System

Abstract

A system is provided.SOLUTION: The system includes a means for receiving a question from a user, a means for distributing the received question to a plurality of generation models, a means for generating an answer from each generation model, a means for evaluating a plurality of generated answers and selecting an optimum answer, and a means for providing the selected optimum answer to the user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] In recent years, generative models have been utilized in a variety of fields, increasing the opportunities for users to obtain information using them. However, different generative models can produce different answers, and the quality of answers from the same generative model can vary depending on how the question is asked. Generative models can also sometimes produce incorrect answers, making it difficult for users to obtain reliable information. Furthermore, specialized knowledge and operations are required for users to use multiple generative models, making them difficult for average consumers to use. There is a need for a system that can solve these issues and enable anyone to easily obtain reliable answers. [Means for solving the problem]

[0005] To solve the above problems, the present invention provides a system that accepts a question from a user, distributes it to multiple generative models, generates answers from each of the generative models, evaluates the generated answers to select an optimal answer, and provides the selected answer to the user. Specifically, the system includes: (1) a means for accepting a question from a user; (2) a means for distributing the received question to multiple generative models; (3) a means for generating an answer from each of the generative models; (4) a means for evaluating the generated answers to select an optimal answer; and (5) a means for providing the selected optimal answer to the user. This system not only improves the quality of answers to questions but also enables users to easily obtain high-quality and reliable information. Furthermore, by selecting an optimal generative model based on the results of analyzing the user's question and double-checking the answer as necessary, the risk of an incorrect answer being generated by the generative model can be reduced.

[0006] A "user" is a person or organization that uses a generative model to enter a question and obtain an answer.

[0007] A "question" is an inquiry that a user inputs to a generative model about a problem they want to solve or information they want to know.

[0008] "Generative model" is a general term for algorithms and systems that use artificial intelligence technology to automatically generate answers to input questions.

[0009] An "answer" is information generated by a generative model in response to a question and provided to the user.

[0010] "Evaluation" is the process of comparing and analyzing multiple answers generated by a generative model and selecting the most appropriate answer.

[0011] The "best answer" is the most accurate and reliable answer selected from multiple answers based on the evaluation criteria.

[0012] "Double checking" is a technique for reconfirming the accuracy of answers using multiple generative models and algorithms to increase the reliability of generated answers. [Brief explanation of the drawings]

[0013] [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

[0014] 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.

[0015] First, the terms used in the following description will be explained.

[0016] 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).

[0017] 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.

[0018] 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.

[0019] 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.

[0020] 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."

[0021] [First embodiment]

[0022] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0023] 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.

[0024] 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).

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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.

[0029] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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."

[0034] The present invention is a system that accepts a question from a user, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user. The program processing of this system is explained below in natural language.

[0035] System configuration

[0036] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[0037] Program processing flow

[0038] 1. Accepting user questions:

[0039] The user enters a question into the terminal interface.

[0040] The terminal transmits the question received from the user to the server.

[0041] 2. Question analysis and distribution:

[0042] The server analyzes the received question and selects an appropriate generative model (e.g., a text generation model or a data analysis model) based on its content.

[0043] The questions are distributed among a number of selected generative models.

[0044] 3. Generate answers:

[0045] Each generative model (a module within the server) generates an answer to the question it receives in its own way.

[0046] The generated answer is sent back to the server.

[0047] 4. Evaluating and selecting answers:

[0048] The server collects the answers returned by each generative model and evaluates them using an evaluation algorithm.

[0049] Select the most appropriate answer based on evaluation criteria (e.g., accuracy, reliability, applicability, etc.).

[0050] If necessary, generative models brainstorm and double-check with each other to produce reliable answers.

[0051] 5. Providing answers to users:

[0052] The server transmits the selected optimal answer to the user terminal.

[0053] The terminal displays the received answer to the user.

[0054] Specific examples

[0055] Question: "What is the market size for AI platforms?"

[0056] 1. Accepting user questions:

[0057] The user types into the device interface, "Please tell me about the market size of AI platforms."

[0058] The terminal transmits this input content to the server.

[0059] 2. Question analysis and distribution:

[0060] The server receives the question and analyzes it using natural language processing techniques.

[0061] The question is determined to be related to "market research" and "economic data," and is distributed to a text generation model, an economic data analysis model, and a trend prediction model.

[0062] 3. Generate answers:

[0063] The text generation model generates relevant market size forecast articles.

[0064] The economic data analysis model analyzes the current market size based on past market data.

[0065] The trend forecasting model generates data for predicting future market size.

[0066] Each model sends its answer back to the server.

[0067] 4. Evaluating and selecting answers:

[0068] The server evaluates and compares the answers from each model using an evaluation algorithm.

[0069] The text from the text generation model, the analysis results from the economic data analysis model, and the predicted values ​​from the trend prediction model are integrated to generate the optimal answer.

[0070] If necessary, the results of each generative model are checked again to select the final answer with the highest reliability.

[0071] 5. Providing answers to users:

[0072] The server sends the best answer to the user terminal.

[0073] The terminal will display the answer to the user in the following format:

[0074] The AI ​​platform market size is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026.

[0075] This system allows users to easily obtain reliable information and generates answers to questions from multiple perspectives by utilizing different generative models, which allows users to obtain more accurate and comprehensive information.

[0076] The processing flow will be explained below.

[0077] Step 1: Accepting user questions

[0078] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[0079] The device receives the entered question and sends it to the server along with metadata (question acceptance time, user ID, etc.).

[0080] Step 2: Question analysis and generative model selection

[0081] The server analyzes the received question and uses natural language processing algorithms to extract the question's topic and keywords.

[0082] Based on the analysis results, the server determines which generative model (text generation model, economic data analysis model, trend prediction model, etc.) is appropriate.

[0083] Step 3: Distributing questions

[0084] The server distributes the questions to the selected generative models, and sends the questions to each generative model depending on the analysis results.

[0085] Step 4: Generate an answer

[0086] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[0087] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0088] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[0089] Each generative model sends its answer back to the server.

[0090] Step 5: Evaluate and select responses

[0091] The server receives the answers returned from each generative model.

[0092] The server evaluates each answer using a rating algorithm, with criteria including accuracy, reliability, and applicability.

[0093] The server selects the most appropriate answer, and if necessary, brainstorms with other generative models to generate a better answer.

[0094] Step 6: Providing answers to users

[0095] The server transmits the selected optimal answer to the user terminal.

[0096] The device displays the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[0097] Through these steps, users can easily obtain reliable information. By performing multifaceted analysis and evaluation, the system can compensate for weaknesses in the generative model and provide optimal answers.

[0098] Example 1

[0099] 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."

[0100] Conventional systems typically generate answers to user questions using a single generative AI model. However, there are limitations to the answers provided by a single model, making it difficult to provide multifaceted and accurate information. Another issue is that the reliability of answers cannot be guaranteed because cross-validation between models is not performed. This can reduce the accuracy and reliability of the information users receive, potentially leading to incorrect decision-making. This invention aims to solve these issues and provide users with more reliable information.

[0101] 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.

[0102] In this invention, the server includes means for accepting questions from users, means for analyzing the accepted questions and selecting generative models based on the analysis results, means for distributing the questions to the selected generative models, means for generating answers from each generative model, means for evaluating the generated answers using an evaluation algorithm, and means for double-checking the answers from each generative model. This enables the server to analyze the questions from multiple angles and mutually verify the answers from each generative model, thereby providing more reliable information to the user.

[0103] The "means for accepting a question from a user" refers to a means for receiving a question input by a user via the interface of a terminal.

[0104] The "means for analyzing the received question and selecting a generative model based on the analysis results" refers to the means by which the server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the content of the question.

[0105] The "means for distributing questions to a plurality of selected generative models" is a means for appropriately allocating questions to generative models selected from the analysis results.

[0106] "Means for generating an answer from each generative model" refers to means for each generative model to generate an answer based on a specific prompt sentence in response to a question received.

[0107] The "means for evaluating a plurality of generated answers using an evaluation algorithm" is a means for quantitatively or qualitatively evaluating the answers returned from each generative model.

[0108] "Means for double-checking answers from each generative model" refers to a means for cross-checking answers from multiple generative models to reinforce accuracy and reliability.

[0109] The "means for providing the selected optimal answer to the user" refers to a means for transmitting the most appropriate answer that has been evaluated and double-checked to the user terminal and presenting it to the user.

[0110] This invention is a system that accepts questions from users, distributes them to multiple generative AI models, has each generative AI model generate an answer, evaluates the multiple generated answers to select the optimal answer, and provides the selected answer to the user.

[0111] System configuration

[0112] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to appropriate generative AI models, has each generate an answer, evaluates these answers using an evaluation algorithm, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[0113] Hardware and software used

[0114] User device: A computer or smart device (e.g., PC, smartphone, tablet) with an interface for inputting user questions.

[0115] Server: A server machine with a high-performance processor (e.g., x86 server, cloud-based virtual machine).

[0116] Natural language processing engine: Software to parse the question (e.g. SpaCy, NLTK).

[0117] Generative AI models: Models that generate answers to questions (e.g., GPT-3 (registered trademark), BERT, Prophet).

[0118] Evaluation algorithm: Software for evaluating the generated answers (e.g. TF-IDF, BERTScore).

[0119] Specific examples

[0120] Question: "What is the market size for AI platforms?"

[0121] 1. Accepting user questions:

[0122] The user types, "Please tell me about the market size of AI platforms" into the interface of the user device.

[0123] The terminal transmits this input content to the server.

[0124] 2. Question Analysis:

[0125] The server analyzes the received question using a natural language processing engine and extracts the question's topic and keywords (e.g., "AI platform," "market size").

[0126] 3. Generative model selection and distribution:

[0127] Based on the extracted keywords, the server selects a text generation model (GPT-3), an economic data analysis model (BERT), and a trend prediction model (Prophet).

[0128] Distribute questions to each generative AI model.

[0129] 4. Generate answers:

[0130] The text generation model (GPT-3) generates answers based on the prompt: "Please provide the latest data on the current market size of AI platforms."

[0131] The Economic Data Analysis Model (BERT) generates answers based on the prompt, "Analyze the current market size based on past market data."

[0132] The trend prediction model (Prophet) generates answers based on the prompt, "Please provide data to predict future market size."

[0133] 5. Evaluating and selecting answers:

[0134] The server evaluates the answers from each model using an evaluation algorithm (e.g., BERTScore) to score them for accuracy, reliability, and applicability.

[0135] The answer with the highest score or the best answer is selected by combining the results of each generative AI model.

[0136] 6. Providing answers to users:

[0137] The server sends the best answer to the user terminal.

[0138] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0139] This system allows users to quickly obtain reliable information, and it is capable of analyzing information from multiple angles using multiple generative AI models to provide highly accurate answers.

[0140] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0141] Step 1: Accepting user questions

[0142] The user inputs a question into the interface of the user device. For example, they might input, "Please tell me about the market size of AI platforms."

[0143] The terminal sends the question received from the user to the server. The input data is a text question, and is sent to the server using an HTTP request.

[0144] The output is the textual question data sent to the server.

[0145] Step 2: Parsing the Question

[0146] The server parses the received question using a natural language processing engine (e.g., SpaCy, NLTK). The input is the user's textual question.

[0147] The server tokenizes the question content and extracts important keywords (e.g., "AI platform" and "market size"), and breaks down the text data into words to identify nouns, verbs, etc.

[0148] The output is a list of analyzed keywords and topic information.

[0149] Step 3: Selecting a generative model and distributing questions

[0150] The server selects an appropriate generative AI model based on the extracted keywords and topic information, and prepares to deliver a specific prompt to the generative AI model.

[0151] The input is a keyword list and topic information.

[0152] The server generates appropriate prompts and distributes questions to multiple selected generative AI models (e.g., GPT-3, BERT, Prophet).

[0153] The output is multiple prompt sentences that are sent to the generative AI model.

[0154] Step 4: Generate an answer

[0155] Each generative AI model (GPT-3, BERT, Prophet) generates an answer based on a prompt received from the server. The input is the prompt sent to the generative AI model.

[0156] For example, GPT-3 receives the prompt "Please provide the latest data on the current market size of AI platforms" and generates relevant text, while Prophet receives the prompt "Please provide data to predict future market size" and generates predictive data.

[0157] The output is the answer data sent back to the server from each generative AI model.

[0158] Step 5: Evaluate responses

[0159] The server collects the answers returned by the generative AI model and evaluates them using an evaluation algorithm (e.g., TF-IDF, BERTScore). The input is multiple answer data.

[0160] The server scores answers based on criteria such as accuracy, reliability, and applicability, and compares answers between different models.

[0161] The output is the scoring results and the evaluated answer data.

[0162] Step 6: Selecting the best answer

[0163] The server selects the most appropriate answer based on the evaluation results, and if necessary, brainstorms and double-checks with other generative models.

[0164] The input is the evaluated response data.

[0165] The server determines the best answer and selects it as the final answer.

[0166] The output is the selected optimal answer data.

[0167] Step 7: Providing answers to users

[0168] The server transmits the selected optimal answer to the user terminal, and the input is the selected optimal answer data.

[0169] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0170] The output is the answer information displayed to the user.

[0171] (Application example 1)

[0172] 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."

[0173] In recent years, there has been a demand for faster and more efficient customer service in brick-and-mortar stores. However, conventional systems have difficulty in providing appropriate and prompt answers to customer questions, and it takes a long time to provide accurate answers, especially to complex questions. In addition, it is difficult to provide specific data such as inventory information immediately, which leads to a decrease in customer satisfaction.

[0174] 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.

[0175] In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, means for generating answers from each generative model, means for evaluating the generated answers and selecting an optimal answer, means for providing the selected optimal answer to the user, means for analyzing the user's question and selecting a generative model based on the analysis results, and means for making the server accessible via a smartphone application in a physical store environment. This allows users to easily input questions using their smartphones in the physical store and receive quick and accurate answers. Furthermore, specific data such as inventory information can be provided immediately, improving customer satisfaction.

[0176] The "means for accepting questions from the user" refers to an interface and mechanism for receiving questions entered by the user and processing the content of the questions appropriately.

[0177] The "means for distributing received questions to multiple generative models" is a system for analyzing received questions and distributing them to appropriate generative models based on the questions.

[0178] "Means for generating an answer from each generative model" refers to a process for generating an answer to a question using each generative model and returning the result to the server.

[0179] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for comparing answers obtained from multiple generative models using an evaluation algorithm and selecting the single optimal answer.

[0180] The "means for providing the selected optimal answer to the user" is an interface that transmits the selected optimal answer to the user terminal and presents it to the user.

[0181] "Means for analyzing a user's question and selecting a generative model based on the analysis results" refers to a mechanism that analyzes a user's question using natural language processing technology and determines the most appropriate generative model based on the results.

[0182] "Means for enabling access via smartphone applications within a brick-and-mortar store environment" refers to applications and infrastructure that enable access to the system and use of services using smartphones within a brick-and-mortar store.

[0183] "Means for double-checking answers from each generative model" is a process for reconfirming the accuracy and reliability of answers obtained from multiple generative models.

[0184] The "means for providing inventory information based on specific questions at a physical store" is a system that queries the inventory database of a physical store in response to a user's questions about inventory and provides an immediate answer.

[0185] The present invention is a system that provides quick and accurate answers to user questions in a physical store. This system accepts user questions, distributes them to multiple generative models, generates and evaluates answers from each generative model, and selects the optimal answer to provide to the user.

[0186] System configuration

[0187] The system mainly consists of a user device and a server. The user device uses a smartphone application to provide an interface for receiving questions from users and sending them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The selected answer is then provided to the user via the user device.

[0188] Hardware and software used

[0189] Hardware:

[0190] Smartphone

[0191] Servers (cloud-based servers or on-premise setups)

[0192] software:

[0193] Smartphone application (iOS / ANDROID (registered trademark))

[0194] Server-side processing is based on Python (registered trademark)

[0195] Natural language processing libraries (spaCy, NLTK)

[0196] Generative AI library (OpenAI (registered trademark) GPT-3, BERT)

[0197] Data evaluation libraries (SciPy, NumPy)

[0198] Program processing

[0199] A user enters a question using a smartphone application: The user enters a question into the application interface. For example, the user enters "Do you have this item in stock?" This question is sent to the server.

[0200] Question analysis and distribution: The server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the question content. The question is then distributed to multiple selected generative models.

[0201] Each generative model generates an answer: Each generative model generates an answer in its own way to the question it receives. For example, an AI model for inventory checking queries an inventory database and generates an answer. The generated answer is sent back to the server.

[0202] Answer evaluation and selection: The server collects the answers returned by each generative model, evaluates and compares them using an evaluation algorithm (e.g., evaluation score), and selects the best answer as the final answer.

[0203] Providing an answer to the user: The server sends the selected optimal answer to the user's device. For example, the answer "This item is in stock" is displayed on the user's smartphone.

[0204] Specific examples

[0205] A user enters a question in a store, such as "Do you have this item in stock?" The application sends this question to a server, which analyzes the question and distributes it to a generative model for inventory confirmation. The generative model queries an inventory database and generates inventory information. The server evaluates and selects the most appropriate answer from the generated answers and provides it to the user as "In stock."

[0206] Example prompt sentence:

[0207] User Question: "Do you have this item in stock?"

[0208] Prompt for generative AI model: "The user is asking if this item is in stock. Please provide stock information for the following product: Product Name: XXXX"

[0209] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0210] Step 1:

[0211] A user inputs a question using a smartphone application. Specifically, the user inputs the question into the application interface and presses the submit button. The input at this time is the user's question text, and the output is request data to the server.

[0212] Step 2:

[0213] The device sends the question received from the user to the server. Specifically, the smartphone application sends the question text to the server as an HTTP POST request. The input is the user's question text, and the output is a request packet to the server.

[0214] Step 3:

[0215] The server analyzes the received question using a natural language processing library (e.g., spaCy, NLTK) to identify the intent and category of the question. The input is the user's question text, and the output is the analyzed question metadata (intent, category).

[0216] Step 4:

[0217] The server distributes the question to an appropriate generative model based on the analysis results. Specifically, it selects a generative model based on the question category (e.g., inventory check model, customer service model), and sends the question to the appropriate generative model. The input is the parsed question metadata, and the output is a request to multiple generative models.

[0218] Step 5:

[0219] Each generative model generates an answer based on a question. For example, an inventory check model queries an inventory database, and a customer service model generates answers from pre-trained data. The input is the question distributed to each generative model, and the output is the answer text from each generative model.

[0220] Step 6:

[0221] The server evaluates the answers returned by each generative model. Evaluation algorithms (e.g., rating scores, confidence scores) are used to compare the accuracy and reliability of each answer. The input is multiple answer texts from the generative models, and the output is the answer that is judged to be the best.

[0222] Step 7:

[0223] The server selects the optimal answer and sends it to the user's device. Specifically, it sends the selected answer as an HTTP response to the user's smartphone application. The input is the optimal answer text, and the output is a response packet to the user's device.

[0224] Step 8:

[0225] The user device displays the optimal answer received from the server to the user. Specifically, the answer text is displayed on the interface of the smartphone application. The input is the response data from the server, and the output is the answer text displayed to the user.

[0226] 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.

[0227] The present invention is a system that accepts a user's question, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user.Furthermore, by combining it with an emotion engine that recognizes and analyzes the user's emotions, it provides an appropriate answer according to the user's emotions.

[0228] System configuration

[0229] The system mainly consists of a user terminal, a server, and an emotion engine. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each generative model, and selects the optimal answer. The emotion engine recognizes emotions from the user's input and actions, and adjusts the answer based on the analysis results. The final answer is then provided to the user again via the user terminal.

[0230] Program processing flow

[0231] 1. Accepting user questions:

[0232] The user types a question into the device interface, for example, "What is the market size for AI platforms?"

[0233] The device receives the user's question and sends the input data to the server, and also sends the user's input and actions to the emotion engine.

[0234] 2. Emotion Recognition and Analysis:

[0235] The emotion engine recognizes emotions from user input and behavioral data and analyzes the results, specifically using text analysis, voice analysis, and eye tracking.

[0236] The emotion engine sends the analysis results to the server.

[0237] 3. Question analysis and generative model selection:

[0238] The server takes the analysis results from the emotion engine and takes them into account when analyzing the question.

[0239] The server selects an appropriate generation model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the question and the results of sentiment analysis.

[0240] 4. Question distribution:

[0241] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[0242] 5. Generate answers:

[0243] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[0244] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0245] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[0246] Each generative model sends its answer back to the server.

[0247] 6. Evaluating and selecting answers:

[0248] The server receives the answers from each generative model and evaluates them using an evaluation algorithm, with criteria including accuracy, reliability, and applicability.

[0249] The server selects the most appropriate answer, taking into account the analysis results of the emotion engine.

[0250] If necessary, the generative models brainstorm with each other to generate better answers again.

[0251] 7. Providing answers to users:

[0252] The server sends the best answer to the user terminal.

[0253] The device will then display the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[0254] Specific examples

[0255] Question: "What is the market size for AI platforms?" (if the user is nervous)

[0256] 1. Accepting user questions:

[0257] A user types into the device interface, "Please tell me about the market size of AI platforms." Because the user is nervous, they type quickly and make many typos.

[0258] The terminal sends this input content to the emotion engine.

[0259] 2. Emotion Recognition and Analysis:

[0260] The emotion engine analyzes the user's input patterns and concludes that the user is nervous.

[0261] The emotion engine sends the analysis results to the server.

[0262] 3. Question analysis and generative model selection:

[0263] The server analyzes the question content and the sentiment analysis results, and selects an appropriate generative model to provide an answer in a format that the user can feel comfortable with.

[0264] 4. Question distribution:

[0265] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[0266] 5. Generate answers:

[0267] The text generation model generates simple and reassuring sentences for nervous users.

[0268] The economic data analysis model and trend forecasting model generate answers based on their respective roles.

[0269] Each generative model sends an answer back to the server.

[0270] 6. Evaluating and selecting answers:

[0271] The server evaluates the received answers and selects the most appropriate answer, taking into consideration, among other things, how it will ease the user's tension.

[0272] If necessary, make adjustments again.

[0273] 7. Providing answers to users:

[0274] The server sends the best answer to the terminal in a comfortable format.

[0275] The device displays the answer in a friendly tone to ease the tension.

[0276] This system allows users to obtain appropriate and reliable information according to their emotions. By combining it with an emotion engine, it is possible to further increase user satisfaction and provide more personalized services.

[0277] The processing flow will be explained below.

[0278] Step 1: Accepting user questions

[0279] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[0280] When a question is entered, the device collects the user's input patterns, voice, gaze data, etc., and sends this information to the server and emotion engine.

[0281] Step 2: Emotion Recognition and Analysis

[0282] The emotion engine analyzes the data sent from the device and recognizes the user's emotional state (tension, relief, excitement, etc.).

[0283] The emotion engine sends the results of the recognized emotional state to the server.

[0284] Step 3: Question analysis and generative model selection

[0285] The server analyzes the emotional state information and question it receives, and uses natural language processing technology to extract the topic and keywords of the question.

[0286] The server selects an appropriate generative model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the problem and the user's emotional state.

[0287] Step 4: Distributing questions

[0288] The server distributes questions to selected generative models, and sends them to each model based on the analysis and sentiment analysis results.

[0289] Step 5: Generate an answer

[0290] The text generation model receives a question and uses related text data to generate an answer, such as creating a commentary on market size forecasts for 2023.

[0291] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0292] The trend forecasting model predicts future market size and generates a forecast value for market size three years from now.

[0293] Each model sends the answer it generates back to the server.

[0294] Step 6: Evaluate and select responses

[0295] The server receives the answers from each generative model and evaluates each answer using an evaluation algorithm, with evaluation criteria including accuracy, reliability, and applicability.

[0296] The server takes into account the analysis results of the emotion engine and selects the most appropriate answer that matches the user's emotional state.

[0297] If necessary, the generative models brainstorm with each other to generate better answers again.

[0298] Step 7: Providing answers to users

[0299] The server transmits the selected optimal answer to the user terminal.

[0300] The device will adjust the tone and wording of the response to match the user's emotional state and display it to the user. For example, if a user is feeling nervous, it will display simple, reassuring language such as, "The market size of AI platforms is predicted to reach X billion dollars in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach Z billion dollars in 2026."

[0301] Through this process, users can obtain highly reliable information tailored to their emotions. This system, combined with an emotion engine, will further improve user satisfaction and realize personalized services that allow for individual responses.

[0302] Example 2

[0303] 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."

[0304] Conventional question-answering systems often fail to provide appropriate answers because they struggle to fully consider the user's emotions and input intent. Furthermore, they lack a means to incorporate the results of sentiment analysis when selecting the optimal answer from multiple generative models. This creates a risk of lowering user satisfaction.

[0305] 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.

[0306] In this invention, the server includes: [means for accepting questions from a user]; [means for distributing the accepted questions and user emotion data to a plurality of generative models]; [means for generating answers from each generative model]; [means for evaluating the plurality of generated answers and selecting the optimum answer taking into account the emotion analysis results]; and [means for providing the selected optimum answer to the user.] This makes it possible to provide appropriate and reliable answers that take into account the user's emotions.

[0307] "Means for accepting questions from a user" refers to a function that receives questions entered by a user into the interface and sends the data to the server for processing.

[0308] "Means for distributing received questions and user emotion data to multiple generative models" refers to the function of distributing the user question and emotion data obtained from the emotion engine to appropriate generative models and using that data for analysis.

[0309] "Means for generating answers from each generative model" refers to the function by which each generative AI model generates answers based on the distributed question data.

[0310] "Means for evaluating multiple generated answers and selecting the most appropriate answer taking into account the results of sentiment analysis" refers to a function that compares multiple generated answers based on certain evaluation criteria and selects the most appropriate answer taking into account the results of sentiment analysis.

[0311] The "means for providing the selected optimal answer to the user" refers to a function for transmitting the selected answer to the user terminal and displaying it to the user.

[0312] "Emotion analysis results" are data that indicate the emotional state of the user analyzed by the emotion engine.

[0313] "Generative AI model" is a general term for artificial intelligence models that generate responses to given questions.

[0314] An "emotion engine" refers to a system that analyzes a user's input data and behavioral patterns to recognize and evaluate the user's emotional state.

[0315] The present invention is a system that accepts questions from users, generates optimal answers using a generative AI model, and provides answers that correspond to the user's emotions using an emotion engine. Specific embodiments of the system are described in detail below.

[0316] Hardware and software used

[0317] Hardware

[0318] User terminal: A device through which a user inputs a question, such as a PC or smartphone.

[0319] Server: A device that processes question data and emotion data and runs generative AI models.

[0320] software

[0321] Interface application: An application that is installed on a user's device and accepts input from the user.

[0322] Emotion engine: Software that recognizes emotions by analyzing user input data and behavioral patterns.

[0323] Generative AI model: Software that generates answers to given questions. Examples include text generation models, economic data analysis models, and trend forecasting models.

[0324] Data processing and calculation methods

[0325] 1. Accepting user questions:

[0326] The user types a question into the device interface, for example, "Please tell me about the market size of AI platforms."

[0327] The device sends this input data to the server, and at the same time sends emotion data such as input speed and typos to the emotion engine.

[0328] 2. Emotion Recognition and Analysis:

[0329] The emotion engine analyzes the received emotion data and recognizes the user's emotional state, such as whether they are tense or relaxed.

[0330] The emotion engine sends the analysis results to the server.

[0331] 3. Question analysis and generative model selection:

[0332] The server obtains the emotion analysis results from the emotion engine and analyzes the question content.

[0333] Based on the analysis results, the server selects the optimal generative AI model (text generation model, economic data analysis model, trend prediction model, etc.).

[0334] 4. Question distribution:

[0335] The server sends the question text and sentiment analysis results to the selected generative AI model.

[0336] Each generative AI model generates an answer based on the question data.

[0337] 5. Generate and evaluate answers:

[0338] Each generative AI model generates an answer based on the data it receives and sends the result back to the server.

[0339] The server evaluates answers from multiple generative models and selects the optimal answer taking into account the results of sentiment analysis.

[0340] 6. Providing answers to users:

[0341] The server transmits the selected optimal answer to the user terminal.

[0342] The device will display the answer in a format that is easy for the user to understand, such as, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0343] Specific examples

[0344] Specific questions

[0345] Question: "What is the market size for AI platforms?"

[0346] Specific examples of sentiment analysis

[0347] When a user is nervous, their typing speed slows down and they make more errors. The emotion engine analyzes this and recognizes that the user is nervous. Based on the results of this analysis, the server selects a generative model to reassure the user and generates an appropriate answer.

[0348] Based on these steps and specific examples, the system of the present invention can provide appropriate and reliable answers that correspond to the user's emotions, thereby increasing user satisfaction.

[0349] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0350] Step 1:

[0351] Accepting user questions

[0352] The user types into the device interface, "Please tell me about the market size of AI platforms."

[0353] Input: Text data that a user types into a device's interface.

[0354] The terminal receives a question text from the user and transmits the text data to the server.

[0355] Output: The action of sending the received question text data to the server.

[0356] Step 2:

[0357] Emotion Recognition and Analysis

[0358] The terminal also sends emotion data, such as the user's input speed and typo error patterns, to the emotion engine in parallel.

[0359] Input: Emotional data such as user typing speed, typos, etc.

[0360] The emotion engine analyzes the received emotion data and recognizes the user's emotional state (e.g., tense, relaxed).

[0361] Output: Send the result of the emotion analysis (e.g., the user is nervous) to the server.

[0362] Step 3:

[0363] Question analysis and generative model selection

[0364] The server analyzes the emotion analysis results obtained from the emotion engine and the user's question text.

[0365] Input: User question text and sentiment analysis results.

[0366] The server analyzes the question and selects a corresponding generative AI model (e.g., text generation model, economic data analysis model, trend prediction model).

[0367] Output: Question text and sentiment analysis results are distributed to each generative AI model.

[0368] Step 4:

[0369] Question distribution

[0370] The server distributes the question text to the selected generative AI models (text generation model, economic data analysis model, trend prediction model) and also attaches the sentiment analysis results.

[0371] Input: Question text and sentiment analysis results.

[0372] Output: The question and sentiment analysis results are sent to each generative AI model.

[0373] Step 5:

[0374] Generate answers

[0375] The text generation model generates an appropriate answer based on the question it receives, for example, "The market size of AI platforms in 2023 will be $X billion."

[0376] The economic data analysis model analyzes historical market data and generates data on the current market size.

[0377] The trend forecasting model predicts future market trends and generates forecast data for the market size for the next three years.

[0378] Input: Question text and sentiment analysis results.

[0379] Output: The answer data generated by each generative AI model.

[0380] Step 6:

[0381] Evaluation and selection of answers

[0382] The server aggregates the response data received from each generative AI model and selects the optimal response based on an evaluation algorithm.

[0383] Input: Answer data from each generative AI model.

[0384] The evaluation algorithm includes accuracy, reliability, applicability, etc., and also takes into account sentiment analysis results.

[0385] If necessary, the generative models brainstorm with each other to generate better answers.

[0386] Output: The best answer.

[0387] Step 7:

[0388] Providing answers to users

[0389] The server transmits the selected optimal answer to the user terminal.

[0390] Input: Best answer.

[0391] The device will then display the answer it receives to the user in a friendly and easy-to-understand format, such as "The AI ​​platform market size is predicted to reach $X in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0392] Output: Display the best answer to the user.

[0393] (Application example 2)

[0394] 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."

[0395] Current factory robots respond to questions or instructions from operators by providing uniform answers without considering the operator's emotional state, which can lead to reduced usability and difficulty in effective communication. There is a need for a method to facilitate smooth communication between humans and robots, especially when appropriate responses are required under emotional circumstances.

[0396] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, and means for generating answers from each generative model. This makes it possible to evaluate multiple generated answers, select the optimal answer, and provide the optimal answer to the operator in a format that corresponds to the operator's emotion. In addition, by providing means for analyzing the operator's emotional state using an emotion recognition engine and taking the operator's emotional state into consideration when selecting a generative model, a more appropriate and human-like response is possible.

[0397] The "means for accepting questions from users" is an interface that allows an operator to input questions or instructions to a factory robot, which then recognizes them and receives them as data.

[0398] "Means for distributing received questions to multiple generative models" is a function for analyzing received questions and distributing them to appropriate generative models.

[0399] The "means for generating an answer from each generative model" is a system for generating a corresponding answer based on a question received by the generative model.

[0400] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for evaluating answers provided by multiple generative models and selecting the most appropriate one.

[0401] The "means for providing the operator with the optimal answer in a format that corresponds to their emotions" is a system for providing the selected optimal answer in a format that is appropriate to the operator's emotional state.

[0402] "Means for analyzing an operator's emotional state using an emotion recognition engine" refers to a technology for analyzing an operator's tone of voice, speed, facial expressions, etc. to analyze their emotional state.

[0403] "Means for considering the emotional state of the operator when selecting a generative model" is a function that uses the emotional state of the operator as an important factor in the process of selecting a generative model.

[0404] This invention is a system that enables a factory robot to provide appropriate answers to questions and instructions from an operator. The system is mainly composed of a user terminal, a server, and an emotion engine.

[0405] System configuration

[0406] 1. User Device

[0407] Operators input questions and instructions to the robot using voice commands or a tablet. The voice recognition system uses Google® Cloud Speech-to-Text, and the tablet is rugged and suitable for factory environments.

[0408] The user terminal receives the input question and sends the data to the server. It also sends emotional data such as the operator's voice tone, speed, and facial expressions to the emotion engine.

[0409] 2. Server

[0410] The server has a means for analyzing the query content and distributing it to multiple generative models (e.g., production data analysis model, machine maintenance model, quality control model, etc.).

[0411] The generative model generates corresponding answers by utilizing a text generation model (e.g., OpenAI's GPT-3), a production data analysis model (e.g., Azure® Machine Learning), etc.

[0412] The server evaluates the generated answers and selects the best one based on accuracy, reliability, and applicability, taking into account the analysis results of the emotion engine.

[0413] After the optimal answer is selected, it is provided to the user terminal in a format that corresponds to the emotional state of the operator.

[0414] 3. Emotion Engine

[0415] The emotion engine is a technology that analyzes the operator's emotional state by analyzing their voice tone, speed, facial expressions, etc. Emotion analysis uses emotion-recognition cameras and microphones (e.g., Logitech Brio camera, Shure microphone).

[0416] The analysis results are sent to the server and reflected in the generative model selection process.

[0417] Specific examples

[0418] Let's assume that a user asks, "Please tell me the current production efficiency," and the operator becomes impatient.

[0419] 1. The user terminal converts the operator's question into text using a voice recognition system and sends the text to the server.

[0420] 2. The emotion engine analyzes the operator's tone of voice, speed, and facial expressions and determines that the operator is impatient.

[0421] 3. The server selects a generative model based on the question content and the results of sentiment analysis. For example, a production data analysis model is selected.

[0422] 4. The generative model generates an answer like this: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for the next production line check."

[0423] 5. The server evaluates the generated answers and selects the best answer.

[0424] 6. The user terminal provides this response to the operator.

[0425] Prompt Sentence Examples

[0426] User: "What is your current production efficiency?"

[0427] User Emotion: Impatience

[0428] Generative model: Production data analysis model

[0429] Sample response: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for our next production line check."

[0430] This invention allows factory robots to support operators in a more efficient and human-like manner, and by using an emotion engine, it is possible to provide appropriate and reliable information according to the operator's emotional state.

[0431] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0432] Step 1: Accepting user questions

[0433] Input: The operator inputs the question using voice or a tablet.

[0434] Processing: The user device uses a speech recognition system (e.g., Google Cloud Speech-to-Text) to convert the speech into text data. If the input is from a tablet, it is accepted as text data. Emotional data such as speech tone and speed are also acquired.

[0435] Output: Textualized question data and sentiment data.

[0436] Step 2: Emotion Recognition and Analysis

[0437] Input: Emotional data such as speech tone, rate, and facial expressions.

[0438] Processing: The emotion engine analyzes the acquired emotion data and identifies the operator's emotional state (e.g., impatience, tension, joy, etc.). Emotion analysis is performed using emotion-recognition cameras and microphones (e.g., Logitech Brio cameras, Shure microphones).

[0439] Output: Parsed emotion data.

[0440] Step 3: Question analysis and generative model selection

[0441] Input: Translated question data and parsed sentiment data.

[0442] Processing: The server analyzes the question content and the emotional state and selects an appropriate generative model (e.g., production data analysis model, machine maintenance model, quality control model). The selection criteria include the question content and the operator's emotional state.

[0443] Output: The selected generative model.

[0444] Step 4: Distributing questions

[0445] Input: Textualized question data, parsed sentiment data, and a selected generative model.

[0446] Processing: The server distributes the questions to the selected generative models. It sends the question data and emotion data to each generative model.

[0447] Output: Question data and sentiment data sent to the generative model.

[0448] Step 5: Generate an answer

[0449] Input: Question data and sentiment data sent to the generative model.

[0450] Processing: Each generative model (e.g., text generation model, production data analysis model, machine maintenance model, etc.) generates an answer based on the received question. For example, a text generation model creates an answer in Q&A format.

[0451] Output: The generated answers.

[0452] Step 6: Evaluate and select responses

[0453] Input: Multiple generated answers.

[0454] Processing: The server evaluates the answers from each generative model and selects the best answer based on evaluation criteria (accuracy, reliability, applicability, etc.), also taking into account the analysis results of the emotion engine.

[0455] Output: The best answer selected.

[0456] Step 7: Providing answers to users

[0457] Input: The selected best answer.

[0458] Processing: The server sends the selected optimal answer to the user terminal in a format that corresponds to the operator's emotional state. For example, it displays the answer in a friendly tone or in concise sentences.

[0459] Output: Answers provided in a sentiment-sensitive format.

[0460] 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.

[0461] 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.

[0462] 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.

[0463] [Second embodiment]

[0464] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0465] 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.

[0466] 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).

[0467] 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.

[0468] 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.

[0469] 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).

[0470] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0471] 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.

[0472] 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.

[0473] 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.

[0474] 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.

[0475] 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."

[0476] The present invention is a system that accepts a question from a user, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user. The program processing of this system is explained below in natural language.

[0477] System configuration

[0478] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[0479] Program processing flow

[0480] 1. Accepting user questions:

[0481] The user enters a question into the terminal interface.

[0482] The terminal transmits the question received from the user to the server.

[0483] 2. Question analysis and distribution:

[0484] The server analyzes the received question and selects an appropriate generative model (e.g., a text generation model or a data analysis model) based on its content.

[0485] The questions are distributed among a number of selected generative models.

[0486] 3. Generate answers:

[0487] Each generative model (a module within the server) generates an answer to the question it receives in its own way.

[0488] The generated answer is sent back to the server.

[0489] 4. Evaluating and selecting answers:

[0490] The server collects the answers returned by each generative model and evaluates them using an evaluation algorithm.

[0491] Select the most appropriate answer based on evaluation criteria (e.g., accuracy, reliability, applicability, etc.).

[0492] If necessary, generative models brainstorm and double-check with each other to produce reliable answers.

[0493] 5. Providing answers to users:

[0494] The server transmits the selected optimal answer to the user terminal.

[0495] The terminal displays the received answer to the user.

[0496] Specific examples

[0497] Question: "What is the market size for AI platforms?"

[0498] 1. Accepting user questions:

[0499] The user types into the device interface, "Please tell me about the market size of AI platforms."

[0500] The terminal transmits this input content to the server.

[0501] 2. Question analysis and distribution:

[0502] The server receives the question and analyzes it using natural language processing techniques.

[0503] The question is determined to be related to "market research" and "economic data," and is distributed to a text generation model, an economic data analysis model, and a trend prediction model.

[0504] 3. Generate answers:

[0505] The text generation model generates relevant market size forecast articles.

[0506] The economic data analysis model analyzes the current market size based on past market data.

[0507] The trend forecasting model generates data for predicting future market size.

[0508] Each model sends its answer back to the server.

[0509] 4. Evaluating and selecting answers:

[0510] The server evaluates and compares the answers from each model using an evaluation algorithm.

[0511] The text from the text generation model, the analysis results from the economic data analysis model, and the predicted values ​​from the trend prediction model are integrated to generate the optimal answer.

[0512] If necessary, the results of each generative model are checked again to select the final answer with the highest reliability.

[0513] 5. Providing answers to users:

[0514] The server sends the best answer to the user terminal.

[0515] The terminal will display the answer to the user in the following format:

[0516] The AI ​​platform market size is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026.

[0517] This system allows users to easily obtain reliable information and generates answers to questions from multiple perspectives by utilizing different generative models, which allows users to obtain more accurate and comprehensive information.

[0518] The processing flow will be explained below.

[0519] Step 1: Accepting user questions

[0520] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[0521] The device receives the entered question and sends it to the server along with metadata (question acceptance time, user ID, etc.).

[0522] Step 2: Question analysis and generative model selection

[0523] The server analyzes the received question and uses natural language processing algorithms to extract the question's topic and keywords.

[0524] Based on the analysis results, the server determines which generative model (text generation model, economic data analysis model, trend prediction model, etc.) is appropriate.

[0525] Step 3: Distributing questions

[0526] The server distributes the questions to the selected generative models, and sends the questions to each generative model depending on the analysis results.

[0527] Step 4: Generate an answer

[0528] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[0529] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0530] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[0531] Each generative model sends its answer back to the server.

[0532] Step 5: Evaluate and select responses

[0533] The server receives the answers returned from each generative model.

[0534] The server evaluates each answer using a rating algorithm, with criteria including accuracy, reliability, and applicability.

[0535] The server selects the most appropriate answer, and if necessary, brainstorms with other generative models to generate a better answer.

[0536] Step 6: Providing answers to users

[0537] The server transmits the selected optimal answer to the user terminal.

[0538] The device displays the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[0539] Through these steps, users can easily obtain reliable information. By performing multifaceted analysis and evaluation, the system can compensate for weaknesses in the generative model and provide optimal answers.

[0540] Example 1

[0541] 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."

[0542] Conventional systems typically generate answers to user questions using a single generative AI model. However, there are limitations to the answers provided by a single model, making it difficult to provide multifaceted and accurate information. Another issue is that the reliability of answers cannot be guaranteed because cross-validation between models is not performed. This can reduce the accuracy and reliability of the information users receive, potentially leading to incorrect decision-making. This invention aims to solve these issues and provide users with more reliable information.

[0543] 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.

[0544] In this invention, the server includes means for accepting questions from users, means for analyzing the accepted questions and selecting generative models based on the analysis results, means for distributing the questions to the selected generative models, means for generating answers from each generative model, means for evaluating the generated answers using an evaluation algorithm, and means for double-checking the answers from each generative model. This enables the server to analyze the questions from multiple angles and mutually verify the answers from each generative model, thereby providing more reliable information to the user.

[0545] The "means for accepting a question from a user" refers to a means for receiving a question input by a user via the interface of a terminal.

[0546] The "means for analyzing the received question and selecting a generative model based on the analysis results" refers to the means by which the server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the content of the question.

[0547] The "means for distributing questions to a plurality of selected generative models" is a means for appropriately allocating questions to generative models selected from the analysis results.

[0548] "Means for generating an answer from each generative model" refers to means for each generative model to generate an answer based on a specific prompt sentence in response to a question received.

[0549] The "means for evaluating a plurality of generated answers using an evaluation algorithm" is a means for quantitatively or qualitatively evaluating the answers returned from each generative model.

[0550] "Means for double-checking answers from each generative model" refers to a means for cross-checking answers from multiple generative models to reinforce accuracy and reliability.

[0551] The "means for providing the selected optimal answer to the user" refers to a means for transmitting the most appropriate answer that has been evaluated and double-checked to the user terminal and presenting it to the user.

[0552] This invention is a system that accepts questions from users, distributes them to multiple generative AI models, has each generative AI model generate an answer, evaluates the multiple generated answers to select the optimal answer, and provides the selected answer to the user.

[0553] System configuration

[0554] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to appropriate generative AI models, has each generate an answer, evaluates these answers using an evaluation algorithm, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[0555] Hardware and software used

[0556] User device: A computer or smart device (e.g., PC, smartphone, tablet) with an interface for inputting user questions.

[0557] Server: A server machine with a high-performance processor (e.g., x86 server, cloud-based virtual machine).

[0558] Natural language processing engine: Software to parse the question (e.g. SpaCy, NLTK).

[0559] Generative AI models: Models that generate answers to questions (e.g., GPT-3, BERT, Prophet).

[0560] Evaluation algorithm: Software for evaluating the generated answers (e.g. TF-IDF, BERTScore).

[0561] Specific examples

[0562] Question: "What is the market size for AI platforms?"

[0563] 1. Accepting user questions:

[0564] The user types, "Please tell me about the market size of AI platforms" into the interface of the user device.

[0565] The terminal transmits this input content to the server.

[0566] 2. Question Analysis:

[0567] The server analyzes the received question using a natural language processing engine and extracts the question's topic and keywords (e.g., "AI platform," "market size").

[0568] 3. Generative model selection and distribution:

[0569] Based on the extracted keywords, the server selects a text generation model (GPT-3), an economic data analysis model (BERT), and a trend prediction model (Prophet).

[0570] Distribute questions to each generative AI model.

[0571] 4. Generate answers:

[0572] The text generation model (GPT-3) generates answers based on the prompt: "Please provide the latest data on the current market size of AI platforms."

[0573] The Economic Data Analysis Model (BERT) generates answers based on the prompt, "Analyze the current market size based on past market data."

[0574] The trend prediction model (Prophet) generates answers based on the prompt, "Please provide data to predict future market size."

[0575] 5. Evaluating and selecting answers:

[0576] The server evaluates the answers from each model using an evaluation algorithm (e.g., BERTScore) to score them for accuracy, reliability, and applicability.

[0577] The answer with the highest score or the best answer is selected by combining the results of each generative AI model.

[0578] 6. Providing answers to users:

[0579] The server sends the best answer to the user terminal.

[0580] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0581] This system allows users to quickly obtain reliable information, and it is capable of analyzing information from multiple angles using multiple generative AI models to provide highly accurate answers.

[0582] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0583] Step 1: Accepting user questions

[0584] The user inputs a question into the interface of the user device. For example, they might input, "Please tell me about the market size of AI platforms."

[0585] The terminal sends the question received from the user to the server. The input data is a text question, and is sent to the server using an HTTP request.

[0586] The output is the textual question data sent to the server.

[0587] Step 2: Parsing the Question

[0588] The server parses the received question using a natural language processing engine (e.g., SpaCy, NLTK). The input is the user's textual question.

[0589] The server tokenizes the question content and extracts important keywords (e.g., "AI platform" and "market size"), and breaks down the text data into words to identify nouns, verbs, etc.

[0590] The output is a list of analyzed keywords and topic information.

[0591] Step 3: Selecting a generative model and distributing questions

[0592] The server selects an appropriate generative AI model based on the extracted keywords and topic information, and prepares to deliver a specific prompt to the generative AI model.

[0593] The input is a keyword list and topic information.

[0594] The server generates appropriate prompts and distributes questions to multiple selected generative AI models (e.g., GPT-3, BERT, Prophet).

[0595] The output is multiple prompt sentences that are sent to the generative AI model.

[0596] Step 4: Generate an answer

[0597] Each generative AI model (GPT-3, BERT, Prophet) generates an answer based on a prompt received from the server. The input is the prompt sent to the generative AI model.

[0598] For example, GPT-3 receives the prompt "Please provide the latest data on the current market size of AI platforms" and generates relevant text, while Prophet receives the prompt "Please provide data to predict future market size" and generates predictive data.

[0599] The output is the answer data sent back to the server from each generative AI model.

[0600] Step 5: Evaluate responses

[0601] The server collects the answers returned by the generative AI model and evaluates them using an evaluation algorithm (e.g., TF-IDF, BERTScore). The input is multiple answer data.

[0602] The server scores answers based on criteria such as accuracy, reliability, and applicability, and compares answers between different models.

[0603] The output is the scoring results and the evaluated answer data.

[0604] Step 6: Selecting the best answer

[0605] The server selects the most appropriate answer based on the evaluation results, and if necessary, brainstorms and double-checks with other generative models.

[0606] The input is the evaluated response data.

[0607] The server determines the best answer and selects it as the final answer.

[0608] The output is the selected optimal answer data.

[0609] Step 7: Providing answers to users

[0610] The server transmits the selected optimal answer to the user terminal, and the input is the selected optimal answer data.

[0611] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0612] The output is the answer information displayed to the user.

[0613] (Application example 1)

[0614] 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."

[0615] In recent years, there has been a demand for faster and more efficient customer service in brick-and-mortar stores. However, conventional systems have difficulty in providing appropriate and prompt answers to customer questions, and it takes a long time to provide accurate answers, especially to complex questions. In addition, it is difficult to provide specific data such as inventory information immediately, which leads to a decrease in customer satisfaction.

[0616] 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.

[0617] In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, means for generating answers from each generative model, means for evaluating the generated answers and selecting an optimal answer, means for providing the selected optimal answer to the user, means for analyzing the user's question and selecting a generative model based on the analysis results, and means for making the server accessible via a smartphone application in a physical store environment. This allows users to easily input questions using their smartphones in the physical store and receive quick and accurate answers. Furthermore, specific data such as inventory information can be provided immediately, improving customer satisfaction.

[0618] The "means for accepting questions from the user" refers to an interface and mechanism for receiving questions entered by the user and processing the content of the questions appropriately.

[0619] The "means for distributing received questions to multiple generative models" is a system for analyzing received questions and distributing them to appropriate generative models based on the questions.

[0620] "Means for generating an answer from each generative model" refers to a process for generating an answer to a question using each generative model and returning the result to the server.

[0621] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for comparing answers obtained from multiple generative models using an evaluation algorithm and selecting the single optimal answer.

[0622] The "means for providing the selected optimal answer to the user" is an interface that transmits the selected optimal answer to the user terminal and presents it to the user.

[0623] "Means for analyzing a user's question and selecting a generative model based on the analysis results" refers to a mechanism that analyzes a user's question using natural language processing technology and determines the most appropriate generative model based on the results.

[0624] "Means for enabling access via smartphone applications within a brick-and-mortar store environment" refers to applications and infrastructure that enable access to the system and use of services using smartphones within a brick-and-mortar store.

[0625] "Means for double-checking answers from each generative model" is a process for reconfirming the accuracy and reliability of answers obtained from multiple generative models.

[0626] The "means for providing inventory information based on specific questions at a physical store" is a system that queries the inventory database of a physical store in response to a user's questions about inventory and provides an immediate answer.

[0627] The present invention is a system that provides quick and accurate answers to user questions in a physical store. This system accepts user questions, distributes them to multiple generative models, generates and evaluates answers from each generative model, and selects the optimal answer to provide to the user.

[0628] System configuration

[0629] The system mainly consists of a user device and a server. The user device uses a smartphone application to provide an interface for receiving questions from users and sending them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The selected answer is then provided to the user via the user device.

[0630] Hardware and software used

[0631] Hardware:

[0632] Smartphone

[0633] Servers (cloud-based servers or on-premise setups)

[0634] software:

[0635] Smartphone application (iOS / Android)

[0636] Server-side processing is based on Python

[0637] Natural language processing libraries (spaCy, NLTK)

[0638] Generative AI libraries (OpenAI GPT-3, BERT)

[0639] Data evaluation libraries (SciPy, NumPy)

[0640] Program processing

[0641] A user enters a question using a smartphone application: The user enters a question into the application interface. For example, the user enters "Do you have this item in stock?" This question is sent to the server.

[0642] Question analysis and distribution: The server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the question content. The question is then distributed to multiple selected generative models.

[0643] Each generative model generates an answer: Each generative model generates an answer in its own way to the question it receives. For example, an AI model for inventory checking queries an inventory database and generates an answer. The generated answer is sent back to the server.

[0644] Answer evaluation and selection: The server collects the answers returned by each generative model, evaluates and compares them using an evaluation algorithm (e.g., evaluation score), and selects the best answer as the final answer.

[0645] Providing an answer to the user: The server sends the selected optimal answer to the user's device. For example, the answer "This item is in stock" is displayed on the user's smartphone.

[0646] Specific examples

[0647] A user enters a question in a store, such as "Do you have this item in stock?" The application sends this question to a server, which analyzes the question and distributes it to a generative model for inventory confirmation. The generative model queries an inventory database and generates inventory information. The server evaluates and selects the most appropriate answer from the generated answers and provides it to the user as "In stock."

[0648] Example prompt sentence:

[0649] User Question: "Do you have this item in stock?"

[0650] Prompt for generative AI model: "The user is asking if this item is in stock. Please provide stock information for the following product: Product Name: XXXX"

[0651] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0652] Step 1:

[0653] A user inputs a question using a smartphone application. Specifically, the user inputs the question into the application interface and presses the submit button. The input at this time is the user's question text, and the output is request data to the server.

[0654] Step 2:

[0655] The device sends the question received from the user to the server. Specifically, the smartphone application sends the question text to the server as an HTTP POST request. The input is the user's question text, and the output is a request packet to the server.

[0656] Step 3:

[0657] The server analyzes the received question using a natural language processing library (e.g., spaCy, NLTK) to identify the intent and category of the question. The input is the user's question text, and the output is the analyzed question metadata (intent, category).

[0658] Step 4:

[0659] The server distributes the question to an appropriate generative model based on the analysis results. Specifically, it selects a generative model based on the question category (e.g., inventory check model, customer service model), and sends the question to the appropriate generative model. The input is the parsed question metadata, and the output is a request to multiple generative models.

[0660] Step 5:

[0661] Each generative model generates an answer based on a question. For example, an inventory check model queries an inventory database, and a customer service model generates answers from pre-trained data. The input is the question distributed to each generative model, and the output is the answer text from each generative model.

[0662] Step 6:

[0663] The server evaluates the answers returned by each generative model. Evaluation algorithms (e.g., rating scores, confidence scores) are used to compare the accuracy and reliability of each answer. The input is multiple answer texts from the generative models, and the output is the answer that is judged to be the best.

[0664] Step 7:

[0665] The server selects the optimal answer and sends it to the user's device. Specifically, it sends the selected answer as an HTTP response to the user's smartphone application. The input is the optimal answer text, and the output is a response packet to the user's device.

[0666] Step 8:

[0667] The user device displays the optimal answer received from the server to the user. Specifically, the answer text is displayed on the interface of the smartphone application. The input is the response data from the server, and the output is the answer text displayed to the user.

[0668] 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.

[0669] The present invention is a system that accepts a user's question, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user.Furthermore, by combining it with an emotion engine that recognizes and analyzes the user's emotions, it provides an appropriate answer according to the user's emotions.

[0670] System configuration

[0671] The system mainly consists of a user terminal, a server, and an emotion engine. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each generative model, and selects the optimal answer. The emotion engine recognizes emotions from the user's input and actions, and adjusts the answer based on the analysis results. The final answer is then provided to the user again via the user terminal.

[0672] Program processing flow

[0673] 1. Accepting user questions:

[0674] The user types a question into the device interface, for example, "What is the market size for AI platforms?"

[0675] The device receives the user's question and sends the input data to the server, and also sends the user's input and actions to the emotion engine.

[0676] 2. Emotion Recognition and Analysis:

[0677] The emotion engine recognizes emotions from user input and behavioral data and analyzes the results, specifically using text analysis, voice analysis, and eye tracking.

[0678] The emotion engine sends the analysis results to the server.

[0679] 3. Question analysis and generative model selection:

[0680] The server takes the analysis results from the emotion engine and takes them into account when analyzing the question.

[0681] The server selects an appropriate generation model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the question and the results of sentiment analysis.

[0682] 4. Question distribution:

[0683] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[0684] 5. Generate answers:

[0685] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[0686] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0687] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[0688] Each generative model sends its answer back to the server.

[0689] 6. Evaluating and selecting answers:

[0690] The server receives the answers from each generative model and evaluates them using an evaluation algorithm, with criteria including accuracy, reliability, and applicability.

[0691] The server selects the most appropriate answer, taking into account the analysis results of the emotion engine.

[0692] If necessary, the generative models brainstorm with each other to generate better answers again.

[0693] 7. Providing answers to users:

[0694] The server sends the best answer to the user terminal.

[0695] The device will then display the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[0696] Specific examples

[0697] Question: "What is the market size for AI platforms?" (if the user is nervous)

[0698] 1. Accepting user questions:

[0699] A user types into the device interface, "Please tell me about the market size of AI platforms." Because the user is nervous, they type quickly and make many typos.

[0700] The terminal sends this input content to the emotion engine.

[0701] 2. Emotion Recognition and Analysis:

[0702] The emotion engine analyzes the user's input patterns and concludes that the user is nervous.

[0703] The emotion engine sends the analysis results to the server.

[0704] 3. Question analysis and generative model selection:

[0705] The server analyzes the question content and the sentiment analysis results, and selects an appropriate generative model to provide an answer in a format that the user can feel comfortable with.

[0706] 4. Question distribution:

[0707] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[0708] 5. Generate answers:

[0709] The text generation model generates simple and reassuring sentences for nervous users.

[0710] The economic data analysis model and trend forecasting model generate answers based on their respective roles.

[0711] Each generative model sends an answer back to the server.

[0712] 6. Evaluating and selecting answers:

[0713] The server evaluates the received answers and selects the most appropriate answer, taking into consideration, among other things, how it will ease the user's tension.

[0714] If necessary, make adjustments again.

[0715] 7. Providing answers to users:

[0716] The server sends the best answer to the terminal in a comfortable format.

[0717] The device displays the answer in a friendly tone to ease the tension.

[0718] This system allows users to obtain appropriate and reliable information according to their emotions. By combining it with an emotion engine, it is possible to further increase user satisfaction and provide more personalized services.

[0719] The processing flow will be explained below.

[0720] Step 1: Accepting user questions

[0721] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[0722] When a question is entered, the device collects the user's input patterns, voice, gaze data, etc., and sends this information to the server and emotion engine.

[0723] Step 2: Emotion Recognition and Analysis

[0724] The emotion engine analyzes the data sent from the device and recognizes the user's emotional state (tension, relief, excitement, etc.).

[0725] The emotion engine sends the results of the recognized emotional state to the server.

[0726] Step 3: Question analysis and generative model selection

[0727] The server analyzes the emotional state information and question it receives, and uses natural language processing technology to extract the topic and keywords of the question.

[0728] The server selects an appropriate generative model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the problem and the user's emotional state.

[0729] Step 4: Distributing questions

[0730] The server distributes questions to selected generative models, and sends them to each model based on the analysis and sentiment analysis results.

[0731] Step 5: Generate an answer

[0732] The text generation model receives a question and uses related text data to generate an answer, such as creating a commentary on market size forecasts for 2023.

[0733] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0734] The trend forecasting model predicts future market size and generates a forecast value for market size three years from now.

[0735] Each model sends the answer it generates back to the server.

[0736] Step 6: Evaluate and select responses

[0737] The server receives the answers from each generative model and evaluates each answer using an evaluation algorithm, with evaluation criteria including accuracy, reliability, and applicability.

[0738] The server takes into account the analysis results of the emotion engine and selects the most appropriate answer that matches the user's emotional state.

[0739] If necessary, the generative models brainstorm with each other to generate better answers again.

[0740] Step 7: Providing answers to users

[0741] The server transmits the selected optimal answer to the user terminal.

[0742] The device will adjust the tone and wording of the response to match the user's emotional state and display it to the user. For example, if a user is feeling nervous, it will display simple, reassuring language such as, "The market size of AI platforms is predicted to reach X billion dollars in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach Z billion dollars in 2026."

[0743] Through this process, users can obtain highly reliable information tailored to their emotions. This system, combined with an emotion engine, will further improve user satisfaction and realize personalized services that allow for individual responses.

[0744] Example 2

[0745] 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."

[0746] Conventional question-answering systems often fail to provide appropriate answers because they struggle to fully consider the user's emotions and input intent. Furthermore, they lack a means to incorporate the results of sentiment analysis when selecting the optimal answer from multiple generative models. This creates a risk of lowering user satisfaction.

[0747] 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.

[0748] In this invention, the server includes: [means for accepting questions from a user]; [means for distributing the accepted questions and user emotion data to a plurality of generative models]; [means for generating answers from each generative model]; [means for evaluating the plurality of generated answers and selecting the optimum answer taking into account the emotion analysis results]; and [means for providing the selected optimum answer to the user.] This makes it possible to provide appropriate and reliable answers that take into account the user's emotions.

[0749] "Means for accepting questions from a user" refers to a function that receives questions entered by a user into the interface and sends the data to the server for processing.

[0750] "Means for distributing received questions and user emotion data to multiple generative models" refers to the function of distributing the user question and emotion data obtained from the emotion engine to appropriate generative models and using that data for analysis.

[0751] "Means for generating answers from each generative model" refers to the function by which each generative AI model generates answers based on the distributed question data.

[0752] "Means for evaluating multiple generated answers and selecting the most appropriate answer taking into account the results of sentiment analysis" refers to a function that compares multiple generated answers based on certain evaluation criteria and selects the most appropriate answer taking into account the results of sentiment analysis.

[0753] The "means for providing the selected optimal answer to the user" refers to a function for transmitting the selected answer to the user terminal and displaying it to the user.

[0754] "Emotion analysis results" are data that indicate the emotional state of the user analyzed by the emotion engine.

[0755] "Generative AI model" is a general term for artificial intelligence models that generate responses to given questions.

[0756] An "emotion engine" refers to a system that analyzes a user's input data and behavioral patterns to recognize and evaluate the user's emotional state.

[0757] The present invention is a system that accepts questions from users, generates optimal answers using a generative AI model, and provides answers that correspond to the user's emotions using an emotion engine. Specific embodiments of the system are described in detail below.

[0758] Hardware and software used

[0759] Hardware

[0760] User terminal: A device through which a user inputs a question, such as a PC or smartphone.

[0761] Server: A device that processes question data and emotion data and runs generative AI models.

[0762] software

[0763] Interface application: An application that is installed on a user's device and accepts input from the user.

[0764] Emotion engine: Software that recognizes emotions by analyzing user input data and behavioral patterns.

[0765] Generative AI model: Software that generates answers to given questions. Examples include text generation models, economic data analysis models, and trend forecasting models.

[0766] Data processing and calculation methods

[0767] 1. Accepting user questions:

[0768] The user types a question into the device interface, for example, "Please tell me about the market size of AI platforms."

[0769] The device sends this input data to the server, and at the same time sends emotion data such as input speed and typos to the emotion engine.

[0770] 2. Emotion Recognition and Analysis:

[0771] The emotion engine analyzes the received emotion data and recognizes the user's emotional state, such as whether they are tense or relaxed.

[0772] The emotion engine sends the analysis results to the server.

[0773] 3. Question analysis and generative model selection:

[0774] The server obtains the emotion analysis results from the emotion engine and analyzes the question content.

[0775] Based on the analysis results, the server selects the optimal generative AI model (text generation model, economic data analysis model, trend prediction model, etc.).

[0776] 4. Question distribution:

[0777] The server sends the question text and sentiment analysis results to the selected generative AI model.

[0778] Each generative AI model generates an answer based on the question data.

[0779] 5. Generate and evaluate answers:

[0780] Each generative AI model generates an answer based on the data it receives and sends the result back to the server.

[0781] The server evaluates answers from multiple generative models and selects the optimal answer taking into account the results of sentiment analysis.

[0782] 6. Providing answers to users:

[0783] The server transmits the selected optimal answer to the user terminal.

[0784] The device will display the answer in a format that is easy for the user to understand, such as, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0785] Specific examples

[0786] Specific questions

[0787] Question: "What is the market size for AI platforms?"

[0788] Specific examples of sentiment analysis

[0789] When a user is nervous, their typing speed slows down and they make more errors. The emotion engine analyzes this and recognizes that the user is nervous. Based on the results of this analysis, the server selects a generative model to reassure the user and generates an appropriate answer.

[0790] Based on these steps and specific examples, the system of the present invention can provide appropriate and reliable answers that correspond to the user's emotions, thereby increasing user satisfaction.

[0791] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0792] Step 1:

[0793] Accepting user questions

[0794] The user types into the device interface, "Please tell me about the market size of AI platforms."

[0795] Input: Text data that a user types into a device's interface.

[0796] The terminal receives a question text from the user and transmits the text data to the server.

[0797] Output: The action of sending the received question text data to the server.

[0798] Step 2:

[0799] Emotion Recognition and Analysis

[0800] The terminal also sends emotion data, such as the user's input speed and typo error patterns, to the emotion engine in parallel.

[0801] Input: Emotional data such as user typing speed, typos, etc.

[0802] The emotion engine analyzes the received emotion data and recognizes the user's emotional state (e.g., tense, relaxed).

[0803] Output: Send the result of the emotion analysis (e.g., the user is nervous) to the server.

[0804] Step 3:

[0805] Question analysis and generative model selection

[0806] The server analyzes the emotion analysis results obtained from the emotion engine and the user's question text.

[0807] Input: User question text and sentiment analysis results.

[0808] The server analyzes the question and selects a corresponding generative AI model (e.g., text generation model, economic data analysis model, trend prediction model).

[0809] Output: Question text and sentiment analysis results are distributed to each generative AI model.

[0810] Step 4:

[0811] Question distribution

[0812] The server distributes the question text to the selected generative AI models (text generation model, economic data analysis model, trend prediction model) and also attaches the sentiment analysis results.

[0813] Input: Question text and sentiment analysis results.

[0814] Output: The question and sentiment analysis results are sent to each generative AI model.

[0815] Step 5:

[0816] Generate answers

[0817] The text generation model generates an appropriate answer based on the question it receives, for example, "The market size of AI platforms in 2023 will be $X billion."

[0818] The economic data analysis model analyzes historical market data and generates data on the current market size.

[0819] The trend forecasting model predicts future market trends and generates forecast data for the market size for the next three years.

[0820] Input: Question text and sentiment analysis results.

[0821] Output: The answer data generated by each generative AI model.

[0822] Step 6:

[0823] Evaluation and selection of answers

[0824] The server aggregates the response data received from each generative AI model and selects the optimal response based on an evaluation algorithm.

[0825] Input: Answer data from each generative AI model.

[0826] The evaluation algorithm includes accuracy, reliability, applicability, etc., and also takes into account sentiment analysis results.

[0827] If necessary, the generative models brainstorm with each other to generate better answers.

[0828] Output: The best answer.

[0829] Step 7:

[0830] Providing answers to users

[0831] The server transmits the selected optimal answer to the user terminal.

[0832] Input: Best answer.

[0833] The device will then display the answer it receives to the user in a friendly and easy-to-understand format, such as "The AI ​​platform market size is predicted to reach $X in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[0834] Output: Display the best answer to the user.

[0835] (Application example 2)

[0836] 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."

[0837] Current factory robots respond to questions or instructions from operators by providing uniform answers without considering the operator's emotional state, which can lead to reduced usability and difficulty in effective communication. There is a need for a method to facilitate smooth communication between humans and robots, especially when appropriate responses are required under emotional circumstances.

[0838] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, and means for generating answers from each generative model. This makes it possible to evaluate multiple generated answers, select the optimal answer, and provide the optimal answer to the operator in a format that corresponds to the operator's emotion. In addition, by providing means for analyzing the operator's emotional state using an emotion recognition engine and taking the operator's emotional state into consideration when selecting a generative model, a more appropriate and human-like response is possible.

[0839] The "means for accepting questions from users" is an interface that allows an operator to input questions or instructions to a factory robot, which then recognizes them and receives them as data.

[0840] "Means for distributing received questions to multiple generative models" is a function for analyzing received questions and distributing them to appropriate generative models.

[0841] The "means for generating an answer from each generative model" is a system for generating a corresponding answer based on a question received by the generative model.

[0842] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for evaluating answers provided by multiple generative models and selecting the most appropriate one.

[0843] The "means for providing the operator with the optimal answer in a format that corresponds to their emotions" is a system for providing the selected optimal answer in a format that is appropriate to the operator's emotional state.

[0844] "Means for analyzing an operator's emotional state using an emotion recognition engine" refers to a technology for analyzing an operator's tone of voice, speed, facial expressions, etc. to analyze their emotional state.

[0845] "Means for considering the emotional state of the operator when selecting a generative model" is a function that uses the emotional state of the operator as an important factor in the process of selecting a generative model.

[0846] This invention is a system that enables a factory robot to provide appropriate answers to questions and instructions from an operator. The system is mainly composed of a user terminal, a server, and an emotion engine.

[0847] System configuration

[0848] 1. User Device

[0849] Operators can input questions and instructions to the robot using voice commands or a tablet, which uses Google Cloud Speech-to-Text as its voice recognition system and is rugged and suitable for factory environments.

[0850] The user terminal receives the input question and sends the data to the server. It also sends emotional data such as the operator's voice tone, speed, and facial expressions to the emotion engine.

[0851] 2. Server

[0852] The server has a means for analyzing the query content and distributing it to multiple generative models (e.g., production data analysis model, machine maintenance model, quality control model, etc.).

[0853] The generative model generates corresponding answers by utilizing a text generation model (e.g., OpenAI's GPT-3), a production data analysis model (e.g., Azure Machine Learning), etc.

[0854] The server evaluates the generated answers and selects the best one based on accuracy, reliability, and applicability, taking into account the analysis results of the emotion engine.

[0855] After the optimal answer is selected, it is provided to the user terminal in a format that corresponds to the emotional state of the operator.

[0856] 3. Emotion Engine

[0857] The emotion engine is a technology that analyzes the operator's emotional state by analyzing their voice tone, speed, facial expressions, etc. Emotion analysis uses emotion-recognition cameras and microphones (e.g., Logitech Brio camera, Shure microphone).

[0858] The analysis results are sent to the server and reflected in the generative model selection process.

[0859] Specific examples

[0860] Let's assume that a user asks, "Please tell me the current production efficiency," and the operator becomes impatient.

[0861] 1. The user terminal converts the operator's question into text using a voice recognition system and sends the text to the server.

[0862] 2. The emotion engine analyzes the operator's tone of voice, speed, and facial expressions and determines that the operator is impatient.

[0863] 3. The server selects a generative model based on the question content and the results of sentiment analysis. For example, a production data analysis model is selected.

[0864] 4. The generative model generates an answer like this: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for the next production line check."

[0865] 5. The server evaluates the generated answers and selects the best answer.

[0866] 6. The user terminal provides this response to the operator.

[0867] Prompt Sentence Examples

[0868] User: "What is your current production efficiency?"

[0869] User Emotion: Impatience

[0870] Generative model: Production data analysis model

[0871] Sample response: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for our next production line check."

[0872] This invention allows factory robots to support operators in a more efficient and human-like manner, and by using an emotion engine, it is possible to provide appropriate and reliable information according to the operator's emotional state.

[0873] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0874] Step 1: Accepting user questions

[0875] Input: The operator inputs the question using voice or a tablet.

[0876] Processing: The user device uses a speech recognition system (e.g., Google Cloud Speech-to-Text) to convert the speech into text data. If the input is from a tablet, it is accepted as text data. Emotional data such as speech tone and speed are also acquired.

[0877] Output: Textualized question data and sentiment data.

[0878] Step 2: Emotion Recognition and Analysis

[0879] Input: Emotional data such as speech tone, rate, and facial expressions.

[0880] Processing: The emotion engine analyzes the acquired emotion data and identifies the operator's emotional state (e.g., impatience, tension, joy, etc.). Emotion analysis is performed using emotion-recognition cameras and microphones (e.g., Logitech Brio cameras, Shure microphones).

[0881] Output: Parsed emotion data.

[0882] Step 3: Question analysis and generative model selection

[0883] Input: Translated question data and parsed sentiment data.

[0884] Processing: The server analyzes the question content and the emotional state and selects an appropriate generative model (e.g., production data analysis model, machine maintenance model, quality control model). The selection criteria include the question content and the operator's emotional state.

[0885] Output: The selected generative model.

[0886] Step 4: Distributing questions

[0887] Input: Textualized question data, parsed sentiment data, and a selected generative model.

[0888] Processing: The server distributes the questions to the selected generative models. It sends the question data and emotion data to each generative model.

[0889] Output: Question data and sentiment data sent to the generative model.

[0890] Step 5: Generate an answer

[0891] Input: Question data and sentiment data sent to the generative model.

[0892] Processing: Each generative model (e.g., text generation model, production data analysis model, machine maintenance model, etc.) generates an answer based on the received question. For example, a text generation model creates an answer in Q&A format.

[0893] Output: The generated answers.

[0894] Step 6: Evaluate and select responses

[0895] Input: Multiple generated answers.

[0896] Processing: The server evaluates the answers from each generative model and selects the best answer based on evaluation criteria (accuracy, reliability, applicability, etc.), also taking into account the analysis results of the emotion engine.

[0897] Output: The best answer selected.

[0898] Step 7: Providing answers to users

[0899] Input: The selected best answer.

[0900] Processing: The server sends the selected optimal answer to the user terminal in a format that corresponds to the operator's emotional state. For example, it displays the answer in a friendly tone or in concise sentences.

[0901] Output: Answers provided in a sentiment-sensitive format.

[0902] 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.

[0903] 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.

[0904] 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.

[0905] [Third embodiment]

[0906] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0907] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0908] 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).

[0909] 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.

[0910] 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.

[0911] 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).

[0912] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0913] 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.

[0914] 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.

[0915] 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.

[0916] 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.

[0917] 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."

[0918] The present invention is a system that accepts a question from a user, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user. The program processing of this system is explained below in natural language.

[0919] System configuration

[0920] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[0921] Program processing flow

[0922] 1. Accepting user questions:

[0923] The user enters a question into the terminal interface.

[0924] The terminal transmits the question received from the user to the server.

[0925] 2. Question analysis and distribution:

[0926] The server analyzes the received question and selects an appropriate generative model (e.g., a text generation model or a data analysis model) based on its content.

[0927] The questions are distributed among a number of selected generative models.

[0928] 3. Generate answers:

[0929] Each generative model (a module within the server) generates an answer to the question it receives in its own way.

[0930] The generated answer is sent back to the server.

[0931] 4. Evaluating and selecting answers:

[0932] The server collects the answers returned by each generative model and evaluates them using an evaluation algorithm.

[0933] Select the most appropriate answer based on evaluation criteria (e.g., accuracy, reliability, applicability, etc.).

[0934] If necessary, generative models brainstorm and double-check with each other to produce reliable answers.

[0935] 5. Providing answers to users:

[0936] The server transmits the selected optimal answer to the user terminal.

[0937] The terminal displays the received answer to the user.

[0938] Specific examples

[0939] Question: "What is the market size for AI platforms?"

[0940] 1. Accepting user questions:

[0941] The user types into the device interface, "Please tell me about the market size of AI platforms."

[0942] The terminal transmits this input content to the server.

[0943] 2. Question analysis and distribution:

[0944] The server receives the question and analyzes it using natural language processing techniques.

[0945] The question is determined to be related to "market research" and "economic data," and is distributed to a text generation model, an economic data analysis model, and a trend prediction model.

[0946] 3. Generate answers:

[0947] The text generation model generates relevant market size forecast articles.

[0948] The economic data analysis model analyzes the current market size based on past market data.

[0949] The trend forecasting model generates data for predicting future market size.

[0950] Each model sends its answer back to the server.

[0951] 4. Evaluating and selecting answers:

[0952] The server evaluates and compares the answers from each model using an evaluation algorithm.

[0953] The text from the text generation model, the analysis results from the economic data analysis model, and the predicted values ​​from the trend prediction model are integrated to generate the optimal answer.

[0954] If necessary, the results of each generative model are checked again to select the final answer with the highest reliability.

[0955] 5. Providing answers to users:

[0956] The server sends the best answer to the user terminal.

[0957] The terminal will display the answer to the user in the following format:

[0958] The AI ​​platform market size is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026.

[0959] This system allows users to easily obtain reliable information and generates answers to questions from multiple perspectives by utilizing different generative models, which allows users to obtain more accurate and comprehensive information.

[0960] The processing flow will be explained below.

[0961] Step 1: Accepting user questions

[0962] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[0963] The device receives the entered question and sends it to the server along with metadata (question acceptance time, user ID, etc.).

[0964] Step 2: Question analysis and generative model selection

[0965] The server analyzes the received question and uses natural language processing algorithms to extract the question's topic and keywords.

[0966] Based on the analysis results, the server determines which generative model (text generation model, economic data analysis model, trend prediction model, etc.) is appropriate.

[0967] Step 3: Distributing questions

[0968] The server distributes the questions to the selected generative models, and sends the questions to each generative model depending on the analysis results.

[0969] Step 4: Generate an answer

[0970] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[0971] The economic data analysis model analyzes past market data and calculates data on the current market size.

[0972] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[0973] Each generative model sends its answer back to the server.

[0974] Step 5: Evaluate and select responses

[0975] The server receives the answers returned from each generative model.

[0976] The server evaluates each answer using a rating algorithm, with criteria including accuracy, reliability, and applicability.

[0977] The server selects the most appropriate answer, and if necessary, brainstorms with other generative models to generate a better answer.

[0978] Step 6: Providing answers to users

[0979] The server transmits the selected optimal answer to the user terminal.

[0980] The device displays the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[0981] Through these steps, users can easily obtain reliable information. By performing multifaceted analysis and evaluation, the system can compensate for weaknesses in the generative model and provide optimal answers.

[0982] Example 1

[0983] 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."

[0984] Conventional systems typically generate answers to user questions using a single generative AI model. However, there are limitations to the answers provided by a single model, making it difficult to provide multifaceted and accurate information. Another issue is that the reliability of answers cannot be guaranteed because cross-validation between models is not performed. This can reduce the accuracy and reliability of the information users receive, potentially leading to incorrect decision-making. This invention aims to solve these issues and provide users with more reliable information.

[0985] 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.

[0986] In this invention, the server includes means for accepting questions from users, means for analyzing the accepted questions and selecting generative models based on the analysis results, means for distributing the questions to the selected generative models, means for generating answers from each generative model, means for evaluating the generated answers using an evaluation algorithm, and means for double-checking the answers from each generative model. This enables the server to analyze the questions from multiple angles and mutually verify the answers from each generative model, thereby providing more reliable information to the user.

[0987] The "means for accepting a question from a user" refers to a means for receiving a question input by a user via the interface of a terminal.

[0988] The "means for analyzing the received question and selecting a generative model based on the analysis results" refers to the means by which the server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the content of the question.

[0989] The "means for distributing questions to a plurality of selected generative models" is a means for appropriately allocating questions to generative models selected from the analysis results.

[0990] "Means for generating an answer from each generative model" refers to means for each generative model to generate an answer based on a specific prompt sentence in response to a question received.

[0991] The "means for evaluating a plurality of generated answers using an evaluation algorithm" is a means for quantitatively or qualitatively evaluating the answers returned from each generative model.

[0992] "Means for double-checking answers from each generative model" refers to a means for cross-checking answers from multiple generative models to reinforce accuracy and reliability.

[0993] The "means for providing the selected optimal answer to the user" refers to a means for transmitting the most appropriate answer that has been evaluated and double-checked to the user terminal and presenting it to the user.

[0994] This invention is a system that accepts questions from users, distributes them to multiple generative AI models, has each generative AI model generate an answer, evaluates the multiple generated answers to select the optimal answer, and provides the selected answer to the user.

[0995] System configuration

[0996] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to appropriate generative AI models, has each generate an answer, evaluates these answers using an evaluation algorithm, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[0997] Hardware and software used

[0998] User device: A computer or smart device (e.g., PC, smartphone, tablet) with an interface for inputting user questions.

[0999] Server: A server machine with a high-performance processor (e.g., x86 server, cloud-based virtual machine).

[1000] Natural language processing engine: Software to parse the question (e.g. SpaCy, NLTK).

[1001] Generative AI models: Models that generate answers to questions (e.g., GPT-3, BERT, Prophet).

[1002] Evaluation algorithm: Software for evaluating the generated answers (e.g. TF-IDF, BERTScore).

[1003] Specific examples

[1004] Question: "What is the market size for AI platforms?"

[1005] 1. Accepting user questions:

[1006] The user types, "Please tell me about the market size of AI platforms" into the interface of the user device.

[1007] The terminal transmits this input content to the server.

[1008] 2. Question Analysis:

[1009] The server analyzes the received question using a natural language processing engine and extracts the question's topic and keywords (e.g., "AI platform," "market size").

[1010] 3. Generative model selection and distribution:

[1011] Based on the extracted keywords, the server selects a text generation model (GPT-3), an economic data analysis model (BERT), and a trend prediction model (Prophet).

[1012] Distribute questions to each generative AI model.

[1013] 4. Generate answers:

[1014] The text generation model (GPT-3) generates answers based on the prompt: "Please provide the latest data on the current market size of AI platforms."

[1015] The Economic Data Analysis Model (BERT) generates answers based on the prompt, "Analyze the current market size based on past market data."

[1016] The trend prediction model (Prophet) generates answers based on the prompt, "Please provide data to predict future market size."

[1017] 5. Evaluating and selecting answers:

[1018] The server evaluates the answers from each model using an evaluation algorithm (e.g., BERTScore) to score them for accuracy, reliability, and applicability.

[1019] The answer with the highest score or the best answer is selected by combining the results of each generative AI model.

[1020] 6. Providing answers to users:

[1021] The server sends the best answer to the user terminal.

[1022] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1023] This system allows users to quickly obtain reliable information, and it is capable of analyzing information from multiple angles using multiple generative AI models to provide highly accurate answers.

[1024] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1025] Step 1: Accepting user questions

[1026] The user inputs a question into the interface of the user device. For example, they might input, "Please tell me about the market size of AI platforms."

[1027] The terminal sends the question received from the user to the server. The input data is a text question, and is sent to the server using an HTTP request.

[1028] The output is the textual question data sent to the server.

[1029] Step 2: Parsing the Question

[1030] The server parses the received question using a natural language processing engine (e.g., SpaCy, NLTK). The input is the user's textual question.

[1031] The server tokenizes the question content and extracts important keywords (e.g., "AI platform" and "market size"), and breaks down the text data into words to identify nouns, verbs, etc.

[1032] The output is a list of analyzed keywords and topic information.

[1033] Step 3: Selecting a generative model and distributing questions

[1034] The server selects an appropriate generative AI model based on the extracted keywords and topic information, and prepares to deliver a specific prompt to the generative AI model.

[1035] The input is a keyword list and topic information.

[1036] The server generates appropriate prompts and distributes questions to multiple selected generative AI models (e.g., GPT-3, BERT, Prophet).

[1037] The output is multiple prompt sentences that are sent to the generative AI model.

[1038] Step 4: Generate an answer

[1039] Each generative AI model (GPT-3, BERT, Prophet) generates an answer based on a prompt received from the server. The input is the prompt sent to the generative AI model.

[1040] For example, GPT-3 receives the prompt "Please provide the latest data on the current market size of AI platforms" and generates relevant text, while Prophet receives the prompt "Please provide data to predict future market size" and generates predictive data.

[1041] The output is the answer data sent back to the server from each generative AI model.

[1042] Step 5: Evaluate responses

[1043] The server collects the answers returned by the generative AI model and evaluates them using an evaluation algorithm (e.g., TF-IDF, BERTScore). The input is multiple answer data.

[1044] The server scores answers based on criteria such as accuracy, reliability, and applicability, and compares answers between different models.

[1045] The output is the scoring results and the evaluated answer data.

[1046] Step 6: Selecting the best answer

[1047] The server selects the most appropriate answer based on the evaluation results, and if necessary, brainstorms and double-checks with other generative models.

[1048] The input is the evaluated response data.

[1049] The server determines the best answer and selects it as the final answer.

[1050] The output is the selected optimal answer data.

[1051] Step 7: Providing answers to users

[1052] The server transmits the selected optimal answer to the user terminal, and the input is the selected optimal answer data.

[1053] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1054] The output is the answer information displayed to the user.

[1055] (Application example 1)

[1056] 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."

[1057] In recent years, there has been a demand for faster and more efficient customer service in brick-and-mortar stores. However, conventional systems have difficulty in providing appropriate and prompt answers to customer questions, and it takes a long time to provide accurate answers, especially to complex questions. In addition, it is difficult to provide specific data such as inventory information immediately, which leads to a decrease in customer satisfaction.

[1058] 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.

[1059] In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, means for generating answers from each generative model, means for evaluating the generated answers and selecting an optimal answer, means for providing the selected optimal answer to the user, means for analyzing the user's question and selecting a generative model based on the analysis results, and means for making the server accessible via a smartphone application in a physical store environment. This allows users to easily input questions using their smartphones in the physical store and receive quick and accurate answers. Furthermore, specific data such as inventory information can be provided immediately, improving customer satisfaction.

[1060] The "means for accepting questions from the user" refers to an interface and mechanism for receiving questions entered by the user and processing the content of the questions appropriately.

[1061] The "means for distributing received questions to multiple generative models" is a system for analyzing received questions and distributing them to appropriate generative models based on the questions.

[1062] "Means for generating an answer from each generative model" refers to a process for generating an answer to a question using each generative model and returning the result to the server.

[1063] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for comparing answers obtained from multiple generative models using an evaluation algorithm and selecting the single optimal answer.

[1064] The "means for providing the selected optimal answer to the user" is an interface that transmits the selected optimal answer to the user terminal and presents it to the user.

[1065] "Means for analyzing a user's question and selecting a generative model based on the analysis results" refers to a mechanism that analyzes a user's question using natural language processing technology and determines the most appropriate generative model based on the results.

[1066] "Means for enabling access via smartphone applications within a brick-and-mortar store environment" refers to applications and infrastructure that enable access to the system and use of services using smartphones within a brick-and-mortar store.

[1067] "Means for double-checking answers from each generative model" is a process for reconfirming the accuracy and reliability of answers obtained from multiple generative models.

[1068] The "means for providing inventory information based on specific questions at a physical store" is a system that queries the inventory database of a physical store in response to a user's questions about inventory and provides an immediate answer.

[1069] The present invention is a system that provides quick and accurate answers to user questions in a physical store. This system accepts user questions, distributes them to multiple generative models, generates and evaluates answers from each generative model, and selects the optimal answer to provide to the user.

[1070] System configuration

[1071] The system mainly consists of a user device and a server. The user device uses a smartphone application to provide an interface for receiving questions from users and sending them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The selected answer is then provided to the user via the user device.

[1072] Hardware and software used

[1073] Hardware:

[1074] Smartphone

[1075] Servers (cloud-based servers or on-premise setups)

[1076] software:

[1077] Smartphone application (iOS / Android)

[1078] Server-side processing is based on Python

[1079] Natural language processing libraries (spaCy, NLTK)

[1080] Generative AI libraries (OpenAI GPT-3, BERT)

[1081] Data evaluation libraries (SciPy, NumPy)

[1082] Program processing

[1083] A user enters a question using a smartphone application: The user enters a question into the application interface. For example, the user enters "Do you have this item in stock?" This question is sent to the server.

[1084] Question analysis and distribution: The server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the question content. The question is then distributed to multiple selected generative models.

[1085] Each generative model generates an answer: Each generative model generates an answer in its own way to the question it receives. For example, an AI model for inventory checking queries an inventory database and generates an answer. The generated answer is sent back to the server.

[1086] Answer evaluation and selection: The server collects the answers returned by each generative model, evaluates and compares them using an evaluation algorithm (e.g., evaluation score), and selects the best answer as the final answer.

[1087] Providing an answer to the user: The server sends the selected optimal answer to the user's device. For example, the answer "This item is in stock" is displayed on the user's smartphone.

[1088] Specific examples

[1089] A user enters a question in a store, such as "Do you have this item in stock?" The application sends this question to a server, which analyzes the question and distributes it to a generative model for inventory confirmation. The generative model queries an inventory database and generates inventory information. The server evaluates and selects the most appropriate answer from the generated answers and provides it to the user as "In stock."

[1090] Example prompt sentence:

[1091] User Question: "Do you have this item in stock?"

[1092] Prompt for generative AI model: "The user is asking if this item is in stock. Please provide stock information for the following product: Product Name: XXXX"

[1093] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1094] Step 1:

[1095] A user inputs a question using a smartphone application. Specifically, the user inputs the question into the application interface and presses the submit button. The input at this time is the user's question text, and the output is request data to the server.

[1096] Step 2:

[1097] The device sends the question received from the user to the server. Specifically, the smartphone application sends the question text to the server as an HTTP POST request. The input is the user's question text, and the output is a request packet to the server.

[1098] Step 3:

[1099] The server analyzes the received question using a natural language processing library (e.g., spaCy, NLTK) to identify the intent and category of the question. The input is the user's question text, and the output is the analyzed question metadata (intent, category).

[1100] Step 4:

[1101] The server distributes the question to an appropriate generative model based on the analysis results. Specifically, it selects a generative model based on the question category (e.g., inventory check model, customer service model), and sends the question to the appropriate generative model. The input is the parsed question metadata, and the output is a request to multiple generative models.

[1102] Step 5:

[1103] Each generative model generates an answer based on a question. For example, an inventory check model queries an inventory database, and a customer service model generates answers from pre-trained data. The input is the question distributed to each generative model, and the output is the answer text from each generative model.

[1104] Step 6:

[1105] The server evaluates the answers returned by each generative model. Evaluation algorithms (e.g., rating scores, confidence scores) are used to compare the accuracy and reliability of each answer. The input is multiple answer texts from the generative models, and the output is the answer that is judged to be the best.

[1106] Step 7:

[1107] The server selects the optimal answer and sends it to the user's device. Specifically, it sends the selected answer as an HTTP response to the user's smartphone application. The input is the optimal answer text, and the output is a response packet to the user's device.

[1108] Step 8:

[1109] The user device displays the optimal answer received from the server to the user. Specifically, the answer text is displayed on the interface of the smartphone application. The input is the response data from the server, and the output is the answer text displayed to the user.

[1110] 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.

[1111] The present invention is a system that accepts a user's question, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user.Furthermore, by combining it with an emotion engine that recognizes and analyzes the user's emotions, it provides an appropriate answer according to the user's emotions.

[1112] System configuration

[1113] The system mainly consists of a user terminal, a server, and an emotion engine. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each generative model, and selects the optimal answer. The emotion engine recognizes emotions from the user's input and actions, and adjusts the answer based on the analysis results. The final answer is then provided to the user again via the user terminal.

[1114] Program processing flow

[1115] 1. Accepting user questions:

[1116] The user types a question into the device interface, for example, "What is the market size for AI platforms?"

[1117] The device receives the user's question and sends the input data to the server, and also sends the user's input and actions to the emotion engine.

[1118] 2. Emotion Recognition and Analysis:

[1119] The emotion engine recognizes emotions from user input and behavioral data and analyzes the results, specifically using text analysis, voice analysis, and eye tracking.

[1120] The emotion engine sends the analysis results to the server.

[1121] 3. Question analysis and generative model selection:

[1122] The server takes the analysis results from the emotion engine and takes them into account when analyzing the question.

[1123] The server selects an appropriate generation model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the question and the results of sentiment analysis.

[1124] 4. Question distribution:

[1125] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[1126] 5. Generate answers:

[1127] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[1128] The economic data analysis model analyzes past market data and calculates data on the current market size.

[1129] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[1130] Each generative model sends its answer back to the server.

[1131] 6. Evaluating and selecting answers:

[1132] The server receives the answers from each generative model and evaluates them using an evaluation algorithm, with criteria including accuracy, reliability, and applicability.

[1133] The server selects the most appropriate answer, taking into account the analysis results of the emotion engine.

[1134] If necessary, the generative models brainstorm with each other to generate better answers again.

[1135] 7. Providing answers to users:

[1136] The server sends the best answer to the user terminal.

[1137] The device will then display the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[1138] Specific examples

[1139] Question: "What is the market size for AI platforms?" (if the user is nervous)

[1140] 1. Accepting user questions:

[1141] A user types into the device interface, "Please tell me about the market size of AI platforms." Because the user is nervous, they type quickly and make many typos.

[1142] The terminal sends this input content to the emotion engine.

[1143] 2. Emotion Recognition and Analysis:

[1144] The emotion engine analyzes the user's input patterns and concludes that the user is nervous.

[1145] The emotion engine sends the analysis results to the server.

[1146] 3. Question analysis and generative model selection:

[1147] The server analyzes the question content and the sentiment analysis results, and selects an appropriate generative model to provide an answer in a format that the user can feel comfortable with.

[1148] 4. Question distribution:

[1149] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[1150] 5. Generate answers:

[1151] The text generation model generates simple and reassuring sentences for nervous users.

[1152] The economic data analysis model and trend forecasting model generate answers based on their respective roles.

[1153] Each generative model sends an answer back to the server.

[1154] 6. Evaluating and selecting answers:

[1155] The server evaluates the received answers and selects the most appropriate answer, taking into consideration, among other things, how it will ease the user's tension.

[1156] If necessary, make adjustments again.

[1157] 7. Providing answers to users:

[1158] The server sends the best answer to the terminal in a comfortable format.

[1159] The device displays the answer in a friendly tone to ease the tension.

[1160] This system allows users to obtain appropriate and reliable information according to their emotions. By combining it with an emotion engine, it is possible to further increase user satisfaction and provide more personalized services.

[1161] The processing flow will be explained below.

[1162] Step 1: Accepting user questions

[1163] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[1164] When a question is entered, the device collects the user's input patterns, voice, gaze data, etc., and sends this information to the server and emotion engine.

[1165] Step 2: Emotion Recognition and Analysis

[1166] The emotion engine analyzes the data sent from the device and recognizes the user's emotional state (tension, relief, excitement, etc.).

[1167] The emotion engine sends the results of the recognized emotional state to the server.

[1168] Step 3: Question analysis and generative model selection

[1169] The server analyzes the emotional state information and question it receives, and uses natural language processing technology to extract the topic and keywords of the question.

[1170] The server selects an appropriate generative model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the problem and the user's emotional state.

[1171] Step 4: Distributing questions

[1172] The server distributes questions to selected generative models, and sends them to each model based on the analysis and sentiment analysis results.

[1173] Step 5: Generate an answer

[1174] The text generation model receives a question and uses related text data to generate an answer, such as creating a commentary on market size forecasts for 2023.

[1175] The economic data analysis model analyzes past market data and calculates data on the current market size.

[1176] The trend forecasting model predicts future market size and generates a forecast value for market size three years from now.

[1177] Each model sends the answer it generates back to the server.

[1178] Step 6: Evaluate and select responses

[1179] The server receives the answers from each generative model and evaluates each answer using an evaluation algorithm, with evaluation criteria including accuracy, reliability, and applicability.

[1180] The server takes into account the analysis results of the emotion engine and selects the most appropriate answer that matches the user's emotional state.

[1181] If necessary, the generative models brainstorm with each other to generate better answers again.

[1182] Step 7: Providing answers to users

[1183] The server transmits the selected optimal answer to the user terminal.

[1184] The device will adjust the tone and wording of the response to match the user's emotional state and display it to the user. For example, if a user is feeling nervous, it will display simple, reassuring language such as, "The market size of AI platforms is predicted to reach X billion dollars in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach Z billion dollars in 2026."

[1185] Through this process, users can obtain highly reliable information tailored to their emotions. This system, combined with an emotion engine, will further improve user satisfaction and realize personalized services that allow for individual responses.

[1186] Example 2

[1187] 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."

[1188] Conventional question-answering systems often fail to provide appropriate answers because they struggle to fully consider the user's emotions and input intent. Furthermore, they lack a means to incorporate the results of sentiment analysis when selecting the optimal answer from multiple generative models. This creates a risk of lowering user satisfaction.

[1189] 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.

[1190] In this invention, the server includes: [means for accepting questions from a user]; [means for distributing the accepted questions and user emotion data to a plurality of generative models]; [means for generating answers from each generative model]; [means for evaluating the plurality of generated answers and selecting the optimum answer taking into account the emotion analysis results]; and [means for providing the selected optimum answer to the user.] This makes it possible to provide appropriate and reliable answers that take into account the user's emotions.

[1191] "Means for accepting questions from a user" refers to a function that receives questions entered by a user into the interface and sends the data to the server for processing.

[1192] "Means for distributing received questions and user emotion data to multiple generative models" refers to the function of distributing the user question and emotion data obtained from the emotion engine to appropriate generative models and using that data for analysis.

[1193] "Means for generating answers from each generative model" refers to the function by which each generative AI model generates answers based on the distributed question data.

[1194] "Means for evaluating multiple generated answers and selecting the most appropriate answer taking into account the results of sentiment analysis" refers to a function that compares multiple generated answers based on certain evaluation criteria and selects the most appropriate answer taking into account the results of sentiment analysis.

[1195] The "means for providing the selected optimal answer to the user" refers to a function for transmitting the selected answer to the user terminal and displaying it to the user.

[1196] "Emotion analysis results" are data that indicate the emotional state of the user analyzed by the emotion engine.

[1197] "Generative AI model" is a general term for artificial intelligence models that generate responses to given questions.

[1198] An "emotion engine" refers to a system that analyzes a user's input data and behavioral patterns to recognize and evaluate the user's emotional state.

[1199] The present invention is a system that accepts questions from users, generates optimal answers using a generative AI model, and provides answers that correspond to the user's emotions using an emotion engine. Specific embodiments of the system are described in detail below.

[1200] Hardware and software used

[1201] Hardware

[1202] User terminal: A device through which a user inputs a question, such as a PC or smartphone.

[1203] Server: A device that processes question data and emotion data and runs generative AI models.

[1204] software

[1205] Interface application: An application that is installed on a user's device and accepts input from the user.

[1206] Emotion engine: Software that recognizes emotions by analyzing user input data and behavioral patterns.

[1207] Generative AI model: Software that generates answers to given questions. Examples include text generation models, economic data analysis models, and trend forecasting models.

[1208] Data processing and calculation methods

[1209] 1. Accepting user questions:

[1210] The user types a question into the device interface, for example, "Please tell me about the market size of AI platforms."

[1211] The device sends this input data to the server, and at the same time sends emotion data such as input speed and typos to the emotion engine.

[1212] 2. Emotion Recognition and Analysis:

[1213] The emotion engine analyzes the received emotion data and recognizes the user's emotional state, such as whether they are tense or relaxed.

[1214] The emotion engine sends the analysis results to the server.

[1215] 3. Question analysis and generative model selection:

[1216] The server obtains the emotion analysis results from the emotion engine and analyzes the question content.

[1217] Based on the analysis results, the server selects the optimal generative AI model (text generation model, economic data analysis model, trend prediction model, etc.).

[1218] 4. Question distribution:

[1219] The server sends the question text and sentiment analysis results to the selected generative AI model.

[1220] Each generative AI model generates an answer based on the question data.

[1221] 5. Generate and evaluate answers:

[1222] Each generative AI model generates an answer based on the data it receives and sends the result back to the server.

[1223] The server evaluates answers from multiple generative models and selects the optimal answer taking into account the results of sentiment analysis.

[1224] 6. Providing answers to users:

[1225] The server transmits the selected optimal answer to the user terminal.

[1226] The device will display the answer in a format that is easy for the user to understand, such as, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1227] Specific examples

[1228] Specific questions

[1229] Question: "What is the market size for AI platforms?"

[1230] Specific examples of sentiment analysis

[1231] When a user is nervous, their typing speed slows down and they make more errors. The emotion engine analyzes this and recognizes that the user is nervous. Based on the results of this analysis, the server selects a generative model to reassure the user and generates an appropriate answer.

[1232] Based on these steps and specific examples, the system of the present invention can provide appropriate and reliable answers that correspond to the user's emotions, thereby increasing user satisfaction.

[1233] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1234] Step 1:

[1235] Accepting user questions

[1236] The user types into the device interface, "Please tell me about the market size of AI platforms."

[1237] Input: Text data that a user types into a device's interface.

[1238] The terminal receives a question text from the user and transmits the text data to the server.

[1239] Output: The action of sending the received question text data to the server.

[1240] Step 2:

[1241] Emotion Recognition and Analysis

[1242] The terminal also sends emotion data, such as the user's input speed and typo error patterns, to the emotion engine in parallel.

[1243] Input: Emotional data such as user typing speed, typos, etc.

[1244] The emotion engine analyzes the received emotion data and recognizes the user's emotional state (e.g., tense, relaxed).

[1245] Output: Send the result of the emotion analysis (e.g., the user is nervous) to the server.

[1246] Step 3:

[1247] Question analysis and generative model selection

[1248] The server analyzes the emotion analysis results obtained from the emotion engine and the user's question text.

[1249] Input: User question text and sentiment analysis results.

[1250] The server analyzes the question and selects a corresponding generative AI model (e.g., text generation model, economic data analysis model, trend prediction model).

[1251] Output: Question text and sentiment analysis results are distributed to each generative AI model.

[1252] Step 4:

[1253] Question distribution

[1254] The server distributes the question text to the selected generative AI models (text generation model, economic data analysis model, trend prediction model) and also attaches the sentiment analysis results.

[1255] Input: Question text and sentiment analysis results.

[1256] Output: The question and sentiment analysis results are sent to each generative AI model.

[1257] Step 5:

[1258] Generate answers

[1259] The text generation model generates an appropriate answer based on the question it receives, for example, "The market size of AI platforms in 2023 will be $X billion."

[1260] The economic data analysis model analyzes historical market data and generates data on the current market size.

[1261] The trend forecasting model predicts future market trends and generates forecast data for the market size for the next three years.

[1262] Input: Question text and sentiment analysis results.

[1263] Output: The answer data generated by each generative AI model.

[1264] Step 6:

[1265] Evaluation and selection of answers

[1266] The server aggregates the response data received from each generative AI model and selects the optimal response based on an evaluation algorithm.

[1267] Input: Answer data from each generative AI model.

[1268] The evaluation algorithm includes accuracy, reliability, applicability, etc., and also takes into account sentiment analysis results.

[1269] If necessary, the generative models brainstorm with each other to generate better answers.

[1270] Output: The best answer.

[1271] Step 7:

[1272] Providing answers to users

[1273] The server transmits the selected optimal answer to the user terminal.

[1274] Input: Best answer.

[1275] The device will then display the answer it receives to the user in a friendly and easy-to-understand format, such as "The AI ​​platform market size is predicted to reach $X in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1276] Output: Display the best answer to the user.

[1277] (Application example 2)

[1278] 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."

[1279] Current factory robots respond to questions or instructions from operators by providing uniform answers without considering the operator's emotional state, which can lead to reduced usability and difficulty in effective communication. There is a need for a method to facilitate smooth communication between humans and robots, especially when appropriate responses are required under emotional circumstances.

[1280] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, and means for generating answers from each generative model. This makes it possible to evaluate multiple generated answers, select the optimal answer, and provide the optimal answer to the operator in a format that corresponds to the operator's emotion. In addition, by providing means for analyzing the operator's emotional state using an emotion recognition engine and taking the operator's emotional state into consideration when selecting a generative model, a more appropriate and human-like response is possible.

[1281] The "means for accepting questions from users" is an interface that allows an operator to input questions or instructions to a factory robot, which then recognizes them and receives them as data.

[1282] "Means for distributing received questions to multiple generative models" is a function for analyzing received questions and distributing them to appropriate generative models.

[1283] The "means for generating an answer from each generative model" is a system for generating a corresponding answer based on a question received by the generative model.

[1284] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for evaluating answers provided by multiple generative models and selecting the most appropriate one.

[1285] The "means for providing the operator with the optimal answer in a format that corresponds to their emotions" is a system for providing the selected optimal answer in a format that is appropriate to the operator's emotional state.

[1286] "Means for analyzing an operator's emotional state using an emotion recognition engine" refers to a technology for analyzing an operator's tone of voice, speed, facial expressions, etc. to analyze their emotional state.

[1287] "Means for considering the emotional state of the operator when selecting a generative model" is a function that uses the emotional state of the operator as an important factor in the process of selecting a generative model.

[1288] This invention is a system that enables a factory robot to provide appropriate answers to questions and instructions from an operator. The system is mainly composed of a user terminal, a server, and an emotion engine.

[1289] System configuration

[1290] 1. User Device

[1291] Operators can input questions and instructions to the robot using voice commands or a tablet, which uses Google Cloud Speech-to-Text as its voice recognition system and is rugged and suitable for factory environments.

[1292] The user terminal receives the input question and sends the data to the server. It also sends emotional data such as the operator's voice tone, speed, and facial expressions to the emotion engine.

[1293] 2. Server

[1294] The server has a means for analyzing the query content and distributing it to multiple generative models (e.g., production data analysis model, machine maintenance model, quality control model, etc.).

[1295] The generative model generates corresponding answers by utilizing a text generation model (e.g., OpenAI's GPT-3), a production data analysis model (e.g., Azure Machine Learning), etc.

[1296] The server evaluates the generated answers and selects the best one based on accuracy, reliability, and applicability, taking into account the analysis results of the emotion engine.

[1297] After the optimal answer is selected, it is provided to the user terminal in a format that corresponds to the emotional state of the operator.

[1298] 3. Emotion Engine

[1299] The emotion engine is a technology that analyzes the operator's emotional state by analyzing their voice tone, speed, facial expressions, etc. Emotion analysis uses emotion-recognition cameras and microphones (e.g., Logitech Brio camera, Shure microphone).

[1300] The analysis results are sent to the server and reflected in the generative model selection process.

[1301] Specific examples

[1302] Let's assume that a user asks, "Please tell me the current production efficiency," and the operator becomes impatient.

[1303] 1. The user terminal converts the operator's question into text using a voice recognition system and sends the text to the server.

[1304] 2. The emotion engine analyzes the operator's tone of voice, speed, and facial expressions and determines that the operator is impatient.

[1305] 3. The server selects a generative model based on the question content and the results of sentiment analysis. For example, a production data analysis model is selected.

[1306] 4. The generative model generates an answer like this: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for the next production line check."

[1307] 5. The server evaluates the generated answers and selects the best answer.

[1308] 6. The user terminal provides this response to the operator.

[1309] Prompt Sentence Examples

[1310] User: "What is your current production efficiency?"

[1311] User Emotion: Impatience

[1312] Generative model: Production data analysis model

[1313] Sample response: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for our next production line check."

[1314] This invention allows factory robots to support operators in a more efficient and human-like manner, and by using an emotion engine, it is possible to provide appropriate and reliable information according to the operator's emotional state.

[1315] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1316] Step 1: Accepting user questions

[1317] Input: The operator inputs the question using voice or a tablet.

[1318] Processing: The user device uses a speech recognition system (e.g., Google Cloud Speech-to-Text) to convert the speech into text data. If the input is from a tablet, it is accepted as text data. Emotional data such as speech tone and speed are also acquired.

[1319] Output: Textualized question data and sentiment data.

[1320] Step 2: Emotion Recognition and Analysis

[1321] Input: Emotional data such as speech tone, rate, and facial expressions.

[1322] Processing: The emotion engine analyzes the acquired emotion data and identifies the operator's emotional state (e.g., impatience, tension, joy, etc.). Emotion analysis is performed using emotion-recognition cameras and microphones (e.g., Logitech Brio cameras, Shure microphones).

[1323] Output: Parsed emotion data.

[1324] Step 3: Question analysis and generative model selection

[1325] Input: Translated question data and parsed sentiment data.

[1326] Processing: The server analyzes the question content and the emotional state and selects an appropriate generative model (e.g., production data analysis model, machine maintenance model, quality control model). The selection criteria include the question content and the operator's emotional state.

[1327] Output: The selected generative model.

[1328] Step 4: Distributing questions

[1329] Input: Textualized question data, parsed sentiment data, and a selected generative model.

[1330] Processing: The server distributes the questions to the selected generative models. It sends the question data and emotion data to each generative model.

[1331] Output: Question data and sentiment data sent to the generative model.

[1332] Step 5: Generate an answer

[1333] Input: Question data and sentiment data sent to the generative model.

[1334] Processing: Each generative model (e.g., text generation model, production data analysis model, machine maintenance model, etc.) generates an answer based on the received question. For example, a text generation model creates an answer in Q&A format.

[1335] Output: The generated answers.

[1336] Step 6: Evaluate and select responses

[1337] Input: Multiple generated answers.

[1338] Processing: The server evaluates the answers from each generative model and selects the best answer based on evaluation criteria (accuracy, reliability, applicability, etc.), also taking into account the analysis results of the emotion engine.

[1339] Output: The best answer selected.

[1340] Step 7: Providing answers to users

[1341] Input: The selected best answer.

[1342] Processing: The server sends the selected optimal answer to the user terminal in a format that corresponds to the operator's emotional state. For example, it displays the answer in a friendly tone or in concise sentences.

[1343] Output: Answers provided in a sentiment-sensitive format.

[1344] 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.

[1345] 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.

[1346] 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.

[1347] [Fourth embodiment]

[1348] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1349] 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.

[1350] 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).

[1351] 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.

[1352] 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.

[1353] 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).

[1354] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1355] 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.

[1356] 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.

[1357] 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.

[1358] 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.

[1359] 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.

[1360] 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."

[1361] The present invention is a system that accepts a question from a user, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user. The program processing of this system is explained below in natural language.

[1362] System configuration

[1363] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[1364] Program processing flow

[1365] 1. Accepting user questions:

[1366] The user enters a question into the terminal interface.

[1367] The terminal transmits the question received from the user to the server.

[1368] 2. Question analysis and distribution:

[1369] The server analyzes the received question and selects an appropriate generative model (e.g., a text generation model or a data analysis model) based on its content.

[1370] The questions are distributed among a number of selected generative models.

[1371] 3. Generate answers:

[1372] Each generative model (a module within the server) generates an answer to the question it receives in its own way.

[1373] The generated answer is sent back to the server.

[1374] 4. Evaluating and selecting answers:

[1375] The server collects the answers returned by each generative model and evaluates them using an evaluation algorithm.

[1376] Select the most appropriate answer based on evaluation criteria (e.g., accuracy, reliability, applicability, etc.).

[1377] If necessary, generative models brainstorm and double-check with each other to produce reliable answers.

[1378] 5. Providing answers to users:

[1379] The server transmits the selected optimal answer to the user terminal.

[1380] The terminal displays the received answer to the user.

[1381] Specific examples

[1382] Question: "What is the market size for AI platforms?"

[1383] 1. Accepting user questions:

[1384] The user types into the device interface, "Please tell me about the market size of AI platforms."

[1385] The terminal transmits this input content to the server.

[1386] 2. Question analysis and distribution:

[1387] The server receives the question and analyzes it using natural language processing techniques.

[1388] The question is determined to be related to "market research" and "economic data," and is distributed to a text generation model, an economic data analysis model, and a trend prediction model.

[1389] 3. Generate answers:

[1390] The text generation model generates relevant market size forecast articles.

[1391] The economic data analysis model analyzes the current market size based on past market data.

[1392] The trend forecasting model generates data for predicting future market size.

[1393] Each model sends its answer back to the server.

[1394] 4. Evaluating and selecting answers:

[1395] The server evaluates and compares the answers from each model using an evaluation algorithm.

[1396] The text from the text generation model, the analysis results from the economic data analysis model, and the predicted values ​​from the trend prediction model are integrated to generate the optimal answer.

[1397] If necessary, the results of each generative model are checked again to select the final answer with the highest reliability.

[1398] 5. Providing answers to users:

[1399] The server sends the best answer to the user terminal.

[1400] The terminal will display the answer to the user in the following format:

[1401] The AI ​​platform market size is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026.

[1402] This system allows users to easily obtain reliable information and generates answers to questions from multiple perspectives by utilizing different generative models, which allows users to obtain more accurate and comprehensive information.

[1403] The processing flow will be explained below.

[1404] Step 1: Accepting user questions

[1405] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[1406] The device receives the entered question and sends it to the server along with metadata (question acceptance time, user ID, etc.).

[1407] Step 2: Question analysis and generative model selection

[1408] The server analyzes the received question and uses natural language processing algorithms to extract the question's topic and keywords.

[1409] Based on the analysis results, the server determines which generative model (text generation model, economic data analysis model, trend prediction model, etc.) is appropriate.

[1410] Step 3: Distributing questions

[1411] The server distributes the questions to the selected generative models, and sends the questions to each generative model depending on the analysis results.

[1412] Step 4: Generate an answer

[1413] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[1414] The economic data analysis model analyzes past market data and calculates data on the current market size.

[1415] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[1416] Each generative model sends its answer back to the server.

[1417] Step 5: Evaluate and select responses

[1418] The server receives the answers returned from each generative model.

[1419] The server evaluates each answer using a rating algorithm, with criteria including accuracy, reliability, and applicability.

[1420] The server selects the most appropriate answer, and if necessary, brainstorms with other generative models to generate a better answer.

[1421] Step 6: Providing answers to users

[1422] The server transmits the selected optimal answer to the user terminal.

[1423] The device displays the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[1424] Through these steps, users can easily obtain reliable information. By performing multifaceted analysis and evaluation, the system can compensate for weaknesses in the generative model and provide optimal answers.

[1425] Example 1

[1426] 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."

[1427] Conventional systems typically generate answers to user questions using a single generative AI model. However, there are limitations to the answers provided by a single model, making it difficult to provide multifaceted and accurate information. Another issue is that the reliability of answers cannot be guaranteed because cross-validation between models is not performed. This can reduce the accuracy and reliability of the information users receive, potentially leading to incorrect decision-making. This invention aims to solve these issues and provide users with more reliable information.

[1428] 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.

[1429] In this invention, the server includes means for accepting questions from users, means for analyzing the accepted questions and selecting generative models based on the analysis results, means for distributing the questions to the selected generative models, means for generating answers from each generative model, means for evaluating the generated answers using an evaluation algorithm, and means for double-checking the answers from each generative model. This enables the server to analyze the questions from multiple angles and mutually verify the answers from each generative model, thereby providing more reliable information to the user.

[1430] The "means for accepting a question from a user" refers to a means for receiving a question input by a user via the interface of a terminal.

[1431] The "means for analyzing the received question and selecting a generative model based on the analysis results" refers to the means by which the server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the content of the question.

[1432] The "means for distributing questions to a plurality of selected generative models" is a means for appropriately allocating questions to generative models selected from the analysis results.

[1433] "Means for generating an answer from each generative model" refers to means for each generative model to generate an answer based on a specific prompt sentence in response to a question received.

[1434] The "means for evaluating a plurality of generated answers using an evaluation algorithm" is a means for quantitatively or qualitatively evaluating the answers returned from each generative model.

[1435] "Means for double-checking answers from each generative model" refers to a means for cross-checking answers from multiple generative models to reinforce accuracy and reliability.

[1436] The "means for providing the selected optimal answer to the user" refers to a means for transmitting the most appropriate answer that has been evaluated and double-checked to the user terminal and presenting it to the user.

[1437] This invention is a system that accepts questions from users, distributes them to multiple generative AI models, has each generative AI model generate an answer, evaluates the multiple generated answers to select the optimal answer, and provides the selected answer to the user.

[1438] System configuration

[1439] The system mainly consists of a user terminal and a server. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to appropriate generative AI models, has each generate an answer, evaluates these answers using an evaluation algorithm, and selects the optimal answer. The final answer is then provided to the user via the user terminal.

[1440] Hardware and software used

[1441] User device: A computer or smart device (e.g., PC, smartphone, tablet) with an interface for inputting user questions.

[1442] Server: A server machine with a high-performance processor (e.g., x86 server, cloud-based virtual machine).

[1443] Natural language processing engine: Software to parse the question (e.g. SpaCy, NLTK).

[1444] Generative AI models: Models that generate answers to questions (e.g., GPT-3, BERT, Prophet).

[1445] Evaluation algorithm: Software for evaluating the generated answers (e.g. TF-IDF, BERTScore).

[1446] Specific examples

[1447] Question: "What is the market size for AI platforms?"

[1448] 1. Accepting user questions:

[1449] The user types, "Please tell me about the market size of AI platforms" into the interface of the user device.

[1450] The terminal transmits this input content to the server.

[1451] 2. Question Analysis:

[1452] The server analyzes the received question using a natural language processing engine and extracts the question's topic and keywords (e.g., "AI platform," "market size").

[1453] 3. Generative model selection and distribution:

[1454] Based on the extracted keywords, the server selects a text generation model (GPT-3), an economic data analysis model (BERT), and a trend prediction model (Prophet).

[1455] Distribute questions to each generative AI model.

[1456] 4. Generate answers:

[1457] The text generation model (GPT-3) generates answers based on the prompt: "Please provide the latest data on the current market size of AI platforms."

[1458] The Economic Data Analysis Model (BERT) generates answers based on the prompt, "Analyze the current market size based on past market data."

[1459] The trend prediction model (Prophet) generates answers based on the prompt, "Please provide data to predict future market size."

[1460] 5. Evaluating and selecting answers:

[1461] The server evaluates the answers from each model using an evaluation algorithm (e.g., BERTScore) to score them for accuracy, reliability, and applicability.

[1462] The answer with the highest score or the best answer is selected by combining the results of each generative AI model.

[1463] 6. Providing answers to users:

[1464] The server sends the best answer to the user terminal.

[1465] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1466] This system allows users to quickly obtain reliable information, and it is capable of analyzing information from multiple angles using multiple generative AI models to provide highly accurate answers.

[1467] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1468] Step 1: Accepting user questions

[1469] The user inputs a question into the interface of the user device. For example, they might input, "Please tell me about the market size of AI platforms."

[1470] The terminal sends the question received from the user to the server. The input data is a text question, and is sent to the server using an HTTP request.

[1471] The output is the textual question data sent to the server.

[1472] Step 2: Parsing the Question

[1473] The server parses the received question using a natural language processing engine (e.g., SpaCy, NLTK). The input is the user's textual question.

[1474] The server tokenizes the question content and extracts important keywords (e.g., "AI platform" and "market size"), and breaks down the text data into words to identify nouns, verbs, etc.

[1475] The output is a list of analyzed keywords and topic information.

[1476] Step 3: Selecting a generative model and distributing questions

[1477] The server selects an appropriate generative AI model based on the extracted keywords and topic information, and prepares to deliver a specific prompt to the generative AI model.

[1478] The input is a keyword list and topic information.

[1479] The server generates appropriate prompts and distributes questions to multiple selected generative AI models (e.g., GPT-3, BERT, Prophet).

[1480] The output is multiple prompt sentences that are sent to the generative AI model.

[1481] Step 4: Generate an answer

[1482] Each generative AI model (GPT-3, BERT, Prophet) generates an answer based on a prompt received from the server. The input is the prompt sent to the generative AI model.

[1483] For example, GPT-3 receives the prompt "Please provide the latest data on the current market size of AI platforms" and generates relevant text, while Prophet receives the prompt "Please provide data to predict future market size" and generates predictive data.

[1484] The output is the answer data sent back to the server from each generative AI model.

[1485] Step 5: Evaluate responses

[1486] The server collects the answers returned by the generative AI model and evaluates them using an evaluation algorithm (e.g., TF-IDF, BERTScore). The input is multiple answer data.

[1487] The server scores answers based on criteria such as accuracy, reliability, and applicability, and compares answers between different models.

[1488] The output is the scoring results and the evaluated answer data.

[1489] Step 6: Selecting the best answer

[1490] The server selects the most appropriate answer based on the evaluation results, and if necessary, brainstorms and double-checks with other generative models.

[1491] The input is the evaluated response data.

[1492] The server determines the best answer and selects it as the final answer.

[1493] The output is the selected optimal answer data.

[1494] Step 7: Providing answers to users

[1495] The server transmits the selected optimal answer to the user terminal, and the input is the selected optimal answer data.

[1496] The device will then display the received answer to the user, for example, in the form of "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1497] The output is the answer information displayed to the user.

[1498] (Application example 1)

[1499] 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."

[1500] In recent years, there has been a demand for faster and more efficient customer service in brick-and-mortar stores. However, conventional systems have difficulty in providing appropriate and prompt answers to customer questions, and it takes a long time to provide accurate answers, especially to complex questions. In addition, it is difficult to provide specific data such as inventory information immediately, which leads to a decrease in customer satisfaction.

[1501] 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.

[1502] In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, means for generating answers from each generative model, means for evaluating the generated answers and selecting an optimal answer, means for providing the selected optimal answer to the user, means for analyzing the user's question and selecting a generative model based on the analysis results, and means for making the server accessible via a smartphone application in a physical store environment. This allows users to easily input questions using their smartphones in the physical store and receive quick and accurate answers. Furthermore, specific data such as inventory information can be provided immediately, improving customer satisfaction.

[1503] The "means for accepting questions from the user" refers to an interface and mechanism for receiving questions entered by the user and processing the content of the questions appropriately.

[1504] The "means for distributing received questions to multiple generative models" is a system for analyzing received questions and distributing them to appropriate generative models based on the questions.

[1505] "Means for generating an answer from each generative model" refers to a process for generating an answer to a question using each generative model and returning the result to the server.

[1506] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for comparing answers obtained from multiple generative models using an evaluation algorithm and selecting the single optimal answer.

[1507] The "means for providing the selected optimal answer to the user" is an interface that transmits the selected optimal answer to the user terminal and presents it to the user.

[1508] "Means for analyzing a user's question and selecting a generative model based on the analysis results" refers to a mechanism that analyzes a user's question using natural language processing technology and determines the most appropriate generative model based on the results.

[1509] "Means for enabling access via smartphone applications within a brick-and-mortar store environment" refers to applications and infrastructure that enable access to the system and use of services using smartphones within a brick-and-mortar store.

[1510] "Means for double-checking answers from each generative model" is a process for reconfirming the accuracy and reliability of answers obtained from multiple generative models.

[1511] The "means for providing inventory information based on specific questions at a physical store" is a system that queries the inventory database of a physical store in response to a user's questions about inventory and provides an immediate answer.

[1512] The present invention is a system that provides quick and accurate answers to user questions in a physical store. This system accepts user questions, distributes them to multiple generative models, generates and evaluates answers from each generative model, and selects the optimal answer to provide to the user.

[1513] System configuration

[1514] The system mainly consists of a user device and a server. The user device uses a smartphone application to provide an interface for receiving questions from users and sending them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each model, and selects the optimal answer. The selected answer is then provided to the user via the user device.

[1515] Hardware and software used

[1516] Hardware:

[1517] Smartphone

[1518] Servers (cloud-based servers or on-premise setups)

[1519] software:

[1520] Smartphone application (iOS / Android)

[1521] Server-side processing is based on Python

[1522] Natural language processing libraries (spaCy, NLTK)

[1523] Generative AI libraries (OpenAI GPT-3, BERT)

[1524] Data evaluation libraries (SciPy, NumPy)

[1525] Program processing

[1526] A user enters a question using a smartphone application: The user enters a question into the application interface. For example, the user enters "Do you have this item in stock?" This question is sent to the server.

[1527] Question analysis and distribution: The server analyzes the received question using natural language processing technology and selects an appropriate generative model based on the question content. The question is then distributed to multiple selected generative models.

[1528] Each generative model generates an answer: Each generative model generates an answer in its own way to the question it receives. For example, an AI model for inventory checking queries an inventory database and generates an answer. The generated answer is sent back to the server.

[1529] Answer evaluation and selection: The server collects the answers returned by each generative model, evaluates and compares them using an evaluation algorithm (e.g., evaluation score), and selects the best answer as the final answer.

[1530] Providing an answer to the user: The server sends the selected optimal answer to the user's device. For example, the answer "This item is in stock" is displayed on the user's smartphone.

[1531] Specific examples

[1532] A user enters a question in a store, such as "Do you have this item in stock?" The application sends this question to a server, which analyzes the question and distributes it to a generative model for inventory confirmation. The generative model queries an inventory database and generates inventory information. The server evaluates and selects the most appropriate answer from the generated answers and provides it to the user as "In stock."

[1533] Example prompt sentence:

[1534] User Question: "Do you have this item in stock?"

[1535] Prompt for generative AI model: "The user is asking if this item is in stock. Please provide stock information for the following product: Product Name: XXXX"

[1536] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1537] Step 1:

[1538] A user inputs a question using a smartphone application. Specifically, the user inputs the question into the application interface and presses the submit button. The input at this time is the user's question text, and the output is request data to the server.

[1539] Step 2:

[1540] The device sends the question received from the user to the server. Specifically, the smartphone application sends the question text to the server as an HTTP POST request. The input is the user's question text, and the output is a request packet to the server.

[1541] Step 3:

[1542] The server analyzes the received question using a natural language processing library (e.g., spaCy, NLTK) to identify the intent and category of the question. The input is the user's question text, and the output is the analyzed question metadata (intent, category).

[1543] Step 4:

[1544] The server distributes the question to an appropriate generative model based on the analysis results. Specifically, it selects a generative model based on the question category (e.g., inventory check model, customer service model), and sends the question to the appropriate generative model. The input is the parsed question metadata, and the output is a request to multiple generative models.

[1545] Step 5:

[1546] Each generative model generates an answer based on a question. For example, an inventory check model queries an inventory database, and a customer service model generates answers from pre-trained data. The input is the question distributed to each generative model, and the output is the answer text from each generative model.

[1547] Step 6:

[1548] The server evaluates the answers returned by each generative model. Evaluation algorithms (e.g., rating scores, confidence scores) are used to compare the accuracy and reliability of each answer. The input is multiple answer texts from the generative models, and the output is the answer that is judged to be the best.

[1549] Step 7:

[1550] The server selects the optimal answer and sends it to the user's device. Specifically, it sends the selected answer as an HTTP response to the user's smartphone application. The input is the optimal answer text, and the output is a response packet to the user's device.

[1551] Step 8:

[1552] The user device displays the optimal answer received from the server to the user. Specifically, the answer text is displayed on the interface of the smartphone application. The input is the response data from the server, and the output is the answer text displayed to the user.

[1553] 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.

[1554] The present invention is a system that accepts a user's question, distributes it to multiple generative models, generates an answer from each of the generative models, evaluates the generated answers to select the optimal answer, and provides the selected answer to the user.Furthermore, by combining it with an emotion engine that recognizes and analyzes the user's emotions, it provides an appropriate answer according to the user's emotions.

[1555] System configuration

[1556] The system mainly consists of a user terminal, a server, and an emotion engine. The user terminal provides an interface that receives questions from users and sends them to the server. The server analyzes the questions, distributes them to multiple generative models, aggregates and evaluates the answers from each generative model, and selects the optimal answer. The emotion engine recognizes emotions from the user's input and actions, and adjusts the answer based on the analysis results. The final answer is then provided to the user again via the user terminal.

[1557] Program processing flow

[1558] 1. Accepting user questions:

[1559] The user types a question into the device interface, for example, "What is the market size for AI platforms?"

[1560] The device receives the user's question and sends the input data to the server, and also sends the user's input and actions to the emotion engine.

[1561] 2. Emotion Recognition and Analysis:

[1562] The emotion engine recognizes emotions from user input and behavioral data and analyzes the results, specifically using text analysis, voice analysis, and eye tracking.

[1563] The emotion engine sends the analysis results to the server.

[1564] 3. Question analysis and generative model selection:

[1565] The server takes the analysis results from the emotion engine and takes them into account when analyzing the question.

[1566] The server selects an appropriate generation model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the question and the results of sentiment analysis.

[1567] 4. Question distribution:

[1568] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[1569] 5. Generate answers:

[1570] The text generation model generates relevant sentences based on the questions it receives, for example, creating text about market size forecasts for 2023.

[1571] The economic data analysis model analyzes past market data and calculates data on the current market size.

[1572] The trend forecasting model predicts future market trends and generates a forecast of market size three years from now.

[1573] Each generative model sends its answer back to the server.

[1574] 6. Evaluating and selecting answers:

[1575] The server receives the answers from each generative model and evaluates them using an evaluation algorithm, with criteria including accuracy, reliability, and applicability.

[1576] The server selects the most appropriate answer, taking into account the analysis results of the emotion engine.

[1577] If necessary, the generative models brainstorm with each other to generate better answers again.

[1578] 7. Providing answers to users:

[1579] The server sends the best answer to the user terminal.

[1580] The device will then display the received answer to the user, for example, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach $Z billion in 2026."

[1581] Specific examples

[1582] Question: "What is the market size for AI platforms?" (if the user is nervous)

[1583] 1. Accepting user questions:

[1584] A user types into the device interface, "Please tell me about the market size of AI platforms." Because the user is nervous, they type quickly and make many typos.

[1585] The terminal sends this input content to the emotion engine.

[1586] 2. Emotion Recognition and Analysis:

[1587] The emotion engine analyzes the user's input patterns and concludes that the user is nervous.

[1588] The emotion engine sends the analysis results to the server.

[1589] 3. Question analysis and generative model selection:

[1590] The server analyzes the question content and the sentiment analysis results, and selects an appropriate generative model to provide an answer in a format that the user can feel comfortable with.

[1591] 4. Question distribution:

[1592] The server distributes the questions to the selected generative models, and sends the analysis results and sentiment analysis results to each of them.

[1593] 5. Generate answers:

[1594] The text generation model generates simple and reassuring sentences for nervous users.

[1595] The economic data analysis model and trend forecasting model generate answers based on their respective roles.

[1596] Each generative model sends an answer back to the server.

[1597] 6. Evaluating and selecting answers:

[1598] The server evaluates the received answers and selects the most appropriate answer, taking into consideration, among other things, how it will ease the user's tension.

[1599] If necessary, make adjustments again.

[1600] 7. Providing answers to users:

[1601] The server sends the best answer to the terminal in a comfortable format.

[1602] The device displays the answer in a friendly tone to ease the tension.

[1603] This system allows users to obtain appropriate and reliable information according to their emotions. By combining it with an emotion engine, it is possible to further increase user satisfaction and provide more personalized services.

[1604] The processing flow will be explained below.

[1605] Step 1: Accepting user questions

[1606] The user inputs a question through the device interface. For example, "Please tell me the market size of AI platforms."

[1607] When a question is entered, the device collects the user's input patterns, voice, gaze data, etc., and sends this information to the server and emotion engine.

[1608] Step 2: Emotion Recognition and Analysis

[1609] The emotion engine analyzes the data sent from the device and recognizes the user's emotional state (tension, relief, excitement, etc.).

[1610] The emotion engine sends the results of the recognized emotional state to the server.

[1611] Step 3: Question analysis and generative model selection

[1612] The server analyzes the emotional state information and question it receives, and uses natural language processing technology to extract the topic and keywords of the question.

[1613] The server selects an appropriate generative model (such as a text generation model, economic data analysis model, or trend prediction model) based on the content of the problem and the user's emotional state.

[1614] Step 4: Distributing questions

[1615] The server distributes questions to selected generative models, and sends them to each model based on the analysis and sentiment analysis results.

[1616] Step 5: Generate an answer

[1617] The text generation model receives a question and uses related text data to generate an answer, such as creating a commentary on market size forecasts for 2023.

[1618] The economic data analysis model analyzes past market data and calculates data on the current market size.

[1619] The trend forecasting model predicts future market size and generates a forecast value for market size three years from now.

[1620] Each model sends the answer it generates back to the server.

[1621] Step 6: Evaluate and select responses

[1622] The server receives the answers from each generative model and evaluates each answer using an evaluation algorithm, with evaluation criteria including accuracy, reliability, and applicability.

[1623] The server takes into account the analysis results of the emotion engine and selects the most appropriate answer that matches the user's emotional state.

[1624] If necessary, the generative models brainstorm with each other to generate better answers again.

[1625] Step 7: Providing answers to users

[1626] The server transmits the selected optimal answer to the user terminal.

[1627] The device will adjust the tone and wording of the response to match the user's emotional state and display it to the user. For example, if a user is feeling nervous, it will display simple, reassuring language such as, "The market size of AI platforms is predicted to reach X billion dollars in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years. The market size is predicted to reach Z billion dollars in 2026."

[1628] Through this process, users can obtain highly reliable information tailored to their emotions. This system, combined with an emotion engine, will further improve user satisfaction and realize personalized services that allow for individual responses.

[1629] Example 2

[1630] 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."

[1631] Conventional question-answering systems often fail to provide appropriate answers because they struggle to fully consider the user's emotions and input intent. Furthermore, they lack a means to incorporate the results of sentiment analysis when selecting the optimal answer from multiple generative models. This creates a risk of lowering user satisfaction.

[1632] 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.

[1633] In this invention, the server includes: [means for accepting questions from a user]; [means for distributing the accepted questions and user emotion data to a plurality of generative models]; [means for generating answers from each generative model]; [means for evaluating the plurality of generated answers and selecting the optimum answer taking into account the emotion analysis results]; and [means for providing the selected optimum answer to the user.] This makes it possible to provide appropriate and reliable answers that take into account the user's emotions.

[1634] "Means for accepting questions from a user" refers to a function that receives questions entered by a user into the interface and sends the data to the server for processing.

[1635] "Means for distributing received questions and user emotion data to multiple generative models" refers to the function of distributing the user question and emotion data obtained from the emotion engine to appropriate generative models and using that data for analysis.

[1636] "Means for generating answers from each generative model" refers to the function by which each generative AI model generates answers based on the distributed question data.

[1637] "Means for evaluating multiple generated answers and selecting the most appropriate answer taking into account the results of sentiment analysis" refers to a function that compares multiple generated answers based on certain evaluation criteria and selects the most appropriate answer taking into account the results of sentiment analysis.

[1638] The "means for providing the selected optimal answer to the user" refers to a function for transmitting the selected answer to the user terminal and displaying it to the user.

[1639] "Emotion analysis results" are data that indicate the emotional state of the user analyzed by the emotion engine.

[1640] "Generative AI model" is a general term for artificial intelligence models that generate responses to given questions.

[1641] An "emotion engine" refers to a system that analyzes a user's input data and behavioral patterns to recognize and evaluate the user's emotional state.

[1642] The present invention is a system that accepts questions from users, generates optimal answers using a generative AI model, and provides answers that correspond to the user's emotions using an emotion engine. Specific embodiments of the system are described in detail below.

[1643] Hardware and software used

[1644] Hardware

[1645] User terminal: A device through which a user inputs a question, such as a PC or smartphone.

[1646] Server: A device that processes question data and emotion data and runs generative AI models.

[1647] software

[1648] Interface application: An application that is installed on a user's device and accepts input from the user.

[1649] Emotion engine: Software that recognizes emotions by analyzing user input data and behavioral patterns.

[1650] Generative AI model: Software that generates answers to given questions. Examples include text generation models, economic data analysis models, and trend forecasting models.

[1651] Data processing and calculation methods

[1652] 1. Accepting user questions:

[1653] The user types a question into the device interface, for example, "Please tell me about the market size of AI platforms."

[1654] The device sends this input data to the server, and at the same time sends emotion data such as input speed and typos to the emotion engine.

[1655] 2. Emotion Recognition and Analysis:

[1656] The emotion engine analyzes the received emotion data and recognizes the user's emotional state, such as whether they are tense or relaxed.

[1657] The emotion engine sends the analysis results to the server.

[1658] 3. Question analysis and generative model selection:

[1659] The server obtains the emotion analysis results from the emotion engine and analyzes the question content.

[1660] Based on the analysis results, the server selects the optimal generative AI model (text generation model, economic data analysis model, trend prediction model, etc.).

[1661] 4. Question distribution:

[1662] The server sends the question text and sentiment analysis results to the selected generative AI model.

[1663] Each generative AI model generates an answer based on the question data.

[1664] 5. Generate and evaluate answers:

[1665] Each generative AI model generates an answer based on the data it receives and sends the result back to the server.

[1666] The server evaluates answers from multiple generative models and selects the optimal answer taking into account the results of sentiment analysis.

[1667] 6. Providing answers to users:

[1668] The server transmits the selected optimal answer to the user terminal.

[1669] The device will display the answer in a format that is easy for the user to understand, such as, "The market size of AI platforms is predicted to reach $X billion in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1670] Specific examples

[1671] Specific questions

[1672] Question: "What is the market size for AI platforms?"

[1673] Specific examples of sentiment analysis

[1674] When a user is nervous, their typing speed slows down and they make more errors. The emotion engine analyzes this and recognizes that the user is nervous. Based on the results of this analysis, the server selects a generative model to reassure the user and generates an appropriate answer.

[1675] Based on these steps and specific examples, the system of the present invention can provide appropriate and reliable answers that correspond to the user's emotions, thereby increasing user satisfaction.

[1676] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1677] Step 1:

[1678] Accepting user questions

[1679] The user types into the device interface, "Please tell me about the market size of AI platforms."

[1680] Input: Text data that a user types into a device's interface.

[1681] The terminal receives a question text from the user and transmits the text data to the server.

[1682] Output: The action of sending the received question text data to the server.

[1683] Step 2:

[1684] Emotion Recognition and Analysis

[1685] The terminal also sends emotion data, such as the user's input speed and typo error patterns, to the emotion engine in parallel.

[1686] Input: Emotional data such as user typing speed, typos, etc.

[1687] The emotion engine analyzes the received emotion data and recognizes the user's emotional state (e.g., tense, relaxed).

[1688] Output: Send the result of the emotion analysis (e.g., the user is nervous) to the server.

[1689] Step 3:

[1690] Question analysis and generative model selection

[1691] The server analyzes the emotion analysis results obtained from the emotion engine and the user's question text.

[1692] Input: User question text and sentiment analysis results.

[1693] The server analyzes the question and selects a corresponding generative AI model (e.g., text generation model, economic data analysis model, trend prediction model).

[1694] Output: Question text and sentiment analysis results are distributed to each generative AI model.

[1695] Step 4:

[1696] Question distribution

[1697] The server distributes the question text to the selected generative AI models (text generation model, economic data analysis model, trend prediction model) and also attaches the sentiment analysis results.

[1698] Input: Question text and sentiment analysis results.

[1699] Output: The question and sentiment analysis results are sent to each generative AI model.

[1700] Step 5:

[1701] Generate answers

[1702] The text generation model generates an appropriate answer based on the question it receives, for example, "The market size of AI platforms in 2023 will be $X billion."

[1703] The economic data analysis model analyzes historical market data and generates data on the current market size.

[1704] The trend forecasting model predicts future market trends and generates forecast data for the market size for the next three years.

[1705] Input: Question text and sentiment analysis results.

[1706] Output: The answer data generated by each generative AI model.

[1707] Step 6:

[1708] Evaluation and selection of answers

[1709] The server aggregates the response data received from each generative AI model and selects the optimal response based on an evaluation algorithm.

[1710] Input: Answer data from each generative AI model.

[1711] The evaluation algorithm includes accuracy, reliability, applicability, etc., and also takes into account sentiment analysis results.

[1712] If necessary, the generative models brainstorm with each other to generate better answers.

[1713] Output: The best answer.

[1714] Step 7:

[1715] Providing answers to users

[1716] The server transmits the selected optimal answer to the user terminal.

[1717] Input: Best answer.

[1718] The device will then display the answer it receives to the user in a friendly and easy-to-understand format, such as "The AI ​​platform market size is predicted to reach $X in 2023. The average annual growth rate over the past five years has been Y%, and further growth is expected over the next three years."

[1719] Output: Display the best answer to the user.

[1720] (Application example 2)

[1721] 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."

[1722] Current factory robots respond to questions or instructions from operators by providing uniform answers without considering the operator's emotional state, which can lead to reduced usability and difficulty in effective communication. There is a need for a method to facilitate smooth communication between humans and robots, especially when appropriate responses are required under emotional circumstances.

[1723] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for accepting questions from users, means for distributing the accepted questions to multiple generative models, and means for generating answers from each generative model. This makes it possible to evaluate multiple generated answers, select the optimal answer, and provide the optimal answer to the operator in a format that corresponds to the operator's emotion. In addition, by providing means for analyzing the operator's emotional state using an emotion recognition engine and taking the operator's emotional state into consideration when selecting a generative model, a more appropriate and human-like response is possible.

[1724] The "means for accepting questions from users" is an interface that allows an operator to input questions or instructions to a factory robot, which then recognizes them and receives them as data.

[1725] "Means for distributing received questions to multiple generative models" is a function for analyzing received questions and distributing them to appropriate generative models.

[1726] The "means for generating an answer from each generative model" is a system for generating a corresponding answer based on a question received by the generative model.

[1727] The "means for evaluating multiple generated answers and selecting the optimal answer" is an algorithm for evaluating answers provided by multiple generative models and selecting the most appropriate one.

[1728] The "means for providing the operator with the optimal answer in a format that corresponds to their emotions" is a system for providing the selected optimal answer in a format that is appropriate to the operator's emotional state.

[1729] "Means for analyzing an operator's emotional state using an emotion recognition engine" refers to a technology for analyzing an operator's tone of voice, speed, facial expressions, etc. to analyze their emotional state.

[1730] "Means for considering the emotional state of the operator when selecting a generative model" is a function that uses the emotional state of the operator as an important factor in the process of selecting a generative model.

[1731] This invention is a system that enables a factory robot to provide appropriate answers to questions and instructions from an operator. The system is mainly composed of a user terminal, a server, and an emotion engine.

[1732] System configuration

[1733] 1. User Device

[1734] Operators can input questions and instructions to the robot using voice commands or a tablet, which uses Google Cloud Speech-to-Text as its voice recognition system and is rugged and suitable for factory environments.

[1735] The user terminal receives the input question and sends the data to the server. It also sends emotional data such as the operator's voice tone, speed, and facial expressions to the emotion engine.

[1736] 2. Server

[1737] The server has a means for analyzing the query content and distributing it to multiple generative models (e.g., production data analysis model, machine maintenance model, quality control model, etc.).

[1738] The generative model generates corresponding answers by utilizing a text generation model (e.g., OpenAI's GPT-3), a production data analysis model (e.g., Azure Machine Learning), etc.

[1739] The server evaluates the generated answers and selects the best one based on accuracy, reliability, and applicability, taking into account the analysis results of the emotion engine.

[1740] After the optimal answer is selected, it is provided to the user terminal in a format that corresponds to the emotional state of the operator.

[1741] 3. Emotion Engine

[1742] The emotion engine is a technology that analyzes the operator's emotional state by analyzing their voice tone, speed, facial expressions, etc. Emotion analysis uses emotion-recognition cameras and microphones (e.g., Logitech Brio camera, Shure microphone).

[1743] The analysis results are sent to the server and reflected in the generative model selection process.

[1744] Specific examples

[1745] Let's assume that a user asks, "Please tell me the current production efficiency," and the operator becomes impatient.

[1746] 1. The user terminal converts the operator's question into text using a voice recognition system and sends the text to the server.

[1747] 2. The emotion engine analyzes the operator's tone of voice, speed, and facial expressions and determines that the operator is impatient.

[1748] 3. The server selects a generative model based on the question content and the results of sentiment analysis. For example, a production data analysis model is selected.

[1749] 4. The generative model generates an answer like this: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for the next production line check."

[1750] 5. The server evaluates the generated answers and selects the best answer.

[1751] 6. The user terminal provides this response to the operator.

[1752] Prompt Sentence Examples

[1753] User: "What is your current production efficiency?"

[1754] User Emotion: Impatience

[1755] Generative model: Production data analysis model

[1756] Sample response: "The current production line is 90% efficient. There may be some delays due to recent maintenance. Please be patient and wait for our next production line check."

[1757] This invention allows factory robots to support operators in a more efficient and human-like manner, and by using an emotion engine, it is possible to provide appropriate and reliable information according to the operator's emotional state.

[1758] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1759] Step 1: Accepting user questions

[1760] Input: The operator inputs the question using voice or a tablet.

[1761] Processing: The user device uses a speech recognition system (e.g., Google Cloud Speech-to-Text) to convert the speech into text data. If the input is from a tablet, it is accepted as text data. Emotional data such as speech tone and speed are also acquired.

[1762] Output: Textualized question data and sentiment data.

[1763] Step 2: Emotion Recognition and Analysis

[1764] Input: Emotional data such as speech tone, rate, and facial expressions.

[1765] Processing: The emotion engine analyzes the acquired emotion data and identifies the operator's emotional state (e.g., impatience, tension, joy, etc.). Emotion analysis is performed using emotion-recognition cameras and microphones (e.g., Logitech Brio cameras, Shure microphones).

[1766] Output: Parsed emotion data.

[1767] Step 3: Question analysis and generative model selection

[1768] Input: Translated question data and parsed sentiment data.

[1769] Processing: The server analyzes the question content and the emotional state and selects an appropriate generative model (e.g., production data analysis model, machine maintenance model, quality control model). The selection criteria include the question content and the operator's emotional state.

[1770] Output: The selected generative model.

[1771] Step 4: Distributing questions

[1772] Input: Textualized question data, parsed sentiment data, and a selected generative model.

[1773] Processing: The server distributes the questions to the selected generative models. It sends the question data and emotion data to each generative model.

[1774] Output: Question data and sentiment data sent to the generative model.

[1775] Step 5: Generate an answer

[1776] Input: Question data and sentiment data sent to the generative model.

[1777] Processing: Each generative model (e.g., text generation model, production data analysis model, machine maintenance model, etc.) generates an answer based on the received question. For example, a text generation model creates an answer in Q&A format.

[1778] Output: The generated answers.

[1779] Step 6: Evaluate and select responses

[1780] Input: Multiple generated answers.

[1781] Processing: The server evaluates the answers from each generative model and selects the best answer based on evaluation criteria (accuracy, reliability, applicability, etc.), also taking into account the analysis results of the emotion engine.

[1782] Output: The best answer selected.

[1783] Step 7: Providing answers to users

[1784] Input: The selected best answer.

[1785] Processing: The server sends the selected optimal answer to the user terminal in a format that corresponds to the operator's emotional state. For example, it displays the answer in a friendly tone or in concise sentences.

[1786] Output: Answers provided in a sentiment-sensitive format.

[1787] 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.

[1788] 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.

[1789] 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.

[1790] 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.

[1791] 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.

[1792] 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.

[1793] 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).

[1794] 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.

[1795] 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."

[1796] 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.

[1797] 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).

[1798] 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.

[1799] 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.

[1800] 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.

[1801] 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.

[1802] 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.

[1803] 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.

[1804] 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.

[1805] 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.

[1806] 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.

[1807] 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.

[1808] The following is further disclosed regarding the above embodiment.

[1809] (Claim 1)

[1810] [means for accepting questions from users;

[1811] [Means for distributing received queries to multiple generative models;

[1812] [Means for generating answers from each generative model; and

[1813] [Means for evaluating the generated multiple answers and selecting the best answer;

[1814] [Means for providing the user with the selected optimal answer;

[1815] A system including:

[1816] (Claim 2)

[1817] The system of claim 1, further comprising: means for double-checking answers from each generative model.

[1818] (Claim 3)

[1819] The system of claim 1, further comprising: means for analyzing a user's question and selecting a generative model based on the analysis result.

[1820] "Example 1"

[1821] (Claim 1)

[1822] [means for accepting questions from users;

[1823] [Means for analyzing the received question and selecting a generative model based on the analysis result;

[1824] [Means for distributing questions to the selected generative models; and

[1825] [Means for generating answers from each generative model; and

[1826] [Means for evaluating the generated multiple answers and selecting the best answer;

[1827] [Means for providing the user with the selected optimal answer;

[1828] A system including:

[1829] (Claim 2)

[1830] The system of claim 1, further comprising: means for evaluating the answer from each generative model using an evaluation algorithm.

[1831] (Claim 3)

[1832] The system of claim 1, further comprising: means for double-checking answers from each generative model.

[1833] "Application Example 1"

[1834] (Claim 1)

[1835] [means for accepting questions from users;

[1836] [Means for distributing received queries to multiple generative models;

[1837] [Means for generating answers from each generative model; and

[1838] [Means for evaluating the generated multiple answers and selecting the best answer;

[1839] [Means for providing the user with the selected optimal answer;

[1840] [Means for analyzing a user's question and selecting a generative model based on the analysis results;

[1841] [means for making it accessible through a smartphone application within a brick-and-mortar store environment;

[1842] A system including:

[1843] (Claim 2)

[1844] The system of claim 1, further comprising: means for double-checking answers from each generative model.

[1845] (Claim 3)

[1846] The system of claim 1, further comprising means for providing inventory information based on a specific question in a physical store.

[1847] "Example 2: Combining Emotion Engines"

[1848] (Claim 1)

[1849] [means for accepting questions from users;

[1850] [Means for distributing received questions and user emotion data to multiple generative models;

[1851] [Means for generating answers from each generative model; and

[1852] [Means for evaluating the multiple generated answers and selecting the optimal answer taking into account the results of sentiment analysis;

[1853] [Means for providing the user with the selected optimal answer;

[1854] A system including:

[1855] (Claim 2)

[1856] The system of claim 1, further comprising means for double-checking answers from each generative model.

[1857] (Claim 3)

[1858] The system according to claim 1, further comprising means for analyzing a user's question and selecting a generative model based on the analysis result and the user's emotion data.

[1859] "Application example 2 when combining emotion engines"

[1860] (Claim 1)

[1861] [means for accepting questions from users;

[1862] [Means for distributing received queries to multiple generative models;

[1863] [Means for generating answers from each generative model; and

[1864] [Means for evaluating the generated multiple answers and selecting the best answer;

[1865] [A system including a means for providing an optimal answer to an operator in an emotion-sensitive format.

[1866] (Claim 2)

[1867] The system of claim 1, further comprising means for analyzing the emotional state of the operator using an emotion recognition engine.

[1868] (Claim 3)

[1869] The system of claim 1, further comprising means for taking into account the emotional state of the operator when selecting a generative model. [Explanation of symbols]

[1870] 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 accepting questions from users; a means for distributing received queries to a plurality of generative models; a means for generating an answer from each generative model; A means for evaluating the generated answers and selecting the best answer; a means for providing the selected optimal answer to the user; A system including:

2. The system of claim 1 , further comprising: means for double-checking answers from each generative model.

3. The system of claim 1 , further comprising means for analyzing a user's question and selecting a generative model based on the analysis result.

Images