is the conversation history data, and the output is an updated database and a trained model.

[1189] This process will enable a smart concierge service to be realized in physical stores, allowing customers to smoothly obtain product information and have questions answered using their smartphones or smart glasses.

[1190] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1191] The present invention combines an emotion engine with a system that uses generative AI models to resolve conversational shortcomings in nursing homes, thereby achieving more natural and emotionally sensitive dialogue. The following describes specific embodiments.

[1192] Server Action:

[1193] 1. Initializing and loading the model

[1194] At system startup, the server initializes the generative AI model and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[1195] Example: A server loads a pre-trained generative AI model (e.g., GPT-4) and an emotion engine.

[1196] 2. Request acceptance

[1197] The server receives a conversation request from the terminal, which includes a message or question entered by the user.

[1198] Example: The server receives user input saying "The weather is nice today."

[1199] 3. Natural Language Understanding and Sentiment Analysis

[1200] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, and also analyzes the user's emotions using an emotion engine, thereby accurately understanding the user's intentions and emotions.

[1201] Example: The server reads positive sentiment from the message "The weather is nice."

[1202] 4. Generating the Response

[1203] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analyzed information and the results of sentiment analysis. The response is formulated in natural language and takes into account the user's sentiment.

[1204] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[1205] 5. Sending the Response

[1206] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[1207] Example: The server generates a response and sends it to the device.

[1208] 6. Conversation history storage and continuous learning

[1209] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn user characteristics, preferences, and sentiment trends over time.

[1210] Example: The server stores the conversation history in a database to help with the next interaction.

[1211] Terminal handling:

[1212] 1. Displaying the user interface

[1213] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1214] Example: A tablet device displays a chat app.

[1215] 2. Getting and Sending User Input

[1216] The terminal receives input from the user (such as a message or question) and sends the content to the server. This operation transfers the user's message to the server.

[1217] Example: A user types "What day is it today?" and the device sends the message to the server.

[1218] 3. View the response

[1219] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[1220] Example: The device displays the response from the server: "Today is World Environment Day."

[1221] User Action:

[1222] 1. Start a conversation

[1223] Users initiate conversations through the device interface and can start conversations at any time.

[1224] Example: A user types "Good morning" into a terminal.

[1225] 2. Enter your message

[1226] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[1227] Example: A user types "What's in the news today?" and presses the send button.

[1228] 3. Check the response and enter the next message

[1229] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[1230] Example: User continues with "Thank you, is there anything else?"

[1231] As described above, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural, emotion-sensitive conversations, which can simultaneously reduce the burden on caregivers and provide emotional support to users.

[1232] The processing flow will be explained below.

[1233] Server processing steps:

[1234] Step 1:

[1235] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, so that the AI ​​and emotion engine are ready to converse with the user.

[1236] Step 2:

[1237] The server receives the user's conversation request sent from the device, which includes a message or question typed by the user.

[1238] Step 3:

[1239] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, thereby understanding the user's intent and determining how to respond.

[1240] Step 4:

[1241] At the same time, the server uses an emotion engine to analyze the emotions from the user's message, which clarifies whether the user has positive, negative, or neutral emotions.

[1242] Step 5:

[1243] The server combines the analysis results of the NLU module and the emotion engine and passes them to the natural language generation (NLG) module, which provides information for generating emotion-sensitive responses.

[1244] Step 6:

[1245] The server uses the NLG module to generate an appropriate response based on the integrated information, which is formulated in natural language and is emotionally relevant to the user.

[1246] Step 7:

[1247] The server generates a response and sends it to the device, allowing the conversation to continue.

[1248] Step 8:

[1249] The server stores the conversation history and sentiment analysis results in a database, allowing the generative AI model and sentiment engine to continuously learn and be more effective in the next conversation.

[1250] Terminal processing steps:

[1251] Step 1:

[1252] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1253] Step 2:

[1254] The device receives messages and questions entered by the user and sends them to the server. This operation transfers the user's messages to the server.

[1255] Step 3:

[1256] The terminal receives the response from the server and displays it on the user interface, allowing the user to see the system's response and enter the next message.

[1257] User steps:

[1258] Step 1:

[1259] The user initiates the conversation through the terminal interface, entering the initial message and beginning a dialogue with the system.

[1260] Step 2:

[1261] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[1262] Step 3:

[1263] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[1264] Through the above processing steps, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural and emotion-sensitive conversations. This system can simultaneously reduce the burden on caregivers and provide emotional support to users.

[1265] Example 2

[1266] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1267] Conventional communication systems in nursing care facilities have difficulty in communicating with users while taking their emotions into full consideration, which makes it difficult to increase the psychological satisfaction of users. Furthermore, there are few systems that support continuous learning by effectively utilizing conversation history, so long-term improvement cannot be expected.

[1268] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for initializing and loading the generative AI model and emotion engine, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, means for analyzing emotions based on the analysis results, natural language generation means for generating a response based on the analysis results and the emotion analysis results, means for sending the generated response to the user, and means for saving the conversation history and the emotion analysis results in a database and using them for continuous learning. This enables natural conversation that takes user emotions into consideration, and continuous learning can be expected to improve the quality of the conversation.

[1269] A "generative AI model" is an artificial intelligence model that generates responses in natural language based on input from the user.

[1270] An "emotion engine" is a means of analyzing emotions from user input data and generating a response according to those emotions.

[1271] "Initialization" refers to the operation of setting a program or system into an operational state.

[1272] "Load" refers to the operation of reading a program or data into memory.

[1273] A "conversation request" refers to a message or question that a user inputs to the system.

[1274] "Natural language understanding" is a technology that analyzes the content and context of received conversation requests and understands their meaning.

[1275] "Natural language generation" is a technology that generates responses in natural language based on analyzed information.

[1276] "Send" refers to the act of transferring generated data or responses to another device or terminal.

[1277] "Conversation history" refers to a record of all interactions between a user and a system.

[1278] "Continuous learning" is a learning process that improves the system's performance based on past conversation history and new data.

[1279] "User interface" refers to the display screen and input devices that allow a user to interact with a system.

[1280] "Terminal" refers to a device for communicating with the system and displaying responses.

[1281] MODE FOR CARRYING OUT THE INVENTION

[1282] This invention is a system that combines a generative AI model and an emotion engine to resolve the lack of conversation in nursing care facilities and realize natural, emotion-sensitive dialogue.

[1283] Server Processing

[1284] When the system starts up, the server initializes the generative AI model and emotion engine and loads the necessary data and settings into memory. Specifically, it loads the parameters of a pre-trained generative AI model such as GPT-4 and the emotion engine that performs emotion analysis into memory. This prepares the server to respond immediately to conversations with users.

[1285] The server receives requests from users. These requests include messages or questions entered by the user, which are sent from the device to the server. The received messages are parsed by a natural language understanding (NLU) module, which tokenizes the messages and analyzes their content and context. At the same time, an emotion engine calculates an emotion value (e.g., positive, negative, neutral) based on the message tokens.

[1286] Based on the analysis results, the server generates a response using a natural language generation (NLG) module. The generative AI model uses the user's message and sentiment value as prompts to predict the next conversation flow and generate a response. The generated response is formatted in natural language and sent to the user.

[1287] Conversation history and sentiment analysis results are stored in a database and used as input for continuous learning. This process allows the generative AI model and emotion engine to improve performance with each use, resulting in more appropriate and natural responses.

[1288] Terminal handling

[1289] The terminal displays an interface through which the user can interact with the system, including a chat window and a text entry field where the user can type a message. After the user types a message, the terminal sends it to the server.

[1290] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to enter the next message based on the displayed response and progress the conversation.

[1291] User operations

[1292] A user can initiate a conversation at any time through the device's interface. For example, the user can type "Good morning" into the device and press the send button. The user can then use the interface to type and send questions or messages. For example, the user can type and send "Please tell me the news today."

[1293] The response generated by the server is displayed on the terminal, and after checking it, the user can input the next message, enabling continuous dialogue. For example, the user can input a message such as "Thank you, is there anything else?"

[1294] Specific examples

[1295] Specific examples are shown below.

[1296] When a user types "Are there any events today?" into their device, the server receives the message and analyzes it. Based on the analysis results and the emotion value, the generative AI model generates a response such as "There will be music activities this afternoon." This response is sent to the device and displayed to the user.

[1297] An example prompt is:

[1298] "Is there any event today?"

[1299] "Tell me about what happened yesterday."

[1300] "Please tell me the latest news."

[1301] As described above, this system, which includes a generative AI model, an emotion engine, and a natural language understanding and generation module, is designed to facilitate conversations in nursing homes. By conducting conversations in a natural and emotionally sensitive manner, it can improve the psychological satisfaction of users and reduce the burden on caregivers.

[1302] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1303] Step 1: Initialize and load the model

[1304] When the system starts up, the server initializes the parameters of the generative AI model (e.g., GPT-4) and emotion engine and loads them into memory. Specifically, it calls the initialize_model() function. The input of this step is the trained data for the generative AI model and emotion engine, and the output is the initialized model and engine deployed in memory.

[1305] Step 2: Request acceptance

[1306] The server receives a conversation request from the terminal. It uses the receive_request(connection) method to receive a data packet containing a message or question entered by the user. The input is the message from the user, and the output is the received conversation request.

[1307] Step 3: Natural Language Understanding and Sentiment Analysis

[1308] The server passes the received message to the Natural Language Understanding (NLU) module for content and context analysis. It also performs sentiment analysis using the emotion engine. Specifically, nlu_module.analyze(text) performs tokenization and syntactic analysis, and emotion_engine.analyze(text) calculates the sentiment value. The input to this step is the received conversation request, and the output is the analysis result and sentiment value.

[1309] Step 4: Generate a response

[1310] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analysis result and emotion value. It generates a response by calling the nlg_module.generate_response(context, emotion) method. The input is the analysis result and emotion value, and the output is the generated response sentence.

[1311] Step 5: Sending a Response

[1312] The server sends the generated response to the terminal using the send_response(response, client) method. The input is the generated response text, and the output is the response sent to the terminal.

[1313] Step 6: Save conversation history and continue learning

[1314] The server saves the conversation history and sentiment analysis results in a database for continuous learning. Specifically, it uses the save_to_database(user_id, conversation_data) method. The input of this step is the conversation history and sentiment analysis results, and the output is the data saved in the database.

[1315] Step 7: Displaying the User Interface

[1316] The terminal displays an interface for the user to interact with the system. Specifically, the display_interface() function is called to display the chat window. The input is the interface design loaded when the system started, and the output is the chat screen displayed to the user.

[1317] Step 8: Getting and Sending User Input

[1318] The terminal gets input from the user and sends the contents to the server. get_user_input() receives user input and send_to_server(input_text) sends it to the server. The input is the text entered by the user and the output is the message sent to the server.

[1319] Step 9: View the response

[1320] The terminal receives the response sent from the server and displays it on the user interface. The response is received with receive_from_server() and updated with update_interface(response). The input is the response from the server, and the output is the response displayed to the user.

[1321] Through these steps, natural, emotion-sensitive dialogue is realized using a generative AI model and emotion engine. Appropriate processing is performed based on the input data at each step, and the resulting output data is passed on to the next step. The user can engage in natural dialogue through their device, and the system can generate responses that are in tune with the user's emotions.

[1322] (Application example 2)

[1323] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1324] Customer service is a major issue in modern brick-and-mortar stores. In particular, when customers ask about products, store clerks may lack sufficient knowledge. While a response that takes into consideration the customer's feelings is required, not all store clerks have the skills to do so. This can lead to a decline in customer satisfaction. The present invention aims to solve these issues and realize natural, emotionally sensitive conversations in brick-and-mortar stores.

[1325] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1326] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, means for extracting information about products and explaining the products in a natural conversational format, means for generating responses according to emotions and taking customer emotions into consideration, and means for analyzing customer reactions and questions in real time. This enables improvement in the quality of customer service in physical stores and improvement in customer satisfaction.

[1327] A "generative AI model" refers to an artificial intelligence technology that is trained and learned in advance using large amounts of data to understand human language and generate responses.

[1328] "Means to take emotions into consideration" refers to the function of analyzing emotions from the user's input message and generating an appropriate response based on those emotions.

[1329] "Natural language understanding means" refers to a function for analyzing input messages from users and understanding their content and intent.

[1330] "Natural language generator" refers to a function that generates responses in natural language based on analyzed user input.

[1331] "User interface" refers to the collection of screens and input devices that allow a user to interact with a system.

[1332] A "conversation request" refers to a question or message that a user inputs to the system.

[1333] "Means for receiving" refers to a function for taking in conversation requests input from the outside.

[1334] "Means for generating a response" refers to the function of generating an appropriate reply based on a received conversation request.

[1335] "Means for sending" refers to a function for sending the generated response back to the user.

[1336] "Conversation history" refers to data that stores the contents of past conversations exchanged between a user and the system.

[1337] A "database" refers to a digital storage device that systematically stores information such as conversation history.

[1338] "Means for continuous learning" refers to a function that allows the AI ​​model to further learn based on saved conversation history, improving the accuracy and quality of responses.

[1339] "Product Information" refers to detailed information about a specific product, such as its characteristics, usage, and price.

[1340] "Means of analysis" refers to the function of analyzing input data in various ways and understanding its meaning and intent.

[1341] The present invention relates to a system that uses a generative AI model to improve the efficiency of customer service in brick-and-mortar stores and provide natural conversations that take emotions into consideration. Specific embodiments are described below.

[1342] Server Processing

[1343] The server performs the following process.

[1344] 1. Initializing and loading the model

[1345] At system startup, the server initializes the generative AI model (e.g., GPT-4) and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[1346] 2. Receiving a conversation request

[1347] The server receives a conversation request from the terminal, which includes a message or question entered by the customer.

[1348] 3. Natural Language Understanding and Sentiment Analysis

[1349] The server passes the received message to a natural language understanding module to analyze the content and context, and also analyzes the customer's emotions using an emotion engine, thereby accurately understanding the customer's intentions and emotions.

[1350] 4. Generating the Response

[1351] The server uses a natural language generation module to generate an appropriate response based on the analyzed information and the sentiment analysis results. The generated response is written in natural language and takes into consideration the customer's sentiment.

[1352] 5. Sending the Response

[1353] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[1354] 6. Conversation history storage and continuous learning

[1355] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn customer characteristics, preferences, and sentiment trends over time.

[1356] Terminal handling

[1357] The terminal performs the following process.

[1358] 1. Displaying the user interface

[1359] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the customer to type messages.

[1360] 2. Capture and submit customer input

[1361] The terminal receives input from the customer (such as a message or question) and sends the content to the server. This operation transfers the customer's message to the server.

[1362] 3. View the response

[1363] The terminal receives the response sent from the server and displays it on the user interface, allowing the customer to check the system's response and enter the next message.

[1364] Hardware and software used

[1365] The system is implemented using the following hardware and software:

[1366] Hardware: Smartphones, smart glasses, head-mounted displays

[1367] Software: Flask (web server), OpenAI GPT-4 (generative AI model), TextBlob (sentiment analysis)

[1368] Specific examples

[1369] 1. A customer asks through smart glasses, "Tell me about this shirt."

[1370] The server generates a response such as "This shirt has a modern design and is highly breathable" and sends it to the device, which displays the response on the smart glasses.

[1371] 2. A customer types, "I'm tired from work today."

[1372] The server generates a response that takes into account the user's emotions, such as "Thank you for your hard work. Would you like something to help you relax?", and sends it to the device. The device then displays this response on its smartphone.

[1373] Prompt Sentence Examples

[1374] Customer: Tell me about this shirt.

[1375] clerk:

[1376] Customer: I'm tired from work today.

[1377] clerk:

[1378] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1379] Step 1:

[1380] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, which prepares the generative AI model and emotion engine to generate customer responses. The input is data from the pre-trained AI model, and the output is the system configuration information loaded into memory.

[1381] Step 2:

[1382] A user inputs a conversation request through a terminal. Specifically, the user inputs a question or message into the interface (e.g., a chat window) of a smartphone or smart glasses. The input is the user's text message, and the output is the transfer of the request from the terminal to the server.

[1383] Step 3:

[1384] The server passes the conversation request received from the device to a natural language understanding module, which analyzes the content and context. At the same time, it uses an emotion engine to analyze the user's emotions. The input is the user's text message, and the output is the analysis results of the message's content, context, and emotions. Natural language processing and emotion analysis technologies are used for the analysis.

[1385] Step 4:

[1386] The server uses a natural language generation module based on the analysis results to generate an appropriate response. A generative AI model (e.g., GPT-4) is used to form a natural-language response based on the user's intent and sentiment. The input is the analysis results of the message's content, context, and sentiment, and the output is the generated text response.

[1387] Step 5:

[1388] The server sends the generated response to the terminal, which allows the interaction to reach the user and proceed to the next step. The input is the generated text response, and the output is the display of the response in the user interface.

[1389] Step 6:

[1390] The terminal displays the response received from the server on the user interface. The user checks the displayed response and prepares to enter the next message. The input is the text response sent from the server, and the output is the text displayed on the screen of the smartphone or smart glasses.

[1391] Step 7:

[1392] Through the user interface, the user inputs the next message and sends it back to the server. This cycle continues, creating a continuous dialogue: the input is the user's new message, and the output is again a request to be parsed and a response generated.

[1393] Step 8:

[1394] The server stores the conversation history and emotion analysis results in a database and uses them for continuous learning of the generative AI model and emotion engine. This allows the system to learn customer characteristics, preferences, and emotional trends over time and use them for the next interaction. The input is the conversation history and emotion data to be saved, and the output is the information accumulated in the database.

[1395] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1396] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1397] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1398] [Fourth embodiment]

[1399] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1400] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1401] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1402] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1403] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1404] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1405] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1406] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1407] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1408] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1409] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1410] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1411] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1412] The present invention is a service system dedicated to conversation using a generative AI model, and specific embodiments are described below.

[1413] Server Action:

[1414] 1. Initializing and loading the model

[1415] The server initializes the generative AI model at system startup and loads the necessary data and settings into memory, making the AI ​​ready to converse with the user.

[1416] Example: A server loads a pre-trained generative AI model (e.g., GPT-4).

[1417] 2. Request acceptance

[1418] The server receives a conversation request sent from the terminal, which includes a message or question entered by the user.

[1419] Example: The server receives user input saying "The weather is nice today."

[1420] 3. Natural language understanding

[1421] The server passes the received message to a natural language understanding module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[1422] Example: The server reads positive sentiment from the message "The weather is nice."

[1423] 4. Generating the Response

[1424] The server uses a natural language generation module to generate appropriate responses based on the analyzed information. The responses are formulated in natural language, enabling smooth conversation with the user.

[1425] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[1426] 5. Sending the Response

[1427] The server generates a response and sends it back to the terminal, where it is displayed to the user and the dialogue proceeds.

[1428] Example: The server generates a response and sends it to the device.

[1429] 6. Conversation history storage and continuous learning

[1430] The server stores the conversation history in a database and uses it to continuously train the AI ​​model, allowing it to learn about the user's characteristics and preferences over time.

[1431] Example: The server stores the conversation history in a database to help with the next interaction.

[1432] Terminal handling:

[1433] 1. Displaying the user interface

[1434] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1435] Example: A tablet device displays a chat app.

[1436] 2. Getting and Sending User Input

[1437] The terminal receives input from the user (such as a message or question) and sends the content to the server.

[1438] Example: A user types "What day is it today?" and the device sends the message to the server.

[1439] 3. View the response

[1440] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[1441] Example: The device displays the response from the server: "Today is World Environment Day."

[1442] User Action:

[1443] 1. Start a conversation

[1444] Users initiate conversations through the device interface and can start conversations at any time.

[1445] Example: A user types "Good morning" into a terminal.

[1446] 2. Enter your message

[1447] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[1448] Example: A user types "What's in the news today?" and presses the send button.

[1449] 3. Check the response and enter the next message

[1450] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[1451] Example: User continues with "Thank you, is there anything else?"

[1452] As described above, by combining a generative AI model with multiple processing means, the present invention enables natural and continuous conversation, thereby reducing the burden on caregivers and eliminating feelings of loneliness among users.

[1453] The processing flow will be explained below.

[1454] Server processing steps:

[1455] Step 1:

[1456] The server initializes the generative AI model and loads the necessary data and settings into memory, so the AI ​​is ready to converse with the user.

[1457] Step 2:

[1458] The server receives a conversation request from the device, which includes a message or question typed by the user.

[1459] Step 3:

[1460] The server passes the received message to a natural language understanding (NLU) module that analyzes the content, sentiment, and context, thereby understanding the user's intent and emotions.

[1461] Step 4:

[1462] Based on the analysis results, the server uses a natural language generation (NLG) module to generate an appropriate response, which is formulated in natural language.

[1463] Step 5:

[1464] The server then sends the generated response to the terminal, allowing the conversation to proceed.

[1465] Step 6:

[1466] The server stores the conversation history in a database and uses it to continuously train the generative AI model, allowing the model to learn about the user's characteristics and preferences over time.

[1467] Terminal processing steps:

[1468] Step 1:

[1469] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1470] Step 2:

[1471] The device receives input from the user (such as a message or question) and sends it to the server, which then forwards the user's message to the server.

[1472] Step 3:

[1473] The terminal receives the response sent by the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[1474] User steps:

[1475] Step 1:

[1476] Users initiate conversations through the device interface, and conversations can begin at any time.

[1477] Step 2:

[1478] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[1479] Step 3:

[1480] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[1481] Example 1

[1482] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1483] Existing conversational systems using generative AI models have difficulty accurately understanding the user's intentions and emotions and continuously learning to achieve more natural dialogue. Furthermore, they lack a means to effectively utilize dialogue history and smoothly advance ongoing dialogue with the user. As a result, users sometimes experience unnatural and fragmented dialogue, resulting in reduced satisfaction with the system.

[1484] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1485] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, natural language generation means for generating a response based on the analysis result, means for sending the generated response to the user, means for saving the conversation history in a database and using it for continuous learning, means for loading necessary data and settings of the generative AI model into memory when the system starts up, means for analyzing the emotion and context of a message from the user to understand the user's intention, and means for generating a response written in natural language. This makes it possible to accurately understand the user's intention and emotion, realize natural conversations, and improve the quality of conversations with the user through continuous learning.

[1486] A "generative AI model" is a model that can generate text using artificial intelligence, and is used for response generation and text completion in natural language processing.

[1487] "Initialization and loading means" refers to the methods or processes for initializing a generative AI model at system startup and loading the necessary data and settings from storage into memory.

[1488] "Means for receiving conversation requests" refers to the method or protocol by which the system receives text messages or questions entered by the user.

[1489] "Natural language understanding" refers to technologies and algorithms that analyze the content, sentiment, and context of received text messages to understand the user's intent.

[1490] "Natural language generation means" refers to technologies and algorithms that generate appropriate responses based on information obtained through natural language understanding.

[1491] "Means for sending a response to a user" refers to a data transmission method or protocol for sending a generated response back to a user to continue the dialogue.

[1492] "Means for storing conversation history in a database" refers to technologies and methods for persistently storing the content of conversations with users and using them to improve the quality of future interactions.

[1493] "Means used for continuous learning" refers to the learning process of retraining the generative AI model based on stored conversation history to improve response accuracy and the naturalness of the dialogue.

[1494] "Means for loading the necessary data and settings for the generative AI model into memory when the system starts up" refers to a method for loading the files and setting information necessary for the generative AI model into memory space when the system starts up.

[1495] "Means for understanding a user's intent by analyzing the emotion and context of messages from the user" refers to technologies and algorithms for analyzing the emotion and context from the user's input text and understanding their intent.

[1496] "Means for generating responses formulated in natural language" refers to techniques or processes for generating responses to a user in a natural language expression.

[1497] This invention relates to a conversation system that uses a generative AI model, and specifically has a configuration in which three elements - a server, a terminal, and a user - function in cooperation with each other.

[1498] 1. Server Processing

[1499] The server initializes the generative AI model (e.g., GPT-4, a model specialized for natural language processing) at system startup and loads the necessary data and settings into memory. The server listens to and receives conversation requests sent from the device, and when a message is received, it is passed to a natural language understanding (NLU) module for analysis of content, sentiment, and context.

[1500] Based on the analyzed information, the server uses a natural language generation (NLG) module to generate an appropriate response. The response is then sent back to the device via the server. Furthermore, the conversation history is stored in a database and used for continuous learning. This allows the generative AI model to improve its accuracy over time, enabling it to provide answers that are tailored to the user's characteristics and preferences.

[1501] Specific examples

[1502] When the server starts up, it loads the GPT-4 model and reads the configuration file into memory.

[1503] The server receives a message from the user saying, "The weather is nice today," and the NLU module analyzes the positive emotion as "the weather is nice."

[1504] The server generates a response saying, "Today is indeed a perfect day to go out," and sends it to the terminal.

[1505] The server stores the conversation history in a database for future use.

[1506] 2. Terminal Processing

[1507] The terminal displays an interface such as a chat window that allows the user to interact with the system. The user can enter messages through the interface. The terminal obtains input from the user and sends the content to the server. The terminal receives responses sent from the server and displays them in the user interface to provide feedback to the user.

[1508] Specific examples

[1509] The tablet device launches a chat app and displays a chat window.

[1510] The user types "What day is it today?" and the device sends the message to the server.

[1511] The terminal displays the response "Today is World Environment Day" received from the server and prompts the user to confirm it.

[1512] 3. User Operation

[1513] Users can freely start a conversation through the interface on their device. A conversation begins by entering a message or question in the chat window and sending it. By checking the response sent back from the server and entering the next message based on that, a natural, continuous conversation is realized.

[1514] Specific examples

[1515] A user opens a chat app and types "Good morning."

[1516] Check the server's response, "Good morning! How can I help you today?", then type, "What's the news today?"

[1517] Look at the generated response, "Today's news is..." and continue with more detailed questions.

[1518] Prompt Sentence Examples

[1519] Below are some specific examples of prompt sentences to input to the generative AI model.

[1520] "Tell me the news today."

[1521] "How's the weather?"

[1522] "What are some popular movies these days?"

[1523] "Tell me your ideas for our next vacation."

[1524] "Please recommend some books."

[1525] This system, which uses a generative AI model (e.g., GPT-4) to enable natural, continuous conversations with users, was developed with the aim of reducing the burden on caregivers and eliminating feelings of loneliness among users.

[1526] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1527] Server processing steps:

[1528] Step 1:

[1529] Initializing and Loading the Model

[1530] Input: System Start Trigger

[1531] Processing: The server initializes the generative AI model (e.g., GPT-4), which involves loading the model file and configuration file from disk into memory.

[1532] Output: An initialized generative AI model

[1533] Specific operation: The server "loads the generative AI model file from disk." For example, "checks the model parameter settings."

[1534] Step 2:

[1535] Request reception

[1536] Input: conversation request from device

[1537] Processing: The server listens on a port for requests and receives conversational requests via HTTP or WebSocket protocols.

[1538] Output: Received user message

[1539] Specific action: The server "listens for requests on port 8080." For example, "receives a message from the user saying 'The weather is nice today.'"

[1540] Step 3:

[1541] natural language understanding

[1542] Input: Received user message

[1543] Processing: The server passes the message to a Natural Language Understanding (NLU) module for content, sentiment, and context analysis, which helps understand the user's intent and sentiment.

[1544] Output: Analysis results (intent, sentiment, context)

[1545] Specific operation: The server sends the user's message to the NLU engine. For example, it determines the positive sentiment from the message 'The weather is good'.

[1546] Step 4:

[1547] Generating a response

[1548] Input: Analysis result of NLU module

[1549] Processing: The server uses a natural language generation (NLG) module to generate an appropriate response based on the parsed information.

[1550] Output: The generated response message

[1551] Specific action: The server generates a response using an NLG engine. For example, it generates a response such as, "Today is indeed a perfect day to go out."

[1552] Step 5:

[1553] Sending a Response

[1554] Input: The generated response message

[1555] Processing: The server generates a response and sends it to the terminal as an HTTP response.

[1556] Output: Response message sent to the terminal

[1557] Specific action: The server "returns the generated response as an HTTP response."

[1558] Step 6:

[1559] Conversation history storage and continuous learning

[1560] Input: Conversation history (user messages and server responses)

[1561] Processing: The server stores the conversation history in a database and uses it to continuously train the generative AI model.

[1562] Output: Saved conversation history, updated generative AI model

[1563] Specific operation: The server "records the conversation history in a database." For example, "saves the content of the conversation with the user."

[1564] Terminal processing steps:

[1565] Step 1:

[1566] User Interface Display

[1567] Input: User-initiated interface launch

[1568] Processing: The terminal displays an interface (such as a chat window) that allows the user to interact with the system.

[1569] Output: Chat window displayed

[1570] Specific action: The device "launches a chat app and displays the screen." For example, "displays a chat input field."

[1571] Step 2:

[1572] Getting and Sending User Input

[1573] Input: Message input from the user

[1574] Processing: The terminal detects input from the user and sends a message to the server.

[1575] Output: User message sent

[1576] Specific action: The device "detects input from the user." For example, "sends the message 'Hello' to the server."

[1577] Step 3:

[1578] Viewing the response

[1579] Input: Response message from the server

[1580] Processing: The terminal receives the response sent by the server and displays it on the user interface.

[1581] Output: The displayed response message

[1582] Specific action: The device "receives the response from the server and displays it on the screen." For example, "displays the response 'What a great day today' in the chat window."

[1583] User steps:

[1584] Step 1:

[1585] Start a conversation

[1586] Input: Chat interface launched by user action

[1587] Process: The user initiates a conversation through the interface on the device.

[1588] Output: User's first message

[1589] Specific action: The user "opens a chat app." For example, "types 'Good morning.'"

[1590] Step 2:

[1591] Enter your message

[1592] Input: Message entered into the user interface

[1593] Process: The user enters a message or question through the interface and presses the send button.

[1594] Output: Message sent to the server

[1595] Specific action: The user "types a question into the chat window." For example, "types 'What's the weather like today?' and presses the send button."

[1596] Step 3:

[1597] Check the response and enter the next message

[1598] Input: Response from the server displayed on the terminal

[1599] Action: The user checks the response from the server and enters the next message based on it.

[1600] Output: The following message

[1601] Specific action: The user "reads the server's response." For example, "after seeing the response 'It's sunny today,' he / she re-enters the question."

[1602] (Application example 1)

[1603] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1604] Physical stores require systems that allow customers to quickly and efficiently obtain product information and ask questions. However, conventional methods require in-store staff to respond to each customer individually, which can lead to delayed responses or insufficient explanations. In addition, it can be difficult for customers to instantly obtain the information they want, which can lead to a decrease in customer satisfaction. To solve these issues, it is necessary to develop a system that uses AI technology to quickly and accurately respond to customers' questions.

[1605] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1606] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, and means for customers to obtain product information and ask questions using a user interface. This allows customers to instantly obtain the product information they need, reducing the burden on in-store staff and improving customer satisfaction.

[1607] A "generative AI model" is an artificial intelligence model that generates natural responses based on sentences entered by the user.

[1608] "Means for initializing and loading" refers to the function of loading the generated AI model and necessary data and settings into memory when the system starts up.

[1609] The "means for receiving conversation requests from users" is a function for sending messages and questions entered by users to the server.

[1610] "Natural language understanding means" is a function that analyzes received user messages and understands their content, emotions, and context.

[1611] The "natural language generation means" is a function that generates an appropriate response in natural language based on the analyzed information.

[1612] The "means for sending to the user" is a function for sending the generated response back to the user's terminal.

[1613] "Means for saving conversation history in a database and using it for continuous learning" is a function that saves conversation history with the user and uses it to improve the accuracy of the model.

[1614] A "user interface" is a function that displays screens and widgets that allow users to interact with the system and accepts user input.

[1615] "Means for customers to obtain product information and ask questions" refers to functions that allow customers to obtain product information and ask questions via smartphones or smart glasses.

[1616] In this invention, we build a system that uses generative AI models to realize smart concierge services in physical stores.

[1617] Server Processing

[1618] The server has the means to initialize and load the generative AI model, loading the model and necessary data into memory. This step prepares the server for a conversation with the user.

[1619] A conversation request from a user is received by the server using a receiving means. The received request includes a message or question, which the server analyzes using a natural language understanding means to understand the content, emotion, and context. Then, based on the analysis results, a natural language generation means generates an appropriate response, which is then sent to the user using a transmission means. This response is displayed on the user's device.

[1620] Additionally, by storing conversation history in a database and providing a means for continuous learning, the server can learn about the user's characteristics and preferences over time and generate more appropriate responses.

[1621] Terminal handling

[1622] The terminal has a means for displaying a user interface, allowing the user to input messages. The terminal receives input from the user and sends it to the server. The response sent from the server is again displayed to the user through the terminal's interface.

[1623] This user interface is used by customers to obtain product information and ask questions, allowing customers to instantly obtain the information they need and reducing the burden on store staff.

[1624] Specific Examples

[1625] The server uses a server machine equipped with a high-performance GPU, and the terminals can be smartphones or smart glasses. The software library used is the transformers library from Hugging Face.

[1626] For example, if a customer uses their smartphone to type, "Good morning. What products do you recommend?", the generative AI model will analyze the information and respond with, "Good morning. Today's recommendation is our newly released smartwatch. Would you like to hear more about it?" In this way, customers can easily obtain information, making in-store customer service more efficient.

[1627] Prompt Sentence Examples

[1628] Here is an example of a real prompt:

[1629] "Good morning. What products do you recommend?"

[1630] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1631] Step 1:

[1632] The server initializes the generative AI model at system startup, loading the necessary data and configuration into memory. At this stage, the AI ​​model is ready to converse with the user. Specifically, a server machine equipped with a powerful GPU loads a pre-trained generative AI model (e.g., GPT-4) using Hugging Face's transformers library. The input is the model and associated configuration data, and the output is the model in its ready state.

[1633] Step 2:

[1634] The user initiates a dialogue with the system using a user interface on a smartphone or smart glasses. Specifically, the user types "Good morning. What products do you recommend?" into the interface. The device receives this message in real time and sends it to the server. The input is the user's message, and the output is a transmission request to the server.

[1635] Step 3:

[1636] The server receives a conversation request from the user. Specifically, the server receives a message sent by the user and passes the content to the natural language understanding module for analysis. The input is the message sent by the user, and the output is data to the understanding module.

[1637] Step 4:

[1638] The server analyzes the received message using natural language understanding. Specifically, the server analyzes the message content, sentiment, and context using a natural language understanding module (e.g., BERT) to understand the user's intent. The input is the user's message data, and the output is the analyzed information.

[1639] Step 5:

[1640] The server generates an appropriate response using natural language generation means based on the analysis results. Specifically, the server uses a natural language generation module (e.g., GPT-4) to generate a natural response from the analyzed information. The input is the analysis results, and the output is the generated response sentence.

[1641] Step 6:

[1642] The server sends the generated response to the user. Specifically, the server sends the generated response to the terminal, which displays it on the user interface. The input is the generated response sentence, and the output is a message displayed on the user's terminal.

[1643] Step 7:

[1644] The terminal displays the response received from the server on the user interface. Specifically, the terminal displays the message received from the server on the interface and notifies the user. The input is the response message from the server, and the output is the display on the user interface.

[1645] Step 8:

[1646] The server stores the conversation history in a database and uses it for continuous learning. Specifically, the server records all conversation history in a database and uses it as learning data to improve the accuracy of future conversations. The input is the conversation history data, and the output is an updated database and a trained model.

[1647] This process will enable a smart concierge service to be realized in physical stores, allowing customers to smoothly obtain product information and have questions answered using their smartphones or smart glasses.

[1648] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1649] The present invention combines an emotion engine with a system that uses generative AI models to resolve conversational shortcomings in nursing homes, thereby achieving more natural and emotionally sensitive dialogue. The following describes specific embodiments.

[1650] Server Action:

[1651] 1. Initializing and loading the model

[1652] At system startup, the server initializes the generative AI model and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[1653] Example: A server loads a pre-trained generative AI model (e.g., GPT-4) and an emotion engine.

[1654] 2. Request acceptance

[1655] The server receives a conversation request from the terminal, which includes a message or question entered by the user.

[1656] Example: The server receives user input saying "The weather is nice today."

[1657] 3. Natural Language Understanding and Sentiment Analysis

[1658] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, and also analyzes the user's emotions using an emotion engine, thereby accurately understanding the user's intentions and emotions.

[1659] Example: The server reads positive sentiment from the message "The weather is nice."

[1660] 4. Generating the Response

[1661] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analyzed information and the results of sentiment analysis. The response is formulated in natural language and takes into account the user's sentiment.

[1662] Example: The server generates a response such as "Today is indeed a perfect day to go out."

[1663] 5. Sending the Response

[1664] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[1665] Example: The server generates a response and sends it to the device.

[1666] 6. Conversation history storage and continuous learning

[1667] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn user characteristics, preferences, and sentiment trends over time.

[1668] Example: The server stores the conversation history in a database to help with the next interaction.

[1669] Terminal handling:

[1670] 1. Displaying the user interface

[1671] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1672] Example: A tablet device displays a chat app.

[1673] 2. Getting and Sending User Input

[1674] The terminal receives input from the user (such as a message or question) and sends the content to the server. This operation transfers the user's message to the server.

[1675] Example: A user types "What day is it today?" and the device sends the message to the server.

[1676] 3. View the response

[1677] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to check the system's response and enter the next message.

[1678] Example: The device displays the response from the server: "Today is World Environment Day."

[1679] User Action:

[1680] 1. Start a conversation

[1681] Users initiate conversations through the device interface and can start conversations at any time.

[1682] Example: A user types "Good morning" into a terminal.

[1683] 2. Enter your message

[1684] The user uses the interface to input a message or question, and when they are finished, they press the send button to send the message to the server.

[1685] Example: A user types "What's in the news today?" and presses the send button.

[1686] 3. Check the response and enter the next message

[1687] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, thereby enabling continuous conversation.

[1688] Example: User continues with "Thank you, is there anything else?"

[1689] As described above, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural, emotion-sensitive conversations, which can simultaneously reduce the burden on caregivers and provide emotional support to users.

[1690] The processing flow will be explained below.

[1691] Server processing steps:

[1692] Step 1:

[1693] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, so that the AI ​​and emotion engine are ready to converse with the user.

[1694] Step 2:

[1695] The server receives the user's conversation request sent from the device, which includes a message or question typed by the user.

[1696] Step 3:

[1697] The server passes the received message to a natural language understanding (NLU) module to analyze the content and context, thereby understanding the user's intent and determining how to respond.

[1698] Step 4:

[1699] At the same time, the server uses an emotion engine to analyze the emotions from the user's message, which clarifies whether the user has positive, negative, or neutral emotions.

[1700] Step 5:

[1701] The server combines the analysis results of the NLU module and the emotion engine and passes them to the natural language generation (NLG) module, which provides information for generating emotion-sensitive responses.

[1702] Step 6:

[1703] The server uses the NLG module to generate an appropriate response based on the integrated information, which is formulated in natural language and is emotionally relevant to the user.

[1704] Step 7:

[1705] The server generates a response and sends it to the device, allowing the conversation to continue.

[1706] Step 8:

[1707] The server stores the conversation history and sentiment analysis results in a database, allowing the generative AI model and sentiment engine to continuously learn and be more effective in the next conversation.

[1708] Terminal processing steps:

[1709] Step 1:

[1710] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the user to type messages.

[1711] Step 2:

[1712] The device receives messages and questions entered by the user and sends them to the server. This operation transfers the user's messages to the server.

[1713] Step 3:

[1714] The terminal receives the response from the server and displays it on the user interface, allowing the user to see the system's response and enter the next message.

[1715] User steps:

[1716] Step 1:

[1717] The user initiates the conversation through the terminal interface, entering the initial message and beginning a dialogue with the system.

[1718] Step 2:

[1719] Users use the interface to enter messages or questions, and when they're done, they press the send button, which sends the message to the server.

[1720] Step 3:

[1721] The user checks the response from the server displayed on the terminal and inputs the next message based on that response, allowing for continuous conversation.

[1722] Through the above processing steps, the present invention combines a generative AI model, an emotion engine, and multiple processing means to realize natural and emotion-sensitive conversations. This system can simultaneously reduce the burden on caregivers and provide emotional support to users.

[1723] Example 2

[1724] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1725] Conventional communication systems in nursing care facilities have difficulty in communicating with users while taking their emotions into full consideration, which makes it difficult to increase the psychological satisfaction of users. Furthermore, there are few systems that support continuous learning by effectively utilizing conversation history, so long-term improvement cannot be expected.

[1726] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for initializing and loading the generative AI model and emotion engine, means for receiving a conversation request from a user, natural language understanding means for analyzing the received conversation request, means for analyzing emotions based on the analysis results, natural language generation means for generating a response based on the analysis results and the emotion analysis results, means for sending the generated response to the user, and means for saving the conversation history and the emotion analysis results in a database and using them for continuous learning. This enables natural conversation that takes user emotions into consideration, and continuous learning can be expected to improve the quality of the conversation.

[1727] A "generative AI model" is an artificial intelligence model that generates responses in natural language based on input from the user.

[1728] An "emotion engine" is a means of analyzing emotions from user input data and generating a response according to those emotions.

[1729] "Initialization" refers to the operation of setting a program or system into an operational state.

[1730] "Load" refers to the operation of reading a program or data into memory.

[1731] A "conversation request" refers to a message or question that a user inputs to the system.

[1732] "Natural language understanding" is a technology that analyzes the content and context of received conversation requests and understands their meaning.

[1733] "Natural language generation" is a technology that generates responses in natural language based on analyzed information.

[1734] "Send" refers to the act of transferring generated data or responses to another device or terminal.

[1735] "Conversation history" refers to a record of all interactions between a user and a system.

[1736] "Continuous learning" is a learning process that improves the system's performance based on past conversation history and new data.

[1737] "User interface" refers to the display screen and input devices that allow a user to interact with a system.

[1738] "Terminal" refers to a device for communicating with the system and displaying responses.

[1739] MODE FOR CARRYING OUT THE INVENTION

[1740] This invention is a system that combines a generative AI model and an emotion engine to resolve the lack of conversation in nursing care facilities and realize natural, emotion-sensitive dialogue.

[1741] Server Processing

[1742] When the system starts up, the server initializes the generative AI model and emotion engine and loads the necessary data and settings into memory. Specifically, it loads the parameters of a pre-trained generative AI model such as GPT-4 and the emotion engine that performs emotion analysis into memory. This prepares the server to respond immediately to conversations with users.

[1743] The server receives requests from users. These requests include messages or questions entered by the user, which are sent from the device to the server. The received messages are parsed by a natural language understanding (NLU) module, which tokenizes the messages and analyzes their content and context. At the same time, an emotion engine calculates an emotion value (e.g., positive, negative, neutral) based on the message tokens.

[1744] Based on the analysis results, the server generates a response using a natural language generation (NLG) module. The generative AI model uses the user's message and sentiment value as prompts to predict the next conversation flow and generate a response. The generated response is formatted in natural language and sent to the user.

[1745] Conversation history and sentiment analysis results are stored in a database and used as input for continuous learning. This process allows the generative AI model and emotion engine to improve performance with each use, resulting in more appropriate and natural responses.

[1746] Terminal handling

[1747] The terminal displays an interface through which the user can interact with the system, including a chat window and a text entry field where the user can type a message. After the user types a message, the terminal sends it to the server.

[1748] The terminal receives the response sent from the server and displays it on the user interface, allowing the user to enter the next message based on the displayed response and progress the conversation.

[1749] User operations

[1750] A user can initiate a conversation at any time through the device's interface. For example, the user can type "Good morning" into the device and press the send button. The user can then use the interface to type and send questions or messages. For example, the user can type and send "Please tell me the news today."

[1751] The response generated by the server is displayed on the terminal, and after checking it, the user can input the next message, enabling continuous dialogue. For example, the user can input a message such as "Thank you, is there anything else?"

[1752] Specific examples

[1753] Specific examples are shown below.

[1754] When a user types "Are there any events today?" into their device, the server receives the message and analyzes it. Based on the analysis results and the emotion value, the generative AI model generates a response such as "There will be music activities this afternoon." This response is sent to the device and displayed to the user.

[1755] An example prompt is:

[1756] "Is there any event today?"

[1757] "Tell me about what happened yesterday."

[1758] "Please tell me the latest news."

[1759] As described above, this system, which includes a generative AI model, an emotion engine, and a natural language understanding and generation module, is designed to facilitate conversations in nursing homes. By conducting conversations in a natural and emotionally sensitive manner, it can improve the psychological satisfaction of users and reduce the burden on caregivers.

[1760] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1761] Step 1: Initialize and load the model

[1762] When the system starts up, the server initializes the parameters of the generative AI model (e.g., GPT-4) and emotion engine and loads them into memory. Specifically, it calls the initialize_model() function. The input of this step is the trained data for the generative AI model and emotion engine, and the output is the initialized model and engine deployed in memory.

[1763] Step 2: Request acceptance

[1764] The server receives a conversation request from the terminal. It uses the receive_request(connection) method to receive a data packet containing a message or question entered by the user. The input is the message from the user, and the output is the received conversation request.

[1765] Step 3: Natural Language Understanding and Sentiment Analysis

[1766] The server passes the received message to the Natural Language Understanding (NLU) module for content and context analysis. It also performs sentiment analysis using the emotion engine. Specifically, nlu_module.analyze(text) performs tokenization and syntactic analysis, and emotion_engine.analyze(text) calculates the sentiment value. The input to this step is the received conversation request, and the output is the analysis result and sentiment value.

[1767] Step 4: Generate a response

[1768] The server uses a natural language generation (NLG) module to generate an appropriate response based on the analysis result and emotion value. It generates a response by calling the nlg_module.generate_response(context, emotion) method. The input is the analysis result and emotion value, and the output is the generated response sentence.

[1769] Step 5: Sending a Response

[1770] The server sends the generated response to the terminal using the send_response(response, client) method. The input is the generated response text, and the output is the response sent to the terminal.

[1771] Step 6: Save conversation history and continue learning

[1772] The server saves the conversation history and sentiment analysis results in a database for continuous learning. Specifically, it uses the save_to_database(user_id, conversation_data) method. The input of this step is the conversation history and sentiment analysis results, and the output is the data saved in the database.

[1773] Step 7: Displaying the User Interface

[1774] The terminal displays an interface for the user to interact with the system. Specifically, the display_interface() function is called to display the chat window. The input is the interface design loaded when the system started, and the output is the chat screen displayed to the user.

[1775] Step 8: Getting and Sending User Input

[1776] The terminal gets input from the user and sends the contents to the server. get_user_input() receives user input and send_to_server(input_text) sends it to the server. The input is the text entered by the user and the output is the message sent to the server.

[1777] Step 9: View the response

[1778] The terminal receives the response sent from the server and displays it on the user interface. The response is received with receive_from_server() and updated with update_interface(response). The input is the response from the server, and the output is the response displayed to the user.

[1779] Through these steps, natural, emotion-sensitive dialogue is realized using a generative AI model and emotion engine. Appropriate processing is performed based on the input data at each step, and the resulting output data is passed on to the next step. The user can engage in natural dialogue through their device, and the system can generate responses that are in tune with the user's emotions.

[1780] (Application example 2)

[1781] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1782] Customer service is a major issue in modern brick-and-mortar stores. In particular, when customers ask about products, store clerks may lack sufficient knowledge. While a response that takes into consideration the customer's feelings is required, not all store clerks have the skills to do so. This can lead to a decline in customer satisfaction. The present invention aims to solve these issues and realize natural, emotionally sensitive conversations in brick-and-mortar stores.

[1783] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1784] In this invention, the server includes means for initializing and loading the generative AI model, means for receiving conversation requests from users, natural language understanding means for analyzing the received conversation requests, natural language generation means for generating responses based on the analysis results, means for sending the generated responses to the users, means for saving the conversation history in a database and using it for continuous learning, means for extracting information about products and explaining the products in a natural conversational format, means for generating responses according to emotions and taking customer emotions into consideration, and means for analyzing customer reactions and questions in real time. This enables improvement in the quality of customer service in physical stores and improvement in customer satisfaction.

[1785] A "generative AI model" refers to an artificial intelligence technology that is trained and learned in advance using large amounts of data to understand human language and generate responses.

[1786] "Means to take emotions into consideration" refers to the function of analyzing emotions from the user's input message and generating an appropriate response based on those emotions.

[1787] "Natural language understanding means" refers to a function for analyzing input messages from users and understanding their content and intent.

[1788] "Natural language generator" refers to a function that generates responses in natural language based on analyzed user input.

[1789] "User interface" refers to the collection of screens and input devices that allow a user to interact with a system.

[1790] A "conversation request" refers to a question or message that a user inputs to the system.

[1791] "Means for receiving" refers to a function for taking in conversation requests input from the outside.

[1792] "Means for generating a response" refers to the function of generating an appropriate reply based on a received conversation request.

[1793] "Means for sending" refers to a function for sending the generated response back to the user.

[1794] "Conversation history" refers to data that stores the contents of past conversations exchanged between a user and the system.

[1795] A "database" refers to a digital storage device that systematically stores information such as conversation history.

[1796] "Means for continuous learning" refers to a function that allows the AI ​​model to further learn based on saved conversation history, improving the accuracy and quality of responses.

[1797] "Product Information" refers to detailed information about a specific product, such as its characteristics, usage, and price.

[1798] "Means of analysis" refers to the function of analyzing input data in various ways and understanding its meaning and intent.

[1799] The present invention relates to a system that uses a generative AI model to improve the efficiency of customer service in brick-and-mortar stores and provide natural conversations that take emotions into consideration. Specific embodiments are described below.

[1800] Server Processing

[1801] The server performs the following process.

[1802] 1. Initializing and loading the model

[1803] At system startup, the server initializes the generative AI model (e.g., GPT-4) and emotion engine, loading the necessary data and settings into memory, so that the AI ​​is ready to converse with the user.

[1804] 2. Receiving a conversation request

[1805] The server receives a conversation request from the terminal, which includes a message or question entered by the customer.

[1806] 3. Natural Language Understanding and Sentiment Analysis

[1807] The server passes the received message to a natural language understanding module to analyze the content and context, and also analyzes the customer's emotions using an emotion engine, thereby accurately understanding the customer's intentions and emotions.

[1808] 4. Generating the Response

[1809] The server uses a natural language generation module to generate an appropriate response based on the analyzed information and the sentiment analysis results. The generated response is written in natural language and takes into consideration the customer's sentiment.

[1810] 5. Sending the Response

[1811] The server sends the generated response to the terminal, thereby proceeding with the dialogue.

[1812] 6. Conversation history storage and continuous learning

[1813] The server stores the conversation history and sentiment analysis results in a database and uses them to continuously train the generative AI model and sentiment engine, allowing the model to learn customer characteristics, preferences, and sentiment trends over time.

[1814] Terminal handling

[1815] The terminal performs the following process.

[1816] 1. Displaying the user interface

[1817] The terminal displays an interface (such as a chat window) through which the user can interact with the system, allowing the customer to type messages.

[1818] 2. Capture and submit customer input

[1819] The terminal receives input from the customer (such as a message or question) and sends the content to the server. This operation transfers the customer's message to the server.

[1820] 3. View the response

[1821] The terminal receives the response sent from the server and displays it on the user interface, allowing the customer to check the system's response and enter the next message.

[1822] Hardware and software used

[1823] The system is implemented using the following hardware and software:

[1824] Hardware: Smartphones, smart glasses, head-mounted displays

[1825] Software: Flask (web server), OpenAI GPT-4 (generative AI model), TextBlob (sentiment analysis)

[1826] Specific examples

[1827] 1. A customer asks through smart glasses, "Tell me about this shirt."

[1828] The server generates a response such as "This shirt has a modern design and is highly breathable" and sends it to the device, which displays the response on the smart glasses.

[1829] 2. A customer types, "I'm tired from work today."

[1830] The server generates a response that takes into account the user's emotions, such as "Thank you for your hard work. Would you like something to help you relax?", and sends it to the device. The device then displays this response on its smartphone.

[1831] Prompt Sentence Examples

[1832] Customer: Tell me about this shirt.

[1833] clerk:

[1834] Customer: I'm tired from work today.

[1835] clerk:

[1836] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1837] Step 1:

[1838] The server initializes the generative AI model and emotion engine at system startup and loads the necessary data and settings into memory, which prepares the generative AI model and emotion engine to generate customer responses. The input is data from the pre-trained AI model, and the output is the system configuration information loaded into memory.

[1839] Step 2:

[1840] A user inputs a conversation request through a terminal. Specifically, the user inputs a question or message into the interface (e.g., a chat window) of a smartphone or smart glasses. The input is the user's text message, and the output is the transfer of the request from the terminal to the server.

[1841] Step 3:

[1842] The server passes the conversation request received from the device to a natural language understanding module, which analyzes the content and context. At the same time, it uses an emotion engine to analyze the user's emotions. The input is the user's text message, and the output is the analysis results of the message's content, context, and emotions. Natural language processing and emotion analysis technologies are used for the analysis.

[1843] Step 4:

[1844] The server uses a natural language generation module based on the analysis results to generate an appropriate response. A generative AI model (e.g., GPT-4) is used to form a natural-language response based on the user's intent and sentiment. The input is the analysis results of the message's content, context, and sentiment, and the output is the generated text response.

[1845] Step 5:

[1846] The server sends the generated response to the terminal, which allows the interaction to reach the user and proceed to the next step. The input is the generated text response, and the output is the display of the response in the user interface.

[1847] Step 6:

[1848] The terminal displays the response received from the server on the user interface. The user checks the displayed response and prepares to enter the next message. The input is the text response sent from the server, and the output is the text displayed on the screen of the smartphone or smart glasses.

[1849] Step 7:

[1850] Through the user interface, the user inputs the next message and sends it back to the server. This cycle continues, creating a continuous dialogue: the input is the user's new message, and the output is again a request to be parsed and a response generated.

[1851] Step 8:

[1852] The server stores the conversation history and emotion analysis results in a database and uses them for continuous learning of the generative AI model and emotion engine. This allows the system to learn customer characteristics, preferences, and emotional trends over time and use them for the next interaction. The input is the conversation history and emotion data to be saved, and the output is the information accumulated in the database.

[1853] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1854] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1855] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1856] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1857] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1858] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1859] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1860] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1861] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1862] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1863] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1864] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1865] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1866] 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.

[1867] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1868] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1869] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1870] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1871] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1872] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1873] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1874] The following is further disclosed regarding the above embodiment.

[1875] (Claim 1)

[1876] a means for initializing and loading the generative AI model;

[1877] means for receiving a conversation request from a user;

[1878] natural language understanding means for analyzing the received conversation request;

[1879] natural language generation means for generating a response based on the analysis result;

[1880] means for transmitting the generated response to a user;

[1881] A means to store conversation history in a database and use it for continuous learning;

[1882] A system including:

[1883] (Claim 2)

[1884] 10. The system of claim 1, further comprising means for displaying a user interface and obtaining a message from the user.

[1885] (Claim 3)

[1886] 10. The system of claim 1, further comprising means for displaying the response via a terminal in communication with the server.

[1887] "Example 1"

[1888] (Claim 1)

[1889] a means for initializing and loading the generative AI model;

[1890] means for receiving a conversation request from a user;

[1891] natural language understanding means for analyzing the received conversation request;

[1892] natural language generation means for generating a response based on the analysis result;

[1893] means for transmitting the generated response to a user;

[1894] A means to store conversation history in a database and use it for continuous learning;

[1895] a means for loading the necessary data and settings of the generative AI model into memory at system startup;

[1896] A means to understand the user's intentions by analyzing the sentiment and context of the user's message;

[1897] means for generating a response formulated in natural language;

[1898] A system including:

[1899] (Claim 2)

[1900] 10. The system of claim 1, further comprising means for displaying a user interface and obtaining a message from the user.

[1901] (Claim 3)

[1902] 10. The system of claim 1, further comprising means for displaying the response via a terminal in communication with the server.

[1903] "Application Example 1"

[1904] (Claim 1)

[1905] a means for initializing and loading the generative AI model;

[1906] means for receiving a conversation request from a user;

[1907] natural language understanding means for analyzing the received conversation request;

[1908] natural language generation means for generating a response based on the analysis result;

[1909] means for transmitting the generated response to a user;

[1910] A means to store conversation history in a database and use it for continuous learning;

[1911] A means for customers to obtain product information and ask questions using the EuzentoFace,

[1912] A system including:

[1913] (Claim 2)

[1914] 10. The system of claim 1, further comprising means for displaying a user interface and obtaining a message from the user.

[1915] (Claim 3)

[1916] 10. The system of claim 1, further comprising means for displaying the response via a terminal in communication with the server.

[1917] "Example 2: Combining Emotion Engines"

[1918] (Claim 1)

[1919] means for initializing and loading the generative AI model and emotion engine;

[1920] means for receiving a conversation request from a user;

[1921] natural language understanding means for analyzing the received conversation request;

[1922] A means for analyzing emotions based on the analysis results;

[1923] a natural language generation means for generating a response based on the analysis result and the sentiment analysis result;

[1924] means for transmitting the generated response to a user;

[1925] A means of storing conversation history and sentiment analysis results in a database for continuous learning;

[1926] A system including:

[1927] (Claim 2)

[1928] 10. The system of claim 1, further comprising means for displaying a user interface and obtaining a message from the user.

[1929] (Claim 3)

[1930] 10. The system of claim 1, further comprising means for displaying the response via a terminal in communication with the server.

[1931] "Application example 2 when combining emotion engines"

[1932] (Claim 1)

[1933] a means for initializing and loading the generative AI model;

[1934] means for receiving a conversation request from a user;

[1935] natural language understanding means for analyzing the received conversation request;

[1936] natural language generation means for generating a response based on the analysis result;

[1937] means for transmitting the generated response to a user;

[1938] A means to store conversation history in a database and use it for continuous learning;

[1939] A means to extract information about the product and explain it in a natural conversational style.

[1940] A means for generating emotional responses and taking into consideration customer emotions;

[1941] A means of analyzing customer reactions and questions in real time,

[1942] A system including:

[1943] (Claim 2)

[1944] 10. The system of claim 1, further comprising means for displaying a user interface and obtaining a message from the user.

[1945] (Claim 3)

[1946] 10. The system of claim 1, further comprising means for displaying the response via a terminal in communication with the server. [Explanation of symbols]

[1947] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for initializing and loading the generative AI model; means for receiving a conversation request from a user; natural language understanding means for analyzing the received conversation request; natural language generation means for generating a response based on the analysis result; means for transmitting the generated response to a user; A means to store conversation history in a database and use it for continuous learning; A system including:

2. The system of claim 1 , further comprising means for displaying a user interface and obtaining a message from a user.

3. 10. The system of claim 1, further comprising means for displaying the response via a terminal in communication with the server.

Images