system

Abstract

Provide a system. A method for collecting prompts entered by a plurality of users; a means for extracting features from the collected prompts; A means for learning a user's prompt engineering habits based on the extracted features; A means to save and share the learned habit model, means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit model; a means for generating generative AI responses utilizing the tailored prompts; means for displaying the generated answers to the user; A system including:

Description

[Technical Field]

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

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

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

[0004] Conventional generative AI systems rely on a single prompt engineering method, which limits the accuracy and diversity of the answers they provide. As a result, the generated answers tend to be biased, making it difficult to adequately meet the diverse needs of users. Another problem is that they ignore the prompt input styles and question habits of different users, making it impossible to provide answers optimized for each individual user. [Means for solving the problem]

[0005] To solve the above problems, the present invention provides the following means. The present invention includes a means for collecting prompts entered by multiple users and a means for extracting features from the collected prompts. The present invention also includes a means for learning the user's prompt engineering habits based on the extracted features and a means for saving and sharing the learned habit model. Furthermore, the present invention includes a means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit, and a means for generating answers using a generative AI using the adjusted prompt. This system takes into account various prompt engineering habits and can provide multifaceted and accurate answers.

[0006] A "prompt" is a question or instruction entered by a user into the generating AI.

[0007] The "means for collecting" is a function for recording and saving prompts entered by the user.

[0008] "Feature extraction" refers to techniques for analyzing key keywords and grammatical patterns from prompts.

[0009] The "means for learning habits" is a function for learning the user's prompt engineering tendencies and patterns from the extracted features.

[0010] A "habit model" is a numerical representation of each user's learned prompt engineering patterns.

[0011] "Means for saving and sharing" is a function that allows the learned habit model to be saved in a database or the like, making it accessible to other users.

[0012] The "adjustment means" is the ability to reconstruct new prompts based on the associated habit model.

[0013] "Generative AI" is an artificial intelligence technology that automatically generates answers based on input prompts.

[0014] The "means for generating an answer" is the function that enables the generation AI to generate an appropriate answer using tailored prompts.

[0015] The "display means" is a function for presenting the generated answer to the user. [Brief explanation of the drawings]

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

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

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

[0019] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0020] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

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

[0022] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0024] [First embodiment]

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

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

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

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

[0029] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

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

[0031] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

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

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

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

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

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

[0037] This invention is a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[0038] First, a user inputs a prompt into the terminal. For example, if the user inputs the prompt "What will the weather be like tomorrow?", the prompt is recorded by the terminal. Then, the terminal sends the recorded prompt to the server.

[0039] The server then analyzes the received prompt. Specifically, the server uses natural language processing technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[0040] Next, the server learns the user's prompt engineering habits based on the extracted features. For example, if a user has a strong tendency to request specific data, this habit is learned. The learned habits are stored in the brain database as a numerical model.

[0041] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know the weather tomorrow," the prompt is sent to the server, which selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[0042] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[0043] This system can provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, it is possible to improve the overall accuracy of answers. This invention is particularly useful in generative AI systems that require flexible responses to users with different questioning styles.

[0044] The processing flow will be explained below.

[0045] Step 1:

[0046] The user inputs a prompt into the terminal, for example, "What is the weather like tomorrow?"

[0047] Step 2:

[0048] Records the prompts entered at the terminal and saves them in a log file or database.

[0049] Step 3:

[0050] The device sends the recorded prompt to the server, for example, using an HTTP request.

[0051] Step 4:

[0052] The server analyzes the received prompt, specifically extracting keywords and grammatical patterns from the prompt using natural language processing techniques.

[0053] Step 5:

[0054] The server converts the extracted features into feature vectors. For example, words such as "tomorrow," "weather," and "tell me" and their contexts are converted into numerical vectors.

[0055] Step 6:

[0056] The server classifies the feature vectors using a clustering algorithm and learns the user's prompt engineering habits.

[0057] Step 7:

[0058] The server stores the learned habit model in a brain database, which is then shared so that other users can use it.

[0059] Step 8:

[0060] The terminal receives a new prompt. For example, user B inputs the prompt "I want to know the weather tomorrow."

[0061] Step 9:

[0062] The terminal sends a new prompt to the server.

[0063] Step 10:

[0064] The server selects relevant habit models from the brain database, and the most appropriate habit model is chosen and applied to the new prompt.

[0065] Step 11:

[0066] The server reconstructs a new prompt based on the habit model, e.g., "Please tell me the specific weather data for tomorrow."

[0067] Step 12:

[0068] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[0069] Step 13:

[0070] The server sends the generated response to the terminal.

[0071] Step 14:

[0072] The device displays the generated answer to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[0073] Example 1

[0074] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0075] Conventional generative AI systems have difficulty taking into account the different prompting habits of each user, which can result in inappropriate answers being generated. Furthermore, there is a lack of efficient means for learning user habits and reconstructing prompts based on them. Furthermore, there is a need to effectively utilize natural language processing technology in the process of collecting and analyzing user prompts. By resolving these issues, it will be possible to provide more accurate answers for each user.

[0076] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0077] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting a habit model associated with a new prompt and reconstructing the prompt based on the habit, means for generating an answer by a generation AI using the reconstructed prompt, means for displaying the generated answer to the user, means for transmitting the prompt from the user's terminal to the server, means for the server to analyze the received prompt using natural language processing technology, and means for the server to select a habit model from a brain database. This enables the generation of appropriate answers that take into account each user's prompt habits.

[0078] A "prompt" is a sentence containing a question or instruction that a user enters into a generative AI system.

[0079] A "collection means" is a method or device for capturing prompts entered by multiple users into the system.

[0080] A "means for extracting features" is a method or device for identifying and extracting important keywords and grammatical patterns from collected prompts.

[0081] "Prompt engineering habits" refer to the particular patterns or tendencies that users have when entering prompts.

[0082] "Means for learning" refers to a method or device for recognizing and storing a user's prompt engineering habits based on extracted features using a machine learning algorithm or the like.

[0083] "Means for saving and sharing" refers to a method or device for saving the learned habit model in a database or the like and sharing it with other systems or users as needed.

[0084] A "reconstruction means" is a method or device for selecting a relevant habit model for a newly input prompt and improving or adjusting the prompt based on that model.

[0085] A "means for generating" is a method or device for generating responses to reconstructed prompts using generative AI.

[0086] The "displaying means" is a method or device for visually or audibly presenting the generated answer to the user.

[0087] A "means for sending" is a method or device for sending a prompt from a user's terminal to a server.

[0088] The "means for analyzing" refers to a method or device for using natural language processing techniques to examine in detail the prompt received by the server and understand its meaning and characteristics.

[0089] The "brain database" is a database that stores learned prompt engineering habit models of users and can be searched and retrieved as needed.

[0090] This invention is a generative AI system that provides users with multifaceted and accurate answers. The system begins operation when the user inputs a prompt sentence into a terminal. Specifically, when the user inputs the prompt sentence, "Please tell me what the weather will be like tomorrow," the prompt sentence is recorded by the terminal. The terminal then transmits the recorded prompt sentence to a server.

[0091] To analyze the prompt received by the server, natural language processing technology (such as NLTK or SpaCy) is used. The server uses this technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[0092] Next, the server learns the user's prompt engineering habits based on the extracted features. This learning process uses machine learning algorithms, and the learned habits are stored in the brain database as a numerical model. For example, if a user has a habit of requesting specific data, this tendency will be stored in the brain database.

[0093] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know tomorrow's weather," the prompt is sent to the server once, and the server selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[0094] The reconstructed prompt is passed to a generative AI model, which generates an appropriate response. An example of a generative AI model is OpenAI's GPT-4 (registered trademark). This generative AI model has the ability to generate the most appropriate response based on the reconstructed prompt.

[0095] The generated answer is sent from the server to the terminal and displayed to the user through the terminal. For example, an answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed. This system allows users to receive more accurate and multifaceted answers based on their own prompt engineering habits.

[0096] This invention allows a generative AI system to provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, the overall accuracy of answers is improved. This system is particularly useful in generative AI model systems that require flexible responses to users with different questioning styles.

[0097] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0098] Step 1:

[0099] The user inputs a prompt sentence using a terminal. Specifically, the user inputs a prompt sentence such as "Please tell me what the weather will be tomorrow." using a keyboard or voice input. This prompt sentence becomes the input data for the process.

[0100] Step 2:

[0101] The terminal records the input prompt and sends the data to the server. The specific operation of the terminal is to save the prompt in memory and send the data to the server using an HTTP request. The input here is the prompt, and the output is the prompt sent to the server.

[0102] Step 3:

[0103] The server parses the received prompt. Specifically, it uses natural language processing techniques (such as NLTK or SpaCy) to tokenize the prompt and extract keywords and grammatical patterns. The input to this process is the prompt, and the output is the extracted keywords and grammatical patterns.

[0104] Step 4:

[0105] The server learns the user's prompt engineering habits based on the extracted keywords. The server uses a machine learning algorithm (such as a decision tree or random forest) to model the prompt engineering habits. The input is the extracted keywords, and the output is a model that reflects the user's habits. This model is stored in a brain database.

[0106] Step 5:

[0107] When a new prompt sentence is input, the server selects a highly relevant habit model from the brain database. For example, if a new prompt such as "I want to know the weather tomorrow," is input, the server receives the prompt and selects an appropriate habit model from the brain database. The input is the new prompt sentence, and the output is the selected habit model.

[0108] Step 6:

[0109] The server reconstructs the prompt sentence based on the selected habit model. As a specific operation of reconstruction, the server changes the prompt sentence "I want to know the weather tomorrow" to "Please tell me the specific weather data for tomorrow." The input is the selected habit model and the new prompt sentence, and the output is the reconstructed prompt sentence.

[0110] Step 7:

[0111] The reconstructed prompt is passed to a generative AI model, which generates an appropriate answer. GPT-4 is used as the generative AI model, and the answer is generated through an API request. The input is the reconstructed prompt, and the output is the generated answer.

[0112] Step 8:

[0113] The generated answer is sent from the server to the terminal. As a server operation, the generated answer is sent to the terminal using an HTTP response. The input is the generated answer, and the output is the answer sent to the terminal.

[0114] Step 9:

[0115] The user displays the generated answer through the terminal. The specific operation of the terminal is to display the received answer on the screen. The input is the answer received from the server, and the output is the answer displayed to the user. For example, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees."

[0116] (Application example 1)

[0117] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0118] In today's virtual stores, it is extremely important to quickly and accurately provide users with the product information and recommendations they need. However, typical generative AI systems often lack the ability to accurately understand the intent and preferences of users' questions, resulting in inability to provide the information they seek. Furthermore, when using voice prompts, the process of accurately converting the speech into text and adjusting it to match the user's intent is complex, and the entire process of displaying information on a visual device must be carried out efficiently and accurately. Therefore, technology is needed that can learn users' prompt engineering habits and provide answers that adapt to them.

[0119] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0120] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit, means for generating an answer by a generation AI using the adjusted prompt, means for displaying the generated answer to the user, means for allowing the user to obtain information using voice prompts, speech recognition means for converting the voice prompts into text, and means for displaying the generated answer on a visual device, thereby enabling optimal information provision taking into account the user's prompt engineering habits.

[0121] Definitions of important words

[0122] A "user" is an individual or organization that uses the system.

[0123] A "prompt" is a question or instruction that a user enters into a system.

[0124] "Collection means" refers to the software and hardware used to capture and record prompts.

[0125] "Feature extraction methods" refer to techniques for identifying and extracting important elements or keywords from the prompt.

[0126] "Prompt engineering habits" refer to patterns or tendencies that users have when entering prompts.

[0127] "Means of learning" refers to the technology used to analyze prompt engineering habits based on collected feature information and build a model.

[0128] "Means for saving and sharing" refers to saving the trained model in a database or the like and sharing it with other systems or users as needed.

[0129] "Adjustment" refers to the process of optimizing prompts based on user habits, allowing the generative AI to generate more appropriate answers.

[0130] "Generative AI" refers to a system that uses artificial intelligence technology to automatically generate answers based on input data.

[0131] "Displaying means" refers to a device or software for notifying the user of the generated answer.

[0132] "Voice prompt" refers to a question or instruction that the user speaks.

[0133] "Speech recognition means" refers to technology that converts voice prompts into text.

[0134] "Visual device" refers to an electronic device that visually presents information to a user.

[0135] MODE FOR CARRYING OUT THE INVENTION

[0136] In order to implement the present invention, a system is used in which a user wears smart glasses and obtains information using voice prompts. Specific embodiments are described in detail below.

[0137] System Configuration

[0138] 1. Hardware:

[0139] Smart glasses: Devices with built-in voice recognition microphones and displays.

[0140] Server: A server with high-performance computing power that is responsible for prompt analysis and running AI models.

[0141] 2. Software:

[0142] Speech Recognition Library: Use speech_recognition to convert user voice prompts into text.

[0143] Communication library: requests is used to send prompts to the server and retrieve results.

[0144] Natural language processing technology: We use spaCy and Transformers to analyze prompts and extract features.

[0145] Generative AI models: Models that generate answers based on prompts, such as GPT and BERT.

[0146] Data Processing and Computation

[0147] When a user enters a voice prompt, the voice data is converted to text by a speech recognition library, and the converted text is sent to the server, which then performs the following steps:

[0148] 1. Prompt analysis and feature extraction:

[0149] The server uses natural language processing techniques to extract important keywords and grammatical patterns from the text prompts.

[0150] 2. Prompt Engineering Habit Learning:

[0151] The server learns the user's prompt engineering habits based on the extracted features, and stores and shares the habit model in a brain database.

[0152] 3. Prompt adjustment and AI model execution:

[0153] For each new prompt, it selects a relevant habit model and adjusts the prompt accordingly, which is then passed to a generative AI model to generate the optimal answer.

[0154] 4. Show Answer:

[0155] The generated answers are sent to the smart glasses and displayed on the visual device.

[0156] Specific examples

[0157] A user asks a question through the smart glasses. For example, if a user inputs a voice prompt such as "What new sneakers do you recommend?", the prompt is converted into text and sent to the server. The server extracts keywords such as "new," "sneakers," and "recommended," and reconstructs the prompt based on the user's habits to "Please recommend some new sneakers with a flashy design." The generative AI model generates an appropriate answer, which is sent from the server to the smart glasses. The user's field of vision is displayed as follows: "The latest sneakers with a flashy design are the Nike Air Max, which comes in a wide range of color options."

[0158] Prompt Sentence Examples

[0159] "What new sneakers do you recommend?"

[0160] "Can you recommend some new sneakers with flashy designs?"

[0161] This system makes it possible to provide optimal information taking into account the user's prompt engineering habits.

[0162] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0163] Specific explanation of program processing

[0164] Step 1:

[0165] The user inputs a voice prompt into the smart glasses.

[0166] Input: A voice prompt from the user (e.g., "What new sneakers would you recommend?").

[0167] Output: The speech data received by the speech recognizer.

[0168] Specific operation: When the user speaks into the device, the smart glasses' built-in microphone captures the voice data.

[0169] Step 2:

[0170] The device converts the voice prompts into text.

[0171] Input: Audio data from smart glasses.

[0172] Output: A text prompt (e.g., "What new sneakers do you recommend?").

[0173] Specific operation: Uses the speech_recognition library to convert audio data into text data.

[0174] Step 3:

[0175] The terminal sends a text prompt to the server.

[0176] Input: A text prompt.

[0177] Output: The text prompt sent to the server.

[0178] Specific operation: Uses the requests library to send text data to the server via an HTTP POST request.

[0179] Step 4:

[0180] The server parses the prompt and extracts features.

[0181] Input: The text prompt sent from the terminal.

[0182] Output: Extracted feature data (e.g., keywords "new arrivals," "sneakers," "recommended").

[0183] Specific operation: Using the spaCy library and Transformers, it analyzes and extracts important keywords and grammatical patterns from text data.

[0184] Step 5:

[0185] The server learns the user's prompt engineering habits based on the extracted features.

[0186] Input: Extracted feature data.

[0187] Output: Habit model (e.g., patterns of keywords and phrases frequently used by the user).

[0188] Specific operation: The analyzed feature data is input into a machine learning algorithm to learn the user's prompt engineering habits.

[0189] Step 6:

[0190] The server stores and shares the learned habit models.

[0191] Input: The newly trained habit model.

[0192] Output: Habit model stored in the brain database.

[0193] What it does: Saves the habit model in a database so that it can be shared with other systems and users in the future.

[0194] Step 7:

[0195] The server selects the relevant habit model for the new prompt and reconstructs the prompt.

[0196] Input: A new prompt and a saved habit model.

[0197] Output: A tailored prompt (e.g., "Can you recommend some new sneakers with a flashy design?").

[0198] What it does: Refers to the saved habit model and adjusts new prompts appropriately.

[0199] Step 8:

[0200] The server uses the tailored prompts to generate answers using the generative AI.

[0201] Input: Tailored prompt.

[0202] Output: The generated answer (e.g., "The latest, flashiest sneaker design is the Nike Air Max, which comes in a wide variety of colors.").

[0203] What it does: Feeds the tailored prompts into an AI model to generate the best answer.

[0204] Step 9:

[0205] The server sends the generated response to the terminal.

[0206] Input: The generated answer.

[0207] Output: The answer sent to the terminal.

[0208] Specific operation: The generated answer is sent to the terminal as an HTTP response.

[0209] Step 10:

[0210] The terminal displays the generated answer on the visual device of the smart glasses.

[0211] Input: The generated answer received from the server.

[0212] Output: The answer that is displayed to the user.

[0213] Specific behavior: Displays text on the smart glasses display to provide information to the user.

[0214] Through the above steps, the user can efficiently obtain optimal product information in the virtual store using voice prompts.

[0215] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0216] This invention adds an emotion engine to a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[0217] First, the user inputs a prompt into the device, and the device reads the emotion at that time. For example, while the user inputs the prompt "Please tell me what the weather will be like tomorrow," emotional information (e.g., joy, anger, sadness, etc.) is acquired through the camera and microphone. This emotional information is then sent by the device to the server.

[0218] The server analyzes the received prompt and emotional information. Specifically, the server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[0219] The server converts the extracted prompt features and emotion features into feature vectors, for example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[0220] The server then classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns. For example, if a user has a strong tendency to seek specific data and is feeling "impatient," the server learns the habits and emotional patterns. The learned habits and emotional patterns are stored in a brain database as a numerical model.

[0221] Next, when a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server, which then selects an appropriate habit model and emotion model. The prompt is reconstructed based on the habit model and emotion model, and adjusted to a form such as, "Please tell me the specific weather data for tomorrow as soon as possible."

[0222] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0223] This system can provide more appropriate, multifaceted, and emotionally relevant answers by taking into account the user's prompt engineering habits and emotions. Furthermore, by sharing each user's habit model and emotion model, it is possible to improve overall answer accuracy and user satisfaction. This invention is particularly useful in generative AI systems that require flexible responses to users with different question styles and various emotions.

[0224] The processing flow will be explained below.

[0225] Step 1:

[0226] A user inputs a prompt into a terminal. For example, the user inputs the prompt "What will the weather be like tomorrow?"

[0227] Step 2:

[0228] The device uses a camera and microphone to acquire information about the user's emotions. For example, it can recognize the emotion of "anxiety" by analyzing the user's facial expressions and tone of voice.

[0229] Step 3:

[0230] The device records the prompts and emotional information entered, and the recorded data is stored in a log file or database.

[0231] Step 4:

[0232] The device sends the recorded prompts and emotion information to the server using HTTP requests or similar.

[0233] Step 5:

[0234] The server analyzes the received prompt and emotional information. Specifically, it uses natural language processing technology to extract keywords and grammatical patterns from the prompt, and then analyzes the emotional information using an emotional engine.

[0235] Step 6:

[0236] The server converts the extracted prompt features and emotion features into feature vectors. For example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[0237] Step 7:

[0238] The server classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns.

[0239] Step 8:

[0240] The server stores the learned habit models and emotion models in a brain database, which is then shared for other users to use.

[0241] Step 9:

[0242] The terminal receives a new prompt. For example, user B inputs a prompt such as "I want to know the weather tomorrow." At the same time, the emotion "I'm anxious" is acquired.

[0243] Step 10:

[0244] The device sends the new prompt and emotion information to the server.

[0245] Step 11:

[0246] The server selects relevant habit models and emotion models from the brain database, and the most appropriate habit model and emotion model are selected and applied to the new prompt.

[0247] Step 12:

[0248] The server reconstructs a new prompt based on the habit model and emotion model, for example, "Please tell me the specific weather data for tomorrow as soon as possible."

[0249] Step 13:

[0250] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[0251] Step 14:

[0252] The server sends the generated response to the terminal.

[0253] Step 15:

[0254] The device displays the generated answer to the user. For example, the answer may read, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0255] Example 2

[0256] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0257] Conventional generative AI systems have a problem in that simply generating answers to prompts entered by the user makes it difficult to respond flexibly while taking into account the user's emotions and prompt engineering habits. As a result, the information the user seeks is not accurately provided, often resulting in dissatisfaction. The present invention provides a system that generates more accurate and multifaceted answers by taking into account the user's emotional information and prompt engineering habits.

[0258] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0259] In this invention, the server includes: means for collecting prompts input by multiple users; means for acquiring the collected prompts and emotional information; means for transmitting the acquired prompts and emotional information to the server; means for the server to analyze the prompts and emotional information; means for generating feature vectors from the analyzed prompts and emotional information; means for classifying the generated feature vectors using a clustering algorithm and learning the user's prompt engineering habits and emotional patterns; means for saving and sharing the learned habit model and emotional model in a brain database; means for selecting a habit model and emotional model associated with a new prompt and reconstructing the prompt based on the habits and emotions; means for generating an answer using a generation AI using the reconstructed prompt; and means for displaying the generated answer to the user. This enables the generation of appropriate answers that take into account the user's emotional information and prompt engineering habits.

[0260] "User" refers to a person using the system to enter prompts.

[0261] "Terminal" refers to a device that allows a user to input prompts, obtain emotion information, and transmit it to a server.

[0262] A "prompt" is a question or instruction text that a user enters into a system.

[0263] "Emotional information" refers to data indicating the user's emotional state that is acquired by the device through the camera or microphone.

[0264] "Server" refers to the central computer that analyzes prompts and emotional information, generates feature vectors, clusters, and generates answers.

[0265] "Natural language processing" refers to the technology of extracting keywords and grammatical patterns from prompts.

[0266] "Emotion engine" refers to technology that analyzes emotional information.

[0267] "Feature vector" refers to vector-format data that quantifies the analyzed prompt and emotional information.

[0268] A "clustering algorithm" refers to an algorithm for classifying feature vectors.

[0269] "Prompt engineering" refers to techniques for reconstructing the questions or instructions that users enter into a system.

[0270] The "habit model" refers to a model generated by learning the prompt engineering habits of a user.

[0271] An "emotion model" refers to a model generated by learning the user's emotional patterns.

[0272] The "brain database" refers to a database that stores and manages learned habit models and emotion models.

[0273] "Generative AI" refers to artificial intelligence that generates appropriate answers based on reconstructed prompts.

[0274] This invention relates to a system that analyzes prompts entered by users and their emotional information at the time, and generates appropriate answers using generative AI. This system is configured by combining multiple hardware and software components, and is implemented as follows.

[0275] This system starts when a user inputs a prompt into a terminal. For example, the user might input, "Please tell me what the weather will be like tomorrow." At this time, the device's built-in camera and microphone are used to acquire the user's emotional information (e.g., whether they are anxious or happy). The acquired prompt and emotional information are then sent from the terminal to the server.

[0276] The server then analyzes the received prompt and emotional information. The server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, keywords such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[0277] The analyzed prompt and emotion information are converted into numerical vectors. The server classifies these numerical vectors using a clustering algorithm (e.g., K-means clustering) to learn the user's prompt engineering habits and emotion patterns. This generates a habit model and emotion model for each user. The learned habit model and emotion model are stored in the brain database.

[0278] When a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server. The server selects an appropriate habit model and emotion model, and the prompt is reconstructed based on the habit model and emotion model. For example, the prompt may be adjusted to something like, "Please tell me the specific weather data for tomorrow as soon as possible."

[0279] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0280] As a specific example, if a user inputs "Tell me how to prepare for a presentation," and the device detects the emotion "I'm nervous," this prompt and emotional information are sent to the server. Based on the emotion model and habit model, the server reconstructs the prompt into something like, "I'm nervous, but can you briefly tell me how to prepare for a presentation?" An example of a prompt sentence to input to the generative AI model is, "The user is nervous and would like more details about how to prepare for a presentation. Please provide specific and calming instructions."

[0281] This system can generate appropriate answers that take into account the user's emotional information and prompt engineering habits, thereby improving user satisfaction.

[0282] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0283] Step 1:

[0284] The user enters the prompt.

[0285] In a specific operation, the user inputs "What will the weather be like tomorrow?" into the terminal. The input obtained is a prompt "What will the weather be like tomorrow?" This prompt is taken into the terminal as text data.

[0286] Step 2:

[0287] The device acquires emotion information.

[0288] Specifically, the device uses its built-in camera and microphone to perform image recognition and voice analysis. Images of the user's facial expressions and voice data are obtained as input. These data are analyzed to obtain emotional information such as "anxious." The emotional information "anxious" is generated as output.

[0289] Step 3:

[0290] The device sends prompts and emotion information to the server.

[0291] Specifically, the terminal uses the communication module to send a prompt, "What will the weather be like tomorrow?" and emotional information, "I'm anxious," in packet form to the server. The input is the prompt and emotional information, and the output is a data packet.

[0292] Step 4:

[0293] The server parses the prompt and emotion information.

[0294] Specifically, the server uses natural language processing technology and an emotion engine. The prompt "Please tell me what the weather will be tomorrow" and the emotion information "I'm anxious" arrive as input to the server. Natural language processing technology extracts keywords such as "tomorrow," "weather," and "tell me," while the emotion engine simultaneously analyzes the emotion "I'm anxious." The output is the keyword and emotion information.

[0295] Step 5:

[0296] The server generates the feature vector.

[0297] Specifically, the system quantifies the keywords and emotional information extracted from the prompt. The inputs are the keywords "tomorrow," "weather," and "tell me about" and the emotional information "anxious." These are converted into numerical vectors, and the output is a prompt vector [0.6, 0.8, 0.5] and an emotional vector [0.1, 0.9].

[0298] Step 6:

[0299] The server classifies the feature vectors using a clustering algorithm.

[0300] Specifically, it uses algorithms such as K-means clustering. The inputs are a prompt vector [0.6, 0.8, 0.5] and an emotion vector [0.1, 0.9]. Clustering is performed to classify the user's prompt engineering habits and emotion patterns. The output is a feature vector classified into clusters.

[0301] Step 7:

[0302] The server stores the habit model and emotion model in a brain database.

[0303] Specifically, the clustering results are written to the database. The input is a feature vector classified into clusters, which is saved in the brain database as a habit model and an emotion model. The output is an updated database.

[0304] Step 8:

[0305] Select relevant habit models and emotion models for the new prompt.

[0306] Specifically, the server searches the brain database and selects the most relevant habit model and emotion model. The inputs are a new prompt, "I want to know tomorrow's weather," and the emotion information, "I'm anxious." The output is the selection of the relevant habit model and emotion model.

[0307] Step 9:

[0308] The server reconstructs the prompt.

[0309] The specific operation is to adjust the prompt based on the selected habit model and emotion model. The prompt and the associated model are obtained as input. The prompt is reconstructed to "Please tell me the specific weather data for tomorrow as soon as possible." and the reconstructed prompt is generated as output.

[0310] Step 10:

[0311] The server generates an answer using a generation AI.

[0312] Specifically, the reconstructed prompt is input into the generation AI, which then generates an answer. The reconstructed prompt is obtained as input. An answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​is generated, and an answer is generated as output.

[0313] Step 11:

[0314] The server sends the generated answer to the terminal, which displays it to the user.

[0315] Specifically, the server sends the generated answer to the terminal, and the terminal displays the answer to the user. The generated answer is obtained as input. The user can see the answer on the screen: "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​The output is the answer displayed to the user.

[0316] (Application example 2)

[0317] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0318] Conventional generative AI systems generate answers based on user input prompts, but because they do not take the user's emotional state into consideration, they often generate inappropriate answers or answers that do not meet the user's expectations. Furthermore, in emergencies, it is necessary to accurately grasp the user's emotions and provide a quick and appropriate response, but current systems have difficulty in doing so.

[0319] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0320] In this invention, the server includes means for collecting prompts input by multiple users, means for acquiring user emotion information using a camera or microphone, and means for analyzing the acquired emotion information to identify the user emotion, thereby enabling the generation of appropriate answers that take user emotion into consideration and quick and appropriate responses in emergencies.

[0321] "User emotion information" is information that indicates the emotional state of the user, obtained from the user's facial expression, tone of voice, and the like.

[0322] A "feature extraction means" is a means capable of extracting meaningful features or patterns from prompts and emotional information.

[0323] "Prompt engineering habits" refer to the characteristics or patterns that a particular user consistently exhibits when completing prompts.

[0324] The "means for identifying emotions" is a means having a function of analyzing acquired emotion information and recognizing the type of emotion the user has.

[0325] The "means for adjusting the prompt" is a means that has the function of reconstructing the original prompt more appropriately, taking into account the user's emotions and prompt engineering habits.

[0326] A "means for generating answers using generative AI" is a means that has the function of generating appropriate answers based on tailored prompts.

[0327] "Problems to be solved by the invention" refers to the specific problems or obstacles that the invention aims to solve, such as the inability of conventional systems to generate appropriate answers because they do not take into account the user's emotional state or prompt engineering habits, and the lack of rapid and appropriate responses in emergencies.

[0328] "Means for solving the problem" refers to the specific technical means proposed to solve the above problem.

[0329] This invention is realized by incorporating user emotional information into a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments for carrying out this invention are described below.

[0330] First, the user inputs a prompt into the device. At the same time, the user's emotional information is acquired using a camera and microphone. For example, while the user inputs the prompt "What will the weather be like tomorrow?", the camera captures the user's facial expression and the microphone captures the user's tone of voice. This emotional information is sent from the user's device to the server.

[0331] The server analyzes the received prompts and emotional information. Specifically, the server uses natural language processing technology (e.g., a generative AI model such as GPT-3®) to extract keywords and grammatical patterns from the prompts, and simultaneously analyzes the emotional information using an emotional engine. The analyzed data is converted into a numerical vector, and then a clustering algorithm is used to classify and learn the user's prompt engineering habits and emotional patterns. This information is stored in a brain database as a feature vector.

[0332] When a new prompt is input, the server selects the relevant habit model and emotion model from the brain database and adjusts the prompt based on the user's prompt and emotion information. For example, if the user inputs the prompt "I want to know the weather tomorrow," and the emotion "I'm anxious" is detected, the server will reconstruct the prompt as "Please tell me the specific weather data for tomorrow as soon as possible" based on the prompt and emotion information.

[0333] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0334] The hardware used includes the user's device (smartphone, tablet, etc.), camera, and microphone. The software used includes a generative AI model (e.g., GPT-3) for implementing natural language processing technology and an emotion engine for emotion analysis (e.g., an emotion classification model using TENSORFLOW (registered trademark) or Keras).

[0335] For example, if the user types the following prompt:

[0336] 1. "What's the weather going to be like tomorrow?" (User is anxious)

[0337] 2. "I'm in a hurry and need the latest news headlines right away." (Users are impatient)

[0338] The server applies appropriate habit models and emotion models to these prompts to generate answers that match the user's emotional state, allowing the user to obtain more satisfying information.

[0339] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0340] Step 1:

[0341] The user types a prompt into the terminal.

[0342] Input: User prompt (e.g., "What's the weather like tomorrow?")

[0343] Output: The prompt is typed into the terminal.

[0344] Step 2:

[0345] The device uses a camera and microphone to acquire information about the user's emotions.

[0346] Input: User's facial expression (camera), user's tone of voice (microphone)

[0347] Output: Emotional information (e.g., facial image data, audio data)

[0348] Step 3:

[0349] The device transmits the acquired prompt text and emotion information to the server.

[0350] Input: prompt sentence, emotion information

[0351] Output: The prompt and emotion information are converted into a dedicated format and sent to the server.

[0352] Step 4:

[0353] The server analyzes the prompt sentence using natural language processing technology to extract keywords and grammatical patterns.

[0354] Input: Prompt sentence (e.g. "What will the weather be like tomorrow?")

[0355] Output: Extracted features (keywords: "tomorrow", "weather", "tell me")

[0356] Step 5:

[0357] The server uses an emotion engine to analyze the transmitted emotion information.

[0358] Input: Emotion information (face video data, audio data)

[0359] Output: Parsed emotion (e.g. "anxious")

[0360] Step 6:

[0361] The server converts the features and sentiment information of the parsed prompt into a numerical vector.

[0362] Input: extracted features, parsed emotions

[0363] Output: Feature vector (e.g. "tomorrow", "weather", "tell me", "I'm anxious" → numeric vector)

[0364] Step 7:

[0365] The server clusters the feature vectors and learns the user's prompt engineering habits and emotional patterns.

[0366] Input: feature vector

[0367] Output: Learned habit model and emotion model

[0368] Step 8:

[0369] When a new prompt is entered, the server selects the relevant habit model and emotion model from the brain database.

[0370] Input: New prompt, emotion information

[0371] Output: Appropriate habit model and emotion model

[0372] Step 9:

[0373] The server adjusts the prompt sentence based on the selected habit model and emotion model.

[0374] Input: New prompt, habit model, emotion model

[0375] Output: The adjusted prompt (e.g., "Please give me the specific weather data for tomorrow as soon as possible.")

[0376] Step 10:

[0377] The server passes the adjusted prompt to the generative AI model, which generates the optimal answer.

[0378] Input: Adjusted prompt text

[0379] Output: Generated answer (e.g. "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[0380] Step 11:

[0381] The server sends the generated response to the terminal.

[0382] Input: Generated answer

[0383] Output: The generated answer sent by the server

[0384] Step 12:

[0385] The terminal displays the generated answer to the user.

[0386] Input: Generated answer

[0387] Output: The answer shown to the user (e.g., "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[0388] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0389] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0390] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0391] [Second embodiment]

[0392] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0393] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0394] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0395] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0396] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0397] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0398] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0399] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0400] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0401] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0402] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0403] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0404] This invention is a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[0405] First, a user inputs a prompt into the terminal. For example, if the user inputs the prompt "What will the weather be like tomorrow?", the prompt is recorded by the terminal. Then, the terminal sends the recorded prompt to the server.

[0406] The server then analyzes the received prompt. Specifically, the server uses natural language processing technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[0407] Next, the server learns the user's prompt engineering habits based on the extracted features. For example, if a user has a strong tendency to request specific data, this habit is learned. The learned habits are stored in the brain database as a numerical model.

[0408] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know the weather tomorrow," the prompt is sent to the server, which selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[0409] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[0410] This system can provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, it is possible to improve the overall accuracy of answers. This invention is particularly useful in generative AI systems that require flexible responses to users with different questioning styles.

[0411] The processing flow will be explained below.

[0412] Step 1:

[0413] The user inputs a prompt into the terminal, for example, "What is the weather like tomorrow?"

[0414] Step 2:

[0415] Records the prompts entered at the terminal and saves them in a log file or database.

[0416] Step 3:

[0417] The device sends the recorded prompt to the server, for example, using an HTTP request.

[0418] Step 4:

[0419] The server analyzes the received prompt, specifically extracting keywords and grammatical patterns from the prompt using natural language processing techniques.

[0420] Step 5:

[0421] The server converts the extracted features into feature vectors. For example, words such as "tomorrow," "weather," and "tell me" and their contexts are converted into numerical vectors.

[0422] Step 6:

[0423] The server classifies the feature vectors using a clustering algorithm and learns the user's prompt engineering habits.

[0424] Step 7:

[0425] The server stores the learned habit model in a brain database, which is then shared so that other users can use it.

[0426] Step 8:

[0427] The terminal receives a new prompt. For example, user B inputs the prompt "I want to know the weather tomorrow."

[0428] Step 9:

[0429] The terminal sends a new prompt to the server.

[0430] Step 10:

[0431] The server selects relevant habit models from the brain database, and the most appropriate habit model is chosen and applied to the new prompt.

[0432] Step 11:

[0433] The server reconstructs a new prompt based on the habit model, e.g., "Please tell me the specific weather data for tomorrow."

[0434] Step 12:

[0435] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[0436] Step 13:

[0437] The server sends the generated response to the terminal.

[0438] Step 14:

[0439] The device displays the generated answer to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[0440] Example 1

[0441] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0442] Conventional generative AI systems have difficulty taking into account the different prompting habits of each user, which can result in inappropriate answers being generated. Furthermore, there is a lack of efficient means for learning user habits and reconstructing prompts based on them. Furthermore, there is a need to effectively utilize natural language processing technology in the process of collecting and analyzing user prompts. By resolving these issues, it will be possible to provide more accurate answers for each user.

[0443] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0444] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting a habit model associated with a new prompt and reconstructing the prompt based on the habit, means for generating an answer by a generation AI using the reconstructed prompt, means for displaying the generated answer to the user, means for transmitting the prompt from the user's terminal to the server, means for the server to analyze the received prompt using natural language processing technology, and means for the server to select a habit model from a brain database. This enables the generation of appropriate answers that take into account each user's prompt habits.

[0445] A "prompt" is a sentence containing a question or instruction that a user enters into a generative AI system.

[0446] A "collection means" is a method or device for capturing prompts entered by multiple users into the system.

[0447] A "means for extracting features" is a method or device for identifying and extracting important keywords and grammatical patterns from collected prompts.

[0448] "Prompt engineering habits" refer to the particular patterns or tendencies that users have when entering prompts.

[0449] "Means for learning" refers to a method or device for recognizing and storing a user's prompt engineering habits based on extracted features using a machine learning algorithm or the like.

[0450] "Means for saving and sharing" refers to a method or device for saving the learned habit model in a database or the like and sharing it with other systems or users as needed.

[0451] A "reconstruction means" is a method or device for selecting a relevant habit model for a newly input prompt and improving or adjusting the prompt based on that model.

[0452] A "means for generating" is a method or device for generating responses to reconstructed prompts using generative AI.

[0453] The "displaying means" is a method or device for visually or audibly presenting the generated answer to the user.

[0454] A "means for sending" is a method or device for sending a prompt from a user's terminal to a server.

[0455] The "means for analyzing" refers to a method or device for using natural language processing techniques to examine in detail the prompt received by the server and understand its meaning and characteristics.

[0456] The "brain database" is a database that stores learned prompt engineering habit models of users and can be searched and retrieved as needed.

[0457] This invention is a generative AI system that provides users with multifaceted and accurate answers. The system begins operation when the user inputs a prompt sentence into a terminal. Specifically, when the user inputs the prompt sentence, "Please tell me what the weather will be like tomorrow," the prompt sentence is recorded by the terminal. The terminal then transmits the recorded prompt sentence to a server.

[0458] To analyze the prompt received by the server, natural language processing technology (such as NLTK or SpaCy) is used. The server uses this technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[0459] Next, the server learns the user's prompt engineering habits based on the extracted features. This learning process uses machine learning algorithms, and the learned habits are stored in the brain database as a numerical model. For example, if a user has a habit of requesting specific data, this tendency will be stored in the brain database.

[0460] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know tomorrow's weather," the prompt is sent to the server once, and the server selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[0461] The reconstructed prompt is passed to a generative AI model, which generates an appropriate response. Examples of generative AI models include OpenAI's GPT-4. This generative AI model has the ability to generate the most appropriate response based on the reconstructed prompt.

[0462] The generated answer is sent from the server to the terminal and displayed to the user through the terminal. For example, an answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed. This system allows users to receive more accurate and multifaceted answers based on their own prompt engineering habits.

[0463] This invention allows a generative AI system to provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, the overall accuracy of answers is improved. This system is particularly useful in generative AI model systems that require flexible responses to users with different questioning styles.

[0464] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0465] Step 1:

[0466] The user inputs a prompt sentence using a terminal. Specifically, the user inputs a prompt sentence such as "Please tell me what the weather will be tomorrow." using a keyboard or voice input. This prompt sentence becomes the input data for the process.

[0467] Step 2:

[0468] The terminal records the input prompt and sends the data to the server. The specific operation of the terminal is to save the prompt in memory and send the data to the server using an HTTP request. The input here is the prompt, and the output is the prompt sent to the server.

[0469] Step 3:

[0470] The server parses the received prompt. Specifically, it uses natural language processing techniques (such as NLTK or SpaCy) to tokenize the prompt and extract keywords and grammatical patterns. The input to this process is the prompt, and the output is the extracted keywords and grammatical patterns.

[0471] Step 4:

[0472] The server learns the user's prompt engineering habits based on the extracted keywords. The server uses a machine learning algorithm (such as a decision tree or random forest) to model the prompt engineering habits. The input is the extracted keywords, and the output is a model that reflects the user's habits. This model is stored in a brain database.

[0473] Step 5:

[0474] When a new prompt sentence is input, the server selects a highly relevant habit model from the brain database. For example, if a new prompt such as "I want to know the weather tomorrow," is input, the server receives the prompt and selects an appropriate habit model from the brain database. The input is the new prompt sentence, and the output is the selected habit model.

[0475] Step 6:

[0476] The server reconstructs the prompt sentence based on the selected habit model. As a specific operation of reconstruction, the server changes the prompt sentence "I want to know the weather tomorrow" to "Please tell me the specific weather data for tomorrow." The input is the selected habit model and the new prompt sentence, and the output is the reconstructed prompt sentence.

[0477] Step 7:

[0478] The reconstructed prompt is passed to a generative AI model, which generates an appropriate answer. GPT-4 is used as the generative AI model, and the answer is generated through an API request. The input is the reconstructed prompt, and the output is the generated answer.

[0479] Step 8:

[0480] The generated answer is sent from the server to the terminal. As a server operation, the generated answer is sent to the terminal using an HTTP response. The input is the generated answer, and the output is the answer sent to the terminal.

[0481] Step 9:

[0482] The user displays the generated answer through the terminal. The specific operation of the terminal is to display the received answer on the screen. The input is the answer received from the server, and the output is the answer displayed to the user. For example, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees."

[0483] (Application example 1)

[0484] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0485] In today's virtual stores, it is extremely important to quickly and accurately provide users with the product information and recommendations they need. However, typical generative AI systems often lack the ability to accurately understand the intent and preferences of users' questions, resulting in inability to provide the information they seek. Furthermore, when using voice prompts, the process of accurately converting the speech into text and adjusting it to match the user's intent is complex, and the entire process of displaying information on a visual device must be carried out efficiently and accurately. Therefore, technology is needed that can learn users' prompt engineering habits and provide answers that adapt to them.

[0486] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0487] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit, means for generating an answer by a generation AI using the adjusted prompt, means for displaying the generated answer to the user, means for allowing the user to obtain information using voice prompts, speech recognition means for converting the voice prompts into text, and means for displaying the generated answer on a visual device, thereby enabling optimal information provision taking into account the user's prompt engineering habits.

[0488] Definitions of important words

[0489] A "user" is an individual or organization that uses the system.

[0490] A "prompt" is a question or instruction that a user enters into a system.

[0491] "Collection means" refers to the software and hardware used to capture and record prompts.

[0492] "Feature extraction methods" refer to techniques for identifying and extracting important elements or keywords from the prompt.

[0493] "Prompt engineering habits" refer to patterns or tendencies that users have when entering prompts.

[0494] "Means of learning" refers to the technology used to analyze prompt engineering habits based on collected feature information and build a model.

[0495] "Means for saving and sharing" refers to saving the trained model in a database or the like and sharing it with other systems or users as needed.

[0496] "Adjustment" refers to the process of optimizing prompts based on user habits, allowing the generative AI to generate more appropriate answers.

[0497] "Generative AI" refers to a system that uses artificial intelligence technology to automatically generate answers based on input data.

[0498] "Displaying means" refers to a device or software for notifying the user of the generated answer.

[0499] "Voice prompt" refers to a question or instruction that the user speaks.

[0500] "Speech recognition means" refers to technology that converts voice prompts into text.

[0501] "Visual device" refers to an electronic device that visually presents information to a user.

[0502] MODE FOR CARRYING OUT THE INVENTION

[0503] In order to implement the present invention, a system is used in which a user wears smart glasses and obtains information using voice prompts. Specific embodiments are described in detail below.

[0504] System Configuration

[0505] 1. Hardware:

[0506] Smart glasses: Devices with built-in voice recognition microphones and displays.

[0507] Server: A server with high-performance computing power that is responsible for prompt analysis and running AI models.

[0508] 2. Software:

[0509] Speech Recognition Library: Use speech_recognition to convert user voice prompts into text.

[0510] Communication library: requests is used to send prompts to the server and retrieve results.

[0511] Natural language processing technology: We use spaCy and Transformers to analyze prompts and extract features.

[0512] Generative AI models: Models that generate answers based on prompts, such as GPT and BERT.

[0513] Data Processing and Computation

[0514] When a user enters a voice prompt, the voice data is converted to text by a speech recognition library, and the converted text is sent to the server, which then performs the following steps:

[0515] 1. Prompt analysis and feature extraction:

[0516] The server uses natural language processing techniques to extract important keywords and grammatical patterns from the text prompts.

[0517] 2. Prompt Engineering Habit Learning:

[0518] The server learns the user's prompt engineering habits based on the extracted features, and stores and shares the habit model in a brain database.

[0519] 3. Prompt adjustment and AI model execution:

[0520] For each new prompt, it selects a relevant habit model and adjusts the prompt accordingly, which is then passed to a generative AI model to generate the optimal answer.

[0521] 4. Show Answer:

[0522] The generated answers are sent to the smart glasses and displayed on the visual device.

[0523] Specific examples

[0524] A user asks a question through the smart glasses. For example, if a user inputs a voice prompt such as "What new sneakers do you recommend?", the prompt is converted into text and sent to the server. The server extracts keywords such as "new," "sneakers," and "recommended," and reconstructs the prompt based on the user's habits to "Please recommend some new sneakers with a flashy design." The generative AI model generates an appropriate answer, which is sent from the server to the smart glasses. The user's field of vision is displayed as follows: "The latest sneakers with a flashy design are the Nike Air Max, which comes in a wide range of color options."

[0525] Prompt Sentence Examples

[0526] "What new sneakers do you recommend?"

[0527] "Can you recommend some new sneakers with flashy designs?"

[0528] This system makes it possible to provide optimal information taking into account the user's prompt engineering habits.

[0529] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0530] Specific explanation of program processing

[0531] Step 1:

[0532] The user inputs a voice prompt into the smart glasses.

[0533] Input: A voice prompt from the user (e.g., "What new sneakers would you recommend?").

[0534] Output: The speech data received by the speech recognizer.

[0535] Specific operation: When the user speaks into the device, the smart glasses' built-in microphone captures the voice data.

[0536] Step 2:

[0537] The device converts the voice prompts into text.

[0538] Input: Audio data from smart glasses.

[0539] Output: A text prompt (e.g., "What new sneakers do you recommend?").

[0540] Specific operation: Uses the speech_recognition library to convert audio data into text data.

[0541] Step 3:

[0542] The terminal sends a text prompt to the server.

[0543] Input: A text prompt.

[0544] Output: The text prompt sent to the server.

[0545] Specific operation: Uses the requests library to send text data to the server via an HTTP POST request.

[0546] Step 4:

[0547] The server parses the prompt and extracts features.

[0548] Input: The text prompt sent from the terminal.

[0549] Output: Extracted feature data (e.g., keywords "new arrivals," "sneakers," "recommended").

[0550] Specific operation: Using the spaCy library and Transformers, it analyzes and extracts important keywords and grammatical patterns from text data.

[0551] Step 5:

[0552] The server learns the user's prompt engineering habits based on the extracted features.

[0553] Input: Extracted feature data.

[0554] Output: Habit model (e.g., patterns of keywords and phrases frequently used by the user).

[0555] Specific operation: The analyzed feature data is input into a machine learning algorithm to learn the user's prompt engineering habits.

[0556] Step 6:

[0557] The server stores and shares the learned habit models.

[0558] Input: The newly trained habit model.

[0559] Output: Habit model stored in the brain database.

[0560] What it does: Saves the habit model in a database so that it can be shared with other systems and users in the future.

[0561] Step 7:

[0562] The server selects the relevant habit model for the new prompt and reconstructs the prompt.

[0563] Input: A new prompt and a saved habit model.

[0564] Output: A tailored prompt (e.g., "Can you recommend some new sneakers with a flashy design?").

[0565] What it does: Refers to the saved habit model and adjusts new prompts appropriately.

[0566] Step 8:

[0567] The server uses the tailored prompts to generate answers using the generative AI.

[0568] Input: Tailored prompt.

[0569] Output: The generated answer (e.g., "The latest, flashiest sneaker design is the Nike Air Max, which comes in a wide variety of colors.").

[0570] What it does: Feeds the tailored prompts into an AI model to generate the best answer.

[0571] Step 9:

[0572] The server sends the generated response to the terminal.

[0573] Input: The generated answer.

[0574] Output: The answer sent to the terminal.

[0575] Specific operation: The generated answer is sent to the terminal as an HTTP response.

[0576] Step 10:

[0577] The terminal displays the generated answer on the visual device of the smart glasses.

[0578] Input: The generated answer received from the server.

[0579] Output: The answer that is displayed to the user.

[0580] Specific behavior: Displays text on the smart glasses display to provide information to the user.

[0581] Through the above steps, the user can efficiently obtain optimal product information in the virtual store using voice prompts.

[0582] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0583] This invention adds an emotion engine to a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[0584] First, the user inputs a prompt into the device, and the device reads the emotion at that time. For example, while the user inputs the prompt "Please tell me what the weather will be like tomorrow," emotional information (e.g., joy, anger, sadness, etc.) is acquired through the camera and microphone. This emotional information is then sent by the device to the server.

[0585] The server analyzes the received prompt and emotional information. Specifically, the server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[0586] The server converts the extracted prompt features and emotion features into feature vectors, for example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[0587] The server then classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns. For example, if a user has a strong tendency to seek specific data and is feeling "impatient," the server learns the habits and emotional patterns. The learned habits and emotional patterns are stored in a brain database as a numerical model.

[0588] Next, when a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server, which then selects an appropriate habit model and emotion model. The prompt is reconstructed based on the habit model and emotion model, and adjusted to a form such as, "Please tell me the specific weather data for tomorrow as soon as possible."

[0589] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0590] This system can provide more appropriate, multifaceted, and emotionally relevant answers by taking into account the user's prompt engineering habits and emotions. Furthermore, by sharing each user's habit model and emotion model, it is possible to improve overall answer accuracy and user satisfaction. This invention is particularly useful in generative AI systems that require flexible responses to users with different question styles and various emotions.

[0591] The processing flow will be explained below.

[0592] Step 1:

[0593] A user inputs a prompt into a terminal. For example, the user inputs the prompt "What will the weather be like tomorrow?"

[0594] Step 2:

[0595] The device uses a camera and microphone to acquire information about the user's emotions. For example, it can recognize the emotion of "anxiety" by analyzing the user's facial expressions and tone of voice.

[0596] Step 3:

[0597] The device records the prompts and emotional information entered, and the recorded data is stored in a log file or database.

[0598] Step 4:

[0599] The device sends the recorded prompts and emotion information to the server using HTTP requests or similar.

[0600] Step 5:

[0601] The server analyzes the received prompt and emotional information. Specifically, it uses natural language processing technology to extract keywords and grammatical patterns from the prompt, and then analyzes the emotional information using an emotional engine.

[0602] Step 6:

[0603] The server converts the extracted prompt features and emotion features into feature vectors. For example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[0604] Step 7:

[0605] The server classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns.

[0606] Step 8:

[0607] The server stores the learned habit models and emotion models in a brain database, which is then shared for other users to use.

[0608] Step 9:

[0609] The terminal receives a new prompt. For example, user B inputs a prompt such as "I want to know the weather tomorrow." At the same time, the emotion "I'm anxious" is acquired.

[0610] Step 10:

[0611] The device sends the new prompt and emotion information to the server.

[0612] Step 11:

[0613] The server selects relevant habit models and emotion models from the brain database, and the most appropriate habit model and emotion model are selected and applied to the new prompt.

[0614] Step 12:

[0615] The server reconstructs a new prompt based on the habit model and emotion model, for example, "Please tell me the specific weather data for tomorrow as soon as possible."

[0616] Step 13:

[0617] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[0618] Step 14:

[0619] The server sends the generated response to the terminal.

[0620] Step 15:

[0621] The device displays the generated answer to the user. For example, the answer may read, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0622] Example 2

[0623] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0624] Conventional generative AI systems have a problem in that simply generating answers to prompts entered by the user makes it difficult to respond flexibly while taking into account the user's emotions and prompt engineering habits. As a result, the information the user seeks is not accurately provided, often resulting in dissatisfaction. The present invention provides a system that generates more accurate and multifaceted answers by taking into account the user's emotional information and prompt engineering habits.

[0625] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0626] In this invention, the server includes: means for collecting prompts input by multiple users; means for acquiring the collected prompts and emotional information; means for transmitting the acquired prompts and emotional information to the server; means for the server to analyze the prompts and emotional information; means for generating feature vectors from the analyzed prompts and emotional information; means for classifying the generated feature vectors using a clustering algorithm and learning the user's prompt engineering habits and emotional patterns; means for saving and sharing the learned habit model and emotional model in a brain database; means for selecting a habit model and emotional model associated with a new prompt and reconstructing the prompt based on the habits and emotions; means for generating an answer using a generation AI using the reconstructed prompt; and means for displaying the generated answer to the user. This enables the generation of appropriate answers that take into account the user's emotional information and prompt engineering habits.

[0627] "User" refers to a person using the system to enter prompts.

[0628] "Terminal" refers to a device that allows a user to input prompts, obtain emotion information, and transmit it to a server.

[0629] A "prompt" is a question or instruction text that a user enters into a system.

[0630] "Emotional information" refers to data indicating the user's emotional state that is acquired by the device through the camera or microphone.

[0631] "Server" refers to the central computer that analyzes prompts and emotional information, generates feature vectors, clusters, and generates answers.

[0632] "Natural language processing" refers to the technology of extracting keywords and grammatical patterns from prompts.

[0633] "Emotion engine" refers to technology that analyzes emotional information.

[0634] "Feature vector" refers to vector-format data that quantifies the analyzed prompt and emotional information.

[0635] A "clustering algorithm" refers to an algorithm for classifying feature vectors.

[0636] "Prompt engineering" refers to techniques for reconstructing the questions or instructions that users enter into a system.

[0637] The "habit model" refers to a model generated by learning the prompt engineering habits of a user.

[0638] An "emotion model" refers to a model generated by learning the user's emotional patterns.

[0639] The "brain database" refers to a database that stores and manages learned habit models and emotion models.

[0640] "Generative AI" refers to artificial intelligence that generates appropriate answers based on reconstructed prompts.

[0641] This invention relates to a system that analyzes prompts entered by users and their emotional information at the time, and generates appropriate answers using generative AI. This system is configured by combining multiple hardware and software components, and is implemented as follows.

[0642] This system starts when a user inputs a prompt into a terminal. For example, the user might input, "Please tell me what the weather will be like tomorrow." At this time, the device's built-in camera and microphone are used to acquire the user's emotional information (e.g., whether they are anxious or happy). The acquired prompt and emotional information are then sent from the terminal to the server.

[0643] The server then analyzes the received prompt and emotional information. The server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, keywords such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[0644] The analyzed prompt and emotion information are converted into numerical vectors. The server classifies these numerical vectors using a clustering algorithm (e.g., K-means clustering) to learn the user's prompt engineering habits and emotion patterns. This generates a habit model and emotion model for each user. The learned habit model and emotion model are stored in the brain database.

[0645] When a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server. The server selects an appropriate habit model and emotion model, and the prompt is reconstructed based on the habit model and emotion model. For example, the prompt may be adjusted to something like, "Please tell me the specific weather data for tomorrow as soon as possible."

[0646] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0647] As a specific example, if a user inputs "Tell me how to prepare for a presentation," and the device detects the emotion "I'm nervous," this prompt and emotional information are sent to the server. Based on the emotion model and habit model, the server reconstructs the prompt into something like, "I'm nervous, but can you briefly tell me how to prepare for a presentation?" An example of a prompt sentence to input to the generative AI model is, "The user is nervous and would like more details about how to prepare for a presentation. Please provide specific and calming instructions."

[0648] This system can generate appropriate answers that take into account the user's emotional information and prompt engineering habits, thereby improving user satisfaction.

[0649] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0650] Step 1:

[0651] The user enters the prompt.

[0652] In a specific operation, the user inputs "What will the weather be like tomorrow?" into the terminal. The input obtained is a prompt "What will the weather be like tomorrow?" This prompt is taken into the terminal as text data.

[0653] Step 2:

[0654] The device acquires emotion information.

[0655] Specifically, the device uses its built-in camera and microphone to perform image recognition and voice analysis. Images of the user's facial expressions and voice data are obtained as input. These data are analyzed to obtain emotional information such as "anxious." The emotional information "anxious" is generated as output.

[0656] Step 3:

[0657] The device sends prompts and emotion information to the server.

[0658] Specifically, the terminal uses the communication module to send a prompt, "What will the weather be like tomorrow?" and emotional information, "I'm anxious," in packet form to the server. The input is the prompt and emotional information, and the output is a data packet.

[0659] Step 4:

[0660] The server parses the prompt and emotion information.

[0661] Specifically, the server uses natural language processing technology and an emotion engine. The prompt "Please tell me what the weather will be tomorrow" and the emotion information "I'm anxious" arrive as input to the server. Natural language processing technology extracts keywords such as "tomorrow," "weather," and "tell me," while the emotion engine simultaneously analyzes the emotion "I'm anxious." The output is the keyword and emotion information.

[0662] Step 5:

[0663] The server generates the feature vector.

[0664] Specifically, the system quantifies the keywords and emotional information extracted from the prompt. The inputs are the keywords "tomorrow," "weather," and "tell me about" and the emotional information "anxious." These are converted into numerical vectors, and the output is a prompt vector [0.6, 0.8, 0.5] and an emotional vector [0.1, 0.9].

[0665] Step 6:

[0666] The server classifies the feature vectors using a clustering algorithm.

[0667] Specifically, it uses algorithms such as K-means clustering. The inputs are a prompt vector [0.6, 0.8, 0.5] and an emotion vector [0.1, 0.9]. Clustering is performed to classify the user's prompt engineering habits and emotion patterns. The output is a feature vector classified into clusters.

[0668] Step 7:

[0669] The server stores the habit model and emotion model in a brain database.

[0670] Specifically, the clustering results are written to the database. The input is a feature vector classified into clusters, which is saved in the brain database as a habit model and an emotion model. The output is an updated database.

[0671] Step 8:

[0672] Select relevant habit models and emotion models for the new prompt.

[0673] Specifically, the server searches the brain database and selects the most relevant habit model and emotion model. The inputs are a new prompt, "I want to know tomorrow's weather," and the emotion information, "I'm anxious." The output is the selection of the relevant habit model and emotion model.

[0674] Step 9:

[0675] The server reconstructs the prompt.

[0676] The specific operation is to adjust the prompt based on the selected habit model and emotion model. The prompt and the associated model are obtained as input. The prompt is reconstructed to "Please tell me the specific weather data for tomorrow as soon as possible." and the reconstructed prompt is generated as output.

[0677] Step 10:

[0678] The server generates an answer using a generation AI.

[0679] Specifically, the reconstructed prompt is input into the generation AI, which then generates an answer. The reconstructed prompt is obtained as input. An answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​is generated, and an answer is generated as output.

[0680] Step 11:

[0681] The server sends the generated answer to the terminal, which displays it to the user.

[0682] Specifically, the server sends the generated answer to the terminal, and the terminal displays the answer to the user. The generated answer is obtained as input. The user can see the answer on the screen: "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​The output is the answer displayed to the user.

[0683] (Application example 2)

[0684] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0685] Conventional generative AI systems generate answers based on user input prompts, but because they do not take the user's emotional state into consideration, they often generate inappropriate answers or answers that do not meet the user's expectations. Furthermore, in emergencies, it is necessary to accurately grasp the user's emotions and provide a quick and appropriate response, but current systems have difficulty in doing so.

[0686] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0687] In this invention, the server includes means for collecting prompts input by multiple users, means for acquiring user emotion information using a camera or microphone, and means for analyzing the acquired emotion information to identify the user emotion, thereby enabling the generation of appropriate answers that take user emotion into consideration and quick and appropriate responses in emergencies.

[0688] "User emotion information" is information that indicates the emotional state of the user, obtained from the user's facial expression, tone of voice, and the like.

[0689] A "feature extraction means" is a means capable of extracting meaningful features or patterns from prompts and emotional information.

[0690] "Prompt engineering habits" refer to the characteristics or patterns that a particular user consistently exhibits when completing prompts.

[0691] The "means for identifying emotions" is a means having a function of analyzing acquired emotion information and recognizing the type of emotion the user has.

[0692] The "means for adjusting the prompt" is a means that has the function of reconstructing the original prompt more appropriately, taking into account the user's emotions and prompt engineering habits.

[0693] A "means for generating answers using generative AI" is a means that has the function of generating appropriate answers based on tailored prompts.

[0694] "Problems to be solved by the invention" refers to the specific problems or obstacles that the invention aims to solve, such as the inability of conventional systems to generate appropriate answers because they do not take into account the user's emotional state or prompt engineering habits, and the lack of rapid and appropriate responses in emergencies.

[0695] "Means for solving the problem" refers to the specific technical means proposed to solve the above problem.

[0696] This invention is realized by incorporating user emotional information into a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments for carrying out this invention are described below.

[0697] First, the user inputs a prompt into the device. At the same time, the user's emotional information is acquired using a camera and microphone. For example, while the user inputs the prompt "What will the weather be like tomorrow?", the camera captures the user's facial expression and the microphone captures the user's tone of voice. This emotional information is sent from the user's device to the server.

[0698] The server analyzes the received prompts and emotional information. Specifically, the server uses natural language processing technology (e.g., generative AI models such as GPT-3) to extract keywords and grammatical patterns from the prompts, and simultaneously analyzes the emotional information using an emotional engine. The analyzed data is converted into a numerical vector, and then a clustering algorithm is used to classify and learn the user's prompt engineering habits and emotional patterns. This information is stored in a brain database as a feature vector.

[0699] When a new prompt is input, the server selects the relevant habit model and emotion model from the brain database and adjusts the prompt based on the user's prompt and emotion information. For example, if the user inputs the prompt "I want to know the weather tomorrow," and the emotion "I'm anxious" is detected, the server will reconstruct the prompt as "Please tell me the specific weather data for tomorrow as soon as possible" based on the prompt and emotion information.

[0700] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0701] The hardware used includes the user's device (smartphone, tablet, etc.), camera, and microphone. The software used includes a generative AI model (e.g., GPT-3) for implementing natural language processing technology and an emotion engine for emotion analysis (e.g., an emotion classification model using TensorFlow or Keras).

[0702] For example, if the user types the following prompt:

[0703] 1. "What's the weather going to be like tomorrow?" (User is anxious)

[0704] 2. "I'm in a hurry and need the latest news headlines right away." (Users are impatient)

[0705] The server applies appropriate habit models and emotion models to these prompts to generate answers that match the user's emotional state, allowing the user to obtain more satisfying information.

[0706] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0707] Step 1:

[0708] The user types a prompt into the terminal.

[0709] Input: User prompt (e.g., "What's the weather like tomorrow?")

[0710] Output: The prompt is typed into the terminal.

[0711] Step 2:

[0712] The device uses a camera and microphone to acquire information about the user's emotions.

[0713] Input: User's facial expression (camera), user's tone of voice (microphone)

[0714] Output: Emotional information (e.g., facial image data, audio data)

[0715] Step 3:

[0716] The device transmits the acquired prompt text and emotion information to the server.

[0717] Input: prompt sentence, emotion information

[0718] Output: The prompt and emotion information are converted into a dedicated format and sent to the server.

[0719] Step 4:

[0720] The server analyzes the prompt sentence using natural language processing technology to extract keywords and grammatical patterns.

[0721] Input: Prompt sentence (e.g. "What will the weather be like tomorrow?")

[0722] Output: Extracted features (keywords: "tomorrow", "weather", "tell me")

[0723] Step 5:

[0724] The server uses an emotion engine to analyze the transmitted emotion information.

[0725] Input: Emotion information (face video data, audio data)

[0726] Output: Parsed emotion (e.g. "anxious")

[0727] Step 6:

[0728] The server converts the features and sentiment information of the parsed prompt into a numerical vector.

[0729] Input: extracted features, parsed emotions

[0730] Output: Feature vector (e.g. "tomorrow", "weather", "tell me", "I'm anxious" → numeric vector)

[0731] Step 7:

[0732] The server clusters the feature vectors and learns the user's prompt engineering habits and emotional patterns.

[0733] Input: feature vector

[0734] Output: Learned habit model and emotion model

[0735] Step 8:

[0736] When a new prompt is entered, the server selects the relevant habit model and emotion model from the brain database.

[0737] Input: New prompt, emotion information

[0738] Output: Appropriate habit model and emotion model

[0739] Step 9:

[0740] The server adjusts the prompt sentence based on the selected habit model and emotion model.

[0741] Input: New prompt, habit model, emotion model

[0742] Output: The adjusted prompt (e.g., "Please give me the specific weather data for tomorrow as soon as possible.")

[0743] Step 10:

[0744] The server passes the adjusted prompt to the generative AI model, which generates the optimal answer.

[0745] Input: Adjusted prompt text

[0746] Output: Generated answer (e.g. "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[0747] Step 11:

[0748] The server sends the generated response to the terminal.

[0749] Input: Generated answer

[0750] Output: The generated answer sent by the server

[0751] Step 12:

[0752] The terminal displays the generated answer to the user.

[0753] Input: Generated answer

[0754] Output: The answer shown to the user (e.g., "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[0755] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0756] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0757] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0758] [Third embodiment]

[0759] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0760] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0761] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0762] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0763] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0764] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0765] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0766] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0767] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0768] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0769] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0770] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0771] This invention is a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[0772] First, a user inputs a prompt into the terminal. For example, if the user inputs the prompt "What will the weather be like tomorrow?", the prompt is recorded by the terminal. Then, the terminal sends the recorded prompt to the server.

[0773] The server then analyzes the received prompt. Specifically, the server uses natural language processing technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[0774] Next, the server learns the user's prompt engineering habits based on the extracted features. For example, if a user has a strong tendency to request specific data, this habit is learned. The learned habits are stored in the brain database as a numerical model.

[0775] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know the weather tomorrow," the prompt is sent to the server, which selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[0776] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[0777] This system can provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, it is possible to improve the overall accuracy of answers. This invention is particularly useful in generative AI systems that require flexible responses to users with different questioning styles.

[0778] The processing flow will be explained below.

[0779] Step 1:

[0780] The user inputs a prompt into the terminal, for example, "What is the weather like tomorrow?"

[0781] Step 2:

[0782] Records the prompts entered at the terminal and saves them in a log file or database.

[0783] Step 3:

[0784] The device sends the recorded prompt to the server, for example, using an HTTP request.

[0785] Step 4:

[0786] The server analyzes the received prompt, specifically extracting keywords and grammatical patterns from the prompt using natural language processing techniques.

[0787] Step 5:

[0788] The server converts the extracted features into feature vectors. For example, words such as "tomorrow," "weather," and "tell me" and their contexts are converted into numerical vectors.

[0789] Step 6:

[0790] The server classifies the feature vectors using a clustering algorithm and learns the user's prompt engineering habits.

[0791] Step 7:

[0792] The server stores the learned habit model in a brain database, which is then shared so that other users can use it.

[0793] Step 8:

[0794] The terminal receives a new prompt. For example, user B inputs the prompt "I want to know the weather tomorrow."

[0795] Step 9:

[0796] The terminal sends a new prompt to the server.

[0797] Step 10:

[0798] The server selects relevant habit models from the brain database, and the most appropriate habit model is chosen and applied to the new prompt.

[0799] Step 11:

[0800] The server reconstructs a new prompt based on the habit model, e.g., "Please tell me the specific weather data for tomorrow."

[0801] Step 12:

[0802] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[0803] Step 13:

[0804] The server sends the generated response to the terminal.

[0805] Step 14:

[0806] The device displays the generated answer to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[0807] Example 1

[0808] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0809] Conventional generative AI systems have difficulty taking into account the different prompting habits of each user, which can result in inappropriate answers being generated. Furthermore, there is a lack of efficient means for learning user habits and reconstructing prompts based on them. Furthermore, there is a need to effectively utilize natural language processing technology in the process of collecting and analyzing user prompts. By resolving these issues, it will be possible to provide more accurate answers for each user.

[0810] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0811] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting a habit model associated with a new prompt and reconstructing the prompt based on the habit, means for generating an answer by a generation AI using the reconstructed prompt, means for displaying the generated answer to the user, means for transmitting the prompt from the user's terminal to the server, means for the server to analyze the received prompt using natural language processing technology, and means for the server to select a habit model from a brain database. This enables the generation of appropriate answers that take into account each user's prompt habits.

[0812] A "prompt" is a sentence containing a question or instruction that a user enters into a generative AI system.

[0813] A "collection means" is a method or device for capturing prompts entered by multiple users into the system.

[0814] A "means for extracting features" is a method or device for identifying and extracting important keywords and grammatical patterns from collected prompts.

[0815] "Prompt engineering habits" refer to the particular patterns or tendencies that users have when entering prompts.

[0816] "Means for learning" refers to a method or device for recognizing and storing a user's prompt engineering habits based on extracted features using a machine learning algorithm or the like.

[0817] "Means for saving and sharing" refers to a method or device for saving the learned habit model in a database or the like and sharing it with other systems or users as needed.

[0818] A "reconstruction means" is a method or device for selecting a relevant habit model for a newly input prompt and improving or adjusting the prompt based on that model.

[0819] A "means for generating" is a method or device for generating responses to reconstructed prompts using generative AI.

[0820] The "displaying means" is a method or device for visually or audibly presenting the generated answer to the user.

[0821] A "means for sending" is a method or device for sending a prompt from a user's terminal to a server.

[0822] The "means for analyzing" refers to a method or device for using natural language processing techniques to examine in detail the prompt received by the server and understand its meaning and characteristics.

[0823] The "brain database" is a database that stores learned prompt engineering habit models of users and can be searched and retrieved as needed.

[0824] This invention is a generative AI system that provides users with multifaceted and accurate answers. The system begins operation when the user inputs a prompt sentence into a terminal. Specifically, when the user inputs the prompt sentence, "Please tell me what the weather will be like tomorrow," the prompt sentence is recorded by the terminal. The terminal then transmits the recorded prompt sentence to a server.

[0825] To analyze the prompt received by the server, natural language processing technology (such as NLTK or SpaCy) is used. The server uses this technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[0826] Next, the server learns the user's prompt engineering habits based on the extracted features. This learning process uses machine learning algorithms, and the learned habits are stored in the brain database as a numerical model. For example, if a user has a habit of requesting specific data, this tendency will be stored in the brain database.

[0827] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know tomorrow's weather," the prompt is sent to the server once, and the server selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[0828] The reconstructed prompt is passed to a generative AI model, which generates an appropriate response. Examples of generative AI models include OpenAI's GPT-4. This generative AI model has the ability to generate the most appropriate response based on the reconstructed prompt.

[0829] The generated answer is sent from the server to the terminal and displayed to the user through the terminal. For example, an answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed. This system allows users to receive more accurate and multifaceted answers based on their own prompt engineering habits.

[0830] This invention allows a generative AI system to provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, the overall accuracy of answers is improved. This system is particularly useful in generative AI model systems that require flexible responses to users with different questioning styles.

[0831] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0832] Step 1:

[0833] The user inputs a prompt sentence using a terminal. Specifically, the user inputs a prompt sentence such as "Please tell me what the weather will be tomorrow." using a keyboard or voice input. This prompt sentence becomes the input data for the process.

[0834] Step 2:

[0835] The terminal records the input prompt and sends the data to the server. The specific operation of the terminal is to save the prompt in memory and send the data to the server using an HTTP request. The input here is the prompt, and the output is the prompt sent to the server.

[0836] Step 3:

[0837] The server parses the received prompt. Specifically, it uses natural language processing techniques (such as NLTK or SpaCy) to tokenize the prompt and extract keywords and grammatical patterns. The input to this process is the prompt, and the output is the extracted keywords and grammatical patterns.

[0838] Step 4:

[0839] The server learns the user's prompt engineering habits based on the extracted keywords. The server uses a machine learning algorithm (such as a decision tree or random forest) to model the prompt engineering habits. The input is the extracted keywords, and the output is a model that reflects the user's habits. This model is stored in a brain database.

[0840] Step 5:

[0841] When a new prompt sentence is input, the server selects a highly relevant habit model from the brain database. For example, if a new prompt such as "I want to know the weather tomorrow," is input, the server receives the prompt and selects an appropriate habit model from the brain database. The input is the new prompt sentence, and the output is the selected habit model.

[0842] Step 6:

[0843] The server reconstructs the prompt sentence based on the selected habit model. As a specific operation of reconstruction, the server changes the prompt sentence "I want to know the weather tomorrow" to "Please tell me the specific weather data for tomorrow." The input is the selected habit model and the new prompt sentence, and the output is the reconstructed prompt sentence.

[0844] Step 7:

[0845] The reconstructed prompt is passed to a generative AI model, which generates an appropriate answer. GPT-4 is used as the generative AI model, and the answer is generated through an API request. The input is the reconstructed prompt, and the output is the generated answer.

[0846] Step 8:

[0847] The generated answer is sent from the server to the terminal. As a server operation, the generated answer is sent to the terminal using an HTTP response. The input is the generated answer, and the output is the answer sent to the terminal.

[0848] Step 9:

[0849] The user displays the generated answer through the terminal. The specific operation of the terminal is to display the received answer on the screen. The input is the answer received from the server, and the output is the answer displayed to the user. For example, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees."

[0850] (Application example 1)

[0851] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0852] In today's virtual stores, it is extremely important to quickly and accurately provide users with the product information and recommendations they need. However, typical generative AI systems often lack the ability to accurately understand the intent and preferences of users' questions, resulting in inability to provide the information they seek. Furthermore, when using voice prompts, the process of accurately converting the speech into text and adjusting it to match the user's intent is complex, and the entire process of displaying information on a visual device must be carried out efficiently and accurately. Therefore, technology is needed that can learn users' prompt engineering habits and provide answers that adapt to them.

[0853] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0854] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit, means for generating an answer by a generation AI using the adjusted prompt, means for displaying the generated answer to the user, means for allowing the user to obtain information using voice prompts, speech recognition means for converting the voice prompts into text, and means for displaying the generated answer on a visual device, thereby enabling optimal information provision taking into account the user's prompt engineering habits.

[0855] Definitions of important words

[0856] A "user" is an individual or organization that uses the system.

[0857] A "prompt" is a question or instruction that a user enters into a system.

[0858] "Collection means" refers to the software and hardware used to capture and record prompts.

[0859] "Feature extraction methods" refer to techniques for identifying and extracting important elements or keywords from the prompt.

[0860] "Prompt engineering habits" refer to patterns or tendencies that users have when entering prompts.

[0861] "Means of learning" refers to the technology used to analyze prompt engineering habits based on collected feature information and build a model.

[0862] "Means for saving and sharing" refers to saving the trained model in a database or the like and sharing it with other systems or users as needed.

[0863] "Adjustment" refers to the process of optimizing prompts based on user habits, allowing the generative AI to generate more appropriate answers.

[0864] "Generative AI" refers to a system that uses artificial intelligence technology to automatically generate answers based on input data.

[0865] "Displaying means" refers to a device or software for notifying the user of the generated answer.

[0866] "Voice prompt" refers to a question or instruction that the user speaks.

[0867] "Speech recognition means" refers to technology that converts voice prompts into text.

[0868] "Visual device" refers to an electronic device that visually presents information to a user.

[0869] MODE FOR CARRYING OUT THE INVENTION

[0870] In order to implement the present invention, a system is used in which a user wears smart glasses and obtains information using voice prompts. Specific embodiments are described in detail below.

[0871] System Configuration

[0872] 1. Hardware:

[0873] Smart glasses: Devices with built-in voice recognition microphones and displays.

[0874] Server: A server with high-performance computing power that is responsible for prompt analysis and running AI models.

[0875] 2. Software:

[0876] Speech Recognition Library: Use speech_recognition to convert user voice prompts into text.

[0877] Communication library: requests is used to send prompts to the server and retrieve results.

[0878] Natural language processing technology: We use spaCy and Transformers to analyze prompts and extract features.

[0879] Generative AI models: Models that generate answers based on prompts, such as GPT and BERT.

[0880] Data Processing and Computation

[0881] When a user enters a voice prompt, the voice data is converted to text by a speech recognition library, and the converted text is sent to the server, which then performs the following steps:

[0882] 1. Prompt analysis and feature extraction:

[0883] The server uses natural language processing techniques to extract important keywords and grammatical patterns from the text prompts.

[0884] 2. Prompt Engineering Habit Learning:

[0885] The server learns the user's prompt engineering habits based on the extracted features, and stores and shares the habit model in a brain database.

[0886] 3. Prompt adjustment and AI model execution:

[0887] For each new prompt, it selects a relevant habit model and adjusts the prompt accordingly, which is then passed to a generative AI model to generate the optimal answer.

[0888] 4. Show Answer:

[0889] The generated answers are sent to the smart glasses and displayed on the visual device.

[0890] Specific examples

[0891] A user asks a question through the smart glasses. For example, if a user inputs a voice prompt such as "What new sneakers do you recommend?", the prompt is converted into text and sent to the server. The server extracts keywords such as "new," "sneakers," and "recommended," and reconstructs the prompt based on the user's habits to "Please recommend some new sneakers with a flashy design." The generative AI model generates an appropriate answer, which is sent from the server to the smart glasses. The user's field of vision is displayed as follows: "The latest sneakers with a flashy design are the Nike Air Max, which comes in a wide range of color options."

[0892] Prompt Sentence Examples

[0893] "What new sneakers do you recommend?"

[0894] "Can you recommend some new sneakers with flashy designs?"

[0895] This system makes it possible to provide optimal information taking into account the user's prompt engineering habits.

[0896] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0897] Specific explanation of program processing

[0898] Step 1:

[0899] The user inputs a voice prompt into the smart glasses.

[0900] Input: A voice prompt from the user (e.g., "What new sneakers would you recommend?").

[0901] Output: The speech data received by the speech recognizer.

[0902] Specific operation: When the user speaks into the device, the smart glasses' built-in microphone captures the voice data.

[0903] Step 2:

[0904] The device converts the voice prompts into text.

[0905] Input: Audio data from smart glasses.

[0906] Output: A text prompt (e.g., "What new sneakers do you recommend?").

[0907] Specific operation: Uses the speech_recognition library to convert audio data into text data.

[0908] Step 3:

[0909] The terminal sends a text prompt to the server.

[0910] Input: A text prompt.

[0911] Output: The text prompt sent to the server.

[0912] Specific operation: Uses the requests library to send text data to the server via an HTTP POST request.

[0913] Step 4:

[0914] The server parses the prompt and extracts features.

[0915] Input: The text prompt sent from the terminal.

[0916] Output: Extracted feature data (e.g., keywords "new arrivals," "sneakers," "recommended").

[0917] Specific operation: Using the spaCy library and Transformers, it analyzes and extracts important keywords and grammatical patterns from text data.

[0918] Step 5:

[0919] The server learns the user's prompt engineering habits based on the extracted features.

[0920] Input: Extracted feature data.

[0921] Output: Habit model (e.g., patterns of keywords and phrases frequently used by the user).

[0922] Specific operation: The analyzed feature data is input into a machine learning algorithm to learn the user's prompt engineering habits.

[0923] Step 6:

[0924] The server stores and shares the learned habit models.

[0925] Input: The newly trained habit model.

[0926] Output: Habit model stored in the brain database.

[0927] What it does: Saves the habit model in a database so that it can be shared with other systems and users in the future.

[0928] Step 7:

[0929] The server selects the relevant habit model for the new prompt and reconstructs the prompt.

[0930] Input: A new prompt and a saved habit model.

[0931] Output: A tailored prompt (e.g., "Can you recommend some new sneakers with a flashy design?").

[0932] What it does: Refers to the saved habit model and adjusts new prompts appropriately.

[0933] Step 8:

[0934] The server uses the tailored prompts to generate answers using the generative AI.

[0935] Input: Tailored prompt.

[0936] Output: The generated answer (e.g., "The latest, flashiest sneaker design is the Nike Air Max, which comes in a wide variety of colors.").

[0937] What it does: Feeds the tailored prompts into an AI model to generate the best answer.

[0938] Step 9:

[0939] The server sends the generated response to the terminal.

[0940] Input: The generated answer.

[0941] Output: The answer sent to the terminal.

[0942] Specific operation: The generated answer is sent to the terminal as an HTTP response.

[0943] Step 10:

[0944] The terminal displays the generated answer on the visual device of the smart glasses.

[0945] Input: The generated answer received from the server.

[0946] Output: The answer that is displayed to the user.

[0947] Specific behavior: Displays text on the smart glasses display to provide information to the user.

[0948] Through the above steps, the user can efficiently obtain optimal product information in the virtual store using voice prompts.

[0949] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0950] This invention adds an emotion engine to a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[0951] First, the user inputs a prompt into the device, and the device reads the emotion at that time. For example, while the user inputs the prompt "Please tell me what the weather will be like tomorrow," emotional information (e.g., joy, anger, sadness, etc.) is acquired through the camera and microphone. This emotional information is then sent by the device to the server.

[0952] The server analyzes the received prompt and emotional information. Specifically, the server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[0953] The server converts the extracted prompt features and emotion features into feature vectors, for example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[0954] The server then classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns. For example, if a user has a strong tendency to seek specific data and is feeling "impatient," the server learns the habits and emotional patterns. The learned habits and emotional patterns are stored in a brain database as a numerical model.

[0955] Next, when a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server, which then selects an appropriate habit model and emotion model. The prompt is reconstructed based on the habit model and emotion model, and adjusted to a form such as, "Please tell me the specific weather data for tomorrow as soon as possible."

[0956] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0957] This system can provide more appropriate, multifaceted, and emotionally relevant answers by taking into account the user's prompt engineering habits and emotions. Furthermore, by sharing each user's habit model and emotion model, it is possible to improve overall answer accuracy and user satisfaction. This invention is particularly useful in generative AI systems that require flexible responses to users with different question styles and various emotions.

[0958] The processing flow will be explained below.

[0959] Step 1:

[0960] A user inputs a prompt into a terminal. For example, the user inputs the prompt "What will the weather be like tomorrow?"

[0961] Step 2:

[0962] The device uses a camera and microphone to acquire information about the user's emotions. For example, it can recognize the emotion of "anxiety" by analyzing the user's facial expressions and tone of voice.

[0963] Step 3:

[0964] The device records the prompts and emotional information entered, and the recorded data is stored in a log file or database.

[0965] Step 4:

[0966] The device sends the recorded prompts and emotion information to the server using HTTP requests or similar.

[0967] Step 5:

[0968] The server analyzes the received prompt and emotional information. Specifically, it uses natural language processing technology to extract keywords and grammatical patterns from the prompt, and then analyzes the emotional information using an emotional engine.

[0969] Step 6:

[0970] The server converts the extracted prompt features and emotion features into feature vectors. For example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[0971] Step 7:

[0972] The server classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns.

[0973] Step 8:

[0974] The server stores the learned habit models and emotion models in a brain database, which is then shared for other users to use.

[0975] Step 9:

[0976] The terminal receives a new prompt. For example, user B inputs a prompt such as "I want to know the weather tomorrow." At the same time, the emotion "I'm anxious" is acquired.

[0977] Step 10:

[0978] The device sends the new prompt and emotion information to the server.

[0979] Step 11:

[0980] The server selects relevant habit models and emotion models from the brain database, and the most appropriate habit model and emotion model are selected and applied to the new prompt.

[0981] Step 12:

[0982] The server reconstructs a new prompt based on the habit model and emotion model, for example, "Please tell me the specific weather data for tomorrow as soon as possible."

[0983] Step 13:

[0984] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[0985] Step 14:

[0986] The server sends the generated response to the terminal.

[0987] Step 15:

[0988] The device displays the generated answer to the user. For example, the answer may read, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[0989] Example 2

[0990] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0991] Conventional generative AI systems have a problem in that simply generating answers to prompts entered by the user makes it difficult to respond flexibly while taking into account the user's emotions and prompt engineering habits. As a result, the information the user seeks is not accurately provided, often resulting in dissatisfaction. The present invention provides a system that generates more accurate and multifaceted answers by taking into account the user's emotional information and prompt engineering habits.

[0992] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0993] In this invention, the server includes: means for collecting prompts input by multiple users; means for acquiring the collected prompts and emotional information; means for transmitting the acquired prompts and emotional information to the server; means for the server to analyze the prompts and emotional information; means for generating feature vectors from the analyzed prompts and emotional information; means for classifying the generated feature vectors using a clustering algorithm and learning the user's prompt engineering habits and emotional patterns; means for saving and sharing the learned habit model and emotional model in a brain database; means for selecting a habit model and emotional model associated with a new prompt and reconstructing the prompt based on the habits and emotions; means for generating an answer using a generation AI using the reconstructed prompt; and means for displaying the generated answer to the user. This enables the generation of appropriate answers that take into account the user's emotional information and prompt engineering habits.

[0994] "User" refers to a person using the system to enter prompts.

[0995] "Terminal" refers to a device that allows a user to input prompts, obtain emotion information, and transmit it to a server.

[0996] A "prompt" is a question or instruction text that a user enters into a system.

[0997] "Emotional information" refers to data indicating the user's emotional state that is acquired by the device through the camera or microphone.

[0998] "Server" refers to the central computer that analyzes prompts and emotional information, generates feature vectors, clusters, and generates answers.

[0999] "Natural language processing" refers to the technology of extracting keywords and grammatical patterns from prompts.

[1000] "Emotion engine" refers to technology that analyzes emotional information.

[1001] "Feature vector" refers to vector-format data that quantifies the analyzed prompt and emotional information.

[1002] A "clustering algorithm" refers to an algorithm for classifying feature vectors.

[1003] "Prompt engineering" refers to techniques for reconstructing the questions or instructions that users enter into a system.

[1004] The "habit model" refers to a model generated by learning the prompt engineering habits of a user.

[1005] An "emotion model" refers to a model generated by learning the user's emotional patterns.

[1006] The "brain database" refers to a database that stores and manages learned habit models and emotion models.

[1007] "Generative AI" refers to artificial intelligence that generates appropriate answers based on reconstructed prompts.

[1008] This invention relates to a system that analyzes prompts entered by users and their emotional information at the time, and generates appropriate answers using generative AI. This system is configured by combining multiple hardware and software components, and is implemented as follows.

[1009] This system starts when a user inputs a prompt into a terminal. For example, the user might input, "Please tell me what the weather will be like tomorrow." At this time, the device's built-in camera and microphone are used to acquire the user's emotional information (e.g., whether they are anxious or happy). The acquired prompt and emotional information are then sent from the terminal to the server.

[1010] The server then analyzes the received prompt and emotional information. The server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, keywords such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[1011] The analyzed prompt and emotion information are converted into numerical vectors. The server classifies these numerical vectors using a clustering algorithm (e.g., K-means clustering) to learn the user's prompt engineering habits and emotion patterns. This generates a habit model and emotion model for each user. The learned habit model and emotion model are stored in the brain database.

[1012] When a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server. The server selects an appropriate habit model and emotion model, and the prompt is reconstructed based on the habit model and emotion model. For example, the prompt may be adjusted to something like, "Please tell me the specific weather data for tomorrow as soon as possible."

[1013] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[1014] As a specific example, if a user inputs "Tell me how to prepare for a presentation," and the device detects the emotion "I'm nervous," this prompt and emotional information are sent to the server. Based on the emotion model and habit model, the server reconstructs the prompt into something like, "I'm nervous, but can you briefly tell me how to prepare for a presentation?" An example of a prompt sentence to input to the generative AI model is, "The user is nervous and would like more details about how to prepare for a presentation. Please provide specific and calming instructions."

[1015] This system can generate appropriate answers that take into account the user's emotional information and prompt engineering habits, thereby improving user satisfaction.

[1016] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1017] Step 1:

[1018] The user enters the prompt.

[1019] In a specific operation, the user inputs "What will the weather be like tomorrow?" into the terminal. The input obtained is a prompt "What will the weather be like tomorrow?" This prompt is taken into the terminal as text data.

[1020] Step 2:

[1021] The device acquires emotion information.

[1022] Specifically, the device uses its built-in camera and microphone to perform image recognition and voice analysis. Images of the user's facial expressions and voice data are obtained as input. These data are analyzed to obtain emotional information such as "anxious." The emotional information "anxious" is generated as output.

[1023] Step 3:

[1024] The device sends prompts and emotion information to the server.

[1025] Specifically, the terminal uses the communication module to send a prompt, "What will the weather be like tomorrow?" and emotional information, "I'm anxious," in packet form to the server. The input is the prompt and emotional information, and the output is a data packet.

[1026] Step 4:

[1027] The server parses the prompt and emotion information.

[1028] Specifically, the server uses natural language processing technology and an emotion engine. The prompt "Please tell me what the weather will be tomorrow" and the emotion information "I'm anxious" arrive as input to the server. Natural language processing technology extracts keywords such as "tomorrow," "weather," and "tell me," while the emotion engine simultaneously analyzes the emotion "I'm anxious." The output is the keyword and emotion information.

[1029] Step 5:

[1030] The server generates the feature vector.

[1031] Specifically, the system quantifies the keywords and emotional information extracted from the prompt. The inputs are the keywords "tomorrow," "weather," and "tell me about" and the emotional information "anxious." These are converted into numerical vectors, and the output is a prompt vector [0.6, 0.8, 0.5] and an emotional vector [0.1, 0.9].

[1032] Step 6:

[1033] The server classifies the feature vectors using a clustering algorithm.

[1034] Specifically, it uses algorithms such as K-means clustering. The inputs are a prompt vector [0.6, 0.8, 0.5] and an emotion vector [0.1, 0.9]. Clustering is performed to classify the user's prompt engineering habits and emotion patterns. The output is a feature vector classified into clusters.

[1035] Step 7:

[1036] The server stores the habit model and emotion model in a brain database.

[1037] Specifically, the clustering results are written to the database. The input is a feature vector classified into clusters, which is saved in the brain database as a habit model and an emotion model. The output is an updated database.

[1038] Step 8:

[1039] Select relevant habit models and emotion models for the new prompt.

[1040] Specifically, the server searches the brain database and selects the most relevant habit model and emotion model. The inputs are a new prompt, "I want to know tomorrow's weather," and the emotion information, "I'm anxious." The output is the selection of the relevant habit model and emotion model.

[1041] Step 9:

[1042] The server reconstructs the prompt.

[1043] The specific operation is to adjust the prompt based on the selected habit model and emotion model. The prompt and the associated model are obtained as input. The prompt is reconstructed to "Please tell me the specific weather data for tomorrow as soon as possible." and the reconstructed prompt is generated as output.

[1044] Step 10:

[1045] The server generates an answer using a generation AI.

[1046] Specifically, the reconstructed prompt is input into the generation AI, which then generates an answer. The reconstructed prompt is obtained as input. An answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​is generated, and an answer is generated as output.

[1047] Step 11:

[1048] The server sends the generated answer to the terminal, which displays it to the user.

[1049] Specifically, the server sends the generated answer to the terminal, and the terminal displays the answer to the user. The generated answer is obtained as input. The user can see the answer on the screen: "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​The output is the answer displayed to the user.

[1050] (Application example 2)

[1051] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1052] Conventional generative AI systems generate answers based on user input prompts, but because they do not take the user's emotional state into consideration, they often generate inappropriate answers or answers that do not meet the user's expectations. Furthermore, in emergencies, it is necessary to accurately grasp the user's emotions and provide a quick and appropriate response, but current systems have difficulty in doing so.

[1053] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1054] In this invention, the server includes means for collecting prompts input by multiple users, means for acquiring user emotion information using a camera or microphone, and means for analyzing the acquired emotion information to identify the user emotion, thereby enabling the generation of appropriate answers that take user emotion into consideration and quick and appropriate responses in emergencies.

[1055] "User emotion information" is information that indicates the emotional state of the user, obtained from the user's facial expression, tone of voice, and the like.

[1056] A "feature extraction means" is a means capable of extracting meaningful features or patterns from prompts and emotional information.

[1057] "Prompt engineering habits" refer to the characteristics or patterns that a particular user consistently exhibits when completing prompts.

[1058] The "means for identifying emotions" is a means having a function of analyzing acquired emotion information and recognizing the type of emotion the user has.

[1059] The "means for adjusting the prompt" is a means that has the function of reconstructing the original prompt more appropriately, taking into account the user's emotions and prompt engineering habits.

[1060] A "means for generating answers using generative AI" is a means that has the function of generating appropriate answers based on tailored prompts.

[1061] "Problems to be solved by the invention" refers to the specific problems or obstacles that the invention aims to solve, such as the inability of conventional systems to generate appropriate answers because they do not take into account the user's emotional state or prompt engineering habits, and the lack of rapid and appropriate responses in emergencies.

[1062] "Means for solving the problem" refers to the specific technical means proposed to solve the above problem.

[1063] This invention is realized by incorporating user emotional information into a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments for carrying out this invention are described below.

[1064] First, the user inputs a prompt into the device. At the same time, the user's emotional information is acquired using a camera and microphone. For example, while the user inputs the prompt "What will the weather be like tomorrow?", the camera captures the user's facial expression and the microphone captures the user's tone of voice. This emotional information is sent from the user's device to the server.

[1065] The server analyzes the received prompts and emotional information. Specifically, the server uses natural language processing technology (e.g., generative AI models such as GPT-3) to extract keywords and grammatical patterns from the prompts, and simultaneously analyzes the emotional information using an emotional engine. The analyzed data is converted into a numerical vector, and then a clustering algorithm is used to classify and learn the user's prompt engineering habits and emotional patterns. This information is stored in a brain database as a feature vector.

[1066] When a new prompt is input, the server selects the relevant habit model and emotion model from the brain database and adjusts the prompt based on the user's prompt and emotion information. For example, if the user inputs the prompt "I want to know the weather tomorrow," and the emotion "I'm anxious" is detected, the server will reconstruct the prompt as "Please tell me the specific weather data for tomorrow as soon as possible" based on the prompt and emotion information.

[1067] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[1068] The hardware used includes the user's device (smartphone, tablet, etc.), camera, and microphone. The software used includes a generative AI model (e.g., GPT-3) for implementing natural language processing technology and an emotion engine for emotion analysis (e.g., an emotion classification model using TensorFlow or Keras).

[1069] For example, if the user types the following prompt:

[1070] 1. "What's the weather going to be like tomorrow?" (User is anxious)

[1071] 2. "I'm in a hurry and need the latest news headlines right away." (Users are impatient)

[1072] The server applies appropriate habit models and emotion models to these prompts to generate answers that match the user's emotional state, allowing the user to obtain more satisfying information.

[1073] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1074] Step 1:

[1075] The user types a prompt into the terminal.

[1076] Input: User prompt (e.g., "What's the weather like tomorrow?")

[1077] Output: The prompt is typed into the terminal.

[1078] Step 2:

[1079] The device uses a camera and microphone to acquire information about the user's emotions.

[1080] Input: User's facial expression (camera), user's tone of voice (microphone)

[1081] Output: Emotional information (e.g., facial image data, audio data)

[1082] Step 3:

[1083] The device transmits the acquired prompt text and emotion information to the server.

[1084] Input: prompt sentence, emotion information

[1085] Output: The prompt and emotion information are converted into a dedicated format and sent to the server.

[1086] Step 4:

[1087] The server analyzes the prompt sentence using natural language processing technology to extract keywords and grammatical patterns.

[1088] Input: Prompt sentence (e.g. "What will the weather be like tomorrow?")

[1089] Output: Extracted features (keywords: "tomorrow", "weather", "tell me")

[1090] Step 5:

[1091] The server uses an emotion engine to analyze the transmitted emotion information.

[1092] Input: Emotion information (face video data, audio data)

[1093] Output: Parsed emotion (e.g. "anxious")

[1094] Step 6:

[1095] The server converts the features and sentiment information of the parsed prompt into a numerical vector.

[1096] Input: extracted features, parsed emotions

[1097] Output: Feature vector (e.g. "tomorrow", "weather", "tell me", "I'm anxious" → numeric vector)

[1098] Step 7:

[1099] The server clusters the feature vectors and learns the user's prompt engineering habits and emotional patterns.

[1100] Input: feature vector

[1101] Output: Learned habit model and emotion model

[1102] Step 8:

[1103] When a new prompt is entered, the server selects the relevant habit model and emotion model from the brain database.

[1104] Input: New prompt, emotion information

[1105] Output: Appropriate habit model and emotion model

[1106] Step 9:

[1107] The server adjusts the prompt sentence based on the selected habit model and emotion model.

[1108] Input: New prompt, habit model, emotion model

[1109] Output: The adjusted prompt (e.g., "Please give me the specific weather data for tomorrow as soon as possible.")

[1110] Step 10:

[1111] The server passes the adjusted prompt to the generative AI model, which generates the optimal answer.

[1112] Input: Adjusted prompt text

[1113] Output: Generated answer (e.g. "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[1114] Step 11:

[1115] The server sends the generated response to the terminal.

[1116] Input: Generated answer

[1117] Output: The generated answer sent by the server

[1118] Step 12:

[1119] The terminal displays the generated answer to the user.

[1120] Input: Generated answer

[1121] Output: The answer shown to the user (e.g., "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[1122] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1123] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1124] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1125] [Fourth embodiment]

[1126] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1127] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1128] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1129] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1130] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1131] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1132] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1133] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1134] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1135] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1136] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1137] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1138] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1139] This invention is a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[1140] First, a user inputs a prompt into the terminal. For example, if the user inputs the prompt "What will the weather be like tomorrow?", the prompt is recorded by the terminal. Then, the terminal sends the recorded prompt to the server.

[1141] The server then analyzes the received prompt. Specifically, the server uses natural language processing technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[1142] Next, the server learns the user's prompt engineering habits based on the extracted features. For example, if a user has a strong tendency to request specific data, this habit is learned. The learned habits are stored in the brain database as a numerical model.

[1143] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know the weather tomorrow," the prompt is sent to the server, which selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[1144] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[1145] This system can provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, it is possible to improve the overall accuracy of answers. This invention is particularly useful in generative AI systems that require flexible responses to users with different questioning styles.

[1146] The processing flow will be explained below.

[1147] Step 1:

[1148] The user inputs a prompt into the terminal, for example, "What is the weather like tomorrow?"

[1149] Step 2:

[1150] Records the prompts entered at the terminal and saves them in a log file or database.

[1151] Step 3:

[1152] The device sends the recorded prompt to the server, for example, using an HTTP request.

[1153] Step 4:

[1154] The server analyzes the received prompt, specifically extracting keywords and grammatical patterns from the prompt using natural language processing techniques.

[1155] Step 5:

[1156] The server converts the extracted features into feature vectors. For example, words such as "tomorrow," "weather," and "tell me" and their contexts are converted into numerical vectors.

[1157] Step 6:

[1158] The server classifies the feature vectors using a clustering algorithm and learns the user's prompt engineering habits.

[1159] Step 7:

[1160] The server stores the learned habit model in a brain database, which is then shared so that other users can use it.

[1161] Step 8:

[1162] The terminal receives a new prompt. For example, user B inputs the prompt "I want to know the weather tomorrow."

[1163] Step 9:

[1164] The terminal sends a new prompt to the server.

[1165] Step 10:

[1166] The server selects relevant habit models from the brain database, and the most appropriate habit model is chosen and applied to the new prompt.

[1167] Step 11:

[1168] The server reconstructs a new prompt based on the habit model, e.g., "Please tell me the specific weather data for tomorrow."

[1169] Step 12:

[1170] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[1171] Step 13:

[1172] The server sends the generated response to the terminal.

[1173] Step 14:

[1174] The device displays the generated answer to the user. For example, the answer "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed.

[1175] Example 1

[1176] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1177] Conventional generative AI systems have difficulty taking into account the different prompting habits of each user, which can result in inappropriate answers being generated. Furthermore, there is a lack of efficient means for learning user habits and reconstructing prompts based on them. Furthermore, there is a need to effectively utilize natural language processing technology in the process of collecting and analyzing user prompts. By resolving these issues, it will be possible to provide more accurate answers for each user.

[1178] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1179] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting a habit model associated with a new prompt and reconstructing the prompt based on the habit, means for generating an answer by a generation AI using the reconstructed prompt, means for displaying the generated answer to the user, means for transmitting the prompt from the user's terminal to the server, means for the server to analyze the received prompt using natural language processing technology, and means for the server to select a habit model from a brain database. This enables the generation of appropriate answers that take into account each user's prompt habits.

[1180] A "prompt" is a sentence containing a question or instruction that a user enters into a generative AI system.

[1181] A "collection means" is a method or device for capturing prompts entered by multiple users into the system.

[1182] A "means for extracting features" is a method or device for identifying and extracting important keywords and grammatical patterns from collected prompts.

[1183] "Prompt engineering habits" refer to the particular patterns or tendencies that users have when entering prompts.

[1184] "Means for learning" refers to a method or device for recognizing and storing a user's prompt engineering habits based on extracted features using a machine learning algorithm or the like.

[1185] "Means for saving and sharing" refers to a method or device for saving the learned habit model in a database or the like and sharing it with other systems or users as needed.

[1186] A "reconstruction means" is a method or device for selecting a relevant habit model for a newly input prompt and improving or adjusting the prompt based on that model.

[1187] A "means for generating" is a method or device for generating responses to reconstructed prompts using generative AI.

[1188] The "displaying means" is a method or device for visually or audibly presenting the generated answer to the user.

[1189] A "means for sending" is a method or device for sending a prompt from a user's terminal to a server.

[1190] The "means for analyzing" refers to a method or device for using natural language processing techniques to examine in detail the prompt received by the server and understand its meaning and characteristics.

[1191] The "brain database" is a database that stores learned prompt engineering habit models of users and can be searched and retrieved as needed.

[1192] This invention is a generative AI system that provides users with multifaceted and accurate answers. The system begins operation when the user inputs a prompt sentence into a terminal. Specifically, when the user inputs the prompt sentence, "Please tell me what the weather will be like tomorrow," the prompt sentence is recorded by the terminal. The terminal then transmits the recorded prompt sentence to a server.

[1193] To analyze the prompt received by the server, natural language processing technology (such as NLTK or SpaCy) is used. The server uses this technology to extract important keywords and grammatical patterns from the prompt. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the relationships between them are analyzed.

[1194] Next, the server learns the user's prompt engineering habits based on the extracted features. This learning process uses machine learning algorithms, and the learned habits are stored in the brain database as a numerical model. For example, if a user has a habit of requesting specific data, this tendency will be stored in the brain database.

[1195] When a new prompt is entered, the server selects the most relevant habit model from the brain database. For example, if a user enters "I want to know tomorrow's weather," the prompt is sent to the server once, and the server selects an appropriate habit model. Based on that habit model, the prompt is reconstructed and adjusted to, for example, "Please tell me the specific weather data for tomorrow."

[1196] The reconstructed prompt is passed to a generative AI model, which generates an appropriate response. Examples of generative AI models include OpenAI's GPT-4. This generative AI model has the ability to generate the most appropriate response based on the reconstructed prompt.

[1197] The generated answer is sent from the server to the terminal and displayed to the user through the terminal. For example, an answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees" is displayed. This system allows users to receive more accurate and multifaceted answers based on their own prompt engineering habits.

[1198] This invention allows a generative AI system to provide more appropriate and multifaceted answers by taking into account the user's prompt engineering habits. Furthermore, by sharing each user's habit model, the overall accuracy of answers is improved. This system is particularly useful in generative AI model systems that require flexible responses to users with different questioning styles.

[1199] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1200] Step 1:

[1201] The user inputs a prompt sentence using a terminal. Specifically, the user inputs a prompt sentence such as "Please tell me what the weather will be tomorrow." using a keyboard or voice input. This prompt sentence becomes the input data for the process.

[1202] Step 2:

[1203] The terminal records the input prompt and sends the data to the server. The specific operation of the terminal is to save the prompt in memory and send the data to the server using an HTTP request. The input here is the prompt, and the output is the prompt sent to the server.

[1204] Step 3:

[1205] The server parses the received prompt. Specifically, it uses natural language processing techniques (such as NLTK or SpaCy) to tokenize the prompt and extract keywords and grammatical patterns. The input to this process is the prompt, and the output is the extracted keywords and grammatical patterns.

[1206] Step 4:

[1207] The server learns the user's prompt engineering habits based on the extracted keywords. The server uses a machine learning algorithm (such as a decision tree or random forest) to model the prompt engineering habits. The input is the extracted keywords, and the output is a model that reflects the user's habits. This model is stored in a brain database.

[1208] Step 5:

[1209] When a new prompt sentence is input, the server selects a highly relevant habit model from the brain database. For example, if a new prompt such as "I want to know the weather tomorrow," is input, the server receives the prompt and selects an appropriate habit model from the brain database. The input is the new prompt sentence, and the output is the selected habit model.

[1210] Step 6:

[1211] The server reconstructs the prompt sentence based on the selected habit model. As a specific operation of reconstruction, the server changes the prompt sentence "I want to know the weather tomorrow" to "Please tell me the specific weather data for tomorrow." The input is the selected habit model and the new prompt sentence, and the output is the reconstructed prompt sentence.

[1212] Step 7:

[1213] The reconstructed prompt is passed to a generative AI model, which generates an appropriate answer. GPT-4 is used as the generative AI model, and the answer is generated through an API request. The input is the reconstructed prompt, and the output is the generated answer.

[1214] Step 8:

[1215] The generated answer is sent from the server to the terminal. As a server operation, the generated answer is sent to the terminal using an HTTP response. The input is the generated answer, and the output is the answer sent to the terminal.

[1216] Step 9:

[1217] The user displays the generated answer through the terminal. The specific operation of the terminal is to display the received answer on the screen. The input is the answer received from the server, and the output is the answer displayed to the user. For example, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees."

[1218] (Application example 1)

[1219] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1220] In today's virtual stores, it is extremely important to quickly and accurately provide users with the product information and recommendations they need. However, typical generative AI systems often lack the ability to accurately understand the intent and preferences of users' questions, resulting in inability to provide the information they seek. Furthermore, when using voice prompts, the process of accurately converting the speech into text and adjusting it to match the user's intent is complex, and the entire process of displaying information on a visual device must be carried out efficiently and accurately. Therefore, technology is needed that can learn users' prompt engineering habits and provide answers that adapt to them.

[1221] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1222] In this invention, the server includes means for collecting prompts input by multiple users, means for extracting features from the collected prompts, means for learning users' prompt engineering habits based on the extracted features, means for saving and sharing the learned habit model, means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit, means for generating an answer by a generation AI using the adjusted prompt, means for displaying the generated answer to the user, means for allowing the user to obtain information using voice prompts, speech recognition means for converting the voice prompts into text, and means for displaying the generated answer on a visual device, thereby enabling optimal information provision taking into account the user's prompt engineering habits.

[1223] Definitions of important words

[1224] A "user" is an individual or organization that uses the system.

[1225] A "prompt" is a question or instruction that a user enters into a system.

[1226] "Collection means" refers to the software and hardware used to capture and record prompts.

[1227] "Feature extraction methods" refer to techniques for identifying and extracting important elements or keywords from the prompt.

[1228] "Prompt engineering habits" refer to patterns or tendencies that users have when entering prompts.

[1229] "Means of learning" refers to the technology used to analyze prompt engineering habits based on collected feature information and build a model.

[1230] "Means for saving and sharing" refers to saving the trained model in a database or the like and sharing it with other systems or users as needed.

[1231] "Adjustment" refers to the process of optimizing prompts based on user habits, allowing the generative AI to generate more appropriate answers.

[1232] "Generative AI" refers to a system that uses artificial intelligence technology to automatically generate answers based on input data.

[1233] "Displaying means" refers to a device or software for notifying the user of the generated answer.

[1234] "Voice prompt" refers to a question or instruction that the user speaks.

[1235] "Speech recognition means" refers to technology that converts voice prompts into text.

[1236] "Visual device" refers to an electronic device that visually presents information to a user.

[1237] MODE FOR CARRYING OUT THE INVENTION

[1238] In order to implement the present invention, a system is used in which a user wears smart glasses and obtains information using voice prompts. Specific embodiments are described in detail below.

[1239] System Configuration

[1240] 1. Hardware:

[1241] Smart glasses: Devices with built-in voice recognition microphones and displays.

[1242] Server: A server with high-performance computing power that is responsible for prompt analysis and running AI models.

[1243] 2. Software:

[1244] Speech Recognition Library: Use speech_recognition to convert user voice prompts into text.

[1245] Communication library: requests is used to send prompts to the server and retrieve results.

[1246] Natural language processing technology: We use spaCy and Transformers to analyze prompts and extract features.

[1247] Generative AI models: Models that generate answers based on prompts, such as GPT and BERT.

[1248] Data Processing and Computation

[1249] When a user enters a voice prompt, the voice data is converted to text by a speech recognition library, and the converted text is sent to the server, which then performs the following steps:

[1250] 1. Prompt analysis and feature extraction:

[1251] The server uses natural language processing techniques to extract important keywords and grammatical patterns from the text prompts.

[1252] 2. Prompt Engineering Habit Learning:

[1253] The server learns the user's prompt engineering habits based on the extracted features, and stores and shares the habit model in a brain database.

[1254] 3. Prompt adjustment and AI model execution:

[1255] For each new prompt, it selects a relevant habit model and adjusts the prompt accordingly, which is then passed to a generative AI model to generate the optimal answer.

[1256] 4. Show Answer:

[1257] The generated answers are sent to the smart glasses and displayed on the visual device.

[1258] Specific examples

[1259] A user asks a question through the smart glasses. For example, if a user inputs a voice prompt such as "What new sneakers do you recommend?", the prompt is converted into text and sent to the server. The server extracts keywords such as "new," "sneakers," and "recommended," and reconstructs the prompt based on the user's habits to "Please recommend some new sneakers with a flashy design." The generative AI model generates an appropriate answer, which is sent from the server to the smart glasses. The user's field of vision is displayed as follows: "The latest sneakers with a flashy design are the Nike Air Max, which comes in a wide range of color options."

[1260] Prompt Sentence Examples

[1261] "What new sneakers do you recommend?"

[1262] "Can you recommend some new sneakers with flashy designs?"

[1263] This system makes it possible to provide optimal information taking into account the user's prompt engineering habits.

[1264] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1265] Specific explanation of program processing

[1266] Step 1:

[1267] The user inputs a voice prompt into the smart glasses.

[1268] Input: A voice prompt from the user (e.g., "What new sneakers would you recommend?").

[1269] Output: The speech data received by the speech recognizer.

[1270] Specific operation: When the user speaks into the device, the smart glasses' built-in microphone captures the voice data.

[1271] Step 2:

[1272] The device converts the voice prompts into text.

[1273] Input: Audio data from smart glasses.

[1274] Output: A text prompt (e.g., "What new sneakers do you recommend?").

[1275] Specific operation: Uses the speech_recognition library to convert audio data into text data.

[1276] Step 3:

[1277] The terminal sends a text prompt to the server.

[1278] Input: A text prompt.

[1279] Output: The text prompt sent to the server.

[1280] Specific operation: Uses the requests library to send text data to the server via an HTTP POST request.

[1281] Step 4:

[1282] The server parses the prompt and extracts features.

[1283] Input: The text prompt sent from the terminal.

[1284] Output: Extracted feature data (e.g., keywords "new arrivals," "sneakers," "recommended").

[1285] Specific operation: Using the spaCy library and Transformers, it analyzes and extracts important keywords and grammatical patterns from text data.

[1286] Step 5:

[1287] The server learns the user's prompt engineering habits based on the extracted features.

[1288] Input: Extracted feature data.

[1289] Output: Habit model (e.g., patterns of keywords and phrases frequently used by the user).

[1290] Specific operation: The analyzed feature data is input into a machine learning algorithm to learn the user's prompt engineering habits.

[1291] Step 6:

[1292] The server stores and shares the learned habit models.

[1293] Input: The newly trained habit model.

[1294] Output: Habit model stored in the brain database.

[1295] What it does: Saves the habit model in a database so that it can be shared with other systems and users in the future.

[1296] Step 7:

[1297] The server selects the relevant habit model for the new prompt and reconstructs the prompt.

[1298] Input: A new prompt and a saved habit model.

[1299] Output: A tailored prompt (e.g., "Can you recommend some new sneakers with a flashy design?").

[1300] What it does: Refers to the saved habit model and adjusts new prompts appropriately.

[1301] Step 8:

[1302] The server uses the tailored prompts to generate answers using the generative AI.

[1303] Input: Tailored prompt.

[1304] Output: The generated answer (e.g., "The latest, flashiest sneaker design is the Nike Air Max, which comes in a wide variety of colors.").

[1305] What it does: Feeds the tailored prompts into an AI model to generate the best answer.

[1306] Step 9:

[1307] The server sends the generated response to the terminal.

[1308] Input: The generated answer.

[1309] Output: The answer sent to the terminal.

[1310] Specific operation: The generated answer is sent to the terminal as an HTTP response.

[1311] Step 10:

[1312] The terminal displays the generated answer on the visual device of the smart glasses.

[1313] Input: The generated answer received from the server.

[1314] Output: The answer that is displayed to the user.

[1315] Specific behavior: Displays text on the smart glasses display to provide information to the user.

[1316] Through the above steps, the user can efficiently obtain optimal product information in the virtual store using voice prompts.

[1317] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1318] This invention adds an emotion engine to a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments are described below.

[1319] First, the user inputs a prompt into the device, and the device reads the emotion at that time. For example, while the user inputs the prompt "Please tell me what the weather will be like tomorrow," emotional information (e.g., joy, anger, sadness, etc.) is acquired through the camera and microphone. This emotional information is then sent by the device to the server.

[1320] The server analyzes the received prompt and emotional information. Specifically, the server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, words such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[1321] The server converts the extracted prompt features and emotion features into feature vectors, for example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[1322] The server then classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns. For example, if a user has a strong tendency to seek specific data and is feeling "impatient," the server learns the habits and emotional patterns. The learned habits and emotional patterns are stored in a brain database as a numerical model.

[1323] Next, when a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server, which then selects an appropriate habit model and emotion model. The prompt is reconstructed based on the habit model and emotion model, and adjusted to a form such as, "Please tell me the specific weather data for tomorrow as soon as possible."

[1324] The adjusted prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[1325] This system can provide more appropriate, multifaceted, and emotionally relevant answers by taking into account the user's prompt engineering habits and emotions. Furthermore, by sharing each user's habit model and emotion model, it is possible to improve overall answer accuracy and user satisfaction. This invention is particularly useful in generative AI systems that require flexible responses to users with different question styles and various emotions.

[1326] The processing flow will be explained below.

[1327] Step 1:

[1328] A user inputs a prompt into a terminal. For example, the user inputs the prompt "What will the weather be like tomorrow?"

[1329] Step 2:

[1330] The device uses a camera and microphone to acquire information about the user's emotions. For example, it can recognize the emotion of "anxiety" by analyzing the user's facial expressions and tone of voice.

[1331] Step 3:

[1332] The device records the prompts and emotional information entered, and the recorded data is stored in a log file or database.

[1333] Step 4:

[1334] The device sends the recorded prompts and emotion information to the server using HTTP requests or similar.

[1335] Step 5:

[1336] The server analyzes the received prompt and emotional information. Specifically, it uses natural language processing technology to extract keywords and grammatical patterns from the prompt, and then analyzes the emotional information using an emotional engine.

[1337] Step 6:

[1338] The server converts the extracted prompt features and emotion features into feature vectors. For example, the words "tomorrow," "weather," and "tell me" and the emotion "impatient" are converted into numeric vectors.

[1339] Step 7:

[1340] The server classifies the feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns.

[1341] Step 8:

[1342] The server stores the learned habit models and emotion models in a brain database, which is then shared for other users to use.

[1343] Step 9:

[1344] The terminal receives a new prompt. For example, user B inputs a prompt such as "I want to know the weather tomorrow." At the same time, the emotion "I'm anxious" is acquired.

[1345] Step 10:

[1346] The device sends the new prompt and emotion information to the server.

[1347] Step 11:

[1348] The server selects relevant habit models and emotion models from the brain database, and the most appropriate habit model and emotion model are selected and applied to the new prompt.

[1349] Step 12:

[1350] The server reconstructs a new prompt based on the habit model and emotion model, for example, "Please tell me the specific weather data for tomorrow as soon as possible."

[1351] Step 13:

[1352] The server passes the tailored prompts to the generation AI, which generates answers based on them.

[1353] Step 14:

[1354] The server sends the generated response to the terminal.

[1355] Step 15:

[1356] The device displays the generated answer to the user. For example, the answer may read, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[1357] Example 2

[1358] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1359] Conventional generative AI systems have a problem in that simply generating answers to prompts entered by the user makes it difficult to respond flexibly while taking into account the user's emotions and prompt engineering habits. As a result, the information the user seeks is not accurately provided, often resulting in dissatisfaction. The present invention provides a system that generates more accurate and multifaceted answers by taking into account the user's emotional information and prompt engineering habits.

[1360] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1361] In this invention, the server includes: means for collecting prompts input by multiple users; means for acquiring the collected prompts and emotional information; means for transmitting the acquired prompts and emotional information to the server; means for the server to analyze the prompts and emotional information; means for generating feature vectors from the analyzed prompts and emotional information; means for classifying the generated feature vectors using a clustering algorithm and learning the user's prompt engineering habits and emotional patterns; means for saving and sharing the learned habit model and emotional model in a brain database; means for selecting a habit model and emotional model associated with a new prompt and reconstructing the prompt based on the habits and emotions; means for generating an answer using a generation AI using the reconstructed prompt; and means for displaying the generated answer to the user. This enables the generation of appropriate answers that take into account the user's emotional information and prompt engineering habits.

[1362] "User" refers to a person using the system to enter prompts.

[1363] "Terminal" refers to a device that allows a user to input prompts, obtain emotion information, and transmit it to a server.

[1364] A "prompt" is a question or instruction text that a user enters into a system.

[1365] "Emotional information" refers to data indicating the user's emotional state that is acquired by the device through the camera or microphone.

[1366] "Server" refers to the central computer that analyzes prompts and emotional information, generates feature vectors, clusters, and generates answers.

[1367] "Natural language processing" refers to the technology of extracting keywords and grammatical patterns from prompts.

[1368] "Emotion engine" refers to technology that analyzes emotional information.

[1369] "Feature vector" refers to vector-format data that quantifies the analyzed prompt and emotional information.

[1370] A "clustering algorithm" refers to an algorithm for classifying feature vectors.

[1371] "Prompt engineering" refers to techniques for reconstructing the questions or instructions that users enter into a system.

[1372] The "habit model" refers to a model generated by learning the prompt engineering habits of a user.

[1373] An "emotion model" refers to a model generated by learning the user's emotional patterns.

[1374] The "brain database" refers to a database that stores and manages learned habit models and emotion models.

[1375] "Generative AI" refers to artificial intelligence that generates appropriate answers based on reconstructed prompts.

[1376] This invention relates to a system that analyzes prompts entered by users and their emotional information at the time, and generates appropriate answers using generative AI. This system is configured by combining multiple hardware and software components, and is implemented as follows.

[1377] This system starts when a user inputs a prompt into a terminal. For example, the user might input, "Please tell me what the weather will be like tomorrow." At this time, the device's built-in camera and microphone are used to acquire the user's emotional information (e.g., whether they are anxious or happy). The acquired prompt and emotional information are then sent from the terminal to the server.

[1378] The server then analyzes the received prompt and emotional information. The server uses natural language processing technology to extract keywords and grammatical patterns from the prompt, while simultaneously analyzing the emotional information using an emotional engine. For example, keywords such as "tomorrow," "weather," and "tell me" are extracted, and the emotion "the user is anxious" is recognized.

[1379] The analyzed prompt and emotion information are converted into numerical vectors. The server classifies these numerical vectors using a clustering algorithm (e.g., K-means clustering) to learn the user's prompt engineering habits and emotion patterns. This generates a habit model and emotion model for each user. The learned habit model and emotion model are stored in the brain database.

[1380] When a new prompt is entered, the server selects the most relevant habit model and emotion model from the brain database. For example, if a user enters "I want to know tomorrow's weather," accompanied by the emotion of "I'm anxious," the prompt and emotion information are sent to the server. The server selects an appropriate habit model and emotion model, and the prompt is reconstructed based on the habit model and emotion model. For example, the prompt may be adjusted to something like, "Please tell me the specific weather data for tomorrow as soon as possible."

[1381] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[1382] As a specific example, if a user inputs "Tell me how to prepare for a presentation," and the device detects the emotion "I'm nervous," this prompt and emotional information are sent to the server. Based on the emotion model and habit model, the server reconstructs the prompt into something like, "I'm nervous, but can you briefly tell me how to prepare for a presentation?" An example of a prompt sentence to input to the generative AI model is, "The user is nervous and would like more details about how to prepare for a presentation. Please provide specific and calming instructions."

[1383] This system can generate appropriate answers that take into account the user's emotional information and prompt engineering habits, thereby improving user satisfaction.

[1384] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1385] Step 1:

[1386] The user enters the prompt.

[1387] In a specific operation, the user inputs "What will the weather be like tomorrow?" into the terminal. The input obtained is a prompt "What will the weather be like tomorrow?" This prompt is taken into the terminal as text data.

[1388] Step 2:

[1389] The device acquires emotion information.

[1390] Specifically, the device uses its built-in camera and microphone to perform image recognition and voice analysis. Images of the user's facial expressions and voice data are obtained as input. These data are analyzed to obtain emotional information such as "anxious." The emotional information "anxious" is generated as output.

[1391] Step 3:

[1392] The device sends prompts and emotion information to the server.

[1393] Specifically, the terminal uses the communication module to send a prompt, "What will the weather be like tomorrow?" and emotional information, "I'm anxious," in packet form to the server. The input is the prompt and emotional information, and the output is a data packet.

[1394] Step 4:

[1395] The server parses the prompt and emotion information.

[1396] Specifically, the server uses natural language processing technology and an emotion engine. The prompt "Please tell me what the weather will be tomorrow" and the emotion information "I'm anxious" arrive as input to the server. Natural language processing technology extracts keywords such as "tomorrow," "weather," and "tell me," while the emotion engine simultaneously analyzes the emotion "I'm anxious." The output is the keyword and emotion information.

[1397] Step 5:

[1398] The server generates the feature vector.

[1399] Specifically, the system quantifies the keywords and emotional information extracted from the prompt. The inputs are the keywords "tomorrow," "weather," and "tell me about" and the emotional information "anxious." These are converted into numerical vectors, and the output is a prompt vector [0.6, 0.8, 0.5] and an emotional vector [0.1, 0.9].

[1400] Step 6:

[1401] The server classifies the feature vectors using a clustering algorithm.

[1402] Specifically, it uses algorithms such as K-means clustering. The inputs are a prompt vector [0.6, 0.8, 0.5] and an emotion vector [0.1, 0.9]. Clustering is performed to classify the user's prompt engineering habits and emotion patterns. The output is a feature vector classified into clusters.

[1403] Step 7:

[1404] The server stores the habit model and emotion model in a brain database.

[1405] Specifically, the clustering results are written to the database. The input is a feature vector classified into clusters, which is saved in the brain database as a habit model and an emotion model. The output is an updated database.

[1406] Step 8:

[1407] Select relevant habit models and emotion models for the new prompt.

[1408] Specifically, the server searches the brain database and selects the most relevant habit model and emotion model. The inputs are a new prompt, "I want to know tomorrow's weather," and the emotion information, "I'm anxious." The output is the selection of the relevant habit model and emotion model.

[1409] Step 9:

[1410] The server reconstructs the prompt.

[1411] The specific operation is to adjust the prompt based on the selected habit model and emotion model. The prompt and the associated model are obtained as input. The prompt is reconstructed to "Please tell me the specific weather data for tomorrow as soon as possible." and the reconstructed prompt is generated as output.

[1412] Step 10:

[1413] The server generates an answer using a generation AI.

[1414] Specifically, the reconstructed prompt is input into the generation AI, which then generates an answer. The reconstructed prompt is obtained as input. An answer such as "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​is generated, and an answer is generated as output.

[1415] Step 11:

[1416] The server sends the generated answer to the terminal, which displays it to the user.

[1417] Specifically, the server sends the generated answer to the terminal, and the terminal displays the answer to the user. The generated answer is obtained as input. The user can see the answer on the screen: "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush." ​​The output is the answer displayed to the user.

[1418] (Application example 2)

[1419] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1420] Conventional generative AI systems generate answers based on user input prompts, but because they do not take the user's emotional state into consideration, they often generate inappropriate answers or answers that do not meet the user's expectations. Furthermore, in emergencies, it is necessary to accurately grasp the user's emotions and provide a quick and appropriate response, but current systems have difficulty in doing so.

[1421] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1422] In this invention, the server includes means for collecting prompts input by multiple users, means for acquiring user emotion information using a camera or microphone, and means for analyzing the acquired emotion information to identify the user emotion, thereby enabling the generation of appropriate answers that take user emotion into consideration and quick and appropriate responses in emergencies.

[1423] "User emotion information" is information that indicates the emotional state of the user, obtained from the user's facial expression, tone of voice, and the like.

[1424] A "feature extraction means" is a means capable of extracting meaningful features or patterns from prompts and emotional information.

[1425] "Prompt engineering habits" refer to the characteristics or patterns that a particular user consistently exhibits when completing prompts.

[1426] The "means for identifying emotions" is a means having a function of analyzing acquired emotion information and recognizing the type of emotion the user has.

[1427] The "means for adjusting the prompt" is a means that has the function of reconstructing the original prompt more appropriately, taking into account the user's emotions and prompt engineering habits.

[1428] A "means for generating answers using generative AI" is a means that has the function of generating appropriate answers based on tailored prompts.

[1429] "Problems to be solved by the invention" refers to the specific problems or obstacles that the invention aims to solve, such as the inability of conventional systems to generate appropriate answers because they do not take into account the user's emotional state or prompt engineering habits, and the lack of rapid and appropriate responses in emergencies.

[1430] "Means for solving the problem" refers to the specific technical means proposed to solve the above problem.

[1431] This invention is realized by incorporating user emotional information into a generative AI system that provides users with multifaceted and accurate answers. Specific embodiments for carrying out this invention are described below.

[1432] First, the user inputs a prompt into the device. At the same time, the user's emotional information is acquired using a camera and microphone. For example, while the user inputs the prompt "What will the weather be like tomorrow?", the camera captures the user's facial expression and the microphone captures the user's tone of voice. This emotional information is sent from the user's device to the server.

[1433] The server analyzes the received prompts and emotional information. Specifically, the server uses natural language processing technology (e.g., generative AI models such as GPT-3) to extract keywords and grammatical patterns from the prompts, and simultaneously analyzes the emotional information using an emotional engine. The analyzed data is converted into a numerical vector, and then a clustering algorithm is used to classify and learn the user's prompt engineering habits and emotional patterns. This information is stored in a brain database as a feature vector.

[1434] When a new prompt is input, the server selects the relevant habit model and emotion model from the brain database and adjusts the prompt based on the user's prompt and emotion information. For example, if the user inputs the prompt "I want to know the weather tomorrow," and the emotion "I'm anxious" is detected, the server will reconstruct the prompt as "Please tell me the specific weather data for tomorrow as soon as possible" based on the prompt and emotion information.

[1435] The reconstructed prompt is passed to the generation AI, which generates an appropriate answer. The generated answer is sent from the server to the device and displayed to the user. For example, the answer displayed might be, "Tomorrow's weather will be sunny, with a maximum temperature of 25 degrees and a minimum temperature of 18 degrees. There's no need to rush."

[1436] The hardware used includes the user's device (smartphone, tablet, etc.), camera, and microphone. The software used includes a generative AI model (e.g., GPT-3) for implementing natural language processing technology and an emotion engine for emotion analysis (e.g., an emotion classification model using TensorFlow or Keras).

[1437] For example, if the user types the following prompt:

[1438] 1. "What's the weather going to be like tomorrow?" (User is anxious)

[1439] 2. "I'm in a hurry and need the latest news headlines right away." (Users are impatient)

[1440] The server applies appropriate habit models and emotion models to these prompts to generate answers that match the user's emotional state, allowing the user to obtain more satisfying information.

[1441] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1442] Step 1:

[1443] The user types a prompt into the terminal.

[1444] Input: User prompt (e.g., "What's the weather like tomorrow?")

[1445] Output: The prompt is typed into the terminal.

[1446] Step 2:

[1447] The device uses a camera and microphone to acquire information about the user's emotions.

[1448] Input: User's facial expression (camera), user's tone of voice (microphone)

[1449] Output: Emotional information (e.g., facial image data, audio data)

[1450] Step 3:

[1451] The device transmits the acquired prompt text and emotion information to the server.

[1452] Input: prompt sentence, emotion information

[1453] Output: The prompt and emotion information are converted into a dedicated format and sent to the server.

[1454] Step 4:

[1455] The server analyzes the prompt sentence using natural language processing technology to extract keywords and grammatical patterns.

[1456] Input: Prompt sentence (e.g. "What will the weather be like tomorrow?")

[1457] Output: Extracted features (keywords: "tomorrow", "weather", "tell me")

[1458] Step 5:

[1459] The server uses an emotion engine to analyze the transmitted emotion information.

[1460] Input: Emotion information (face video data, audio data)

[1461] Output: Parsed emotion (e.g. "anxious")

[1462] Step 6:

[1463] The server converts the features and sentiment information of the parsed prompt into a numerical vector.

[1464] Input: extracted features, parsed emotions

[1465] Output: Feature vector (e.g. "tomorrow", "weather", "tell me", "I'm anxious" → numeric vector)

[1466] Step 7:

[1467] The server clusters the feature vectors and learns the user's prompt engineering habits and emotional patterns.

[1468] Input: feature vector

[1469] Output: Learned habit model and emotion model

[1470] Step 8:

[1471] When a new prompt is entered, the server selects the relevant habit model and emotion model from the brain database.

[1472] Input: New prompt, emotion information

[1473] Output: Appropriate habit model and emotion model

[1474] Step 9:

[1475] The server adjusts the prompt sentence based on the selected habit model and emotion model.

[1476] Input: New prompt, habit model, emotion model

[1477] Output: The adjusted prompt (e.g., "Please give me the specific weather data for tomorrow as soon as possible.")

[1478] Step 10:

[1479] The server passes the adjusted prompt to the generative AI model, which generates the optimal answer.

[1480] Input: Adjusted prompt text

[1481] Output: Generated answer (e.g. "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[1482] Step 11:

[1483] The server sends the generated response to the terminal.

[1484] Input: Generated answer

[1485] Output: The generated answer sent by the server

[1486] Step 12:

[1487] The terminal displays the generated answer to the user.

[1488] Input: Generated answer

[1489] Output: The answer shown to the user (e.g., "Tomorrow's weather will be sunny, with a high of 25 degrees and a low of 18 degrees. There's no need to rush.")

[1490] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1491] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1492] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1493] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1494] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1495] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1496] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1497] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1498] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1499] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1500] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1501] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1502] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1503] 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.

[1504] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1505] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1506] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1507] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1508] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1509] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1510] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1511] The following is further disclosed regarding the above embodiment.

[1512] (Claim 1)

[1513] means for collecting prompts entered by a plurality of users;

[1514] a means for extracting features from the collected prompts;

[1515] A means for learning a user's prompt engineering habits based on the extracted features;

[1516] A means to save and share the learned habit model,

[1517] means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit model;

[1518] a means for generating generative AI responses utilizing the tailored prompts;

[1519] means for displaying the generated answers to the user;

[1520] A system including:

[1521] (Claim 2)

[1522] The system of claim 1, further comprising means for storing and sharing the learned habit model using the brain database.

[1523] (Claim 3)

[1524] 10. The system of claim 1, further comprising: means for extracting features from the prompt using natural language processing techniques.

[1525] "Example 1"

[1526] (Claim 1)

[1527] means for collecting prompts entered by a plurality of users;

[1528] a means for extracting features from the collected prompts;

[1529] A means for learning a user's prompt engineering habits based on the extracted features;

[1530] A means to save and share the learned habit model,

[1531] means for selecting an associated habit model for a new prompt and reconstructing the prompt based on the habit model;

[1532] a means for generating a generative AI response utilizing the reconstructed prompt; and

[1533] means for displaying the generated answers to the user;

[1534] means for sending a prompt from the user's terminal to the server;

[1535] means for analyzing the prompt received by the server using natural language processing techniques;

[1536] A means for the server to select a habit model from the brain database;

[1537] A system including:

[1538] (Claim 2)

[1539] The system of claim 1, further comprising means for storing and sharing the learned habit model using the brain database.

[1540] (Claim 3)

[1541] 10. The system of claim 1, further comprising: means for extracting features from the prompt using natural language processing techniques.

[1542] "Application Example 1"

[1543] (Claim 1)

[1544] means for collecting prompts entered by a plurality of users;

[1545] a means for extracting features from the collected prompts;

[1546] A means for learning a user's prompt engineering habits based on the extracted features;

[1547] A means to save and share the learned habit model,

[1548] means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit model;

[1549] a means for generating generative AI responses utilizing the tailored prompts;

[1550] means for displaying the generated answers to the user;

[1551] a means by which a user can obtain information using voice prompts;

[1552] a speech recognition means for converting the voice prompts into text;

[1553] means for displaying the generated answers on a visual device;

[1554] A system including:

[1555] (Claim 2)

[1556] The system of claim 1, further comprising means for storing and sharing the learned habit model using the brain database.

[1557] (Claim 3)

[1558] 10. The system of claim 1, further comprising: means for extracting features from the prompt using natural language processing techniques.

[1559] "Example 2: Combining Emotion Engines"

[1560] (Claim 1)

[1561] means for collecting prompts entered by a plurality of users;

[1562] a means of capturing the collected prompt and emotional information;

[1563] means for transmitting the acquired prompt and emotion information to a server;

[1564] means for the server to analyze the prompt and emotion information;

[1565] means for generating a feature vector from the parsed prompt and emotion information;

[1566] A means for classifying the generated feature vectors using a clustering algorithm to learn the user's prompt engineering habits and emotional patterns;

[1567] A means for storing and sharing the learned habit model and emotion model in a brain database;

[1568] means for selecting an associated habit model and emotion model for a new prompt and reconstructing the prompt based on the habit and emotion;

[1569] a means for generating a generative AI response utilizing the reconstructed prompt; and

[1570] means for displaying the generated answers to the user;

[1571] A system including:

[1572] (Claim 2)

[1573] The system according to claim 1, further comprising means for storing and sharing the habit model and emotion model learned using the brain database.

[1574] (Claim 3)

[1575] 10. The system of claim 1, further comprising means for extracting features from the prompt and emotion information using natural language processing techniques and an emotion engine.

[1576] "Application example 2 when combining emotion engines"

[1577] (Claim 1)

[1578] means for collecting prompts entered by a plurality of users;

[1579] a means for extracting features from the collected prompts;

[1580] A means for learning a user's prompt engineering habits based on the extracted features;

[1581] A means for acquiring user emotion information using a camera or microphone;

[1582] means for analyzing the acquired emotion information to identify the emotion of the user;

[1583] means for adjusting prompts based on the user's emotional information;

[1584] A means to save and share the learned habit model,

[1585] means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit and emotion information;

[1586] a means for generating generative AI responses utilizing the tailored prompts;

[1587] means for displaying the generated answers to the user;

[1588] A system including:

[1589] (Claim 2)

[1590] The system of claim 1, further comprising means for storing and sharing the habit model and emotion model learned using the brain database.

[1591] (Claim 3)

[1592] 10. The system of claim 1, further comprising means for extracting features from the prompt using natural language processing techniques and identifying sentiment through sentiment analysis. [Explanation of symbols]

[1593] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. means for collecting prompts entered by a plurality of users; a means for extracting features from the collected prompts; A means for learning a user's prompt engineering habits based on the extracted features; A means to save and share the learned habit model, means for selecting an associated habit model for a new prompt and adjusting the prompt based on the habit model; a means for generating generative AI responses utilizing the tailored prompts; means for displaying the generated answers to the user; A system including:

2. The system according to claim 1 , further comprising means for storing and sharing the learned habit model using the brain database.

3. The system of claim 1 , further comprising means for extracting features from the prompt using natural language processing techniques.

Images