system

Abstract

We provide the system. [Solution] A means by which the user inputs a question and sends it to the server, A means for sending a question received by a server to a generative AI to generate an initial answer, The server presents the initial response it generates to a human expert respondent, and a means of receiving feedback is provided. A means by which the server integrates initial responses and feedback to generate a final response, A means by which the server sends the final answer to the user's terminal, A system that includes this.

Description

Technical Field

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

Background Art

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

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In modern times, there are systems where many users post questions via the Internet and seek answers. However, the following problems exist in conventional systems. First, it is difficult to obtain answers quickly. The response of human answerers is limited in real time, and answers are often delayed. Second, answers using generative AI may lack accuracy, which may cause misunderstandings and confusion to users. There is a need to solve these problems and provide quick and accurate answers.

Means for Solving the Problems

[0005] To solve this problem, the present invention provides the following means: a means for a user to input a question and send it to a server; a means for the server to send the received question to a generative AI to generate an initial answer, and then to present the generated initial answer to a human expert respondent to receive feedback; further, a means for the server to integrate the initial answer and feedback to generate a final answer and send the final answer to the user's terminal; this allows the user to obtain an answer quickly, and its accuracy is guaranteed through checking by a human expert respondent; and the invention also includes a means for the server to add feedback from the human expert respondent to the generative AI's training data to improve the accuracy of the generative AI, thereby further improving the quality of future answers; and the invention also includes a means for analyzing the question input by the user through the terminal using natural language processing and sending the question to the generative AI in an optimal format, thereby improving the accuracy of the answers provided by the generative AI.

[0006] A "user" refers to an individual or group that uses the system to input questions and receive answers.

[0007] "Terminal" refers to electronic devices such as computers, smartphones, and tablets that users use to interface with a system.

[0008] A "server" refers to a computer system on a network that receives questions from users, sends questions to generative AI, presents them to human expert respondents, and integrates and transmits the final answers.

[0009] "Generative AI" refers to software or algorithms that use artificial intelligence technology to generate initial answers to input questions.

[0010] "Initial response" refers to the first response that a generative AI automatically generates based on the input question.

[0011] A "human expert respondent" refers to an individual with the knowledge and experience to verify the initial response generated by the generative AI and to make corrections or additions as needed.

[0012] "Feedback" refers to comments and opinions, including corrections and additional information, that human expert respondents provide to initial answers.

[0013] "Final answer" refers to the final answer generated by integrating the initial response from a generative AI with feedback from human expert respondents.

[0014] "Natural language processing" refers to the technologies and methods used by computers to understand and analyze natural human language.

[0015] "Training data" refers to the dataset that generative AI uses for learning and improvement. [Brief explanation of the drawing]

[0016] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] This is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] This is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] This is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] It shows an emotion map to which a plurality of emotions are mapped. [Figure 10] It shows an emotion map to which a plurality of emotions are mapped. [Figure 11] It is a sequence diagram showing the processing flow of the data processing system in Embodiment 1. [Figure 12] It is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] It is a sequence diagram showing the processing flow of the data processing system in Embodiment 2 when an emotion engine is combined. [Figure 14] It is a sequence diagram showing the processing flow of the data processing system in Application Example 2 when an emotion engine is combined.

Mode for Carrying Out the Invention

[0017] Hereinafter, an example of an embodiment of a system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0018] First, the terms used in the following description will be described.

[0019] In the following embodiments, a processor with a reference number (hereinafter simply referred to as "processor") may be one arithmetic unit or a combination of a plurality of arithmetic units. Also, the processor may be one type of arithmetic unit or a combination of a plurality of types of arithmetic units. Examples of arithmetic units include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), and the like.

[0020] In the following embodiments, signed RAM (Random Access Memory) is a memory that temporarily stores information and is used as work memory by the processor.

[0021] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0022] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0024] [First Embodiment]

[0025] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0026] As shown in Figure 1, the data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0027] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0028] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0029] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0030] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0031] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0032] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] As shown in Figure 2, in the data processing device 12, a specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" related to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0034] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0035] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0036] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0037] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. The processing of the system's program is described below in natural language, with specific examples.

[0038] Submit a question

[0039] The user enters a question on the device. For example, the user might type, "Why hasn't my cat been eating much lately?" The device receives this question and sends it to the server. The server temporarily stores the received question for processing.

[0040] Question analysis and initial response generation

[0041] The server analyzes the received question using natural language processing (NLP) techniques to understand its meaning. The analyzed question is then input into a generative AI. Based on the analysis results, the generative AI generates the optimal initial answer from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are some possible causes."

[0042] Checked by human expert respondents

[0043] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, the expert respondent might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0044] Integration and generation of the final answer

[0045] The server integrates initial responses from the generative AI with feedback from human expert respondents. This generates the final response to be provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[0046] Delivery of responses

[0047] The server sends the final response to the user's device. The user's device then displays the received final response to the user. This allows the user to obtain information quickly and accurately.

[0048] Feedback for improving AI accuracy

[0049] The server adds feedback from human expert respondents to the generative AI's training data. This allows the generative AI to continuously learn, improving the accuracy of its answers to subsequent questions.

[0050] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and also improves the accuracy of generative AI.

[0051] The following describes the processing flow.

[0052] Step 1:

[0053] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[0054] Step 2:

[0055] The terminal retrieves the entered question and sends it to the server. Specifically, it sends the question content to the server as an HTTP request.

[0056] Step 3:

[0057] The server receives a question and analyzes it using natural language processing (NLP) techniques. The analysis extracts the main keywords and context of the question.

[0058] Step 4:

[0059] The server sends the analyzed question to the generative AI. The generative AI generates the optimal initial answer from the training database based on the question.

[0060] Step 5:

[0061] The generative AI generates the initial response. For example, it might generate a response such as, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are possible causes."

[0062] Step 6:

[0063] The server receives the initial response from the generative AI and stores it temporarily.

[0064] Step 7:

[0065] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[0066] Step 8:

[0067] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[0068] Step 9:

[0069] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[0070] Step 10:

[0071] The server integrates the initial response and expert feedback to generate the final answer. For example, it might say, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to have its health checked."

[0072] Step 11:

[0073] The server sends the final response to the user's device. Specifically, it sends the final response to the device as an HTTP response.

[0074] Step 12:

[0075] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[0076] Step 13:

[0077] The server adds feedback from expert respondents to the training data of the generative AI. The generative AI uses the added feedback to update its model and improve the accuracy of its answers to subsequent questions.

[0078] (Example 1)

[0079] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0080] In today's information society, providing quick and accurate answers to the diverse questions users may have is crucial. However, conventional systems can be time-consuming to accurately analyze user questions and generate appropriate answers. Furthermore, the accuracy and reliability of the generated answers are not always high. This invention aims to solve these problems and provide users with quick and reliable answers.

[0081] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0082] In this invention, the server includes means for the user to input and send a question to the server; means for the server to analyze the received question using natural language processing technology and send it to a generative AI model; means for the generative AI model to generate an initial answer based on the analyzed question; means for the server to present the generated initial answer to an expert respondent and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; and means for the server to send the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer to the user's question.

[0083] A "user" is a person or entity that uses the system to input questions and receive answers.

[0084] A "device" is a device used by a user to input questions, and includes, for example, smartphones, tablets, and personal computers.

[0085] A "server" is a central computer system that analyzes questions received from users, generates answers using generative AI models, and sends those answers to users.

[0086] "Natural language processing technology" refers to techniques for analyzing text written in natural language and understanding its meaning, and includes, for example, morphological analysis and contextual analysis.

[0087] A "generative AI model" is an artificial intelligence model that generates appropriate answers based on analyzed questions, and includes, for example, machine learning algorithms and neural networks.

[0088] An "initial response" is the first answer that a generative AI model generates based on the user's question.

[0089] A "specialist respondent" is a person with the expertise to review the initial responses generated by generative AI models and make corrections or additions as needed.

[0090] "Feedback" refers to the corrections and supplementary information that expert respondents provide to their initial answers.

[0091] The "final answer" refers to the final response provided to the user, generated by integrating the initial response from a generative AI model with feedback from expert respondents.

[0092] "Training data" refers to data used by generative AI models to learn and improve the accuracy of their responses, and includes feedback from expert respondents.

[0093] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. This section describes the specific operation of this system and the hardware and software required to implement it.

[0094] First, the user enters a question using a device such as a smartphone or computer. For example, the user might enter, "Why hasn't my cat been eating much lately?" The device receives this input and sends it to the server via the internet.

[0095] The server temporarily stores the received questions and analyzes them using natural language processing (NLP) techniques. Examples of NLP techniques used include SpaCy and Google's Natural Language API. In this analysis step, the question's structure and keywords are extracted and converted into a format suitable for input into a generative AI model.

[0096] Next, the server sends the analyzed question to a generative AI model. For example, OpenAI's GPT-4® is used as the generative AI model. This model generates an initial answer from a training database based on the analysis results. For example, the generative AI model might answer, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are possible causes."

[0097] The generated initial response is presented to the expert respondent by the server. The expert respondent reviews this initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0098] The server integrates initial responses from generative AI models with feedback from expert respondents. This integration process generates the final response provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[0099] Finally, the server sends the final answer to the user's device. The device then displays the received final answer to the user. This allows the user to obtain information quickly and accurately.

[0100] Furthermore, the server adds feedback from expert respondents to the training data of the generative AI model. This addition of feedback allows the generative AI model to continuously learn, improving the accuracy of its answers to subsequent questions.

[0101] For example, the following prompt statement can be used:

[0102] "A user is asking why their cat hasn't been eating much lately. Analyze this question, generate an initial answer, and then have it checked by a human expert to generate the final answer."

[0103] In this way, the system can provide users with fast and reliable answers, and also improve the accuracy of generative AI models.

[0104] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0105] Step 1:

[0106] The user enters the question using a terminal.

[0107] Input: A question from the user. For example, "Why isn't my cat eating much lately?"

[0108] Specific operation: The user enters a question using a keyboard or touchscreen and presses the submit button.

[0109] Output: Input question data.

[0110] Step 2:

[0111] The terminal retrieves the entered question and sends it to the server.

[0112] Input: Question data entered by the user on their device.

[0113] Specific operation: The terminal sends the question data to the server via the network.

[0114] Output: Question data sent to the server.

[0115] Step 3:

[0116] The server temporarily stores the received questions and analyzes them using natural language processing techniques.

[0117] Input: Question data received by the server.

[0118] Specific operation: The server stores the question data in a database and performs analysis using natural language processing techniques (e.g., SpaCy or Google Natural Language API).

[0119] Output: Analyzed question data.

[0120] Step 4:

[0121] The server sends the analyzed question data to the AI ​​model to generate initial answers.

[0122] Input: Analyzed question data.

[0123] Specific operation: The server uses this data as input to a generating AI model (e.g., GPT-4). The generating AI model refers to the training database and generates an initial response. A response such as "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food" is generated.

[0124] Output: The initial response that was generated.

[0125] Step 5:

[0126] The server presents the generated initial answers to expert respondents and receives their feedback.

[0127] Input: The initial generated response.

[0128] Specific operation: The server sends the initial response to the expert respondent's terminal, who then reviews it. Feedback is provided with corrections and additions as needed. For example, feedback such as, "You should take your cat to the vet to have its health checked," might be added.

[0129] Output: Feedback from expert respondents.

[0130] Step 6:

[0131] The server integrates initial responses and feedback from expert respondents to generate the final response.

[0132] Input: Initial responses and feedback from expert respondents.

[0133] Specific operation: The server combines the initial response and feedback to perform an integrated process. This generates the final response to be provided to the user. For example, it might say, "There are various reasons why your cat isn't eating. Stress, illness, or dislike of the food are possible causes. It would be a good idea to take your cat to the vet to check its health."

[0134] Output: The final answer that was generated.

[0135] Step 7:

[0136] The server sends the final response to the user's device.

[0137] Input: The final generated response.

[0138] Specific operation: The server sends the final answer to the user's terminal via the network.

[0139] Output: The final response sent to the user's terminal.

[0140] Step 8:

[0141] The terminal displays the final response received to the user.

[0142] Input: Final response sent from the server.

[0143] Specific action: The final response received will be displayed on the device's screen.

[0144] Output: Users can view the final answer.

[0145] Step 9:

[0146] The server generates feedback from expert respondents, adds it to the training data of the AI ​​model, and improves the accuracy of the AI ​​model.

[0147] Input: Feedback from expert respondents.

[0148] Specific operation: The server adds the feedback data to the training database of the generating AI model and retrains the model. This improves the accuracy of subsequent generated responses.

[0149] Output: Updated generative AI model.

[0150] (Application Example 1)

[0151] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0152] In modern food delivery services, it is difficult to provide reliable information quickly when users ask specific health-related questions. Furthermore, the accuracy and appropriateness of initial responses from generative AI cannot be guaranteed, potentially leading to a decline in service quality for users. In addition, previous systems failed to effectively utilize expert knowledge, resulting in delays in the learning process of generative AI.

[0153] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0154] In this invention, the server includes means for the user to input and send a question to the server; means for the server to send the received question to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a respondent with expert knowledge and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; means for the server to send the final answer to the user's information display device; means for the generative AI to continuously learn based on the feedback and improve the accuracy of the answer; and means for providing information about food in response to a question entered through the user's terminal. As a result, the user can quickly obtain reliable food delivery information and the accuracy of the generative AI is also improved, making it possible to provide a higher quality service.

[0155] A "user" refers to the entity that uses this system to input questions and obtain information.

[0156] A "server" refers to a central processing unit that receives a question, sends it to a generative AI, generates an initial answer, presents it to a respondent with specialized knowledge, receives feedback, generates a final answer, and sends it to the user.

[0157] "Generative AI" refers to artificial intelligence technology that creates initial answers based on user questions.

[0158] A "respondent with specialized knowledge" refers to a human expert who checks the initial response generated by the generative AI and makes corrections or additions.

[0159] "Feedback" refers to the corrections or additions made by respondents with specialized knowledge to the initial answers.

[0160] The "final answer" refers to the final answer generated by integrating the initial answers with feedback from respondents with specialized knowledge.

[0161] An "information display device" refers to a terminal or device used by a user to receive and display their final response.

[0162] "Continuous learning" refers to the process by which the server adds feedback to the generative AI as training data, thereby improving the accuracy of subsequent responses.

[0163] "Dietary information" refers to specific information about meal menus and nutrition related to the user's health.

[0164] In this invention, a user inputs a question about food using an information display device and sends it to a server. The server receives the question and sends it to a generative AI. The generative AI analyzes the received question using natural language processing technology and generates an initial answer. The open-source GPT-3 (registered trademark) (OpenAI's text-davinci-003 engine) is used for natural language processing. The input question is in the format of "I've been wanting to eat healthier lately. What kind of menu do you recommend?"

[0165] The generated initial answers are presented to respondents with expert knowledge. These respondents review the initial answers generated by the generative AI and make appropriate corrections or additions. For example, if the generative AI generates the initial answer, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie," a respondent with expert knowledge might provide feedback such as, "These menu items offer balanced nutrition, are low in calories, and the smoothie is rich in vitamin C."

[0166] The server integrates initial responses from generative AI with feedback from experts to generate a final response. The final response incorporates the initial responses and feedback, for example, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C."

[0167] Finally, the server sends the final answer to the user's information display device. The user can then view the answer on the display device and obtain the necessary information. The server also adds feedback from expert respondents as training data for the generative AI, allowing the generative AI to continuously learn and improve the accuracy of its answers to subsequent questions.

[0168] As a concrete example of its use, consider the following prompt message:

[0169] "Please answer the user's question about healthy menus: I've recently started wanting to eat healthier meals. What kind of menus would you recommend?"

[0170] This invention allows users to quickly obtain reliable food delivery information and improves the accuracy of generative AI, enabling the provision of higher-quality services.

[0171] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0172] Step 1:

[0173] The user enters questions about their meal on an information display device and then submits those questions.

[0174] Input: A question entered by the user into the information display device (e.g., "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[0175] Output: The information display device sends the question content to the server.

[0176] Specific operation: The user operates the interface of the information display device, enters a question, and then presses the confirmation button. The information display device formats the input data and sends an HTTP request to the server endpoint.

[0177] Step 2:

[0178] The server sends the received question to a generative AI to generate an initial answer.

[0179] Input: Question data sent from the information display device (Example: "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[0180] Output: Initial response generated by a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[0181] Specific operation: The server receives an HTTP request and sends the question content to a natural language processing engine (e.g., OpenAI GPT-3). The generative AI analyzes the question and generates an initial answer from the training data. The generated answer is returned to the server and temporarily stored.

[0182] Step 3:

[0183] The server presents the initial generated answers to respondents with specialized knowledge and receives their feedback.

[0184] Input: Initial response from a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[0185] Output: Feedback from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0186] Specific operation: The server displays the initial response from the generative AI to an interface for experts. Experts review the initial response and input corrections or supplementary information. The corrected feedback is returned to the server.

[0187] Step 4:

[0188] The server integrates the initial response and feedback to generate the final response.

[0189] Input: Initial response from a generative AI and feedback from experts (e.g., "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie." + "These menu items offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0190] Output: Final answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C.")

[0191] Specific operation: The server combines and integrates the initial response and feedback text data, and formats it into a single final response in text format.

[0192] Step 5:

[0193] The server sends the final response to the user's information display device.

[0194] Input: Integrated final answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie is especially rich in vitamin C.")

[0195] Output: Final answer displayed on the user's information display device.

[0196] Specific operation: The server sends the final answer back to the user's information display device as an HTTP response, and the terminal displays this answer with an appropriate UI (user interface).

[0197] Step 6:

[0198] The server continuously trains the generative AI based on feedback, improving the accuracy of its responses.

[0199] Input: Feedback data from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0200] Output: Improved generative AI model

[0201] Specific operation: The server adds the feedback data to the training dataset of the generative AI for the next question, and the model is retrained, thereby improving the AI's accuracy.

[0202] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0203] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions. The following describes the program processing of this system in natural language, with specific examples.

[0204] Question submission and sentiment recognition

[0205] When a user enters a question on their device, the emotion engine recognizes the user's emotions. For example, if a user enters "Why hasn't my cat been eating much lately?", the emotion engine analyzes the user's tone of voice and input patterns at this point and recognizes that the user is feeling anxious.

[0206] The device sends the recognized sentiment information along with the question to the server. Specifically, it sends the question and sentiment information to the server as an HTTP request.

[0207] Question analysis and initial response generation

[0208] The server analyzes the received question and sentiment information using natural language processing (NLP) techniques. The analysis extracts key keywords, context, and sentiment information from the question.

[0209] The server sends the analyzed question and sentiment information to the generative AI. Based on this information, the generative AI generates the optimal initial response from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," while also considering the user's anxiety and adding a note at the end, "If you are concerned, we recommend consulting a professional."

[0210] Checked by human expert respondents

[0211] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0212] Integration and generation of the final answer

[0213] The server integrates initial responses from the generative AI with feedback from human expert respondents. Even in this process, user sentiment is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[0214] Delivery of responses

[0215] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response. The device then displays the received final response to the user. This allows the user to obtain quick and accurate information, as well as receive a response that is sensitive to their feelings.

[0216] Feedback for improving AI accuracy

[0217] The server adds feedback from human expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[0218] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[0219] The following describes the processing flow.

[0220] Step 1:

[0221] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[0222] Step 2:

[0223] The emotion engine analyzes user input in real time and recognizes the user's emotions. For example, the emotion engine can detect feelings of anxiety from the user's input speed and word choice.

[0224] Step 3:

[0225] The terminal retrieves the entered question and recognized sentiment information and sends it to the server. Specifically, it sends the question content and sentiment information to the server as an HTTP request.

[0226] Step 4:

[0227] The server receives the question and sentiment information, and analyzes the question using natural language processing (NLP) techniques. As a result of the analysis, the server extracts the main keywords, context, and sentiment information of the question.

[0228] Step 5:

[0229] The server sends the analyzed question and sentiment information to the generative AI. The generative AI considers the question content and sentiment information to generate the optimal initial response from the training database.

[0230] Step 6:

[0231] A generative AI generates an initial response. For example, it might generate a response like, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," and then, based on emotional information, add a supplement at the end such as, "If you are concerned, we recommend consulting a specialist."

[0232] Step 7:

[0233] The server receives the initial response from the generative AI and stores it temporarily.

[0234] Step 8:

[0235] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[0236] Step 9:

[0237] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to have its health checked."

[0238] Step 10:

[0239] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[0240] Step 11:

[0241] The server integrates the initial response and feedback from expert respondents to generate the final response. Even in this process, user sentiment information is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to a veterinary clinic to have its health checked."

[0242] Step 12:

[0243] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response.

[0244] Step 13:

[0245] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[0246] Step 14:

[0247] The server adds feedback from expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[0248] (Example 2)

[0249] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0250] Conventional question-answering systems have struggled to provide answers that take user emotions into account, resulting in insufficient improvement in the user experience. Furthermore, the quality of initial answers relies solely on generative AI, leaving room for improvement in accuracy. Additionally, the generated answers are provided only once, preventing continuous learning of the generative AI, thus limiting long-term accuracy improvement.

[0251] The specific processing performed by the specific 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 the user to input a question and send it to the server; means for the terminal to recognize the user's emotions from the input question and generate emotion information; means for the server to send the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback and generate a final answer that takes emotion information into consideration; and means for the server to send the final answer to the user's terminal. This makes it possible to provide more accurate and reliable answers that take the user's emotions into consideration, and furthermore, by utilizing the feedback from the expert respondent as training data for the generative AI, it is possible to improve the accuracy of the entire system.

[0252] A "user" refers to a person who uses the system to input questions and receive answers.

[0253] A "server" refers to a device that receives questions and sentiment information from users, analyzes, processes, and sends this information to AI systems or expert respondents to generate the final answer.

[0254] A "terminal" refers to a device that a user directly operates to input questions, generate sentiment information, and send it to a server.

[0255] An "emotion engine" refers to software or an algorithm that recognizes emotions from questions entered by a user through a device and generates emotional information.

[0256] "Generative AI" refers to artificial intelligence models that generate initial responses based on questions and sentiment information sent from a server.

[0257] "Initial response" refers to the version of the response generated by a generative AI and provided to expert respondents.

[0258] A "specialist respondent" refers to an expert who reviews the initial response generated by generative AI and makes corrections or additions as needed.

[0259] "Feedback" refers to the corrections and supplementary information that expert respondents provide to the initial responses of generative AI systems.

[0260] "Final answer" refers to the final answer generated by the server, which integrates the initial answers and feedback from expert respondents.

[0261] Natural Language Processing (NLP) refers to a technology that analyzes questions entered by users through their devices and extracts their context and key keywords.

[0262] "Training data" refers to data that includes feedback from expert respondents and is used to improve the accuracy of generative AI.

[0263] "Improved accuracy" refers to the process where a generative AI updates its model using additional training data to increase the accuracy of its answers to subsequent questions.

[0264] The system based on this invention begins with the user inputting a question via a terminal and sending it to a server. The system of this invention involves an emotion engine, a generative AI, and human expert respondents who generate answers to the user's questions and provide the final answer.

[0265] Question input and sentiment recognition

[0266] The user enters the question using a device (smartphone, tablet, computer, etc.). For example, when a user enters "Why hasn't my cat been eating much lately?", an emotion engine installed on the device (e.g., IBM Watson® Natural Language Understanding) recognizes the user's emotions from the entered text. In this case, the emotion engine analyzes the input pattern and context and generates emotion information such as "anxiety."

[0267] Sending and analyzing questions and sentiment information

[0268] The device sends the question and recognized sentiment information to the server as an HTTP request. The server analyzes the received question and sentiment information using natural language processing (NLP) techniques (e.g., spaCy, NLTK). As a result of the analysis, key keywords and context are extracted. For example, keywords such as "cat," "not eating," and "cause," along with the sentiment information "anxiety," are extracted.

[0269] Generating initial responses

[0270] The server sends the analyzed question and sentiment information to a generative AI (e.g., OpenAI GPT-3) to generate an initial response. Based on this input, the generative AI selects the best response from its training database. For example, the generative AI might generate the response, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food," and then address the anxiety by adding a supplementary sentence such as, "If you are concerned, we recommend consulting a specialist."

[0271] Review and correction by expert respondents

[0272] The generated initial response is presented to human expert respondents via the server. The expert respondents review the initial response and make corrections or additions as needed. For example, an expert respondent might add feedback such as, "You should take your cat to the vet to have its health checked."

[0273] Integration and generation of the final answer

[0274] The server integrates initial responses from the generative AI with feedback from expert respondents to generate a final response. User sentiment information is also taken into consideration during this process. For example, a final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[0275] Submit and display the final response.

[0276] The server sends the final response to the user's device as an HTTP response. The device then displays the received final response to the user, allowing the user to obtain quick and accurate information while also receiving an emotionally sensitive response.

[0277] Feedback for improving AI accuracy

[0278] Furthermore, the server adds feedback from expert respondents as training data for the generative AI, updating the AI ​​model to improve the accuracy of answers to subsequent questions. Since user sentiment information is also included in the training data, the accuracy of responses to emotions also improves.

[0279] As a specific example, when a user inputs "My child had a fight with a friend at school and is feeling down. What should I do?", the emotion engine recognizes that the user is worried and assigns "worried" as the emotion information. Based on this, the generative AI generates a response such as "It's natural for children to learn about relationships with friends through fights. It's important to listen calmly. Also, if necessary, it might be a good idea to consult a school counselor."

[0280] Examples of prompt texts for the generative AI model:

[0281] User's question: "My child had a fight with a friend at school and is feeling down. What should I do?"

[0282] Emotion information: "worried"

[0283] The above are specific embodiments of the present invention. With this system, the user can obtain a quick and reliable answer, and by receiving an answer that takes emotions into account, the user experience is improved.

[0284] The flow of the specific process in Example 2 will be described using FIG. 13.

[0285] Step 1: The user inputs a question

[0286] The user uses a terminal to input a question. The input question is captured by the terminal as text data. As an example, the user inputs "Why has my pet cat been eating less recently?"

[0287] Input: The question text input by the user

[0288] Output: The question data captured by the terminal

[0289] Step 2: The terminal recognizes the emotion

[0290] The emotion engine installed on the device recognizes the user's emotions from the entered question text. Through text analysis, the emotion engine recognizes that the user is feeling anxious.

[0291] Input: Captured question text

[0292] Data processing: The emotion engine performs text analysis (e.g., using IBM Watson Natural Language Understanding).

[0293] Output: Recognized emotional information (e.g., anxiety)

[0294] Step 3: Send the question and sentiment information to the server.

[0295] The device sends the question content and recognized sentiment information to the server as an HTTP request.

[0296] Input: Questionnaire data and sentiment information

[0297] Data processing: Converting data into HTTP request format.

[0298] Output: Data sent to the server

[0299] Step 4: The server receives the question and sentiment information.

[0300] The server receives HTTP requests sent from the terminal and retrieves the question content and sentiment information.

[0301] Input: Data sent as an HTTP request

[0302] Data processing: Extracting question content and sentiment information from requests.

[0303] Output: Extracted question content and sentiment information

[0304] Step 5: Analyze the questions and emotional information.

[0305] The server uses natural language processing (NLP) tools to analyze the question content and sentiment information.

[0306] Input: Extracted question content and sentiment information

[0307] Data processing: Perform analysis using NLP tools (e.g., spaCy, NLTK)

[0308] Output: Main keywords, context, and sentiment information

[0309] Step 6: Generate an initial answer

[0310] The server sends the analysis results to the generative AI to generate an initial answer. The generative AI (e.g., OpenAI GPT-3) provides an answer based on the prompt.

[0311] Input: Main keywords, context, and sentiment information as analysis results

[0312] Data processing: The generative AI generates an initial answer based on the prompt

[0313] Output: Generated initial answer

[0314] Step 7: Present the initial answer to the expert answerer

[0315] The server presents the generated initial answer to a human expert answerer.

[0316] Input: Generated initial answer

[0317] Data processing: Send the initial answer to the expert answerer for viewing

[0318] Output: Display data for the expert answerer to review

[0319] Step 8: The expert answerer checks and modifies

[0320] Expert respondents review the initial responses and make corrections or additions as needed. For example, they may provide additional advice or specific instructions.

[0321] Input: Generated initial answer

[0322] Data processing: Expert respondents make corrections and additions.

[0323] Output: Feedback with corrections and additions.

[0324] Step 9: Integrate feedback and generate the final answer

[0325] The server incorporates feedback from expert respondents and integrates it with the initial responses to generate the final answer. User sentiment information is also taken into consideration.

[0326] Input: Initial responses and feedback from expert respondents

[0327] Data processing: Integrate feedback and initial responses to generate final responses that take emotional information into account.

[0328] Output: Final Answer

[0329] Step 10: Send the final response to the user's device.

[0330] The server sends the final response to the user's terminal as an HTTP response.

[0331] Input: Final answer

[0332] Data processing: Convert data to HTTP response format.

[0333] Output: Final response sent to the user's terminal

[0334] Step 11: Display the final answer to the user.

[0335] The device displays the final received response to the user. The user can receive a quick and accurate response.

[0336] Input: Final response sent from the server

[0337] Output: Final answer displayed to the user

[0338] Step 12: Add feedback to the training data of the generative AI.

[0339] The server adds feedback from expert respondents as training data for the generative AI.

[0340] Input: Feedback from expert respondents

[0341] Data processing: Convert the feedback into a format that can be used as training data.

[0342] Output: Add to the training data for generative AI.

[0343] Step 13: Update the generative AI model.

[0344] Generative AI uses the added feedback as training material for its model to improve the accuracy of subsequent response generation.

[0345] Input: Added feedback

[0346] Data processing: Improve accuracy through model retraining.

[0347] Output: Improved accuracy generative AI model

[0348] (Application Example 2)

[0349] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0350] Conventional user question answering systems often provide answers that do not take user emotions into consideration, resulting in an unsatisfactory user experience. Furthermore, the accuracy of initial responses provided by generative AI is inconsistent, frequently requiring human verification. This makes it difficult to provide quick and accurate answers, leading to decreased user satisfaction.

[0351] In Application Example 2, the specific processing performed by the specific processing unit 290 of the data processing device 12 is realized by the following means. In this invention, the server includes means for the user to input a question and recognize emotions; means for transmitting the question and recognized emotion information to the server; means for the server to transmit the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback, and generate a final answer considering the emotion information; and means for the server to transmit the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer that takes the user's emotions into consideration.

[0352] "A means for a user to input a question and recognize their emotions" refers to a device or software that has the function of analyzing the user's emotions when the user inputs a question into the terminal.

[0353] "Means for transmitting questions and recognized sentiment information to a server" refers to a device or software that has the function of transmitting questions entered from a terminal and sentiment information recognized by a sentiment engine to a server via a network.

[0354] "Generative AI" refers to a platform or system that includes artificial intelligence for generating initial responses based on user questions and sentiment information.

[0355] "Means for generating initial answers" refers to software that uses generative AI to create answers based on questions and sentiment information received by the server.

[0356] A "human expert respondent" is a person with specialized knowledge who can review the initial response generated by a generative AI and make corrections or additions as necessary.

[0357] "Means of receiving feedback" refers to a device or software that has the function of receiving corrections and supplementary information from human expert respondents.

[0358] "A means of integrating initial responses and feedback to generate a final response that takes emotional information into account" refers to software that combines the initial responses of a generative AI with feedback from human expert respondents, and takes the user's emotional information into consideration to create a final response.

[0359] "Means for sending the final answer to the user's terminal" refers to a device or software that has the function of sending the generated final answer to the user's terminal via a network and displaying it to the user.

[0360] The present invention is a system that provides answers to questions while taking user emotions into consideration. Detailed embodiments are described below.

[0361] System Configuration

[0362] The system of this invention begins by recognizing the emotions of the user when they input a question through a terminal. The system includes, as its main hardware and software, a terminal, a server, a generative AI, an emotion engine, and human expert respondents.

[0363] Hardware and software to be used

[0364] User terminal: A device such as a smartphone or tablet on which the user enters a question.

[0365] Server: A system that processes information submitted by users and manages data with generative AI and human expert respondents.

[0366] Generative AI models: Artificial intelligence used to generate initial responses based on questions and sentiment information. OpenAI's GPT-3 is used as an example.

[0367] Emotion engine: Software that recognizes emotions from user input. For example, EmotionEngine can be used.

[0368] Human expert respondents: Experts who review initial responses and make corrections or additions as needed.

[0369] Data processing and data calculation

[0370] 1. User question input and sentiment recognition:

[0371] When a user enters a question into the device, the emotion engine simultaneously recognizes that emotion. This is done by analyzing the patterns of text input or, in the case of voice input, the tone of voice.

[0372] 2. Sending questions and sentiment information to the server:

[0373] The recognized emotion information and question content are sent from the terminal to the server. HTTP requests are used for this process.

[0374] 3. Generation of initial answers using generative AI:

[0375] The server analyzes the received question and sentiment information and sends it to the generative AI model. The generative AI model generates an initial response based on this information.

[0376] 4. Feedback from human expert respondents:

[0377] The server presents the initial generated response to a human expert respondent. The expert respondent reviews the content and makes corrections or additions as needed.

[0378] 5. Generating the final answer:

[0379] The server integrates initial responses with feedback from human expert respondents and generates a final response that takes into account the user's emotional information.

[0380] 6. Sending the final response to the user's terminal:

[0381] The generated final response is then sent back to the user's device as an HTTP response.

[0382] Specific example

[0383] If a user enters a question via their smartphone, such as "I'm worried about the camera performance of this new smartphone, what do you think?", the emotion engine recognizes the user's concern. The question and emotion information are then sent to the server, and a generative AI model generates an initial response. In this case, the initial response would be "This smartphone's camera performance is very high and has received high ratings from many users." A human expert respondent then reviews this and makes a final revision, such as "This smartphone's camera performance is very high and has received high ratings from many users. If you have any concerns for specific uses, please let us know in detail."

[0384] Example of a prompt

[0385] User question: I'm worried about the camera performance of this new smartphone. What do you think?

[0386] User's emotion: Anxiety

[0387] Please provide the best answer.

[0388] Such a system makes it possible to provide quick and accurate responses that take into account the user's feelings, thereby improving the user experience.

[0389] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0390] Step 1:

[0391] The user enters a question into the device. Specifically, the user uses a device such as a smartphone or tablet to enter the question in text format. An example of input is, "I'm worried about the camera performance of this new smartphone, what do you think?" The entered question text is stored in the input field.

[0392] Step 2:

[0393] An emotion engine built into the device analyzes the user's questions and recognizes emotional information. Specifically, it determines whether the user is experiencing emotions such as anxiety, joy, or anger based on text analysis. If the input includes the phrase "I'm worried," the emotion engine identifies "anxiety." The output will then include the question text along with the emotional information "anxiety."

[0394] Step 3:

[0395] The question text and recognized sentiment information are sent from the terminal to the server. HTTP requests are used as the means of communication. Specifically, the terminal sends a request to the server containing the question and sentiment information as the payload. The input is the question text and sentiment information, and the output is the completion of the data transmission to the server.

[0396] Step 4:

[0397] The server analyzes the received question text and sentiment information. Specifically, the server uses natural language processing (NLP) techniques to extract key keywords and context from the question and format them, along with the sentiment information, into an input format for the generative AI model. The input is the question text and sentiment information, and the output is a prompt sentence for the generative AI model.

[0398] Step 5:

[0399] The server sends a prompt to a generative AI model to generate an initial response. Specifically, it sends a prompt to a generative AI model such as OpenAI GPT-3, which then generates a response. The input is the prompt, and the output is the initial response text from the generative AI model.

[0400] Step 6:

[0401] The server presents the generated initial response to a human expert respondent. Specifically, the initial response and sentiment information are displayed on a dedicated response confirmation interface. The input is the initial response text and sentiment information, and the output is the feedback from the human expert respondent.

[0402] Step 7:

[0403] Human expert respondents review the generated initial responses and make corrections or additions as needed. Specifically, expert respondents edit the initial responses on the interface and provide final feedback. The input is the initial response text and sentiment information, and the output is the reviewed and corrected response text.

[0404] Step 8:

[0405] The server integrates the initial response and feedback from human expert respondents, and generates a final response that takes sentiment into account. Specifically, the server merges the initial response and revised feedback, reflecting sentiment to create a response sentence that is appropriate for the user. The inputs are the initial response text, revised feedback, and sentiment information, and the output is the final response text.

[0406] Step 9:

[0407] The server sends the final response to the user's terminal. Specifically, it sends the final response as an HTTP response to the user's terminal. The input is the final response text, and the output is the response displayed on the user's terminal.

[0408] In this way, a system is in place to provide quick and accurate responses that take the user's feelings into consideration.

[0409] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0410] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0411] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0412] [Second Embodiment]

[0413] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0414] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0415] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0416] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0417] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0418] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0419] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0420] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0421] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0422] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0423] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0424] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0425] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. The processing of the system's program is described below in natural language, with specific examples.

[0426] Submit a question

[0427] The user enters a question on the device. For example, the user might type, "Why hasn't my cat been eating much lately?" The device receives this question and sends it to the server. The server temporarily stores the received question for processing.

[0428] Question analysis and initial response generation

[0429] The server analyzes the received question using natural language processing (NLP) techniques to understand its meaning. The analyzed question is then input into a generative AI. Based on the analysis results, the generative AI generates the optimal initial answer from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are some possible causes."

[0430] Checked by human expert respondents

[0431] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, the expert respondent might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0432] Integration and generation of the final answer

[0433] The server integrates initial responses from the generative AI with feedback from human expert respondents. This generates the final response to be provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[0434] Delivery of responses

[0435] The server sends the final response to the user's device. The user's device then displays the received final response to the user. This allows the user to obtain information quickly and accurately.

[0436] Feedback for improving AI accuracy

[0437] The server adds feedback from human expert respondents to the generative AI's training data. This allows the generative AI to continuously learn, improving the accuracy of its answers to subsequent questions.

[0438] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and also improves the accuracy of generative AI.

[0439] The following describes the processing flow.

[0440] Step 1:

[0441] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[0442] Step 2:

[0443] The terminal retrieves the entered question and sends it to the server. Specifically, it sends the question content to the server as an HTTP request.

[0444] Step 3:

[0445] The server receives a question and analyzes it using natural language processing (NLP) techniques. The analysis extracts the main keywords and context of the question.

[0446] Step 4:

[0447] The server sends the analyzed question to the generative AI. The generative AI generates the optimal initial answer from the training database based on the question.

[0448] Step 5:

[0449] The generative AI generates the initial response. For example, it might generate a response such as, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are possible causes."

[0450] Step 6:

[0451] The server receives the initial response from the generative AI and stores it temporarily.

[0452] Step 7:

[0453] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[0454] Step 8:

[0455] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[0456] Step 9:

[0457] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[0458] Step 10:

[0459] The server integrates the initial response and expert feedback to generate the final answer. For example, it might say, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to have its health checked."

[0460] Step 11:

[0461] The server sends the final response to the user's device. Specifically, it sends the final response to the device as an HTTP response.

[0462] Step 12:

[0463] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[0464] Step 13:

[0465] The server adds feedback from expert respondents to the training data of the generative AI. The generative AI uses the added feedback to update its model and improve the accuracy of its answers to subsequent questions.

[0466] (Example 1)

[0467] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0468] In today's information society, providing quick and accurate answers to the diverse questions users may have is crucial. However, conventional systems can be time-consuming to accurately analyze user questions and generate appropriate answers. Furthermore, the accuracy and reliability of the generated answers are not always high. This invention aims to solve these problems and provide users with quick and reliable answers.

[0469] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0470] In this invention, the server includes means for the user to input and send a question to the server; means for the server to analyze the received question using natural language processing technology and send it to a generative AI model; means for the generative AI model to generate an initial answer based on the analyzed question; means for the server to present the generated initial answer to an expert respondent and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; and means for the server to send the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer to the user's question.

[0471] A "user" is a person or entity that uses the system to input questions and receive answers.

[0472] A "device" is a device used by a user to input questions, and includes, for example, smartphones, tablets, and personal computers.

[0473] A "server" is a central computer system that analyzes questions received from users, generates answers using generative AI models, and sends those answers to users.

[0474] "Natural language processing technology" refers to techniques for analyzing text written in natural language and understanding its meaning, and includes, for example, morphological analysis and contextual analysis.

[0475] A "generative AI model" is an artificial intelligence model that generates appropriate answers based on analyzed questions, and includes, for example, machine learning algorithms and neural networks.

[0476] An "initial response" is the first answer that a generative AI model generates based on the user's question.

[0477] A "specialist respondent" is a person with the expertise to review the initial responses generated by generative AI models and make corrections or additions as needed.

[0478] "Feedback" refers to the corrections and supplementary information that expert respondents provide to their initial answers.

[0479] The "final answer" refers to the final response provided to the user, generated by integrating the initial response from a generative AI model with feedback from expert respondents.

[0480] "Training data" refers to data used by generative AI models to learn and improve the accuracy of their responses, and includes feedback from expert respondents.

[0481] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. This section describes the specific operation of this system and the hardware and software required to implement it.

[0482] First, the user enters a question using a device such as a smartphone or computer. For example, the user might enter, "Why hasn't my cat been eating much lately?" The device receives this input and sends it to the server via the internet.

[0483] The server temporarily stores the received questions and analyzes them using natural language processing (NLP) techniques. Examples of NLP techniques used include SpaCy and the Google Natural Language API. This analysis step extracts the question's structure and keywords, converting them into a format suitable for input into a generative AI model.

[0484] Next, the server sends the analyzed question to a generative AI model. For example, OpenAI's GPT-4 is used as the generative AI model. This model generates an initial answer from the training database based on the analysis results. For example, the generative AI model might answer, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food."

[0485] The generated initial response is presented to the expert respondent by the server. The expert respondent reviews this initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0486] The server integrates initial responses from generative AI models with feedback from expert respondents. This integration process generates the final response provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[0487] Finally, the server sends the final answer to the user's device. The device then displays the received final answer to the user. This allows the user to obtain information quickly and accurately.

[0488] Furthermore, the server adds feedback from expert respondents to the training data of the generative AI model. This addition of feedback allows the generative AI model to continuously learn, improving the accuracy of its answers to subsequent questions.

[0489] For example, the following prompt statement can be used:

[0490] "A user is asking why their cat hasn't been eating much lately. Analyze this question, generate an initial answer, and then have it checked by a human expert to generate the final answer."

[0491] In this way, the system can provide users with fast and reliable answers, and also improve the accuracy of generative AI models.

[0492] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0493] Step 1:

[0494] The user enters the question using a terminal.

[0495] Input: A question from the user. For example, "Why isn't my cat eating much lately?"

[0496] Specific operation: The user enters a question using a keyboard or touchscreen and presses the submit button.

[0497] Output: Input question data.

[0498] Step 2:

[0499] The terminal retrieves the entered question and sends it to the server.

[0500] Input: Question data entered by the user on their device.

[0501] Specific operation: The terminal sends the question data to the server via the network.

[0502] Output: Question data sent to the server.

[0503] Step 3:

[0504] The server temporarily stores the received questions and analyzes them using natural language processing techniques.

[0505] Input: Question data received by the server.

[0506] Specific operation: The server stores the question data in a database and performs analysis using natural language processing techniques (e.g., SpaCy or Google Natural Language API).

[0507] Output: Analyzed question data.

[0508] Step 4:

[0509] The server sends the analyzed question data to the AI ​​model to generate initial answers.

[0510] Input: Analyzed question data.

[0511] Specific operation: The server uses this data as input to a generating AI model (e.g., GPT-4). The generating AI model refers to the training database and generates an initial response. A response such as "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food" is generated.

[0512] Output: The initial response that was generated.

[0513] Step 5:

[0514] The server presents the generated initial answers to expert respondents and receives their feedback.

[0515] Input: The initial generated response.

[0516] Specific operation: The server sends the initial response to the expert respondent's terminal, who then reviews it. Feedback is provided with corrections and additions as needed. For example, feedback such as, "You should take your cat to the vet to have its health checked," might be added.

[0517] Output: Feedback from expert respondents.

[0518] Step 6:

[0519] The server integrates initial responses and feedback from expert respondents to generate the final response.

[0520] Input: Initial responses and feedback from expert respondents.

[0521] Specific operation: The server combines the initial response and feedback to perform an integrated process. This generates the final response to be provided to the user. For example, it might say, "There are various reasons why your cat isn't eating. Stress, illness, or dislike of the food are possible causes. It would be a good idea to take your cat to the vet to check its health."

[0522] Output: The final answer that was generated.

[0523] Step 7:

[0524] The server sends the final response to the user's device.

[0525] Input: The final generated response.

[0526] Specific operation: The server sends the final answer to the user's terminal via the network.

[0527] Output: The final response sent to the user's terminal.

[0528] Step 8:

[0529] The terminal displays the final response received to the user.

[0530] Input: Final response sent from the server.

[0531] Specific action: The final response received will be displayed on the device's screen.

[0532] Output: Users can view the final answer.

[0533] Step 9:

[0534] The server generates feedback from expert respondents, adds it to the training data of the AI ​​model, and improves the accuracy of the AI ​​model.

[0535] Input: Feedback from expert respondents.

[0536] Specific operation: The server adds the feedback data to the training database of the AI ​​model that generates the data, and then retrains the model. This improves the accuracy of the generated answers in subsequent attempts.

[0537] Output: Updated generative AI model.

[0538] (Application Example 1)

[0539] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0540] In modern food delivery services, it is difficult to provide reliable information quickly when users ask specific health-related questions. Furthermore, the accuracy and appropriateness of initial responses from generative AI cannot be guaranteed, potentially leading to a decline in service quality for users. In addition, previous systems failed to effectively utilize expert knowledge, resulting in delays in the learning process of generative AI.

[0541] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0542] In this invention, the server includes means for the user to input and send a question to the server; means for the server to send the received question to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a respondent with expert knowledge and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; means for the server to send the final answer to the user's information display device; means for the generative AI to continuously learn based on the feedback and improve the accuracy of the answer; and means for providing information about food in response to a question entered through the user's terminal. As a result, the user can quickly obtain reliable food delivery information and the accuracy of the generative AI is also improved, making it possible to provide a higher quality service.

[0543] A "user" refers to the entity that uses this system to input questions and obtain information.

[0544] A "server" refers to a central processing unit that receives a question, sends it to a generative AI, generates an initial answer, presents it to a respondent with specialized knowledge, receives feedback, generates a final answer, and sends it to the user.

[0545] "Generative AI" refers to artificial intelligence technology that creates initial answers based on user questions.

[0546] A "respondent with specialized knowledge" refers to a human expert who checks the initial response generated by the generative AI and makes corrections or additions.

[0547] "Feedback" refers to the corrections or additions made by respondents with specialized knowledge to the initial answers.

[0548] The "final answer" refers to the final answer generated by integrating the initial answers with feedback from respondents with specialized knowledge.

[0549] An "information display device" refers to a terminal or device used by a user to receive and display their final response.

[0550] "Continuous learning" refers to the process by which the server adds feedback to the generative AI as training data, thereby improving the accuracy of subsequent responses.

[0551] "Dietary information" refers to specific information about meal menus and nutrition related to the user's health.

[0552] In this invention, a user inputs a question about food using an information display device and sends it to a server. The server receives the question and sends it to a generative AI. The generative AI analyzes the received question using natural language processing technology and generates an initial answer. The open-source GPT-3 (OpenAI's text-davinci-003 engine) is used for natural language processing. The input question is in the format of "I've been wanting to eat healthier lately. What kind of menu do you recommend?"

[0553] The generated initial answers are presented to respondents with expert knowledge. These respondents review the initial answers generated by the generative AI and make appropriate corrections or additions. For example, if the generative AI generates the initial answer, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie," a respondent with expert knowledge might provide feedback such as, "These menu items offer balanced nutrition, are low in calories, and the smoothie is rich in vitamin C."

[0554] The server integrates initial responses from generative AI with feedback from experts to generate a final response. The final response incorporates the initial responses and feedback, for example, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C."

[0555] Finally, the server sends the final answer to the user's information display device. The user can then view the answer on the display device and obtain the necessary information. The server also adds feedback from expert respondents as training data for the generative AI, allowing the generative AI to continuously learn and improve the accuracy of its answers to subsequent questions.

[0556] As a concrete example of its use, consider the following prompt message:

[0557] "Please answer the user's question about healthy menus: I've recently started wanting to eat healthier meals. What kind of menus would you recommend?"

[0558] This invention allows users to quickly obtain reliable food delivery information and improves the accuracy of generative AI, enabling the provision of higher-quality services.

[0559] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0560] Step 1:

[0561] The user enters questions about their meal on an information display device and then submits those questions.

[0562] Input: A question entered by the user into the information display device (e.g., "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[0563] Output: The information display device sends the question content to the server.

[0564] Specific operation: The user operates the interface of the information display device, enters a question, and then presses the confirmation button. The information display device formats the input data and sends an HTTP request to the server endpoint.

[0565] Step 2:

[0566] The server sends the received question to a generative AI to generate an initial answer.

[0567] Input: Question data sent from the information display device (Example: "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[0568] Output: Initial response generated by a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[0569] Specific operation: The server receives an HTTP request and sends the question content to a natural language processing engine (e.g., OpenAI GPT-3). The generative AI analyzes the question and generates an initial answer from the training data. The generated answer is returned to the server and temporarily stored.

[0570] Step 3:

[0571] The server presents the initial generated answers to respondents with specialized knowledge and receives their feedback.

[0572] Input: Initial response from a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[0573] Output: Feedback from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0574] Specific operation: The server displays the initial response from the generative AI to an interface for experts. Experts review the initial response and input corrections or supplementary information. The corrected feedback is returned to the server.

[0575] Step 4:

[0576] The server integrates the initial response and feedback to generate the final response.

[0577] Input: Initial response from a generative AI and feedback from a knowledgeable respondent (e.g., "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie." + "These menu items offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0578] Output: Final Answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C.")

[0579] Specific operation: The server combines and integrates the initial response and feedback text data, and formats it into a single final response in text format.

[0580] Step 5:

[0581] The server sends the final response to the user's information display device.

[0582] Input: Integrated final answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie is especially rich in vitamin C.")

[0583] Output: Final answer displayed on the user's information display device.

[0584] Specific operation: The server sends the final answer back to the user's information display device as an HTTP response, and the terminal displays this answer with an appropriate UI (user interface).

[0585] Step 6:

[0586] The server continuously trains the generative AI based on feedback, improving the accuracy of its responses.

[0587] Input: Feedback data from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0588] Output: Improved generative AI model

[0589] Specific operation: The server adds the feedback data to the training dataset of the generative AI for the next question, and retrains the model, thereby improving the AI's accuracy.

[0590] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0591] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions. The following describes the program processing of this system in natural language, with specific examples.

[0592] Question submission and sentiment recognition

[0593] When a user enters a question on their device, the emotion engine recognizes the user's emotions. For example, if a user enters "Why hasn't my cat been eating much lately?", the emotion engine analyzes the user's tone of voice and input patterns at this point and recognizes that the user is feeling anxious.

[0594] The device sends the recognized sentiment information along with the question to the server. Specifically, it sends the question and sentiment information to the server as an HTTP request.

[0595] Question analysis and initial response generation

[0596] The server analyzes the received question and sentiment information using natural language processing (NLP) techniques. The analysis extracts key keywords, context, and sentiment information from the question.

[0597] The server sends the analyzed question and sentiment information to the generative AI. Based on this information, the generative AI generates the optimal initial response from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," while also considering the user's anxiety and adding a note at the end, "If you are concerned, we recommend consulting a professional."

[0598] Checked by human expert respondents

[0599] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0600] Integration and generation of the final answer

[0601] The server integrates initial responses from the generative AI with feedback from human expert respondents. Even in this process, user sentiment is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[0602] Delivery of responses

[0603] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response. The device then displays the received final response to the user. This allows the user to obtain quick and accurate information, as well as receive a response that is sensitive to their feelings.

[0604] Feedback for improving AI accuracy

[0605] The server adds feedback from human expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[0606] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[0607] The following describes the processing flow.

[0608] Step 1:

[0609] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[0610] Step 2:

[0611] The emotion engine analyzes user input in real time and recognizes the user's emotions. For example, the emotion engine can detect feelings of anxiety from the user's input speed and word choice.

[0612] Step 3:

[0613] The terminal retrieves the entered question and recognized sentiment information and sends it to the server. Specifically, it sends the question content and sentiment information to the server as an HTTP request.

[0614] Step 4:

[0615] The server receives the question and sentiment information, and analyzes the question using natural language processing (NLP) techniques. As a result of the analysis, the server extracts the main keywords, context, and sentiment information of the question.

[0616] Step 5:

[0617] The server sends the analyzed question and sentiment information to the generative AI. The generative AI considers the question content and sentiment information to generate the optimal initial response from the training database.

[0618] Step 6:

[0619] A generative AI generates an initial response. For example, it might generate a response like, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," and then, based on emotional information, add a supplement at the end of the sentence such as, "If you are concerned, we recommend consulting a specialist."

[0620] Step 7:

[0621] The server receives the initial response from the generative AI and stores it temporarily.

[0622] Step 8:

[0623] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[0624] Step 9:

[0625] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[0626] Step 10:

[0627] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[0628] Step 11:

[0629] The server integrates the initial response and feedback from expert respondents to generate the final response. Even in this process, user sentiment information is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to a veterinary clinic to have its health checked."

[0630] Step 12:

[0631] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response.

[0632] Step 13:

[0633] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[0634] Step 14:

[0635] The server adds feedback from expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[0636] (Example 2)

[0637] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0638] Conventional question-answering systems have struggled to provide answers that take user emotions into account, resulting in insufficient improvement in the user experience. Furthermore, the quality of initial answers relies solely on generative AI, leaving room for improvement in accuracy. Additionally, the generated answers are provided only once, preventing continuous learning of the generative AI, thus limiting long-term accuracy improvement.

[0639] The specific processing performed by the specific 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 the user to input a question and send it to the server; means for the terminal to recognize the user's emotions from the input question and generate emotion information; means for the server to send the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback and generate a final answer that takes emotion information into consideration; and means for the server to send the final answer to the user's terminal. This makes it possible to provide more accurate and reliable answers that take the user's emotions into consideration, and furthermore, by utilizing the feedback from the expert respondent as training data for the generative AI, it is possible to improve the accuracy of the entire system.

[0640] A "user" refers to a person who uses the system to input questions and receive answers.

[0641] A "server" refers to a device that receives questions and sentiment information from users, analyzes, processes, and sends this information to AI systems or expert respondents to generate the final answer.

[0642] A "terminal" refers to a device that a user directly operates to input questions, generate sentiment information, and send it to a server.

[0643] An "emotion engine" refers to software or an algorithm that recognizes emotions from questions entered by a user through a device and generates emotional information.

[0644] "Generative AI" refers to artificial intelligence models that generate initial responses based on questions and sentiment information sent from a server.

[0645] "Initial response" refers to the version of the response generated by a generative AI and provided to expert respondents.

[0646] A "specialist respondent" refers to an expert who reviews the initial response generated by generative AI and makes corrections or additions as needed.

[0647] "Feedback" refers to the corrections and supplementary information that expert respondents provide to the initial responses of generative AI systems.

[0648] "Final answer" refers to the final answer generated by the server, which integrates the initial answers and feedback from expert respondents.

[0649] Natural Language Processing (NLP) refers to a technology that analyzes questions entered by users through their devices and extracts their context and key keywords.

[0650] "Training data" refers to data that includes feedback from expert respondents and is used to improve the accuracy of generative AI.

[0651] "Improved accuracy" refers to the process where a generative AI updates its model using additional training data to increase the accuracy of its answers to subsequent questions.

[0652] The system based on this invention begins with the user inputting a question via a terminal and sending it to a server. The system of this invention involves an emotion engine, a generative AI, and human expert respondents who generate answers to the user's questions and provide the final answer.

[0653] Question input and sentiment recognition

[0654] Users input questions using their devices (smartphones, tablets, computers, etc.). For example, when a user inputs "Why hasn't my cat been eating much lately?", an emotion engine installed on the device (e.g., IBM Watson Natural Language Understanding) recognizes the user's emotions from the input text. In this case, the emotion engine analyzes the input patterns and context and generates emotion information such as "anxiety."

[0655] Sending and analyzing questions and sentiment information

[0656] The device sends the question and recognized sentiment information to the server as an HTTP request. The server analyzes the received question and sentiment information using natural language processing (NLP) techniques (e.g., spaCy, NLTK). As a result of the analysis, key keywords and context are extracted. For example, keywords such as "cat," "not eating," and "cause," along with the sentiment information "anxiety," are extracted.

[0657] Generating initial responses

[0658] The server sends the analyzed question and sentiment information to a generative AI (e.g., OpenAI GPT-3) to generate an initial response. Based on this input, the generative AI selects the best response from its training database. For example, the generative AI might generate the response, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food," and then address the anxiety by adding a supplementary sentence such as, "If you are concerned, we recommend consulting a specialist."

[0659] Review and correction by expert respondents

[0660] The generated initial response is presented to human expert respondents via the server. The expert respondents review the initial response and make corrections or additions as needed. For example, an expert respondent might add feedback such as, "You should take your cat to the vet to have its health checked."

[0661] Integration and generation of the final answer

[0662] The server integrates initial responses from the generative AI with feedback from expert respondents to generate a final response. User sentiment information is also taken into consideration during this process. For example, a final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[0663] Submit and display the final response.

[0664] The server sends the final response to the user's device as an HTTP response. The device then displays the received final response to the user, allowing the user to obtain quick and accurate information while also receiving an emotionally sensitive response.

[0665] Feedback for improving AI accuracy

[0666] Furthermore, the server adds feedback from expert respondents as training data for the generative AI, updating the AI ​​model to improve the accuracy of answers to subsequent questions. Since user sentiment information is also included in the training data, the accuracy of responses to emotions also improves.

[0667] For example, if a user inputs, "My child had a fight with a friend at school and is upset. What should I do?", the emotion engine recognizes the user's concern and assigns "concern" as emotional information. Based on this, the generative AI generates a response such as, "It's natural for your child to learn about friendships through fighting. It's important to listen to them calmly. Also, if necessary, it would be a good idea to talk to the school counselor."

[0668] Examples of prompts for a generative AI model:

[0669] User question: "My child is upset because they had a fight with a friend at school. What should I do?"

[0670] Emotional information: "Worried"

[0671] The above describes specific embodiments of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[0672] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0673] Step 1: The user enters a question.

[0674] The user enters a question using the terminal. The entered question is captured by the terminal as text data. For example, the user enters "Why hasn't my cat been eating much lately?"

[0675] Input: Question text entered by the user

[0676] Output: Question data imported into the terminal

[0677] Step 2: The device recognizes emotions.

[0678] The emotion engine installed on the device recognizes the user's emotions from the entered question text. Through text analysis, the emotion engine recognizes that the user is feeling anxious.

[0679] Input: Captured question text

[0680] Data processing: The emotion engine performs text analysis (e.g., using IBM Watson Natural Language Understanding).

[0681] Output: Recognized emotional information (e.g., anxiety)

[0682] Step 3: Send the question and sentiment information to the server.

[0683] The device sends the question content and recognized sentiment information to the server as an HTTP request.

[0684] Input: Questionnaire data and sentiment information

[0685] Data processing: Converting data into HTTP request format.

[0686] Output: Data sent to the server

[0687] Step 4: The server receives the question and sentiment information.

[0688] The server receives HTTP requests sent from the terminal and retrieves the question content and sentiment information.

[0689] Input: Data sent as an HTTP request

[0690] Data processing: Extracting question content and sentiment information from requests.

[0691] Output: Extracted question content and sentiment information

[0692] Step 5: Analyze the questions and emotional information.

[0693] The server uses natural language processing (NLP) tools to analyze the question content and sentiment information.

[0694] Input: Extracted question content and sentiment information

[0695] Data processing: Analysis is performed using NLP tools (e.g., spaCy, NLTK).

[0696] Output: Key keywords and context, sentiment information

[0697] Step 6: Generate initial response

[0698] The server sends the analysis results to the generative AI, which generates an initial response. The generative AI (e.g., OpenAI GPT-3) then provides the response based on the prompt.

[0699] Input: Key keywords, context, and sentiment information as analysis results.

[0700] Data processing: Generative AI generates initial responses based on prompts.

[0701] Output: Generated initial answer

[0702] Step 7: Present the initial answer to the expert respondent.

[0703] The server presents the generated initial answers to human expert respondents.

[0704] Input: Generated initial answer

[0705] Data processing: Send initial responses to expert respondents and display them.

[0706] Output: Display data for expert respondents to review.

[0707] Step 8: Expert respondents review and make corrections.

[0708] Expert respondents review the initial responses and make corrections or additions as needed. For example, they may provide additional advice or specific instructions.

[0709] Input: Generated initial answer

[0710] Data processing: Expert respondents make corrections and additions.

[0711] Output: Feedback with corrections and additions.

[0712] Step 9: Integrate feedback and generate the final answer

[0713] The server incorporates feedback from expert respondents and integrates it with the initial responses to generate the final answer. User sentiment information is also taken into consideration.

[0714] Input: Initial responses and feedback from expert respondents

[0715] Data processing: Integrate feedback and initial responses to generate final responses that take emotional information into account.

[0716] Output: Final Answer

[0717] Step 10: Send the final response to the user's device.

[0718] The server sends the final response to the user's terminal as an HTTP response.

[0719] Input: Final answer

[0720] Data processing: Convert data to HTTP response format.

[0721] Output: Final response sent to the user's terminal

[0722] Step 11: Display the final answer to the user.

[0723] The device displays the final received response to the user. The user can receive a quick and accurate response.

[0724] Input: Final response sent from the server

[0725] Output: Final answer displayed to the user

[0726] Step 12: Add feedback to the training data of the generative AI.

[0727] The server adds feedback from expert respondents as training data for the generative AI.

[0728] Input: Feedback from expert respondents

[0729] Data processing: Convert the feedback into a format that can be used as training data.

[0730] Output: Add to the training data for generative AI.

[0731] Step 13: Update the generative AI model.

[0732] Generative AI uses the added feedback as training material for its model to improve the accuracy of subsequent response generation.

[0733] Input: Added feedback

[0734] Data processing: Improve accuracy through model retraining.

[0735] Output: Improved accuracy generative AI model

[0736] (Application Example 2)

[0737] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0738] Conventional user question answering systems often provide answers that do not take user emotions into consideration, resulting in an unsatisfactory user experience. Furthermore, the accuracy of initial responses provided by generative AI is inconsistent, frequently requiring human verification. This makes it difficult to provide quick and accurate answers, leading to decreased user satisfaction.

[0739] In Application Example 2, the specific processing performed by the specific processing unit 290 of the data processing device 12 is realized by the following means. In this invention, the server includes means for the user to input a question and recognize emotions; means for transmitting the question and recognized emotion information to the server; means for the server to transmit the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback, and generate a final answer considering the emotion information; and means for the server to transmit the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer that takes the user's emotions into consideration.

[0740] "A means for a user to input a question and recognize their emotions" refers to a device or software that has the function of analyzing the user's emotions when the user inputs a question into the terminal.

[0741] "Means for transmitting questions and recognized sentiment information to a server" refers to a device or software that has the function of transmitting questions entered from a terminal and sentiment information recognized by the sentiment engine to a server via a network.

[0742] "Generative AI" refers to a platform or system that includes artificial intelligence for generating initial responses based on user questions and sentiment information.

[0743] "Means for generating initial answers" refers to software that uses generative AI to create answers based on questions and sentiment information received by the server.

[0744] A "human expert respondent" is a person with specialized knowledge who can review the initial response generated by a generative AI and make corrections or additions as necessary.

[0745] "Means of receiving feedback" refers to a device or software that has the function of receiving corrections and supplementary information from human expert respondents.

[0746] "A means of integrating initial responses and feedback to generate a final response that takes emotional information into account" refers to software that combines the initial responses of a generative AI with feedback from human expert respondents, and takes the user's emotional information into consideration to create a final response.

[0747] "Means for sending the final answer to the user's terminal" refers to a device or software that has the function of sending the generated final answer to the user's terminal via a network and displaying it to the user.

[0748] The present invention is a system that provides answers to questions while taking user emotions into consideration. Detailed embodiments are described below.

[0749] System Configuration

[0750] The system of this invention begins by recognizing the emotions of the user when they input a question through a terminal. The system includes, as its main hardware and software, a terminal, a server, a generative AI, an emotion engine, and human expert respondents.

[0751] Hardware and software to be used

[0752] User terminal: A device such as a smartphone or tablet on which the user enters a question.

[0753] Server: A system that processes information submitted by users and manages data with generative AI and human expert respondents.

[0754] Generative AI models: Artificial intelligence used to generate initial responses based on questions and sentiment information. OpenAI's GPT-3 is used as an example.

[0755] Emotion engine: Software used to recognize emotions from user input. For example, EmotionEngine can be used.

[0756] Human expert respondents: Experts who review initial responses and make corrections or additions as needed.

[0757] Data processing and data calculation

[0758] 1. User question input and sentiment recognition:

[0759] When a user enters a question into the device, the emotion engine simultaneously recognizes that emotion. This is done by analyzing the patterns of text input or, in the case of voice input, the tone of voice.

[0760] 2. Sending questions and sentiment information to the server:

[0761] The recognized emotion information and question content are sent from the terminal to the server. HTTP requests are used for this process.

[0762] 3. Generation of initial answers using generative AI:

[0763] The server analyzes the received question and sentiment information and sends it to the generative AI model. The generative AI model generates an initial response based on this information.

[0764] 4. Feedback from human expert respondents:

[0765] The server presents the initial generated response to a human expert respondent. The expert respondent reviews the content and makes corrections or additions as needed.

[0766] 5. Generating the final answer:

[0767] The server integrates initial responses with feedback from human expert respondents and generates a final response that takes into account the user's emotional information.

[0768] 6. Sending the final response to the user's terminal:

[0769] The generated final response is then sent back to the user's device as an HTTP response.

[0770] Specific example

[0771] If a user enters a question via their smartphone, such as "I'm worried about the camera performance of this new smartphone, what do you think?", the emotion engine recognizes the user's concern. The question and emotion information are then sent to the server, and a generative AI model generates an initial response. In this case, the initial response would be "This smartphone's camera performance is very high and has received high ratings from many users." A human expert respondent then reviews this and makes a final revision, such as "This smartphone's camera performance is very high and has received high ratings from many users. If you have any concerns for specific uses, please let us know in detail."

[0772] Example of a prompt

[0773] User question: I'm worried about the camera performance of this new smartphone. What do you think?

[0774] User's emotion: Anxiety

[0775] Please provide the best answer.

[0776] Such a system makes it possible to provide quick and accurate responses that take into account the user's feelings, thereby improving the user experience.

[0777] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0778] Step 1:

[0779] The user enters a question into the device. Specifically, the user uses a device such as a smartphone or tablet to enter the question in text format. An example of input is, "I'm worried about the camera performance of this new smartphone, what do you think?" The entered question text is stored in the input field.

[0780] Step 2:

[0781] An emotion engine built into the device analyzes the user's questions and recognizes emotional information. Specifically, it determines whether the user is experiencing emotions such as anxiety, joy, or anger based on text analysis. If the input includes the phrase "I'm worried," the emotion engine identifies "anxiety." The output will then include the question text along with the emotional information "anxiety."

[0782] Step 3:

[0783] The question text and recognized sentiment information are sent from the terminal to the server. HTTP requests are used as the means of communication. Specifically, the terminal sends a request to the server containing the question and sentiment information as the payload. The input is the question text and sentiment information, and the output is the completion of the data transmission to the server.

[0784] Step 4:

[0785] The server analyzes the received question text and sentiment information. Specifically, the server uses natural language processing (NLP) techniques to extract key keywords and context from the question and format them, along with the sentiment information, into an input format for the generative AI model. The input is the question text and sentiment information, and the output is a prompt sentence for the generative AI model.

[0786] Step 5:

[0787] The server sends a prompt to a generative AI model to generate an initial response. Specifically, it sends a prompt to a generative AI model such as OpenAI GPT-3, which then generates a response. The input is the prompt, and the output is the initial response text from the generative AI model.

[0788] Step 6:

[0789] The server presents the generated initial response to a human expert respondent. Specifically, the initial response and sentiment information are displayed on a dedicated response confirmation interface. The input is the initial response text and sentiment information, and the output is the feedback from the human expert respondent.

[0790] Step 7:

[0791] Human expert respondents review the generated initial responses and make corrections or additions as needed. Specifically, expert respondents edit the initial responses on the interface and provide final feedback. The input is the initial response text and sentiment information, and the output is the reviewed and corrected response text.

[0792] Step 8:

[0793] The server integrates the initial response and feedback from human expert respondents, and generates a final response that takes sentiment into account. Specifically, the server merges the initial response and revised feedback, reflecting sentiment to create a response sentence that is appropriate for the user. The inputs are the initial response text, revised feedback, and sentiment information, and the output is the final response text.

[0794] Step 9:

[0795] The server sends the final response to the user's terminal. Specifically, it sends the final response as an HTTP response to the user's terminal. The input is the final response text, and the output is the response displayed on the user's terminal.

[0796] In this way, a system is in place to provide quick and accurate responses that take the user's feelings into consideration.

[0797] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0798] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0799] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0800] [Third Embodiment]

[0801] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0802] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0803] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0804] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0805] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0806] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0807] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0808] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0809] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0810] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0811] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0812] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0813] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. The processing of the system's program is described below in natural language, with specific examples.

[0814] Submit a question

[0815] The user enters a question on the device. For example, the user might type, "Why hasn't my cat been eating much lately?" The device receives this question and sends it to the server. The server temporarily stores the received question for processing.

[0816] Question analysis and initial response generation

[0817] The server analyzes the received question using natural language processing (NLP) techniques to understand its meaning. The analyzed question is then input into a generative AI. Based on the analysis results, the generative AI generates the optimal initial answer from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are some possible causes."

[0818] Checked by human expert respondents

[0819] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, the expert respondent might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0820] Integration and generation of the final answer

[0821] The server integrates initial responses from the generative AI with feedback from human expert respondents. This generates the final response to be provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[0822] Delivery of responses

[0823] The server sends the final response to the user's device. The user's device then displays the received final response to the user. This allows the user to obtain information quickly and accurately.

[0824] Feedback for improving AI accuracy

[0825] The server adds feedback from human expert respondents to the generative AI's training data. This allows the generative AI to continuously learn, improving the accuracy of its answers to subsequent questions.

[0826] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and also improves the accuracy of generative AI.

[0827] The following describes the processing flow.

[0828] Step 1:

[0829] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[0830] Step 2:

[0831] The terminal retrieves the entered question and sends it to the server. Specifically, it sends the question content to the server as an HTTP request.

[0832] Step 3:

[0833] The server receives a question and analyzes it using natural language processing (NLP) techniques. The analysis extracts the main keywords and context of the question.

[0834] Step 4:

[0835] The server sends the analyzed question to the generative AI. The generative AI generates the optimal initial answer from the training database based on the question.

[0836] Step 5:

[0837] The generative AI generates the initial response. For example, it might generate a response such as, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are possible causes."

[0838] Step 6:

[0839] The server receives the initial response from the generative AI and stores it temporarily.

[0840] Step 7:

[0841] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[0842] Step 8:

[0843] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[0844] Step 9:

[0845] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[0846] Step 10:

[0847] The server integrates the initial response and expert feedback to generate the final answer. For example, it might say, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to have its health checked."

[0848] Step 11:

[0849] The server sends the final response to the user's device. Specifically, it sends the final response to the device as an HTTP response.

[0850] Step 12:

[0851] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[0852] Step 13:

[0853] The server adds feedback from expert respondents to the training data of the generative AI. The generative AI uses the added feedback to update its model and improve the accuracy of its answers to subsequent questions.

[0854] (Example 1)

[0855] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0856] In today's information society, providing quick and accurate answers to the diverse questions users may have is crucial. However, conventional systems can be time-consuming to accurately analyze user questions and generate appropriate answers. Furthermore, the accuracy and reliability of the generated answers are not always high. This invention aims to solve these problems and provide users with quick and reliable answers.

[0857] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0858] In this invention, the server includes means for the user to input and send a question to the server; means for the server to analyze the received question using natural language processing technology and send it to a generative AI model; means for the generative AI model to generate an initial answer based on the analyzed question; means for the server to present the generated initial answer to an expert respondent and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; and means for the server to send the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer to the user's question.

[0859] A "user" is a person or entity that uses the system to input questions and receive answers.

[0860] A "device" is a device used by a user to input questions, and includes, for example, smartphones, tablets, and personal computers.

[0861] A "server" is a central computer system that analyzes questions received from users, generates answers using generative AI models, and sends those answers to users.

[0862] "Natural language processing technology" refers to techniques for analyzing text written in natural language and understanding its meaning, and includes, for example, morphological analysis and contextual analysis.

[0863] A "generative AI model" is an artificial intelligence model that generates appropriate answers based on analyzed questions, and includes, for example, machine learning algorithms and neural networks.

[0864] An "initial response" is the first answer that a generative AI model generates based on the user's question.

[0865] A "specialist respondent" is a person with the expertise to review the initial responses generated by generative AI models and make corrections or additions as needed.

[0866] "Feedback" refers to the corrections and supplementary information that expert respondents provide to their initial answers.

[0867] The "final answer" refers to the final response provided to the user, generated by integrating the initial response from a generative AI model with feedback from expert respondents.

[0868] "Training data" refers to data used by generative AI models to learn and improve the accuracy of their responses, and includes feedback from expert respondents.

[0869] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. This section describes the specific operation of this system and the hardware and software required to implement it.

[0870] First, the user enters a question using a device such as a smartphone or computer. For example, the user might enter, "Why hasn't my cat been eating much lately?" The device receives this input and sends it to the server via the internet.

[0871] The server temporarily stores the received questions and analyzes them using natural language processing (NLP) techniques. Examples of NLP techniques used include SpaCy and the Google Natural Language API. This analysis step extracts the question's structure and keywords, converting them into a format suitable for input into a generative AI model.

[0872] Next, the server sends the analyzed question to a generative AI model. For example, OpenAI's GPT-4 is used as the generative AI model. This model generates an initial answer from the training database based on the analysis results. For example, the generative AI model might answer, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food."

[0873] The generated initial response is presented to the expert respondent by the server. The expert respondent reviews this initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0874] The server integrates initial responses from generative AI models with feedback from expert respondents. This integration process generates the final response provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[0875] Finally, the server sends the final answer to the user's device. The device then displays the received final answer to the user. This allows the user to obtain information quickly and accurately.

[0876] Furthermore, the server adds feedback from expert respondents to the training data of the generative AI model. This addition of feedback allows the generative AI model to continuously learn, improving the accuracy of its answers to subsequent questions.

[0877] For example, the following prompt statement can be used:

[0878] "A user is asking why their cat hasn't been eating much lately. Analyze this question, generate an initial answer, and then have it checked by a human expert to generate the final answer."

[0879] In this way, the system can provide users with fast and reliable answers, and also improve the accuracy of generative AI models.

[0880] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0881] Step 1:

[0882] The user enters the question using a terminal.

[0883] Input: A question from the user. For example, "Why isn't my cat eating much lately?"

[0884] Specific operation: The user enters a question using a keyboard or touchscreen and presses the submit button.

[0885] Output: Input question data.

[0886] Step 2:

[0887] The terminal retrieves the entered question and sends it to the server.

[0888] Input: Question data entered by the user on their device.

[0889] Specific operation: The terminal sends the question data to the server via the network.

[0890] Output: Question data sent to the server.

[0891] Step 3:

[0892] The server temporarily stores the received questions and analyzes them using natural language processing techniques.

[0893] Input: Question data received by the server.

[0894] Specific operation: The server stores the question data in a database and performs analysis using natural language processing techniques (e.g., SpaCy or Google Natural Language API).

[0895] Output: Analyzed question data.

[0896] Step 4:

[0897] The server sends the analyzed question data to the AI ​​model to generate initial answers.

[0898] Input: Analyzed question data.

[0899] Specific operation: The server uses this data as input to a generating AI model (e.g., GPT-4). The generating AI model refers to the training database and generates an initial response. A response such as "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food" is generated.

[0900] Output: The initial response that was generated.

[0901] Step 5:

[0902] The server presents the generated initial answers to expert respondents and receives their feedback.

[0903] Input: The initial generated response.

[0904] Specific operation: The server sends the initial response to the expert respondent's terminal, who then reviews it. Feedback is provided with corrections and additions as needed. For example, feedback such as, "You should take your cat to the vet to have its health checked," might be added.

[0905] Output: Feedback from expert respondents.

[0906] Step 6:

[0907] The server integrates initial responses and feedback from expert respondents to generate the final response.

[0908] Input: Initial responses and feedback from expert respondents.

[0909] Specific operation: The server combines the initial response and feedback to perform an integrated process. This generates the final response to be provided to the user. For example, it might say, "There are various reasons why your cat isn't eating. Stress, illness, or dislike of the food are possible causes. It would be a good idea to take your cat to the vet to check its health."

[0910] Output: The final answer that was generated.

[0911] Step 7:

[0912] The server sends the final response to the user's device.

[0913] Input: The final generated response.

[0914] Specific operation: The server sends the final answer to the user's terminal via the network.

[0915] Output: The final response sent to the user's terminal.

[0916] Step 8:

[0917] The terminal displays the final response received to the user.

[0918] Input: Final response sent from the server.

[0919] Specific action: The final response received will be displayed on the device's screen.

[0920] Output: Users can view the final answer.

[0921] Step 9:

[0922] The server generates feedback from expert respondents, adds it to the training data of the AI ​​model, and improves the accuracy of the AI ​​model.

[0923] Input: Feedback from expert respondents.

[0924] Specific operation: The server adds the feedback data to the training database of the AI ​​model that generates the data, and then retrains the model. This improves the accuracy of the generated answers in subsequent attempts.

[0925] Output: Updated generative AI model.

[0926] (Application Example 1)

[0927] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0928] In modern food delivery services, it is difficult to provide reliable information quickly when users ask specific health-related questions. Furthermore, the accuracy and appropriateness of initial responses from generative AI cannot be guaranteed, potentially leading to a decline in service quality for users. In addition, previous systems failed to effectively utilize expert knowledge, resulting in delays in the learning process of generative AI.

[0929] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0930] In this invention, the server includes means for the user to input and send a question to the server; means for the server to send the received question to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a respondent with expert knowledge and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; means for the server to send the final answer to the user's information display device; means for the generative AI to continuously learn based on the feedback and improve the accuracy of the answer; and means for providing information about food in response to a question entered through the user's terminal. As a result, the user can quickly obtain reliable food delivery information and the accuracy of the generative AI is also improved, making it possible to provide a higher quality service.

[0931] A "user" refers to the entity that uses this system to input questions and obtain information.

[0932] A "server" refers to a central processing unit that receives a question, sends it to a generative AI, generates an initial answer, presents it to a respondent with specialized knowledge, receives feedback, generates a final answer, and sends it to the user.

[0933] "Generative AI" refers to artificial intelligence technology that creates initial answers based on user questions.

[0934] A "respondent with specialized knowledge" refers to a human expert who checks the initial response generated by the generative AI and makes corrections or additions.

[0935] "Feedback" refers to the corrections or additions made by respondents with specialized knowledge to the initial answers.

[0936] The "final answer" refers to the final answer generated by integrating the initial answers with feedback from respondents with specialized knowledge.

[0937] An "information display device" refers to a terminal or device used by a user to receive and display their final response.

[0938] "Continuous learning" refers to the process by which the server adds feedback to the generative AI as training data, thereby improving the accuracy of subsequent responses.

[0939] "Dietary information" refers to specific information about meal menus and nutrition related to the user's health.

[0940] In this invention, a user inputs a question about food using an information display device and sends it to a server. The server receives the question and sends it to a generative AI. The generative AI analyzes the received question using natural language processing technology and generates an initial answer. The open-source GPT-3 (OpenAI's text-davinci-003 engine) is used for natural language processing. The input question is in the format of "I've been wanting to eat healthier lately. What kind of menu do you recommend?"

[0941] The generated initial answers are presented to respondents with expert knowledge. These respondents review the initial answers generated by the generative AI and make appropriate corrections or additions. For example, if the generative AI generates the initial answer, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie," a respondent with expert knowledge might provide feedback such as, "These menu items offer balanced nutrition, are low in calories, and the smoothie is rich in vitamin C."

[0942] The server integrates initial responses from generative AI with feedback from experts to generate a final response. The final response incorporates the initial responses and feedback, for example, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C."

[0943] Finally, the server sends the final answer to the user's information display device. The user can then view the answer on the display device and obtain the necessary information. The server also adds feedback from expert respondents as training data for the generative AI, allowing the generative AI to continuously learn and improve the accuracy of its answers to subsequent questions.

[0944] As a concrete example of its use, consider the following prompt message:

[0945] "Please answer the user's question about healthy menus: I've recently started wanting to eat healthier meals. What kind of menus would you recommend?"

[0946] This invention allows users to quickly obtain reliable food delivery information and improves the accuracy of generative AI, enabling the provision of higher-quality services.

[0947] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0948] Step 1:

[0949] The user enters questions about their meal on an information display device and then submits those questions.

[0950] Input: A question entered by the user into the information display device (e.g., "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[0951] Output: The information display device sends the question content to the server.

[0952] Specific operation: The user operates the interface of the information display device, enters a question, and then presses the confirmation button. The information display device formats the input data and sends an HTTP request to the server endpoint.

[0953] Step 2:

[0954] The server sends the received question to a generative AI to generate an initial answer.

[0955] Input: Question data sent from the information display device (Example: "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[0956] Output: Initial response generated by a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[0957] Specific operation: The server receives an HTTP request and sends the question content to a natural language processing engine (e.g., OpenAI GPT-3). The generative AI analyzes the question and generates an initial answer from the training data. The generated answer is returned to the server and temporarily stored.

[0958] Step 3:

[0959] The server presents the initial generated answers to respondents with specialized knowledge and receives their feedback.

[0960] Input: Initial response from a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[0961] Output: Feedback from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0962] Specific operation: The server displays the initial response from the generative AI to an interface for experts. Experts review the initial response and input corrections or supplementary information. The corrected feedback is returned to the server.

[0963] Step 4:

[0964] The server integrates the initial response and feedback to generate the final response.

[0965] Input: Initial response from a generative AI and feedback from a knowledgeable respondent (e.g., "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie." + "These menu items offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0966] Output: Final Answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C.")

[0967] Specific operation: The server combines and integrates the initial response and feedback text data, and formats it into a single final response in text format.

[0968] Step 5:

[0969] The server sends the final response to the user's information display device.

[0970] Input: Integrated final answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie is especially rich in vitamin C.")

[0971] Output: Final answer displayed on the user's information display device.

[0972] Specific operation: The server sends the final answer back to the user's information display device as an HTTP response, and the terminal displays this answer with an appropriate UI (user interface).

[0973] Step 6:

[0974] The server continuously trains the generative AI based on feedback, improving the accuracy of its responses.

[0975] Input: Feedback data from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[0976] Output: Improved generative AI model

[0977] Specific operation: The server adds the feedback data to the training dataset of the generative AI for the next question, and retrains the model, thereby improving the AI's accuracy.

[0978] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0979] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions. The following describes the program processing of this system in natural language, with specific examples.

[0980] Question submission and sentiment recognition

[0981] When a user enters a question on their device, the emotion engine recognizes the user's emotions. For example, if a user enters "Why hasn't my cat been eating much lately?", the emotion engine analyzes the user's tone of voice and input patterns at this point and recognizes that the user is feeling anxious.

[0982] The device sends the recognized sentiment information along with the question to the server. Specifically, it sends the question and sentiment information to the server as an HTTP request.

[0983] Question analysis and initial response generation

[0984] The server analyzes the received question and sentiment information using natural language processing (NLP) techniques. The analysis extracts key keywords, context, and sentiment information from the question.

[0985] The server sends the analyzed question and sentiment information to the generative AI. Based on this information, the generative AI generates the optimal initial response from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," while also considering the user's anxiety and adding a note at the end, "If you are concerned, we recommend consulting a professional."

[0986] Checked by human expert respondents

[0987] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[0988] Integration and generation of the final answer

[0989] The server integrates initial responses from the generative AI with feedback from human expert respondents. Even in this process, user sentiment is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[0990] Delivery of responses

[0991] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response. The device then displays the received final response to the user. This allows the user to obtain quick and accurate information, as well as receive a response that is sensitive to their feelings.

[0992] Feedback for improving AI accuracy

[0993] The server adds feedback from human expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[0994] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[0995] The following describes the processing flow.

[0996] Step 1:

[0997] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[0998] Step 2:

[0999] The emotion engine analyzes user input in real time and recognizes the user's emotions. For example, the emotion engine can detect feelings of anxiety from the user's input speed and word choice.

[1000] Step 3:

[1001] The terminal retrieves the entered question and recognized sentiment information and sends it to the server. Specifically, it sends the question content and sentiment information to the server as an HTTP request.

[1002] Step 4:

[1003] The server receives the question and sentiment information, and analyzes the question using natural language processing (NLP) techniques. As a result of the analysis, the server extracts the main keywords, context, and sentiment information of the question.

[1004] Step 5:

[1005] The server sends the analyzed question and sentiment information to the generative AI. The generative AI considers the question content and sentiment information to generate the optimal initial response from the training database.

[1006] Step 6:

[1007] A generative AI generates an initial response. For example, it might generate a response like, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," and then, based on emotional information, add a supplement at the end of the sentence such as, "If you are concerned, we recommend consulting a specialist."

[1008] Step 7:

[1009] The server receives the initial response from the generative AI and stores it temporarily.

[1010] Step 8:

[1011] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[1012] Step 9:

[1013] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[1014] Step 10:

[1015] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[1016] Step 11:

[1017] The server integrates the initial response and feedback from expert respondents to generate the final response. Even in this process, user sentiment information is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to a veterinary clinic to have its health checked."

[1018] Step 12:

[1019] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response.

[1020] Step 13:

[1021] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[1022] Step 14:

[1023] The server adds feedback from expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[1024] (Example 2)

[1025] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[1026] Conventional question-answering systems have struggled to provide answers that take user emotions into account, resulting in insufficient improvement in the user experience. Furthermore, the quality of initial answers relies solely on generative AI, leaving room for improvement in accuracy. Additionally, the generated answers are provided only once, preventing continuous learning of the generative AI, thus limiting long-term accuracy improvement.

[1027] The specific processing performed by the specific 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 the user to input a question and send it to the server; means for the terminal to recognize the user's emotions from the input question and generate emotion information; means for the server to send the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback and generate a final answer that takes emotion information into consideration; and means for the server to send the final answer to the user's terminal. This makes it possible to provide more accurate and reliable answers that take the user's emotions into consideration, and furthermore, by utilizing the feedback from the expert respondent as training data for the generative AI, it is possible to improve the accuracy of the entire system.

[1028] A "user" refers to a person who uses the system to input questions and receive answers.

[1029] A "server" refers to a device that receives questions and sentiment information from users, analyzes, processes, and sends this information to AI systems or expert respondents to generate the final answer.

[1030] A "terminal" refers to a device that a user directly operates to input questions, generate sentiment information, and send it to a server.

[1031] An "emotion engine" refers to software or an algorithm that recognizes emotions from questions entered by a user through a device and generates emotional information.

[1032] "Generative AI" refers to artificial intelligence models that generate initial responses based on questions and sentiment information sent from a server.

[1033] "Initial response" refers to the version of the response generated by a generative AI and provided to expert respondents.

[1034] A "specialist respondent" refers to an expert who reviews the initial response generated by generative AI and makes corrections or additions as needed.

[1035] "Feedback" refers to the corrections and supplementary information that expert respondents provide to the initial responses of generative AI systems.

[1036] "Final answer" refers to the final answer generated by the server, which integrates the initial answers and feedback from expert respondents.

[1037] Natural Language Processing (NLP) refers to a technology that analyzes questions entered by users through their devices and extracts their context and key keywords.

[1038] "Training data" refers to data that includes feedback from expert respondents and is used to improve the accuracy of generative AI.

[1039] "Improved accuracy" refers to the process where a generative AI updates its model using additional training data to increase the accuracy of its answers to subsequent questions.

[1040] The system based on this invention begins with the user inputting a question via a terminal and sending it to a server. The system of this invention involves an emotion engine, a generative AI, and human expert respondents who generate answers to the user's questions and provide the final answer.

[1041] Question input and sentiment recognition

[1042] Users input questions using their devices (smartphones, tablets, computers, etc.). For example, when a user inputs "Why hasn't my cat been eating much lately?", an emotion engine installed on the device (e.g., IBM Watson Natural Language Understanding) recognizes the user's emotions from the input text. In this case, the emotion engine analyzes the input patterns and context and generates emotion information such as "anxiety."

[1043] Sending and analyzing questions and sentiment information

[1044] The device sends the question and recognized sentiment information to the server as an HTTP request. The server analyzes the received question and sentiment information using natural language processing (NLP) techniques (e.g., spaCy, NLTK). As a result of the analysis, key keywords and context are extracted. For example, keywords such as "cat," "not eating," and "cause," along with the sentiment information "anxiety," are extracted.

[1045] Generating initial responses

[1046] The server sends the analyzed question and sentiment information to a generative AI (e.g., OpenAI GPT-3) to generate an initial response. Based on this input, the generative AI selects the best response from its training database. For example, the generative AI might generate the response, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food," and then address the anxiety by adding a supplementary sentence such as, "If you are concerned, we recommend consulting a specialist."

[1047] Review and correction by expert respondents

[1048] The generated initial response is presented to human expert respondents via the server. The expert respondents review the initial response and make corrections or additions as needed. For example, an expert respondent might add feedback such as, "You should take your cat to the vet to have its health checked."

[1049] Integration and generation of the final answer

[1050] The server integrates initial responses from the generative AI with feedback from expert respondents to generate a final response. User sentiment information is also taken into consideration during this process. For example, a final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[1051] Submit and display the final response.

[1052] The server sends the final response to the user's device as an HTTP response. The device then displays the received final response to the user, allowing the user to obtain quick and accurate information while also receiving an emotionally sensitive response.

[1053] Feedback for improving AI accuracy

[1054] Furthermore, the server adds feedback from expert respondents as training data for the generative AI, updating the AI ​​model to improve the accuracy of answers to subsequent questions. Since user sentiment information is also included in the training data, the accuracy of responses to emotions also improves.

[1055] For example, if a user inputs, "My child had a fight with a friend at school and is upset. What should I do?", the emotion engine recognizes the user's concern and assigns "concern" as emotional information. Based on this, the generative AI generates a response such as, "It's natural for your child to learn about friendships through fighting. It's important to listen to them calmly. Also, if necessary, it would be a good idea to talk to the school counselor."

[1056] Examples of prompts for a generative AI model:

[1057] User question: "My child is upset because they had a fight with a friend at school. What should I do?"

[1058] Emotional information: "Worried"

[1059] The above describes specific embodiments of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[1060] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1061] Step 1: The user enters a question.

[1062] The user enters a question using the terminal. The entered question is captured by the terminal as text data. For example, the user enters "Why hasn't my cat been eating much lately?"

[1063] Input: Question text entered by the user

[1064] Output: Question data imported into the terminal

[1065] Step 2: The device recognizes emotions.

[1066] The emotion engine installed on the device recognizes the user's emotions from the entered question text. Through text analysis, the emotion engine recognizes that the user is feeling anxious.

[1067] Input: Captured question text

[1068] Data processing: The emotion engine performs text analysis (e.g., using IBM Watson Natural Language Understanding).

[1069] Output: Recognized emotional information (e.g., anxiety)

[1070] Step 3: Send the question and sentiment information to the server.

[1071] The device sends the question content and recognized sentiment information to the server as an HTTP request.

[1072] Input: Questionnaire data and sentiment information

[1073] Data processing: Converting data into HTTP request format.

[1074] Output: Data sent to the server

[1075] Step 4: The server receives the question and sentiment information.

[1076] The server receives HTTP requests sent from the terminal and retrieves the question content and sentiment information.

[1077] Input: Data sent as an HTTP request

[1078] Data processing: Extracting question content and sentiment information from requests.

[1079] Output: Extracted question content and sentiment information

[1080] Step 5: Analyze the questions and emotional information.

[1081] The server uses natural language processing (NLP) tools to analyze the question content and sentiment information.

[1082] Input: Extracted question content and sentiment information

[1083] Data processing: Analysis is performed using NLP tools (e.g., spaCy, NLTK).

[1084] Output: Key keywords and context, sentiment information

[1085] Step 6: Generate initial response

[1086] The server sends the analysis results to the generative AI, which generates an initial response. The generative AI (e.g., OpenAI GPT-3) then provides the response based on the prompt.

[1087] Input: Key keywords, context, and sentiment information as analysis results.

[1088] Data processing: Generative AI generates initial responses based on prompts.

[1089] Output: Generated initial answer

[1090] Step 7: Present the initial answer to the expert respondent.

[1091] The server presents the generated initial answers to human expert respondents.

[1092] Input: Generated initial answer

[1093] Data processing: Send initial responses to expert respondents and display them.

[1094] Output: Display data for expert respondents to review.

[1095] Step 8: Expert respondents review and make corrections.

[1096] Expert respondents review the initial responses and make corrections or additions as needed. For example, they may provide additional advice or specific instructions.

[1097] Input: Generated initial answer

[1098] Data processing: Expert respondents make corrections and additions.

[1099] Output: Feedback with corrections and additions.

[1100] Step 9: Integrate feedback and generate the final answer

[1101] The server incorporates feedback from expert respondents and integrates it with the initial responses to generate the final answer. User sentiment information is also taken into consideration.

[1102] Input: Initial responses and feedback from expert respondents

[1103] Data processing: Integrate feedback and initial responses to generate final responses that take emotional information into account.

[1104] Output: Final Answer

[1105] Step 10: Send the final response to the user's device.

[1106] The server sends the final response to the user's terminal as an HTTP response.

[1107] Input: Final answer

[1108] Data processing: Convert data to HTTP response format.

[1109] Output: Final response sent to the user's terminal

[1110] Step 11: Display the final answer to the user.

[1111] The device displays the final received response to the user. The user can receive a quick and accurate response.

[1112] Input: Final response sent from the server

[1113] Output: Final answer displayed to the user

[1114] Step 12: Add feedback to the training data of the generative AI.

[1115] The server adds feedback from expert respondents as training data for the generative AI.

[1116] Input: Feedback from expert respondents

[1117] Data processing: Convert the feedback into a format that can be used as training data.

[1118] Output: Add to the training data for generative AI.

[1119] Step 13: Update the generative AI model.

[1120] Generative AI uses the added feedback as training material for its model to improve the accuracy of subsequent response generation.

[1121] Input: Added feedback

[1122] Data processing: Improve accuracy through model retraining.

[1123] Output: Improved accuracy generative AI model

[1124] (Application Example 2)

[1125] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[1126] Conventional user question answering systems often provide answers that do not take user emotions into consideration, resulting in an unsatisfactory user experience. Furthermore, the accuracy of initial responses provided by generative AI is inconsistent, frequently requiring human verification. This makes it difficult to provide quick and accurate answers, leading to decreased user satisfaction.

[1127] In Application Example 2, the specific processing performed by the specific processing unit 290 of the data processing device 12 is realized by the following means. In this invention, the server includes means for the user to input a question and recognize emotions; means for transmitting the question and recognized emotion information to the server; means for the server to transmit the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback, and generate a final answer considering the emotion information; and means for the server to transmit the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer that takes the user's emotions into consideration.

[1128] "A means for a user to input a question and recognize their emotions" refers to a device or software that has the function of analyzing the user's emotions when the user inputs a question into the terminal.

[1129] "Means for transmitting questions and recognized sentiment information to a server" refers to a device or software that has the function of transmitting questions entered from a terminal and sentiment information recognized by the sentiment engine to a server via a network.

[1130] "Generative AI" refers to a platform or system that includes artificial intelligence for generating initial responses based on user questions and sentiment information.

[1131] "Means for generating initial answers" refers to software that uses generative AI to create answers based on questions and sentiment information received by the server.

[1132] A "human expert respondent" is a person with specialized knowledge who can review the initial response generated by a generative AI and make corrections or additions as necessary.

[1133] "Means of receiving feedback" refers to a device or software that has the function of receiving corrections and supplementary information from human expert respondents.

[1134] "A means of integrating initial responses and feedback to generate a final response that takes emotional information into account" refers to software that combines the initial responses of a generative AI with feedback from human expert respondents, and takes the user's emotional information into consideration to create a final response.

[1135] "Means for sending the final answer to the user's terminal" refers to a device or software that has the function of sending the generated final answer to the user's terminal via a network and displaying it to the user.

[1136] The present invention is a system that provides answers to questions while taking user emotions into consideration. Detailed embodiments are described below.

[1137] System Configuration

[1138] The system of this invention begins by recognizing the emotions of the user when they input a question through a terminal. The system includes, as its main hardware and software, a terminal, a server, a generative AI, an emotion engine, and human expert respondents.

[1139] Hardware and software to be used

[1140] User terminal: A device such as a smartphone or tablet on which the user enters a question.

[1141] Server: A system that processes information submitted by users and manages data with generative AI and human expert respondents.

[1142] Generative AI models: Artificial intelligence used to generate initial responses based on questions and sentiment information. OpenAI's GPT-3 is used as an example.

[1143] Emotion engine: Software used to recognize emotions from user input. For example, EmotionEngine can be used.

[1144] Human expert respondents: Experts who review initial responses and make corrections or additions as needed.

[1145] Data processing and data calculation

[1146] 1. User question input and sentiment recognition:

[1147] When a user enters a question into the device, the emotion engine simultaneously recognizes that emotion. This is done by analyzing the patterns of text input or, in the case of voice input, the tone of voice.

[1148] 2. Sending questions and sentiment information to the server:

[1149] The recognized emotion information and question content are sent from the terminal to the server. HTTP requests are used for this process.

[1150] 3. Generation of initial answers using generative AI:

[1151] The server analyzes the received question and sentiment information and sends it to the generative AI model. The generative AI model generates an initial response based on this information.

[1152] 4. Feedback from human expert respondents:

[1153] The server presents the initial generated response to a human expert respondent. The expert respondent reviews the content and makes corrections or additions as needed.

[1154] 5. Generating the final answer:

[1155] The server integrates initial responses with feedback from human expert respondents and generates a final response that takes into account the user's emotional information.

[1156] 6. Sending the final response to the user's terminal:

[1157] The generated final response is then sent back to the user's device as an HTTP response.

[1158] Specific example

[1159] If a user enters a question via their smartphone, such as "I'm worried about the camera performance of this new smartphone, what do you think?", the emotion engine recognizes the user's concern. The question and emotion information are then sent to the server, and a generative AI model generates an initial response. In this case, the initial response would be "This smartphone's camera performance is very high and has received high ratings from many users." A human expert respondent then reviews this and makes a final revision, such as "This smartphone's camera performance is very high and has received high ratings from many users. If you have any concerns for specific uses, please let us know in detail."

[1160] Example of a prompt

[1161] User question: I'm worried about the camera performance of this new smartphone. What do you think?

[1162] User's emotion: Anxiety

[1163] Please provide the best answer.

[1164] Such a system makes it possible to provide quick and accurate responses that take into account the user's feelings, thereby improving the user experience.

[1165] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1166] Step 1:

[1167] The user enters a question into the device. Specifically, the user uses a device such as a smartphone or tablet to enter the question in text format. An example of input is, "I'm worried about the camera performance of this new smartphone, what do you think?" The entered question text is stored in the input field.

[1168] Step 2:

[1169] An emotion engine built into the device analyzes the user's questions and recognizes emotional information. Specifically, it determines whether the user is experiencing emotions such as anxiety, joy, or anger based on text analysis. If the input includes the phrase "I'm worried," the emotion engine identifies "anxiety." The output will then include the question text along with the emotional information "anxiety."

[1170] Step 3:

[1171] The question text and recognized sentiment information are sent from the terminal to the server. HTTP requests are used as the means of communication. Specifically, the terminal sends a request to the server containing the question and sentiment information as the payload. The input is the question text and sentiment information, and the output is the completion of the data transmission to the server.

[1172] Step 4:

[1173] The server analyzes the received question text and sentiment information. Specifically, the server uses natural language processing (NLP) techniques to extract key keywords and context from the question and format them, along with the sentiment information, into an input format for the generative AI model. The input is the question text and sentiment information, and the output is a prompt sentence for the generative AI model.

[1174] Step 5:

[1175] The server sends a prompt to a generative AI model to generate an initial response. Specifically, it sends a prompt to a generative AI model such as OpenAI GPT-3, which then generates a response. The input is the prompt, and the output is the initial response text from the generative AI model.

[1176] Step 6:

[1177] The server presents the generated initial response to a human expert respondent. Specifically, the initial response and sentiment information are displayed on a dedicated response confirmation interface. The input is the initial response text and sentiment information, and the output is the feedback from the human expert respondent.

[1178] Step 7:

[1179] Human expert respondents review the generated initial responses and make corrections or additions as needed. Specifically, expert respondents edit the initial responses on the interface and provide final feedback. The input is the initial response text and sentiment information, and the output is the reviewed and corrected response text.

[1180] Step 8:

[1181] The server integrates the initial response and feedback from human expert respondents, and generates a final response that takes sentiment into account. Specifically, the server merges the initial response and revised feedback, reflecting sentiment to create a response sentence that is appropriate for the user. The inputs are the initial response text, revised feedback, and sentiment information, and the output is the final response text.

[1182] Step 9:

[1183] The server sends the final response to the user's terminal. Specifically, it sends the final response as an HTTP response to the user's terminal. The input is the final response text, and the output is the response displayed on the user's terminal.

[1184] In this way, a system is in place to provide quick and accurate responses that take the user's feelings into consideration.

[1185] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1186] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1187] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[1188] [Fourth Embodiment]

[1189] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[1190] As shown in Figure 7, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1191] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1192] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[1193] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[1194] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[1195] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[1196] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[1197] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[1198] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1199] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1200] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[1201] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1202] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. The processing of the system's program is described below in natural language, with specific examples.

[1203] Submit a question

[1204] The user enters a question on the device. For example, the user might type, "Why hasn't my cat been eating much lately?" The device receives this question and sends it to the server. The server temporarily stores the received question for processing.

[1205] Question analysis and initial response generation

[1206] The server analyzes the received question using natural language processing (NLP) techniques to understand its meaning. The analyzed question is then input into a generative AI. Based on the analysis results, the generative AI generates the optimal initial answer from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are some possible causes."

[1207] Checked by human expert respondents

[1208] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, the expert respondent might provide feedback such as, "You should take your cat to the vet to have its health checked."

[1209] Integration and generation of the final answer

[1210] The server integrates initial responses from the generative AI with feedback from human expert respondents. This generates the final response to be provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[1211] Delivery of responses

[1212] The server sends the final response to the user's device. The user's device then displays the received final response to the user. This allows the user to obtain information quickly and accurately.

[1213] Feedback for improving AI accuracy

[1214] The server adds feedback from human expert respondents to the generative AI's training data. This allows the generative AI to continuously learn, improving the accuracy of its answers to subsequent questions.

[1215] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and also improves the accuracy of generative AI.

[1216] The following describes the processing flow.

[1217] Step 1:

[1218] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[1219] Step 2:

[1220] The terminal retrieves the entered question and sends it to the server. Specifically, it sends the question content to the server as an HTTP request.

[1221] Step 3:

[1222] The server receives a question and analyzes it using natural language processing (NLP) techniques. The analysis extracts the main keywords and context of the question.

[1223] Step 4:

[1224] The server sends the analyzed question to the generative AI. The generative AI generates the optimal initial answer from the training database based on the question.

[1225] Step 5:

[1226] The generative AI generates the initial response. For example, it might generate a response such as, "There are various reasons why a cat might not eat. Stress, illness, or disliking the food are possible causes."

[1227] Step 6:

[1228] The server receives the initial response from the generative AI and stores it temporarily.

[1229] Step 7:

[1230] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[1231] Step 8:

[1232] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[1233] Step 9:

[1234] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[1235] Step 10:

[1236] The server integrates the initial response and expert feedback to generate the final answer. For example, it might say, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to have its health checked."

[1237] Step 11:

[1238] The server sends the final response to the user's device. Specifically, it sends the final response to the device as an HTTP response.

[1239] Step 12:

[1240] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[1241] Step 13:

[1242] The server adds feedback from expert respondents to the training data of the generative AI. The generative AI uses the added feedback to update its model and improve the accuracy of its answers to subsequent questions.

[1243] (Example 1)

[1244] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1245] In today's information society, providing quick and accurate answers to the diverse questions users may have is crucial. However, conventional systems can be time-consuming to accurately analyze user questions and generate appropriate answers. Furthermore, the accuracy and reliability of the generated answers are not always high. This invention aims to solve these problems and provide users with quick and reliable answers.

[1246] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[1247] In this invention, the server includes means for the user to input and send a question to the server; means for the server to analyze the received question using natural language processing technology and send it to a generative AI model; means for the generative AI model to generate an initial answer based on the analyzed question; means for the server to present the generated initial answer to an expert respondent and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; and means for the server to send the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer to the user's question.

[1248] A "user" is a person or entity that uses the system to input questions and receive answers.

[1249] A "device" is a device used by a user to input questions, and includes, for example, smartphones, tablets, and personal computers.

[1250] A "server" is a central computer system that analyzes questions received from users, generates answers using generative AI models, and sends those answers to users.

[1251] "Natural language processing technology" refers to techniques for analyzing text written in natural language and understanding its meaning, and includes, for example, morphological analysis and contextual analysis.

[1252] A "generative AI model" is an artificial intelligence model that generates appropriate answers based on analyzed questions, and includes, for example, machine learning algorithms and neural networks.

[1253] An "initial response" is the first answer that a generative AI model generates based on the user's question.

[1254] A "specialist respondent" is a person with the expertise to review the initial responses generated by generative AI models and make corrections or additions as needed.

[1255] "Feedback" refers to the corrections and supplementary information that expert respondents provide to their initial answers.

[1256] The "final answer" refers to the final response provided to the user, generated by integrating the initial response from a generative AI model with feedback from expert respondents.

[1257] "Training data" refers to data used by generative AI models to learn and improve the accuracy of their responses, and includes feedback from expert respondents.

[1258] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. This section describes the specific operation of this system and the hardware and software required to implement it.

[1259] First, the user enters a question using a device such as a smartphone or computer. For example, the user might enter, "Why hasn't my cat been eating much lately?" The device receives this input and sends it to the server via the internet.

[1260] The server temporarily stores the received questions and analyzes them using natural language processing (NLP) techniques. Examples of NLP techniques used include SpaCy and the Google Natural Language API. This analysis step extracts the question's structure and keywords, converting them into a format suitable for input into a generative AI model.

[1261] Next, the server sends the analyzed question to a generative AI model. For example, OpenAI's GPT-4 is used as the generative AI model. This model generates an initial answer from the training database based on the analysis results. For example, the generative AI model might answer, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food."

[1262] The generated initial response is presented to the expert respondent by the server. The expert respondent reviews this initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[1263] The server integrates initial responses from generative AI models with feedback from expert respondents. This integration process generates the final response provided to the user. For example, the final response might be, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. It would be best to take your cat to the vet to check its health."

[1264] Finally, the server sends the final answer to the user's device. The device then displays the received final answer to the user. This allows the user to obtain information quickly and accurately.

[1265] Furthermore, the server adds feedback from expert respondents to the training data of the generative AI model. This addition of feedback allows the generative AI model to continuously learn, improving the accuracy of its answers to subsequent questions.

[1266] For example, the following prompt statement can be used:

[1267] "A user is asking why their cat hasn't been eating much lately. Analyze this question, generate an initial answer, and then have it checked by a human expert to generate the final answer."

[1268] In this way, the system can provide users with fast and reliable answers, and also improve the accuracy of generative AI models.

[1269] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1270] Step 1:

[1271] The user enters the question using a terminal.

[1272] Input: A question from the user. For example, "Why isn't my cat eating much lately?"

[1273] Specific operation: The user enters a question using a keyboard or touchscreen and presses the submit button.

[1274] Output: Input question data.

[1275] Step 2:

[1276] The terminal retrieves the entered question and sends it to the server.

[1277] Input: Question data entered by the user on their device.

[1278] Specific operation: The terminal sends the question data to the server via the network.

[1279] Output: Question data sent to the server.

[1280] Step 3:

[1281] The server temporarily stores the received questions and analyzes them using natural language processing techniques.

[1282] Input: Question data received by the server.

[1283] Specific operation: The server stores the question data in a database and performs analysis using natural language processing techniques (e.g., SpaCy or Google Natural Language API).

[1284] Output: Analyzed question data.

[1285] Step 4:

[1286] The server sends the analyzed question data to the AI ​​model to generate initial answers.

[1287] Input: Analyzed question data.

[1288] Specific operation: The server uses this data as input to a generating AI model (e.g., GPT-4). The generating AI model refers to the training database and generates an initial response. A response such as "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food" is generated.

[1289] Output: The initial response that was generated.

[1290] Step 5:

[1291] The server presents the generated initial answers to expert respondents and receives their feedback.

[1292] Input: The initial generated response.

[1293] Specific operation: The server sends the initial response to the expert respondent's terminal, who then reviews it. Feedback is provided with corrections and additions as needed. For example, feedback such as, "You should take your cat to the vet to have its health checked," might be added.

[1294] Output: Feedback from expert respondents.

[1295] Step 6:

[1296] The server integrates initial responses and feedback from expert respondents to generate the final response.

[1297] Input: Initial responses and feedback from expert respondents.

[1298] Specific operation: The server combines the initial response and feedback to perform an integrated process. This generates the final response to be provided to the user. For example, it might say, "There are various reasons why your cat isn't eating. Stress, illness, or dislike of the food are possible causes. It would be a good idea to take your cat to the vet to check its health."

[1299] Output: The final answer that was generated.

[1300] Step 7:

[1301] The server sends the final response to the user's device.

[1302] Input: The final generated response.

[1303] Specific operation: The server sends the final answer to the user's terminal via the network.

[1304] Output: The final response sent to the user's terminal.

[1305] Step 8:

[1306] The terminal displays the final response received to the user.

[1307] Input: Final response sent from the server.

[1308] Specific action: The final response received will be displayed on the device's screen.

[1309] Output: Users can view the final answer.

[1310] Step 9:

[1311] The server generates feedback from expert respondents, adds it to the training data of the AI ​​model, and improves the accuracy of the AI ​​model.

[1312] Input: Feedback from expert respondents.

[1313] Specific operation: The server adds the feedback data to the training database of the AI ​​model that generates the data, and then retrains the model. This improves the accuracy of the generated answers in subsequent attempts.

[1314] Output: Updated generative AI model.

[1315] (Application Example 1)

[1316] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1317] In modern food delivery services, it is difficult to provide reliable information quickly when users ask specific health-related questions. Furthermore, the accuracy and appropriateness of initial responses from generative AI cannot be guaranteed, potentially leading to a decline in service quality for users. In addition, previous systems failed to effectively utilize expert knowledge, resulting in delays in the learning process of generative AI.

[1318] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[1319] In this invention, the server includes means for the user to input and send a question to the server; means for the server to send the received question to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a respondent with expert knowledge and receive feedback; means for the server to integrate the initial answer and feedback to generate a final answer; means for the server to send the final answer to the user's information display device; means for the generative AI to continuously learn based on the feedback and improve the accuracy of the answer; and means for providing information about food in response to a question entered through the user's terminal. As a result, the user can quickly obtain reliable food delivery information and the accuracy of the generative AI is also improved, making it possible to provide a higher quality service.

[1320] A "user" refers to the entity that uses this system to input questions and obtain information.

[1321] A "server" refers to a central processing unit that receives a question, sends it to a generative AI, generates an initial answer, presents it to a respondent with specialized knowledge, receives feedback, generates a final answer, and sends it to the user.

[1322] "Generative AI" refers to artificial intelligence technology that creates initial answers based on user questions.

[1323] A "respondent with specialized knowledge" refers to a human expert who checks the initial response generated by the generative AI and makes corrections or additions.

[1324] "Feedback" refers to the corrections or additions made by respondents with specialized knowledge to the initial answers.

[1325] The "final answer" refers to the final answer generated by integrating the initial answers with feedback from respondents with specialized knowledge.

[1326] An "information display device" refers to a terminal or device used by a user to receive and display their final response.

[1327] "Continuous learning" refers to the process by which the server adds feedback to the generative AI as training data, thereby improving the accuracy of subsequent responses.

[1328] "Dietary information" refers to specific information about meal menus and nutrition related to the user's health.

[1329] In this invention, a user inputs a question about food using an information display device and sends it to a server. The server receives the question and sends it to a generative AI. The generative AI analyzes the received question using natural language processing technology and generates an initial answer. The open-source GPT-3 (OpenAI's text-davinci-003 engine) is used for natural language processing. The input question is in the format of "I've been wanting to eat healthier lately. What kind of menu do you recommend?"

[1330] The generated initial answers are presented to respondents with expert knowledge. These respondents review the initial answers generated by the generative AI and make appropriate corrections or additions. For example, if the generative AI generates the initial answer, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie," a respondent with expert knowledge might provide feedback such as, "These menu items offer balanced nutrition, are low in calories, and the smoothie is rich in vitamin C."

[1331] The server integrates initial responses from generative AI with feedback from experts to generate a final response. The final response incorporates the initial responses and feedback, for example, "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C."

[1332] Finally, the server sends the final answer to the user's information display device. The user can then view the answer on the display device and obtain the necessary information. The server also adds feedback from expert respondents as training data for the generative AI, allowing the generative AI to continuously learn and improve the accuracy of its answers to subsequent questions.

[1333] As a concrete example of its use, consider the following prompt message:

[1334] "Please answer the user's question about healthy menus: I've recently started wanting to eat healthier meals. What kind of menus would you recommend?"

[1335] This invention allows users to quickly obtain reliable food delivery information and improves the accuracy of generative AI, enabling the provision of higher-quality services.

[1336] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1337] Step 1:

[1338] The user enters questions about their meal on an information display device and then submits those questions.

[1339] Input: A question entered by the user into the information display device (e.g., "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[1340] Output: The information display device sends the question content to the server.

[1341] Specific operation: The user operates the interface of the information display device, enters a question, and then presses the confirmation button. The information display device formats the input data and sends an HTTP request to the server endpoint.

[1342] Step 2:

[1343] The server sends the received question to a generative AI to generate an initial answer.

[1344] Input: Question data sent from the information display device (Example: "I've been wanting to eat healthier lately. What kind of menu would you recommend?")

[1345] Output: Initial response generated by a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[1346] Specific operation: The server receives an HTTP request and sends the question content to a natural language processing engine (e.g., OpenAI GPT-3). The generative AI analyzes the question and generates an initial answer from the training data. The generated answer is returned to the server and temporarily stored.

[1347] Step 3:

[1348] The server presents the initial generated answers to respondents with specialized knowledge and receives their feedback.

[1349] Input: Initial response from a generative AI (Example: "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie.")

[1350] Output: Feedback from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[1351] Specific operation: The server displays the initial response from the generative AI to an interface for experts. Experts review the initial response and input corrections or supplementary information. The corrected feedback is returned to the server.

[1352] Step 4:

[1353] The server integrates the initial response and feedback to generate the final response.

[1354] Input: Initial response from a generative AI and feedback from a knowledgeable respondent (e.g., "Recommended healthy menu items are grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie." + "These menu items offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[1355] Output: Final Answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie, in particular, is rich in vitamin C.")

[1356] Specific operation: The server combines and integrates the initial response and feedback text data, and formats it into a single final response in text format.

[1357] Step 5:

[1358] The server sends the final response to the user's information display device.

[1359] Input: Integrated final answer (Example: "My recommended healthy menu is grilled chicken salad, mushroom and spinach pasta, and a healthy smoothie. These provide balanced nutrition, are low in calories, and the smoothie is especially rich in vitamin C.")

[1360] Output: Final answer displayed on the user's information display device.

[1361] Specific operation: The server sends the final answer back to the user's information display device as an HTTP response, and the terminal displays this answer with an appropriate UI (user interface).

[1362] Step 6:

[1363] The server continuously trains the generative AI based on feedback, improving the accuracy of its responses.

[1364] Input: Feedback data from respondents with expert knowledge (e.g., "These menus offer balanced nutrition and are low in calories. Smoothies, in particular, are rich in vitamin C.")

[1365] Output: Improved generative AI model

[1366] Specific operation: The server adds the feedback data to the training dataset of the generative AI for the next question, and retrains the model, thereby improving the AI's accuracy.

[1367] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1368] The system based on this invention begins with the user entering a question on a terminal and sending it to the server. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions. The following describes the program processing of this system in natural language, with specific examples.

[1369] Question submission and sentiment recognition

[1370] When a user enters a question on their device, the emotion engine recognizes the user's emotions. For example, if a user enters "Why hasn't my cat been eating much lately?", the emotion engine analyzes the user's tone of voice and input patterns at this point and recognizes that the user is feeling anxious.

[1371] The device sends the recognized sentiment information along with the question to the server. Specifically, it sends the question and sentiment information to the server as an HTTP request.

[1372] Question analysis and initial response generation

[1373] The server analyzes the received question and sentiment information using natural language processing (NLP) techniques. The analysis extracts key keywords, context, and sentiment information from the question.

[1374] The server sends the analyzed question and sentiment information to the generative AI. Based on this information, the generative AI generates the optimal initial response from its training database. For example, the generative AI might respond, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," while also considering the user's anxiety and adding a note at the end, "If you are concerned, we recommend consulting a professional."

[1375] Checked by human expert respondents

[1376] The server presents the generated initial response to a human expert respondent. The expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "You should take your cat to the vet to have its health checked."

[1377] Integration and generation of the final answer

[1378] The server integrates initial responses from the generative AI with feedback from human expert respondents. Even in this process, user sentiment is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[1379] Delivery of responses

[1380] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response. The device then displays the received final response to the user. This allows the user to obtain quick and accurate information, as well as receive a response that is sensitive to their feelings.

[1381] Feedback for improving AI accuracy

[1382] The server adds feedback from human expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[1383] The above describes a specific embodiment of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[1384] The following describes the processing flow.

[1385] Step 1:

[1386] The user enters a question on their device. For example, the user might type, "Why hasn't my cat been eating much lately?"

[1387] Step 2:

[1388] The emotion engine analyzes user input in real time and recognizes the user's emotions. For example, the emotion engine can detect feelings of anxiety from the user's input speed and word choice.

[1389] Step 3:

[1390] The terminal retrieves the entered question and recognized sentiment information and sends it to the server. Specifically, it sends the question content and sentiment information to the server as an HTTP request.

[1391] Step 4:

[1392] The server receives the question and sentiment information, and analyzes the question using natural language processing (NLP) techniques. As a result of the analysis, the server extracts the main keywords, context, and sentiment information of the question.

[1393] Step 5:

[1394] The server sends the analyzed question and sentiment information to the generative AI. The generative AI considers the question content and sentiment information to generate the optimal initial response from the training database.

[1395] Step 6:

[1396] A generative AI generates an initial response. For example, it might generate a response like, "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes," and then, based on emotional information, add a supplement at the end of the sentence such as, "If you are concerned, we recommend consulting a specialist."

[1397] Step 7:

[1398] The server receives the initial response from the generative AI and stores it temporarily.

[1399] Step 8:

[1400] The server presents an initial answer to a human expert respondent. Specifically, it displays the initial answer on an interface for the expert respondent and requests their confirmation.

[1401] Step 9:

[1402] A human expert respondent reviews the initial response and makes corrections or additions as needed. For example, they might provide feedback such as, "It would be a good idea to take your cat to the vet to check its health."

[1403] Step 10:

[1404] The server receives feedback from expert respondents and processes it to integrate the initial response with the feedback.

[1405] Step 11:

[1406] The server integrates the initial response and feedback from expert respondents to generate the final response. Even in this process, user sentiment information is taken into consideration when generating the final response. For example, the final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to a veterinary clinic to have its health checked."

[1407] Step 12:

[1408] The server sends the final response to the user's device. Specifically, it sends the final response as an HTTP response.

[1409] Step 13:

[1410] The device receives the final answer and displays it to the user. The user can then verify the final answer displayed on the device.

[1411] Step 14:

[1412] The server adds feedback from expert respondents to the generative AI's training data. The generative AI uses the added feedback to update its model, improving the accuracy of its answers to subsequent questions. User sentiment information is also added to the training data, improving the accuracy of sentiment handling.

[1413] (Example 2)

[1414] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1415] Conventional question-answering systems have struggled to provide answers that take user emotions into account, resulting in insufficient improvement in the user experience. Furthermore, the quality of initial answers relies solely on generative AI, leaving room for improvement in accuracy. Additionally, the generated answers are provided only once, preventing continuous learning of the generative AI, thus limiting long-term accuracy improvement.

[1416] The specific processing performed by the specific 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 the user to input a question and send it to the server; means for the terminal to recognize the user's emotions from the input question and generate emotion information; means for the server to send the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback and generate a final answer that takes emotion information into consideration; and means for the server to send the final answer to the user's terminal. This makes it possible to provide more accurate and reliable answers that take the user's emotions into consideration, and furthermore, by utilizing the feedback from the expert respondent as training data for the generative AI, it is possible to improve the accuracy of the entire system.

[1417] A "user" refers to a person who uses the system to input questions and receive answers.

[1418] A "server" refers to a device that receives questions and sentiment information from users, analyzes, processes, and sends this information to AI systems or expert respondents to generate the final answer.

[1419] A "terminal" refers to a device that a user directly operates to input questions, generate sentiment information, and send it to a server.

[1420] An "emotion engine" refers to software or an algorithm that recognizes emotions from questions entered by a user through a device and generates emotional information.

[1421] "Generative AI" refers to artificial intelligence models that generate initial responses based on questions and sentiment information sent from a server.

[1422] "Initial response" refers to the version of the response generated by a generative AI and provided to expert respondents.

[1423] A "specialist respondent" refers to an expert who reviews the initial response generated by generative AI and makes corrections or additions as needed.

[1424] "Feedback" refers to the corrections and supplementary information that expert respondents provide to the initial responses of generative AI systems.

[1425] "Final answer" refers to the final answer generated by the server, which integrates the initial answers and feedback from expert respondents.

[1426] Natural Language Processing (NLP) refers to a technology that analyzes questions entered by users through their devices and extracts their context and key keywords.

[1427] "Training data" refers to data that includes feedback from expert respondents and is used to improve the accuracy of generative AI.

[1428] "Improved accuracy" refers to the process where a generative AI updates its model using additional training data to increase the accuracy of its answers to subsequent questions.

[1429] The system based on this invention begins with the user inputting a question via a terminal and sending it to a server. The system of this invention involves an emotion engine, a generative AI, and human expert respondents who generate answers to the user's questions and provide the final answer.

[1430] Question input and sentiment recognition

[1431] Users input questions using their devices (smartphones, tablets, computers, etc.). For example, when a user inputs "Why hasn't my cat been eating much lately?", an emotion engine installed on the device (e.g., IBM Watson Natural Language Understanding) recognizes the user's emotions from the input text. In this case, the emotion engine analyzes the input patterns and context and generates emotion information such as "anxiety."

[1432] Sending and analyzing questions and sentiment information

[1433] The device sends the question and recognized sentiment information to the server as an HTTP request. The server analyzes the received question and sentiment information using natural language processing (NLP) techniques (e.g., spaCy, NLTK). As a result of the analysis, key keywords and context are extracted. For example, keywords such as "cat," "not eating," and "cause," along with the sentiment information "anxiety," are extracted.

[1434] Generating initial responses

[1435] The server sends the analyzed question and sentiment information to a generative AI (e.g., OpenAI GPT-3) to generate an initial response. Based on this input, the generative AI selects the best response from its training database. For example, the generative AI might generate the response, "There are various reasons why a cat might not eat. These could include stress, illness, or disliking the food," and then address the anxiety by adding a supplementary sentence such as, "If you are concerned, we recommend consulting a specialist."

[1436] Review and correction by expert respondents

[1437] The generated initial response is presented to human expert respondents via the server. The expert respondents review the initial response and make corrections or additions as needed. For example, an expert respondent might add feedback such as, "You should take your cat to the vet to have its health checked."

[1438] Integration and generation of the final answer

[1439] The server integrates initial responses from the generative AI with feedback from expert respondents to generate a final response. User sentiment information is also taken into consideration during this process. For example, a final response might be: "There are various reasons why a cat might not eat. Stress, illness, or dislike of the food are possible causes. If you are concerned, it would be best to take your cat to the vet to have its health checked."

[1440] Submit and display the final response.

[1441] The server sends the final response to the user's device as an HTTP response. The device then displays the received final response to the user, allowing the user to obtain quick and accurate information while also receiving an emotionally sensitive response.

[1442] Feedback for improving AI accuracy

[1443] Furthermore, the server adds feedback from expert respondents as training data for the generative AI, updating the AI ​​model to improve the accuracy of answers to subsequent questions. Since user sentiment information is also included in the training data, the accuracy of responses to emotions also improves.

[1444] For example, if a user inputs, "My child had a fight with a friend at school and is upset. What should I do?", the emotion engine recognizes the user's concern and assigns "concern" as emotional information. Based on this, the generative AI generates a response such as, "It's natural for your child to learn about friendships through fighting. It's important to listen to them calmly. Also, if necessary, it would be a good idea to talk to the school counselor."

[1445] Examples of prompts for a generative AI model:

[1446] User question: "My child is upset because they had a fight with a friend at school. What should I do?"

[1447] Emotional information: "Worried"

[1448] The above describes specific embodiments of the present invention. This system allows users to obtain quick and reliable answers, and by receiving emotionally sensitive answers, the user experience is improved.

[1449] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1450] Step 1: The user enters a question.

[1451] The user enters a question using the terminal. The entered question is captured by the terminal as text data. For example, the user enters "Why hasn't my cat been eating much lately?"

[1452] Input: Question text entered by the user

[1453] Output: Question data imported into the terminal

[1454] Step 2: The device recognizes emotions.

[1455] The emotion engine installed on the device recognizes the user's emotions from the entered question text. Through text analysis, the emotion engine recognizes that the user is feeling anxious.

[1456] Input: Captured question text

[1457] Data processing: The emotion engine performs text analysis (e.g., using IBM Watson Natural Language Understanding).

[1458] Output: Recognized emotional information (e.g., anxiety)

[1459] Step 3: Send the question and sentiment information to the server.

[1460] The device sends the question content and recognized sentiment information to the server as an HTTP request.

[1461] Input: Questionnaire data and sentiment information

[1462] Data processing: Converting data into HTTP request format.

[1463] Output: Data sent to the server

[1464] Step 4: The server receives the question and sentiment information.

[1465] The server receives HTTP requests sent from the terminal and retrieves the question content and sentiment information.

[1466] Input: Data sent as an HTTP request

[1467] Data processing: Extracting question content and sentiment information from requests.

[1468] Output: Extracted question content and sentiment information

[1469] Step 5: Analyze the questions and emotional information.

[1470] The server uses natural language processing (NLP) tools to analyze the question content and sentiment information.

[1471] Input: Extracted question content and sentiment information

[1472] Data processing: Analysis is performed using NLP tools (e.g., spaCy, NLTK).

[1473] Output: Key keywords and context, sentiment information

[1474] Step 6: Generate initial response

[1475] The server sends the analysis results to the generative AI, which generates an initial response. The generative AI (e.g., OpenAI GPT-3) then provides the response based on the prompt.

[1476] Input: Key keywords, context, and sentiment information as analysis results.

[1477] Data processing: Generative AI generates initial responses based on prompts.

[1478] Output: Generated initial answer

[1479] Step 7: Present the initial answer to the expert respondent.

[1480] The server presents the generated initial answers to human expert respondents.

[1481] Input: Generated initial answer

[1482] Data processing: Send initial responses to expert respondents and display them.

[1483] Output: Display data for expert respondents to review.

[1484] Step 8: Expert respondents review and make corrections.

[1485] Expert respondents review the initial responses and make corrections or additions as needed. For example, they may provide additional advice or specific instructions.

[1486] Input: Generated initial answer

[1487] Data processing: Expert respondents make corrections and additions.

[1488] Output: Feedback with corrections and additions.

[1489] Step 9: Integrate feedback and generate the final answer

[1490] The server incorporates feedback from expert respondents and integrates it with the initial responses to generate the final answer. User sentiment information is also taken into consideration.

[1491] Input: Initial responses and feedback from expert respondents

[1492] Data processing: Integrate feedback and initial responses to generate final responses that take emotional information into account.

[1493] Output: Final Answer

[1494] Step 10: Send the final response to the user's device.

[1495] The server sends the final response to the user's terminal as an HTTP response.

[1496] Input: Final answer

[1497] Data processing: Convert data to HTTP response format.

[1498] Output: Final response sent to the user's terminal

[1499] Step 11: Display the final answer to the user.

[1500] The device displays the final received response to the user. The user can receive a quick and accurate response.

[1501] Input: Final response sent from the server

[1502] Output: Final answer displayed to the user

[1503] Step 12: Add feedback to the training data of the generative AI.

[1504] The server adds feedback from expert respondents as training data for the generative AI.

[1505] Input: Feedback from expert respondents

[1506] Data processing: Convert the feedback into a format that can be used as training data.

[1507] Output: Add to the training data for generative AI.

[1508] Step 13: Update the generative AI model.

[1509] Generative AI uses the added feedback as training material for its model to improve the accuracy of subsequent response generation.

[1510] Input: Added feedback

[1511] Data processing: Improve accuracy through model retraining.

[1512] Output: Improved accuracy generative AI model

[1513] (Application Example 2)

[1514] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1515] Conventional user question answering systems often provide answers that do not take user emotions into consideration, resulting in an unsatisfactory user experience. Furthermore, the accuracy of initial responses provided by generative AI is inconsistent, frequently requiring human verification. This makes it difficult to provide quick and accurate answers, leading to decreased user satisfaction.

[1516] In Application Example 2, the specific processing performed by the specific processing unit 290 of the data processing device 12 is realized by the following means. In this invention, the server includes means for the user to input a question and recognize emotions; means for transmitting the question and recognized emotion information to the server; means for the server to transmit the received question and emotion information to a generative AI and generate an initial answer; means for the server to present the generated initial answer to a human expert respondent and receive feedback; means for the server to integrate the initial answer and feedback, and generate a final answer considering the emotion information; and means for the server to transmit the final answer to the user's terminal. This makes it possible to provide a quick and accurate answer that takes the user's emotions into consideration.

[1517] "A means for a user to input a question and recognize their emotions" refers to a device or software that has the function of analyzing the user's emotions when the user inputs a question into the terminal.

[1518] "Means for transmitting questions and recognized sentiment information to a server" refers to a device or software that has the function of transmitting questions entered from a terminal and sentiment information recognized by the sentiment engine to a server via a network.

[1519] "Generative AI" refers to a platform or system that includes artificial intelligence for generating initial responses based on user questions and sentiment information.

[1520] "Means for generating initial answers" refers to software that uses generative AI to create answers based on questions and sentiment information received by the server.

[1521] A "human expert respondent" is a person with specialized knowledge who can review the initial response generated by a generative AI and make corrections or additions as necessary.

[1522] "Means of receiving feedback" refers to a device or software that has the function of receiving corrections and supplementary information from human expert respondents.

[1523] "A means of integrating initial responses and feedback to generate a final response that takes emotional information into account" refers to software that combines the initial responses of a generative AI with feedback from human expert respondents, and takes the user's emotional information into consideration to create a final response.

[1524] "Means for sending the final answer to the user's terminal" refers to a device or software that has the function of sending the generated final answer to the user's terminal via a network and displaying it to the user.

[1525] The present invention is a system that provides answers to questions while taking user emotions into consideration. Detailed embodiments are described below.

[1526] System Configuration

[1527] The system of this invention begins by recognizing the emotions of the user when they input a question through a terminal. The system includes, as its main hardware and software, a terminal, a server, a generative AI, an emotion engine, and human expert respondents.

[1528] Hardware and software to be used

[1529] User terminal: A device such as a smartphone or tablet on which the user enters a question.

[1530] Server: A system that processes information submitted by users and manages data with generative AI and human expert respondents.

[1531] Generative AI models: Artificial intelligence used to generate initial responses based on questions and sentiment information. OpenAI's GPT-3 is used as an example.

[1532] Emotion engine: Software used to recognize emotions from user input. For example, EmotionEngine can be used.

[1533] Human expert respondents: Experts who review initial responses and make corrections or additions as needed.

[1534] Data processing and data calculation

[1535] 1. User question input and sentiment recognition:

[1536] When a user enters a question into the device, the emotion engine simultaneously recognizes that emotion. This is done by analyzing the patterns of text input or, in the case of voice input, the tone of voice.

[1537] 2. Sending questions and sentiment information to the server:

[1538] The recognized emotion information and question content are sent from the terminal to the server. HTTP requests are used for this process.

[1539] 3. Generation of initial answers using generative AI:

[1540] The server analyzes the received question and sentiment information and sends it to the generative AI model. The generative AI model generates an initial response based on this information.

[1541] 4. Feedback from human expert respondents:

[1542] The server presents the initial generated response to a human expert respondent. The expert respondent reviews the content and makes corrections or additions as needed.

[1543] 5. Generating the final answer:

[1544] The server integrates initial responses with feedback from human expert respondents and generates a final response that takes into account the user's emotional information.

[1545] 6. Sending the final response to the user's terminal:

[1546] The generated final response is then sent back to the user's device as an HTTP response.

[1547] Specific example

[1548] If a user enters a question via their smartphone, such as "I'm worried about the camera performance of this new smartphone, what do you think?", the emotion engine recognizes the user's concern. The question and emotion information are then sent to the server, and a generative AI model generates an initial response. In this case, the initial response would be "This smartphone's camera performance is very high and has received high ratings from many users." A human expert respondent then reviews this and makes a final revision, such as "This smartphone's camera performance is very high and has received high ratings from many users. If you have any concerns for specific uses, please let us know in detail."

[1549] Example of a prompt

[1550] User question: I'm worried about the camera performance of this new smartphone. What do you think?

[1551] User's emotion: Anxiety

[1552] Please provide the best answer.

[1553] Such a system makes it possible to provide quick and accurate responses that take into account the user's feelings, thereby improving the user experience.

[1554] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1555] Step 1:

[1556] The user enters a question into the device. Specifically, the user uses a device such as a smartphone or tablet to enter the question in text format. An example of input is, "I'm worried about the camera performance of this new smartphone, what do you think?" The entered question text is stored in the input field.

[1557] Step 2:

[1558] An emotion engine built into the device analyzes the user's questions and recognizes emotional information. Specifically, it determines whether the user is experiencing emotions such as anxiety, joy, or anger based on text analysis. If the input includes the phrase "I'm worried," the emotion engine identifies "anxiety." The output will then include the question text along with the emotional information "anxiety."

[1559] Step 3:

[1560] The question text and recognized sentiment information are sent from the terminal to the server. HTTP requests are used as the means of communication. Specifically, the terminal sends a request to the server containing the question and sentiment information as the payload. The input is the question text and sentiment information, and the output is the completion of the data transmission to the server.

[1561] Step 4:

[1562] The server analyzes the received question text and sentiment information. Specifically, the server uses natural language processing (NLP) techniques to extract key keywords and context from the question and format them, along with the sentiment information, into an input format for the generative AI model. The input is the question text and sentiment information, and the output is a prompt sentence for the generative AI model.

[1563] Step 5:

[1564] The server sends a prompt to a generative AI model to generate an initial response. Specifically, it sends a prompt to a generative AI model such as OpenAI GPT-3, which then generates a response. The input is the prompt, and the output is the initial response text from the generative AI model.

[1565] Step 6:

[1566] The server presents the generated initial response to a human expert respondent. Specifically, the initial response and sentiment information are displayed on a dedicated response confirmation interface. The input is the initial response text and sentiment information, and the output is the feedback from the human expert respondent.

[1567] Step 7:

[1568] Human expert respondents review the generated initial responses and make corrections or additions as needed. Specifically, expert respondents edit the initial responses on the interface and provide final feedback. The input is the initial response text and sentiment information, and the output is the reviewed and corrected response text.

[1569] Step 8:

[1570] The server integrates the initial response and feedback from human expert respondents, and generates a final response that takes sentiment into account. Specifically, the server merges the initial response and revised feedback, reflecting sentiment to create a response sentence that is appropriate for the user. The inputs are the initial response text, revised feedback, and sentiment information, and the output is the final response text.

[1571] Step 9:

[1572] The server sends the final response to the user's terminal. Specifically, it sends the final response as an HTTP response to the user's terminal. The input is the final response text, and the output is the response displayed on the user's terminal.

[1573] In this way, a system is in place to provide quick and accurate responses that take the user's feelings into consideration.

[1574] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1575] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1576] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[1577] Furthermore, the emotion identification model 59, acting as an emotion engine, may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1578] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[1579] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[1580] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[1581] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[1582] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[1583] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[1584] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[1585] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[1586] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[1587] 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.

[1588] Furthermore, it is not necessary to store the entirety of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[1589] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[1590] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[1591] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[1592] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[1593] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[1594] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted as being incorporated by reference.

[1595] The following is further disclosed regarding the embodiments described above.

[1596] (Claim 1)

[1597] A means for the user to input a question and send it to the server,

[1598] A means for sending a question received by a server to a generative AI to generate an initial answer,

[1599] The server presents the initial response it generates to a human expert respondent, and a means of receiving feedback is provided.

[1600] A means by which the server integrates initial responses and feedback to generate a final response,

[1601] A means by which the server sends the final answer to the user's terminal,

[1602] A system that includes this.

[1603] (Claim 2)

[1604] The system according to claim 1, further comprising means for the server to add feedback from human expert respondents to the training data of the generative AI in order to improve the accuracy of the generative AI.

[1605] (Claim 3)

[1606] The system according to claim 1, further comprising means for analyzing a question entered by a user through a terminal using natural language processing and sending the question in a format optimized for a generative AI.

[1607] "Example 1"

[1608] (Claim 1)

[1609] A means for the user to input a question and send it to the server,

[1610] A means for analyzing a question received by a server using natural language processing technology and sending it to a generative AI model,

[1611] A means for a generative AI model to generate an initial answer based on an analyzed question,

[1612] The server presents the initial response it generates to expert respondents and provides a means for receiving feedback.

[1613] A means by which the server integrates initial responses and feedback to generate a final response,

[1614] A means by which the server sends the final answer to the user's terminal,

[1615] A system that includes this.

[1616] (Claim 2)

[1617] The system according to claim 1, further comprising means for the server to add feedback from expert respondents to the training data of a generative AI model to improve the accuracy of the generative AI model.

[1618] (Claim 3)

[1619] The system according to claim 1, further comprising means for analyzing a question entered by a user through a terminal using natural language processing technology and sending the question in a format optimized for a generative AI model.

[1620] "Application Example 1"

[1621] (Claim 1)

[1622] A means for the user to input a question and send it to the server,

[1623] A means for sending a question received by a server to a generative AI to generate an initial answer,

[1624] The server presents the initial answers it generates to respondents with specialized knowledge and provides a means of receiving feedback.

[1625] A means by which the server integrates initial responses and feedback to generate a final response,

[1626] A means by which the server transmits the final answer to the user's information display device,

[1627] A means of continuously training a generative AI based on feedback to improve the accuracy of its responses,

[1628] A means of providing information about meals in response to questions entered through the user's terminal,

[1629] A system that includes this.

[1630] (Claim 2)

[1631] The system according to claim 1, further comprising means for the server to add feedback from respondents with specialized knowledge to the training data of the generative AI in order to improve the accuracy of the generative AI.

[1632] (Claim 3)

[1633] The system according to claim 1, further comprising means for analyzing a question entered by a user through an information display device using natural language processing and sending the question in a format optimized for a generative AI.

[1634] "Example 2 of combining an emotion engine"

[1635] (Claim 1)

[1636] A means for the user to input a question and send it to the server,

[1637] A means for a terminal to recognize the user's emotions from the questions entered and generate emotional information,

[1638] A means for the server to send the received question and sentiment information to a generative AI and generate an initial response,

[1639] The server presents the initial response it generates to a human expert respondent, and a means of receiving feedback is provided.

[1640] A means by which the server integrates initial responses and feedback to generate a final response that takes emotional information into account,

[1641] A means by which the server sends the final answer to the user's terminal,

[1642] A system that includes this.

[1643] (Claim 2)

[1644] The system according to claim 1, further comprising means for the server to add feedback from human expert respondents to the training data of the generative AI in order to improve the accuracy of the generative AI.

[1645] (Claim 3)

[1646] The system according to claim 1, further comprising means for analyzing a question entered by a user through a terminal using natural language processing and sending the question in a format optimized for a generative AI.

[1647] "Application example 2 when combining with an emotional engine"

[1648] (Claim 1)

[1649] A means for the user to input a question and recognize emotions,

[1650] Means for sending questions and recognized sentiment information to a server,

[1651] A means for sending the received question and sentiment information to a generative AI and generating an initial response,

[1652] The server presents the initial response it generates to a human expert respondent, and a means of receiving feedback is provided.

[1653] A means by which the server integrates initial responses and feedback, and generates a final response that takes emotional information into consideration,

[1654] A means by which the server sends the final answer to the user's terminal,

[1655] A system that includes this.

[1656] (Claim 2)

[1657] The system according to claim 1, further comprising means for the server to add feedback and sentiment information from human expert respondents to the training data of the generative AI in order to improve the accuracy of the generative AI.

[1658] (Claim 3)

[1659] The system according to claim 1, further comprising means for analyzing a question entered by a user through a terminal using natural language processing and transmitting the question and sentiment information in a format optimized for a generative AI. [Explanation of symbols]

[1660] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means for the user to input a question and send it to the server, A means for sending a question received by a server to a generative AI to generate an initial answer, The server presents the initial response it generates to a human expert respondent, and a means of receiving feedback is provided. A means by which the server integrates initial responses and feedback to generate a final response, A means by which the server sends the final answer to the user's terminal, A system that includes this.

2. The system according to claim 1, further comprising means for the server to add feedback from human expert respondents to the training data of the generative AI in order to improve the accuracy of the generative AI.

3. The system according to claim 1, further comprising means for analyzing a question entered by a user through a terminal using natural language processing and sending the question in a format optimized for a generative AI.

Images