System

Abstract

A system is provided.SOLUTION: A system comprising: means for registering a task desired to be solved by a user; means for selecting a specialized region model suitable for the task; means for allocating a necessary calculation resource to the selected model; means for executing the task by the model using the allocated resource; means for integrating execution results and providing the execution results to the user; and means for collecting feedback from the user and reflecting the feedback in the next and subsequent processes.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] Describe the "problem that the invention aims to solve" and the "means for solving the problem."

[0005] ---

[0006] In recent years, generative artificial intelligence (AI) models have demonstrated their usefulness and practicality in various fields. In particular, approaches that combine multiple domain-specific generative AI models can improve accuracy and efficiency. However, this approach requires large-scale computational resources, and effectively allocating these resources has become a key challenge. Conventional systems have struggled to achieve a fair allocation of computational resources and efficiently integrate inputs from multiple domains. [Means for solving the problem]

[0007] The present invention solves the above problems by providing a system that includes a means for a user to register a task they wish to solve, a means for selecting a specialized domain model suitable for the task, a means for allocating the necessary computational resources to the selected model, a means for the model to execute the task using the allocated resources, a means for integrating the execution results and providing them to the user, and a means for collecting feedback from the user and reflecting it in subsequent processes. This system allows task requesters to utilize the knowledge of specialized models to obtain answers from multiple perspectives, and it also allows AI engineers to use resources effectively and has an incentive to develop specialized models.

[0008] Understood. Below are definitions of important terms contained in the claims.

[0009] ---

[0010] "User" refers to an individual or legal entity that uses the system to solve a specific task.

[0011] A "task" refers to a specific problem or issue that a user wants to solve.

[0012] "Domain-specific model" refers to a generative AI model that is designed and trained specifically for a particular domain of expertise.

[0013] "Computational Resources" refers to the computing resources, such as CPU, GPU, and memory, required to run and train an AI model.

[0014] "Allocation means" refers to the process of appropriately allocating computing resources to each specialized area model and the method of implementing it.

[0015] "Execution" refers to the process by which a model generates a solution to a task using allocated computational resources.

[0016] "Integration" refers to the process of aggregating answers from multiple disciplinary models and presenting them to the user as a single unified result.

[0017] "Feedback" refers to the ratings and comments that users make on the answers provided.

[0018] "Retraining procedure" refers to the retraining process used to improve the performance of each domain model based on collected feedback. [Brief explanation of the drawings]

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

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

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

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

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

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

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

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

[0027] [First embodiment]

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

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

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

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

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

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

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

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

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

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

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

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

[0040] Understood. Below is the "Form for carrying out the invention".

[0041] ---

[0042] The present invention is a system for obtaining highly accurate solutions by selecting a domain model for a task registered by a user and efficiently allocating computational resources. The processing of the system program is described in detail below.

[0043] System Overview

[0044] This system consists of a series of processes in which a user registers a task they wish to solve, selects an appropriate domain model, allocates the necessary computational resources, generates an answer to the task, and provides the user with the integrated answer.

[0045] Task registration phase

[0046] 1. User operations

[0047] Users input the task they want to solve in natural language from their own device and register it in the system.

[0048] As a specific example, when a user requests a component analysis of a new drug, the user inputs "I would like to perform component analysis in new drug development."

[0049] 2. Terminal Processing

[0050] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server.

[0051] Task analysis and model selection phase

[0052] 3. Server Processing

[0053] The server analyzes the received task information and extracts necessary keywords and concepts. For example, keywords such as "new drug development" and "component analysis" are identified.

[0054] 4. Server Processing

[0055] Based on the analysis results, the server selects which specialized model (medical model, chemistry model, pharmaceutical model, etc.) is most appropriate.

[0056] As a specific example, we determine that specialized models in medicine, chemistry, and pharmacy are required for tasks related to new drug development.

[0057] Resource Allocation Phase

[0058] 5. Server Processing

[0059] The server allocates the necessary computing resources (GPU, CPU, memory, etc.) to each selected specialized domain model.

[0060] As a specific example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[0061] 6. Server Processing

[0062] With the allocated resources, each model is prepared to perform its task.

[0063] Answer generation phase

[0064] 7. Server Processing

[0065] Each domain model generates a solution for the task it receives. For example, a chemistry model analyzes the ingredients of a new drug and generates a result.

[0066] 8. Server Processing

[0067] The server collects and integrates the solutions generated by each model, evaluates the contribution of each model, and generates a unified solution.

[0068] As a specific example, component analysis data obtained from a chemical model, data on effects on the human body obtained from a medical model, and efficacy data obtained from a pharmaceutical model are integrated and compiled into a single report format.

[0069] 9. Server Processing

[0070] The integrated answers are converted into an appropriate format and sent to the user's terminal.

[0071] Feedback and Retraining Phase

[0072] 10. User Operations

[0073] The user evaluates the answers received and enters feedback into the system.

[0074] For example, feedback such as "I'm satisfied with the analysis results" or "I need more information" can be entered.

[0075] 11. Terminal Processing

[0076] The user's rating is converted into an appropriate format and sent to the server.

[0077] 12. Server Processing

[0078] The server receives feedback and evaluates the performance of each domain-specific model, retraining it as needed to improve its accuracy the next time the task is performed.

[0079] In this way, the system of the present invention contributes to solving the user's problem by appropriately allocating the user's tasks to the specialized domain model and generating an answer efficiently and accurately.

[0080] The processing flow will be explained below.

[0081] Understood. Below, I will explain the process step by step.

[0082] ---

[0083] Step 1:

[0084] User operations

[0085] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[0086] Step 2:

[0087] Terminal handling

[0088] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis in new drug development\"}".

[0089] Step 3:

[0090] Server Processing

[0091] The server analyzes the received task information. It uses a natural language processing model to understand the task content and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[0092] Step 4:

[0093] Server Processing

[0094] Based on the analysis results, the server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.). For example, it may determine that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[0095] Step 5:

[0096] Server Processing

[0097] The server allocates the necessary computing resources (e.g., GPUs, CPUs, memory, etc.) to each selected specialty model. For example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[0098] Step 6:

[0099] Server Processing

[0100] The server prepares each model to execute the task using the allocated resources, provides each model with detailed task information, and starts the calculation.

[0101] Step 7:

[0102] Server Processing

[0103] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data of a new drug and generates a result.

[0104] Step 8:

[0105] Server Processing

[0106] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[0107] Step 9:

[0108] Server Processing

[0109] The integrated answer is converted into an appropriate format and sent to the user's terminal. For example, a report is generated as a "New Drug Ingredient Analysis Report" and sent to the user.

[0110] Step 10:

[0111] User operations

[0112] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[0113] Step 11:

[0114] Terminal handling

[0115] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[0116] Step 12:

[0117] Server Processing

[0118] The server receives feedback and evaluates the performance of each domain model. If necessary, it retrains the model to improve its accuracy the next time it performs the task. For example, a medical model can be retrained using new clinical data.

[0119] ---

[0120] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[0121] Example 1

[0122] 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."

[0123] For complex tasks that users want to solve, there is a need to select the optimal domain model, appropriately allocate the necessary computational resources, and generate solutions effectively and efficiently. However, current technology does not fully automate the processes of task analysis, model selection, and resource management, placing a heavy burden on users. In addition, the feedback process to improve the accuracy of execution results is not fully integrated. Therefore, a system is needed that reduces the time and effort required for users to solve tasks and improve the accuracy of results.

[0124] 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.

[0125] In this invention, the server includes a means for analyzing task information and extracting related keywords and concepts, a means for selecting a specialized domain model appropriate for the task, and a means for allocating the necessary computational resources to each selected specialized domain model. This enables the selection of an optimal model for the task registered by the user and the efficient allocation of resources. This ultimately results in the rapid generation of highly accurate answers and reduces the burden on the user. Furthermore, by collecting feedback from the user on the answer results and reflecting this feedback in subsequent processes, the accuracy and efficiency of the entire system can be improved.

[0126] The "means for users to register tasks they wish to solve" is an interface that allows users to input specific problems or issues they wish to solve in natural language and register them in the system.

[0127] "Means for analyzing task information and extracting related keywords and concepts" refers to a function that automatically analyzes task information registered by the user and extracts important keywords and related concepts from it.

[0128] The "means for selecting a specialized domain model appropriate for the task" refers to an algorithm or process for selecting the specialized domain model (e.g., medical model, chemical model, pharmaceutical model, etc.) that is optimal for the task based on the extracted keywords and concepts.

[0129] The "means for allocating the necessary computing resources to each selected specialized domain model" is a mechanism for dynamically allocating the necessary computing resources (GPU, CPU, memory, etc.) to efficiently operate the specialized domain model.

[0130] The "means by which each specialized domain model generates an answer for a received task" refers to the process by which a specialized domain model generates an answer or result based on a given task using specialized knowledge or algorithms.

[0131] "Means for collecting and integrating execution results from each specialized model" refers to an integration process that brings together the answers and results generated by each specialized model to derive the optimal answer overall.

[0132] "Means for converting the integrated execution results into an appropriate format and providing them to the user" refers to a function that converts the integrated answers and results into a format that is easy for the user to understand (for example, a report format or graph) and provides them to the user.

[0133] "Means for collecting feedback from users and reflecting it in each process of the system" refers to the process of collecting evaluations and opinions on the answers and results provided by users and reflecting them in improving the system and increasing accuracy the next time the task is performed.

[0134] The present invention provides a system that selects an optimal domain model for a task that a user wants to solve, allocates the necessary computational resources, generates a highly accurate solution based on the model, and provides the solution to the user. Specific embodiments of the system are described in detail below.

[0135] This system mainly consists of three entities: users, terminals, and servers.

[0136] First, the user inputs the task they want to solve in natural language on their own device. For example, they might input, "I want to perform ingredient analysis for new drug development." By registering such a task, they can request the system to solve the problem.

[0137] Next, the device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server, which then transmits the task information throughout the system.

[0138] The server first analyzes the received task information and extracts important keywords and related concepts. For example, keywords such as "new drug development" and "ingredient analysis" are extracted. This analysis is generally performed using a natural language processing (NLP) engine. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK, spaCy).

[0139] Based on the analysis results, the server selects the domain-specific model appropriate for the task. For example, the task of ingredient analysis in new drug development requires a medical model, a chemical model, and a pharmaceutical model. To select these domain-specific models, the server uses Bayesian inference and rule-based systems.

[0140] Next, the necessary computing resources are allocated to each selected domain-specific model. The server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. For example, two GPUs are allocated to a medical model, and three GPUs to a chemistry model.

[0141] Using the allocated resources, each specialized model performs a task and generates an answer based on its own expertise. For example, a chemistry model analyzes input ingredient information and evaluates chemical properties.

[0142] The server collects and integrates the results returned by each model. The integration uses an algorithm that evaluates the contribution of each model to generate the most appropriate answer. For example, it combines component analysis data from a chemical model, data on effects on the human body from a medical model, and drug efficacy data from a pharmaceutical model.

[0143] The integrated answers are converted into an appropriate data format and sent to the user's device. The user then checks and evaluates the answers on their own device. The evaluation is then entered into the system as feedback. For example, the user can send feedback such as "I'm satisfied with the analysis results" or "Additional information is needed."

[0144] The collected feedback is analyzed by the server and used to improve the system's performance, specifically by retraining each domain-specific model to improve its accuracy when performing subsequent tasks.

[0145] Through the above process, the system of the present invention can provide highly accurate and efficient solutions to tasks input by the user, thereby enabling the user to solve problems quickly and reliably.

[0146] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0147] Step 1: The user enters the task they want to solve in natural language into the terminal and registers it in the system. For example, they might enter, "I want to perform ingredient analysis in new drug development." This is the input data. The terminal receives this data and displays it on the screen.

[0148] Step 2: The terminal converts the input task information into an appropriate data format (for example, JSON format). Specifically, the terminal converts the input text into a JSON object (for example, "task": "I would like to perform component analysis in new drug development"). This becomes the output data. This converted data is used in the next step.

[0149] Step 3: The terminal sends the converted data to the server. Specifically, the terminal uses an HTTP request to send JSON formatted data to the server. The input is JSON data, and the output is data transmission to the server.

[0150] Step 4: The server analyzes the received task information. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK or spaCy) to extract important keywords (e.g., "new drug development" and "ingredient analysis"). The input is the received JSON data, and the output is the extracted keywords.

[0151] Step 5: The server selects appropriate domain models based on the analysis results. Specifically, the server uses Bayesian inference and rule-based systems to select medical, chemical, and pharmaceutical models. The input is the extracted keywords, and the output is a list of selected models.

[0152] Step 6: The server allocates the necessary computing resources for each selected domain model. Specifically, the server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. The input is a list of selected models, and the output is the details of the resource allocation. For example, two GPUs are allocated to the medical model, and three GPUs are allocated to the chemistry model.

[0153] Step 7: The server prepares each model to run by adding a task to a queue for each model and launching a script to initialize the execution environment. The input is the resource allocation details, and the output is the ready state of each model.

[0154] Step 8: Each domain model generates a solution for the received task. Specifically, the chemical model analyzes the input ingredient information and evaluates chemical attributes. The input is the readiness status and resources of each model, and the output is the solution generated by each model.

[0155] Step 9: The server collects and integrates the answers from each model. Specifically, it aggregates the data returned from each model and generates a unified answer using a contribution evaluation algorithm. The input is the answer from each model, and the output is the integrated answer.

[0156] Step 10: The server converts the consolidated answer into an appropriate format and sends it to the user's device. Specifically, it converts the consolidated answer data into JSON or report format and returns it to the user using an HTTP response. The input is the consolidated answer, and the output is the answer returned to the user's device.

[0157] Step 11: The user evaluates the answer they received and enters their feedback into the system. For example, they might enter something like "I'm satisfied with the analysis results" or "I need more information." This is the input of feedback data.

[0158] Step 12: The device converts the user's rating into an appropriate format and sends it to the server. Specifically, the device converts the feedback into JSON format and sends it to the server via an HTTP request. The input is the feedback data, and the output is the data transmission to the server.

[0159] Step 13: The server receives the feedback, evaluates the performance of each domain model, and performs retraining. Specifically, it uses the feedback data to update the model's training dataset and retrains it. The input is the received feedback data, and the output is the updated model.

[0160] (Application example 1)

[0161] 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."

[0162] Modern factories have many tasks, and their efficient execution directly leads to improved productivity. However, optimal resource allocation and appropriate utilization of specialized knowledge for each task is difficult. Dynamic improvements based on feedback are also required, making it difficult to address these challenges with traditional static models. Therefore, a new system is needed that can efficiently optimize tasks and improve the performance of factory production lines.

[0163] 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.

[0164] In this invention, the server includes means for a user to register a task to be solved, means for selecting a specialized domain model suitable for the task, means for allocating necessary computational resources to the selected model, means for the model to execute the task using the allocated resources, means for integrating the execution results and providing them to the user, means for collecting feedback from the user and reflecting it in subsequent processes, and means for generating a plan for having robots in the factory execute the task and appropriately allocating resources, thereby enabling efficient optimization of factory tasks and dynamic improvement.

[0165] "User" refers to a person or organization that uses the system.

[0166] A "task" refers to a specific problem or task that a user wants the system to solve.

[0167] "Domain models" refer to machine learning or artificial intelligence algorithms or software that are specialized for a particular field or problem.

[0168] "Computational resources" refers to the computer hardware and software resources required to perform a task, including CPU, GPU, memory, etc.

[0169] "Execution" refers to the process by which a selected domain expertise model processes a task using the allocated computing resources.

[0170] "Synthesis" refers to the process of combining the results of multiple disciplinary models into a single conclusion or answer.

[0171] "Feedback" refers to the evaluations and opinions provided by users, and the system is improved based on that information.

[0172] "Factory robots" refer to autonomous or semi-autonomous machines that perform tasks in a factory.

[0173] "Generating a plan" refers to the process of automatically creating optimal execution procedures for a task.

[0174] "Appropriate resource allocation" refers to efficiently allocating the necessary computing resources and hardware to each task or model.

[0175] The present invention is a system that allows users to register a task they wish to solve, selects an appropriate domain model, efficiently allocates the necessary computational resources, and generates a highly accurate solution. This system is intended to be applied particularly to robots in factories.

[0176] System Program

[0177] First, the user inputs the task they want to solve in natural language and registers it in the system. For example, a task might be, "Please suggest a way to resolve a bottleneck on the production line."

[0178] The server analyzes the input task information and extracts important keywords and concepts. Based on the extracted keywords, it selects the optimal domain model. For example, keywords such as "production line," "bottleneck," "solution," "method," and "proposal" are extracted.

[0179] Next, the server allocates the necessary computing resources to the selected domain model. In this case, CPUs and GPUs are used as computing resources, and specific examples include NVIDIA GPUs and Intel CPUs.

[0180] Using the allocated resources, each domain model performs tasks and generates solutions, such as optimizing a manufacturing process or analyzing the causes of bottlenecks.

[0181] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer, which is provided to the user in the form of a report. This process utilizes natural language generation technology and a generative AI model.

[0182] Users evaluate the answers they receive and provide feedback to the system, which then uses that feedback to improve the task-solving process and, if necessary, retrain the domain model.

[0183] Hardware and software used

[0184] Hardware: Factory robots, dedicated GPU servers, user devices (PCs, smartphones, etc.)

[0185] Software: Natural language processing libraries (NLTK, SpaCy), AI model selection libraries (custom libraries), resource allocation libraries, natural language generation models (GPT-3, etc.)

[0186] Adding specific examples

[0187] As a concrete example, consider the case where a user inputs "Please suggest a method to resolve bottlenecks in the production line." The system analyzes this task, extracts keywords such as "production line," "bottleneck," "resolve," "method," and "proposal," and based on these, selects the most appropriate domain-specific model (e.g., production engineering model, physical simulation model). Each model independently proposes a solution, and finally provides an integrated report.

[0188] An example of a prompt to be input to a generative AI model: "Please suggest a way to resolve a bottleneck on the production line."

[0189] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0190] Step 1:

[0191] Users input the task they want to solve in natural language from their own terminal and register it in the system.

[0192] Input: User task input (e.g., "Please suggest a way to resolve the bottleneck on the production line.")

[0193] Output: Task information is converted to JSON format and sent to the server.

[0194] Specific operation: When a user enters the task details in a dedicated application and presses the send button, the task information is converted into JSON format and sent to the server via the Internet.

[0195] Step 2:

[0196] The server analyzes the received task information and extracts important keywords and concepts.

[0197] Input: Task information sent to the server (JSON format)

[0198] Output: Extracted keywords and concepts (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[0199] How it works: The server parses the task information in JSON format and uses a natural language processing library (such as NLTK or SpaCy) to extract relevant keywords. During this process, it uses a specific algorithm to determine the frequency and importance of the keywords.

[0200] Step 3:

[0201] The server selects the most suitable specialized domain model based on the extracted keywords.

[0202] Input: Extracted keywords (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[0203] Output: List of selected domain-specific models (e.g., industrial engineering models, physical simulation models)

[0204] Specific operation: Based on the extracted keywords, the server selects an appropriate domain-specific model using a model selection library. This library recommends the best model based on past data and experience.

[0205] Step 4:

[0206] The server allocates the necessary computational resources to each selected domain model.

[0207] Input: List of selected domain models

[0208] Output: Allocated computing resources (e.g., 2 GPUs, 4 CPU cores, etc.)

[0209] How it works: The server uses a resource allocation library to allocate optimal resources to each model. For example, a physics simulation model requires a large number of GPUs, so it calculates and allocates appropriate resources.

[0210] Step 5:

[0211] The server uses the allocated resources to have each domain model perform a task and generate an answer.

[0212] Input: Allocated computing resources, selected domain model

[0213] Output: The answer generated by each model (e.g., "How to optimize a manufacturing process?")

[0214] How it works: The server runs each model and makes full use of its resources to generate the information needed to solve the task. Specifically, the production engineering model generates proposals for improving the efficiency of the manufacturing process, and the physical simulation model generates simulation results for eliminating bottlenecks.

[0215] Step 6:

[0216] The server collects the solutions generated from each model and combines them to generate a single unified solution.

[0217] Input: Answers generated by each model

[0218] Output: Unified answer (e.g. unified report)

[0219] How it works: The server uses a natural language generation model to integrate the answers from each model. During this process, it evaluates the contribution of each model and prioritizes the most important information. It then uses a generative AI model (e.g., GPT-3) to compile the results into a readable report.

[0220] Step 7:

[0221] The server sends the integrated answer to the user's terminal.

[0222] Input: Consolidated Answer

[0223] Output: The answer displayed on the user's terminal

[0224] Specific operation: The server converts the integrated answers into JSON format or another appropriate data format and sends them to the user's device. The user can then review the answers and use them to improve the manufacturing process.

[0225] Step 8:

[0226] Users rate the answers they receive and enter their feedback into the system.

[0227] Input: User rating and feedback (e.g., "Satisfied with the analysis results" or "More information needed")

[0228] Output: The feedback information is sent to the server.

[0229] Specific operation: The user uses a dedicated interface to input evaluations of answers and requests for additions, and sends them to the server.

[0230] Step 9:

[0231] The server uses the feedback it receives to improve the task-solving process for future tasks, including retraining the domain expertise model if necessary.

[0232] Input: User feedback

[0233] Output: Improved model and system settings

[0234] What happens: The server analyzes the feedback, identifies areas for improvement in the model and process, and, if necessary, retrains the domain-specific model to reflect it in the next task-solving process.

[0235] 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.

[0236] Understood. Below is the "Form for carrying out the invention".

[0237] ---

[0238] This invention is a system that utilizes a domain model and an emotion engine to efficiently allocate computational resources to obtain highly accurate answers for tasks registered by users. The processing of the system program is described in detail below.

[0239] System Overview

[0240] This system consists of a series of processes: a user registers the task they wish to solve, selects an appropriate domain model and emotion engine, allocates the necessary computing resources, generates an answer to the task, and provides the user with the integrated answer.

[0241] Task registration phase

[0242] 1. User operations

[0243] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[0244] An emotion engine analyzes emotion information from user input to measure, for example, urgency or anxiety.

[0245] 2. Terminal Processing

[0246] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[0247] Task analysis and model selection phase

[0248] 3. Server Processing

[0249] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[0250] Emotional data is also taken into consideration and the task is determined to be urgent.

[0251] 4. Server Processing

[0252] The server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the emotion data and analysis results. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[0253] Resource Allocation Phase

[0254] 5. Server Processing

[0255] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. If the emotional data indicates a high level of urgency, more resources are allocated. For example, a medical model may be allocated twice as many GPUs as usual.

[0256] 6. Server Processing

[0257] Using the allocated resources, prepare each model to execute the task. Provide each model with detailed task information and emotion information, and start the calculation.

[0258] Answer generation phase

[0259] 7. Server Processing

[0260] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data for a new drug and generates a result.

[0261] 8. Server Processing

[0262] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[0263] 9. Server Processing

[0264] The integrated answers are converted into an appropriate format and sent to the user's device. Emotional data is also reflected, and explanations that take the user's feelings into consideration are added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[0265] Feedback and Retraining Phase

[0266] 10. User Operations

[0267] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[0268] 11. Terminal Processing

[0269] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[0270] 12. Server Processing

[0271] The server receives the feedback and evaluates the performance of each specialized model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data. It also takes into account emotional data, reflecting emotional fluctuations due to the feedback.

[0272] In this way, the system of the present invention appropriately allocates the user's tasks to the domain model and emotion engine, and generates answers efficiently and accurately, thereby contributing to the user's problem solving.

[0273] The processing flow will be explained below.

[0274] I understand. I will explain the process in detail in the following format.

[0275] ---

[0276] Step 1:

[0277] User operations

[0278] The user enters the task they want to solve in natural language on their device and registers it in the system. For example, they might enter "I want to perform ingredient analysis for new drug development" and press the send button. At this time, the emotion engine also analyzes emotional information from the user's facial expressions and voice.

[0279] Step 2:

[0280] Terminal handling

[0281] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[0282] Step 3:

[0283] Server Processing

[0284] The server analyzes the received task information and emotional data. It uses a natural language processing model to understand the content of the task and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "ingredient analysis." It also determines the urgency and importance of the task from the emotional data.

[0285] Step 4:

[0286] Server Processing

[0287] The server selects the most appropriate specialized model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the analysis results and emotion data. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug. Furthermore, based on the emotion data, it creates a plan to allocate more resources to high-priority tasks.

[0288] Step 5:

[0289] Server Processing

[0290] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. Based on the emotion data, if the level of urgency is high, the server will increase resource allocation. For example, a medical model will be allocated twice as many GPUs as usual.

[0291] Step 6:

[0292] Server Processing

[0293] The server checks the resource allocation for each model and sends task details to each model, which prepares to execute the task and starts the calculation process.

[0294] Step 7:

[0295] Server Processing

[0296] Each specialized domain model generates a solution for the task it receives. For example, a chemistry model calculates component analysis data for a new drug and sends the results to a server.

[0297] Step 8:

[0298] Server Processing

[0299] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles component analysis data from chemical models, human body effect data from medical models, and drug efficacy data from pharmaceutical models into a single report format.

[0300] Step 9:

[0301] Server Processing

[0302] The integrated answer is converted into an appropriate format for presentation to the user and sent to the user's device. Emotional data is also reflected, and an explanation that takes the user's emotions into consideration is added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[0303] Step 10:

[0304] User operations

[0305] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information," as well as emotional data.

[0306] Step 11:

[0307] Terminal handling

[0308] The user's device converts the feedback and emotion data into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis results\",\"emotion\":\"satisfied\"}\".

[0309] Step 12:

[0310] Server Processing

[0311] The server receives the feedback and emotion data and evaluates the performance of each domain-specific model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data to reflect changes in emotion due to feedback.

[0312] ---

[0313] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[0314] Example 2

[0315] 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."

[0316] Conventional problem-solving systems have difficulty efficiently analyzing user tasks and providing highly accurate solutions. Furthermore, they often fail to adequately consider the user's emotional information, resulting in a poor user experience. Furthermore, they fail to select an expert domain model or optimally allocate computing resources, resulting in the waste of computing resources. The purpose of this invention is to solve these problems.

[0317] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for analyzing task information and extracting appropriate keywords and concepts, means for selecting a specialized domain model based on the task and emotion information, and means for allocating necessary computational resources to the selected specialized domain model. This enables the generation of highly accurate answers to tasks. Furthermore, by taking the user's emotion information into consideration, a more satisfying user experience can be provided. Furthermore, efficient allocation of computational resources improves overall performance and reduces waste.

[0318] A "user" is an individual or group that uses the system to register tasks and receive answers.

[0319] A "task" refers to a specific problem or requirement that a user wants to solve.

[0320] "Emotion information" refers to data obtained by analyzing the user's emotional state at the time of task input.

[0321] A "domain model" is a computational model that includes knowledge and algorithms in a specific field.

[0322] "Computational resources" refers to the hardware and software resources (e.g., CPU, GPU, memory, etc.) required to run a domain-specific model.

[0323] "Answer" refers to the result or conclusion that a domain model generates for a task.

[0324] "Feedback" refers to the act of a user providing the system with an evaluation or opinion on the answer they received.

[0325] "Retraining" refers to the process of updating the domain-specific model based on feedback to improve accuracy the next time the task is performed.

[0326] The "server" refers to a central computing device that analyzes task information and emotional information, selects specialized domain models, allocates computing resources, and generates and integrates answers.

[0327] "Terminal" refers to a device (e.g., PC, smartphone, tablet, etc.) that a user uses to input tasks and receive answers.

[0328] "Data format" refers to a standardized structure (e.g., JSON, XML, etc.) for storing and transmitting information.

[0329] This invention is a system that utilizes domain expertise models and emotion engines to efficiently allocate computational resources to tasks registered by users and obtain highly accurate answers. The program for this system includes a series of processes from task registration to answer generation, feedback collection, and relearning.

[0330] In this system, a user first inputs the task they want to solve in natural language using a device (e.g., PC, smartphone, tablet, etc.). For example, they might input, "I would like you to analyze the ingredients for new drug development." The emotion engine then analyzes the user's input and extracts emotional information (e.g., urgency and anxiety). The device then converts this task information and emotional information into an appropriate data format (e.g., JSON format) and sends it to the server.

[0331] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis." It also takes into account emotional information to determine the urgency of the task. Based on this, the server selects the domain model (e.g., medical model, chemistry model, pharmaceutical model) that is best suited to the task.

[0332] Next, the server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected domain-specific model. If the emotional information indicates a high level of urgency, more computing resources are allocated. For example, a medical model is allocated twice as many GPUs as usual. Using the allocated resources, each model generates an answer to the task.

[0333] The server collects and integrates the generated answers from each specialized model. During integration, the contribution of each model is evaluated and a unified answer is generated. For example, the component analysis data from the chemical model and the human body effect data from the medical model are combined and integrated into a report format. The server then converts the integrated answer into an appropriate format and sends it to the user's device. The server also reflects the user's emotional information and adds an explanation that takes the user's feelings into consideration. For example, a report titled "Component Analysis Report of a New Drug" could be generated, with a comment added such as "We determined that an emergency response was required, so we conducted the analysis quickly."

[0334] Furthermore, the user evaluates the answers they receive and inputs their feedback into the system. For example, they may enter feedback such as "I'm satisfied with the analysis results" or "I need more information." The device converts the feedback into an appropriate format and sends it to the server. The server receives the feedback and evaluates the performance of each specialized domain model. If necessary, retraining is performed to improve accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data.

[0335] Through this series of processes, the system uses the domain model and emotion engine to process user tasks efficiently and with high accuracy, significantly contributing to solving user problems. Furthermore, by taking emotion information into consideration, the system can increase user satisfaction.

[0336] Examples of concrete examples and prompts

[0337] Specific examples

[0338] The user types into the terminal, "I would like to request a component analysis for new drug development."

[0339] The server selects a specialized area model related to new drug development, determines that the task is of high urgency, and allocates more computing resources to it.

[0340] Each specialized model starts its analysis and sends the results to the server.

[0341] The integrated answer is provided to the user, along with a comment such as, "We determined that an emergency response was required, so we conducted a rapid analysis."

[0342] Prompt Sentence Examples

[0343] Input: "Please provide detailed ingredient analysis for my drug development project. It's very important."

[0344] The above is a specific embodiment for carrying out the present invention.

[0345] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0346] System processing steps

[0347] Step 1: Task Input and Sentiment Analysis

[0348] The user inputs the task they want to solve using the terminal in natural language.

[0349] Input: "I would like to request a component analysis for new drug development."

[0350] How it works: The device uses an emotion engine to extract emotion information (e.g., urgency) from the user's input.

[0351] Step 2: Convert to data format and send

[0352] The device converts the input task information and emotion information into an appropriate data format (e.g., JSON format) and sends it to the server.

[0353] Input: User task information and emotion information

[0354] Output: JSON data "{\"task\":\"I would like you to analyze ingredients for new drug development\",\"emotion\":\"high urgency\"}"

[0355] What it does: Serializes input data into JSON format and sends it to the server.

[0356] Step 3: Task analysis and keyword extraction

[0357] The server analyzes the received task information and extracts keywords and concepts.

[0358] Input: JSON data

[0359] Output: Keywords "new drug development" and "component analysis"

[0360] How it works: Extracts keywords using natural language processing algorithms.

[0361] Step 4: Selecting a domain model

[0362] The server selects an appropriate domain model based on task information and emotion information.

[0363] Input: Keywords, emotion information

[0364] Output: Selected domain models (e.g., medical model, chemical model)

[0365] Action: Searches for and selects a corresponding model from the database.

[0366] Step 5: Allocate Computational Resources

[0367] The server allocates the necessary computing resources to the selected domain model.

[0368] Input: Selected domain model, emotion information

[0369] Output: Allocated computing resources (e.g. GPU, CPU specs)

[0370] How it works: A resource management system is used to allocate optimal resources to each model.

[0371] Step 6: Prepare and provide information about the model

[0372] The server prepares each model to run using the allocated resources and provides task and emotion information.

[0373] Input: Liquidation resources, task information, emotion information

[0374] Output: Ready to run

[0375] How it works: Loads the necessary libraries and datasets, and feeds the model with task information.

[0376] Step 7: Generate the answer

[0377] The server uses each domain model to generate an answer to the task.

[0378] Input: Task information provided to the model

[0379] Output: Answer data for each model

[0380] Action: The model performs a calculation and generates an answer (e.g., compositional analysis data).

[0381] Step 8: Consolidating the Answers

[0382] The server collects and integrates the answers generated by each model.

[0383] Input: Answer data for each model

[0384] Output: Consolidated answer

[0385] How it works: Evaluate the contribution of each answer and generate a single unified answer.

[0386] Step 9: Provide the answer

[0387] The server converts the integrated answers into an appropriate format and sends them to the user's device.

[0388] Input: Integrated answers, emotional information

[0389] Output: Answer data to the user

[0390] What it does: Converts the answer data into a user-friendly format (e.g., report format) and adds sentiment-sensitive comments.

[0391] Step 10: Gather feedback

[0392] The user rates the answers they receive and provides feedback.

[0393] Input: Rating and comments on the answer

[0394] Output: Feedback data

[0395] How it works: Use the input form to enter your rating and comments and submit your feedback.

[0396] Step 11: Submit your feedback

[0397] The device converts the feedback into an appropriate format and sends it to the server.

[0398] Input: User feedback

[0399] Output: Feedback data in JSON format

[0400] What it does: Serializes the feedback into JSON format and sends it to the server.

[0401] Step 12: Retraining

[0402] The server evaluates the performance of each specialized domain model based on the feedback and performs re-learning.

[0403] Input: Feedback data

[0404] Output: Updated model

[0405] How it works: Analyzes feedback and applies a re-training algorithm to improve the model.

[0406] The above is the specific program processing flow of this system.

[0407] (Application example 2)

[0408] 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."

[0409] In modern content delivery services, it is extremely important to provide users with the content they need quickly and accurately. However, conventional systems do not dynamically allocate resources based on the user's emotional state or the urgency of the task, limiting the improvement of the user experience. Furthermore, they lack a mechanism for effectively utilizing user feedback to improve accuracy in subsequent task executions. As a result, user satisfaction declines, leading to a diminishing value of the service.

[0410] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0411] In this invention, the server includes: means for analyzing a user's emotional information, determining the urgency based on the information, and adjusting the allocation of computational resources; means for the user to register a task they wish to solve; means for selecting a specialized domain model appropriate for the task; means for allocating the necessary computational resources to the selected model; means for the model to execute the task using the allocated resources; means for integrating and providing the execution results to the user; and means for collecting user feedback and reflecting it in subsequent processes. This enables dynamic resource allocation according to the user's emotional state and the urgency of the task, thereby enabling the rapid and accurate provision of content that meets the user's needs. Furthermore, by effectively utilizing user feedback to retrain the model, accuracy can be improved the next time the task is executed, thereby increasing user satisfaction.

[0412] A "task" refers to a specific request or purpose that a user requests the system to solve.

[0413] A "specialized domain model" is a machine learning model or algorithm built based on knowledge and data in a specific field or area.

[0414] "Emotional information" is emotional data such as emotional state and urgency that is analyzed from the user's input and behavior.

[0415] "Computational resources" refers to the computational resources such as CPU, GPU, and memory required to execute a task.

[0416] "Resource allocation" refers to the process of allocating the necessary computational resources to the selected model.

[0417] "Model execution" refers to the execution of calculations and analysis for a task by a selected domain-specific model using allocated computational resources.

[0418] "Execution result integration" refers to the compilation of results generated by multiple disciplinary models into a single unified answer or report.

[0419] "Feedback" refers to information indicating the user's evaluation and satisfaction with the answer they received.

[0420] "Retraining" is a learning process that improves the performance and accuracy of a domain-specific model based on collected feedback and newly acquired data.

[0421] "Urgency" is an indicator of how quickly a user's task should be resolved.

[0422] This invention is a system for quickly and accurately providing users with content they desire through a content distribution service. The system analyzes users' emotional information, determines the urgency of tasks based on that information, and dynamically adjusts the allocation of computing resources to provide personalized content tailored to the user's needs.

[0423] System Overview

[0424] This system utilizes domain expertise models and emotion engines to efficiently allocate the necessary computing resources to tasks registered by users, and provides highly accurate answers and content to users.

[0425] Hardware and software used

[0426] Hardware: Uses high-speed CPU, GPU and memory.

[0427] Software: Python 3.x, SentimentEngine (for sentiment analysis), and domain-specific models (news model, music model, video model, etc.).

[0428] System processing flow

[0429] 1. User Input and Sentiment Analysis

[0430] The user inputs the task they want to solve in natural language from their device. For example, they might input something like, "I want to know the latest technology news." The server receives this input and uses an emotion engine to analyze the emotional information and measure the urgency and anxiety of the task. The emotional information is treated as data that indicates, for example, the "high urgency" of the task.

[0431] 2. Selecting a Specialty Model and Allocating Resources

[0432] The server extracts necessary keywords and concepts based on the analyzed task information and emotion information, and selects the appropriate domain model. For example, for the keyword "technology news," a news model is selected. Then, the necessary computing resources are allocated based on the emotion data (urgency). If the urgency is high, more resources than usual are allocated.

[0433] 3. Execute the task

[0434] Using the allocated resources, selected domain-specific models perform tasks, such as a technology news model that analyzes the latest news and generates relevant articles.

[0435] 4. Consolidating and Presenting Results

[0436] The server collects and integrates the results generated by each model. This allows it to evaluate the contribution of each model and generate the optimal answer. The integrated answer is converted into an appropriate format and provided to the user along with a comment that takes into account the user's emotional information. For example, it may include a message such as, "Please check the technology news below. Due to its urgency, we analyzed it quickly."

[0437] 5. Feedback and Retraining

[0438] Users provide feedback on the answers they provide, and the server receives this feedback, evaluates the performance of each domain-specific model, and retrains it as needed, improving the model's accuracy the next time the task is performed.

[0439] Specific examples

[0440] For example, if a user requests "I want to know the latest technology news," the following prompt might be generated:

[0441] What's the latest technology news? It's very urgent right now.

[0442] Based on this prompt, the server performs a series of processes, including sentiment analysis, selection of domain expertise models, allocation of computing resources, task execution, integration and provision of results, and feedback collection and retraining, thereby quickly providing users with personalized, up-to-date technology news.

[0443] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0444] Step 1:

[0445] The device accepts the user's natural language input. The user types "I want to know the latest technology news" into the device. The device converts this input data into an appropriate data format (e.g., JSON) to send it to the server.

[0446] Input: User's natural language input

[0447] Data processing: converting natural language into data format

[0448] Output: JSON format data

[0449] Step 2:

[0450] The server receives the user input sent from the device and performs emotion analysis using an emotion engine. The emotion engine analyzes "urgency" and "anxiety" and extracts emotion information such as "high urgency."

[0451] Input: User-supplied data in JSON format

[0452] Data Computing: Sentiment Analysis

[0453] Output: Emotional information (e.g. "High Urgency")

[0454] Step 3:

[0455] The server extracts necessary keywords and concepts from the analyzed emotion information and user input. The keyword "technology news" is identified.

[0456] Input: User input and emotional information

[0457] Data processing: Keyword extraction

[0458] Output: The keyword "technology news"

[0459] Step 4:

[0460] The server selects an appropriate domain model (e.g., a news model) based on keywords and emotion information. At the same time, it dynamically allocates the necessary computing resources according to the emotion information, particularly the urgency. If the urgency is high, more CPU and GPU resources than usual are allocated.

[0461] Input: Extracted keywords and sentiment information

[0462] Data calculation: model selection, computational resource allocation

[0463] Output: Selected model and allocated resources

[0464] Step 5:

[0465] The server uses the allocated resources to run the selected news model. The model analyzes the task (e.g., get the latest technology news) and generates recommended news articles.

[0466] Input: Selected model and allocated resources

[0467] Data calculation: Model execution, recommended content generation

[0468] Output: Generated content (e.g., news article)

[0469] Step 6:

[0470] The server collects and integrates the generated content. If there are results from multiple domain-specific models, it evaluates each and generates a single integrated answer. At this time, it adds comments that take into account the user's emotional information.

[0471] Input: Multiple Content

[0472] Data processing: integrating results and adding comments

[0473] Output: Consolidated content and comments

[0474] Step 7:

[0475] The server transmits the integrated content to the terminal, which displays the content on the user's screen and makes it available to the user.

[0476] Input: Integrated content and comments

[0477] Data output: Send to terminal

[0478] Output: The content that is displayed on the user's screen

[0479] Step 8:

[0480] The user inputs feedback on the provided content, and the terminal converts this feedback into an appropriate data format for transmission to the server.

[0481] Input: User feedback

[0482] Data processing: Convert feedback into data format

[0483] Output: Feedback data in JSON format

[0484] Step 9:

[0485] The server analyzes the feedback received from the device and uses it to evaluate the performance of each specialized domain model and retrain it, and if necessary, retrains the model to improve its accuracy the next time the task is performed.

[0486] Input: Feedback data

[0487] Data calculation: performance evaluation, re-learning

[0488] Output: Improved model

[0489] 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.

[0490] 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.

[0491] 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.

[0492] [Second embodiment]

[0493] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0494] 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.

[0495] 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).

[0496] 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.

[0497] 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.

[0498] 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).

[0499] 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.

[0500] 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.

[0501] 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.

[0502] 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.

[0503] 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.

[0504] 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."

[0505] Understood. Below is the "Form for carrying out the invention".

[0506] ---

[0507] The present invention is a system for obtaining highly accurate solutions by selecting a domain model for a task registered by a user and efficiently allocating computational resources. The processing of the system program is described in detail below.

[0508] System Overview

[0509] This system consists of a series of processes in which a user registers a task they wish to solve, selects an appropriate domain model, allocates the necessary computational resources, generates an answer to the task, and provides the user with the integrated answer.

[0510] Task registration phase

[0511] 1. User operations

[0512] Users input the task they want to solve in natural language from their own device and register it in the system.

[0513] As a specific example, when a user requests a component analysis of a new drug, the user inputs "I would like to perform component analysis in new drug development."

[0514] 2. Terminal Processing

[0515] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server.

[0516] Task analysis and model selection phase

[0517] 3. Server Processing

[0518] The server analyzes the received task information and extracts necessary keywords and concepts. For example, keywords such as "new drug development" and "component analysis" are identified.

[0519] 4. Server Processing

[0520] Based on the analysis results, the server selects which specialized model (medical model, chemistry model, pharmaceutical model, etc.) is most appropriate.

[0521] As a specific example, we determine that specialized models in medicine, chemistry, and pharmacy are required for tasks related to new drug development.

[0522] Resource Allocation Phase

[0523] 5. Server Processing

[0524] The server allocates the necessary computing resources (GPU, CPU, memory, etc.) to each selected specialized domain model.

[0525] As a specific example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[0526] 6. Server Processing

[0527] With the allocated resources, each model is prepared to perform its task.

[0528] Answer generation phase

[0529] 7. Server Processing

[0530] Each domain model generates a solution for the task it receives. For example, a chemistry model analyzes the ingredients of a new drug and generates a result.

[0531] 8. Server Processing

[0532] The server collects and integrates the solutions generated by each model, evaluates the contribution of each model, and generates a unified solution.

[0533] As a specific example, component analysis data obtained from a chemical model, data on effects on the human body obtained from a medical model, and efficacy data obtained from a pharmaceutical model are integrated and compiled into a single report format.

[0534] 9. Server Processing

[0535] The integrated answers are converted into an appropriate format and sent to the user's terminal.

[0536] Feedback and Retraining Phase

[0537] 10. User Operations

[0538] The user evaluates the answers received and enters feedback into the system.

[0539] For example, feedback such as "I'm satisfied with the analysis results" or "I need more information" can be entered.

[0540] 11. Terminal Processing

[0541] The user's rating is converted into an appropriate format and sent to the server.

[0542] 12. Server Processing

[0543] The server receives feedback and evaluates the performance of each domain-specific model, retraining it as needed to improve its accuracy the next time the task is performed.

[0544] In this way, the system of the present invention contributes to solving the user's problem by appropriately allocating the user's tasks to the specialized domain model and generating an answer efficiently and accurately.

[0545] The processing flow will be explained below.

[0546] Understood. Below, I will explain the process step by step.

[0547] ---

[0548] Step 1:

[0549] User operations

[0550] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[0551] Step 2:

[0552] Terminal handling

[0553] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\"}".

[0554] Step 3:

[0555] Server Processing

[0556] The server analyzes the received task information. It uses a natural language processing model to understand the content of the task and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[0557] Step 4:

[0558] Server Processing

[0559] Based on the analysis results, the server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.). For example, it may determine that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[0560] Step 5:

[0561] Server Processing

[0562] The server allocates the necessary computing resources (e.g., GPUs, CPUs, memory, etc.) to each selected specialty model. For example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[0563] Step 6:

[0564] Server Processing

[0565] The server prepares each model to execute the task using the allocated resources, provides each model with detailed task information, and starts the calculation.

[0566] Step 7:

[0567] Server Processing

[0568] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data of a new drug and generates a result.

[0569] Step 8:

[0570] Server Processing

[0571] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[0572] Step 9:

[0573] Server Processing

[0574] The integrated answer is converted into an appropriate format and sent to the user's terminal. For example, a report is generated as a "New Drug Ingredient Analysis Report" and sent to the user.

[0575] Step 10:

[0576] User operations

[0577] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[0578] Step 11:

[0579] Terminal handling

[0580] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[0581] Step 12:

[0582] Server Processing

[0583] The server receives feedback and evaluates the performance of each domain model. If necessary, it retrains the model to improve its accuracy the next time it performs the task. For example, a medical model can be retrained using new clinical data.

[0584] ---

[0585] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[0586] Example 1

[0587] 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."

[0588] For complex tasks that users want to solve, there is a need to select the optimal domain model, appropriately allocate the necessary computational resources, and generate solutions effectively and efficiently. However, current technology does not fully automate the processes of task analysis, model selection, and resource management, placing a heavy burden on users. In addition, the feedback process to improve the accuracy of execution results is not fully integrated. Therefore, a system is needed that reduces the time and effort required for users to solve tasks and improve the accuracy of results.

[0589] 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.

[0590] In this invention, the server includes a means for analyzing task information and extracting related keywords and concepts, a means for selecting a specialized domain model appropriate for the task, and a means for allocating the necessary computational resources to each selected specialized domain model. This enables the selection of an optimal model for the task registered by the user and the efficient allocation of resources. This ultimately results in the rapid generation of highly accurate answers and reduces the burden on the user. Furthermore, by collecting feedback from the user on the answer results and reflecting this feedback in subsequent processes, the accuracy and efficiency of the entire system can be improved.

[0591] The "means for users to register tasks they wish to solve" is an interface that allows users to input specific problems or issues they wish to solve in natural language and register them in the system.

[0592] "Means for analyzing task information and extracting related keywords and concepts" refers to a function for automatically analyzing task information registered by the user and extracting important keywords and related concepts from it.

[0593] The "means for selecting a specialized domain model appropriate for the task" refers to an algorithm or process for selecting the specialized domain model (e.g., medical model, chemical model, pharmaceutical model, etc.) that is optimal for the task based on the extracted keywords and concepts.

[0594] The "means for allocating the necessary computing resources to each selected specialized domain model" is a mechanism for dynamically allocating the necessary computing resources (GPU, CPU, memory, etc.) to efficiently operate the specialized domain model.

[0595] The "means by which each specialized domain model generates an answer for a received task" refers to the process by which a specialized domain model generates an answer or result based on a given task using specialized knowledge or algorithms.

[0596] "Means for collecting and integrating execution results from each specialized model" refers to an integration process that brings together the answers and results generated by each specialized model to derive the optimal answer overall.

[0597] "Means for converting the integrated execution results into an appropriate format and providing them to the user" refers to a function that converts the integrated answers and results into a format that is easy for the user to understand (for example, a report format or graph) and provides them to the user.

[0598] "Means for collecting feedback from users and reflecting it in each process of the system" refers to the process of collecting evaluations and opinions on the answers and results provided by users and reflecting them in improving the system and increasing accuracy the next time the task is performed.

[0599] The present invention provides a system that selects an optimal domain model for a task that a user wants to solve, allocates the necessary computational resources, generates a highly accurate solution based on the model, and provides the solution to the user. Specific embodiments of the system are described in detail below.

[0600] This system mainly consists of three entities: users, terminals, and servers.

[0601] First, the user inputs the task they want to solve in natural language on their own device. For example, they might input, "I want to perform ingredient analysis for new drug development." By registering such a task, they can request the system to solve the problem.

[0602] Next, the device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server, which then transmits the task information throughout the system.

[0603] The server first analyzes the received task information and extracts important keywords and related concepts. For example, keywords such as "new drug development" and "ingredient analysis" are extracted. This analysis is generally performed using a natural language processing (NLP) engine. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK, spaCy).

[0604] Based on the analysis results, the server selects the domain-specific model appropriate for the task. For example, the task of ingredient analysis in new drug development requires a medical model, a chemical model, and a pharmaceutical model. To select these domain-specific models, the server uses Bayesian inference and rule-based systems.

[0605] Next, the necessary computing resources are allocated to each selected domain-specific model. The server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. For example, two GPUs are allocated to a medical model, and three GPUs to a chemistry model.

[0606] Using the allocated resources, each specialized model performs a task and generates an answer based on its own expertise. For example, a chemistry model analyzes input ingredient information and evaluates chemical properties.

[0607] The server collects and integrates the results returned by each model. The integration uses an algorithm that evaluates the contribution of each model to generate the most appropriate answer. For example, it combines component analysis data from a chemical model, data on effects on the human body from a medical model, and drug efficacy data from a pharmaceutical model.

[0608] The integrated answers are converted into an appropriate data format and sent to the user's device. The user then checks and evaluates the answers on their own device. The evaluation is then entered into the system as feedback. For example, the user can send feedback such as "I'm satisfied with the analysis results" or "Additional information is needed."

[0609] The collected feedback is analyzed by the server and used to improve the system's performance, specifically by retraining each domain-specific model to improve its accuracy when performing subsequent tasks.

[0610] Through the above process, the system of the present invention can provide highly accurate and efficient solutions to tasks input by the user, thereby enabling the user to solve problems quickly and reliably.

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

[0612] Step 1: The user enters the task they want to solve in natural language into the terminal and registers it in the system. For example, they might enter, "I want to perform ingredient analysis in new drug development." This is the input data. The terminal receives this data and displays it on the screen.

[0613] Step 2: The terminal converts the input task information into an appropriate data format (for example, JSON format). Specifically, the terminal converts the input text into a JSON object (for example, "task": "I would like to perform component analysis in new drug development"). This becomes the output data. This converted data is used in the next step.

[0614] Step 3: The terminal sends the converted data to the server. Specifically, the terminal uses an HTTP request to send JSON formatted data to the server. The input is JSON data, and the output is data transmission to the server.

[0615] Step 4: The server analyzes the received task information. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK or spaCy) to extract important keywords (e.g., "new drug development" and "ingredient analysis"). The input is the received JSON data, and the output is the extracted keywords.

[0616] Step 5: The server selects appropriate domain models based on the analysis results. Specifically, the server uses Bayesian inference and rule-based systems to select medical, chemical, and pharmaceutical models. The input is the extracted keywords, and the output is a list of selected models.

[0617] Step 6: The server allocates the necessary computing resources for each selected domain model. Specifically, the server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. The input is a list of selected models, and the output is the details of the resource allocation. For example, two GPUs are allocated to the medical model, and three GPUs are allocated to the chemistry model.

[0618] Step 7: The server prepares each model to run by adding a task to a queue for each model and launching a script to initialize the execution environment. The input is the resource allocation details, and the output is the ready state of each model.

[0619] Step 8: Each domain model generates a solution for the received task. Specifically, the chemical model analyzes the input ingredient information and evaluates chemical attributes. The input is the readiness status and resources of each model, and the output is the solution generated by each model.

[0620] Step 9: The server collects and integrates the answers from each model. Specifically, it aggregates the data returned from each model and generates a unified answer using a contribution evaluation algorithm. The input is the answer from each model, and the output is the integrated answer.

[0621] Step 10: The server converts the consolidated answer into an appropriate format and sends it to the user's device. Specifically, it converts the consolidated answer data into JSON or report format and returns it to the user using an HTTP response. The input is the consolidated answer, and the output is the answer returned to the user's device.

[0622] Step 11: The user evaluates the answer they received and enters their feedback into the system. For example, they might enter something like "I'm satisfied with the analysis results" or "I need more information." This is the input of feedback data.

[0623] Step 12: The device converts the user's rating into an appropriate format and sends it to the server. Specifically, the device converts the feedback into JSON format and sends it to the server via an HTTP request. The input is the feedback data, and the output is the data transmission to the server.

[0624] Step 13: The server receives the feedback, evaluates the performance of each domain model, and performs retraining. Specifically, it uses the feedback data to update the model's training dataset and retrains it. The input is the received feedback data, and the output is the updated model.

[0625] (Application example 1)

[0626] 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."

[0627] Modern factories have many tasks, and their efficient execution directly leads to improved productivity. However, optimal resource allocation and appropriate utilization of specialized knowledge for each task is difficult. Dynamic improvements based on feedback are also required, making it difficult to address these challenges with traditional static models. Therefore, a new system is needed that can efficiently optimize tasks and improve the performance of factory production lines.

[0628] 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.

[0629] In this invention, the server includes means for a user to register a task to be solved, means for selecting a specialized domain model suitable for the task, means for allocating necessary computational resources to the selected model, means for the model to execute the task using the allocated resources, means for integrating the execution results and providing them to the user, means for collecting feedback from the user and reflecting it in subsequent processes, and means for generating a plan for having robots in the factory execute the task and appropriately allocating resources, thereby enabling efficient optimization of factory tasks and dynamic improvement.

[0630] "User" refers to a person or organization that uses the system.

[0631] A "task" refers to a specific problem or task that a user wants the system to solve.

[0632] "Domain models" refer to machine learning or artificial intelligence algorithms or software that are specialized for a particular field or problem.

[0633] "Computational resources" refers to the computer hardware and software resources required to perform a task, including CPU, GPU, memory, etc.

[0634] "Execution" refers to the process by which a selected domain expertise model processes a task using the allocated computing resources.

[0635] "Synthesis" refers to the process of combining the results of multiple disciplinary models into a single conclusion or answer.

[0636] "Feedback" refers to the evaluations and opinions provided by users, and the system is improved based on that information.

[0637] "Factory robots" refer to autonomous or semi-autonomous machines that perform tasks in a factory.

[0638] "Generating a plan" refers to the process of automatically creating optimal execution procedures for a task.

[0639] "Appropriate resource allocation" refers to efficiently allocating the necessary computing resources and hardware to each task or model.

[0640] The present invention is a system that allows users to register a task they wish to solve, selects an appropriate domain model, efficiently allocates the necessary computational resources, and generates a highly accurate solution. This system is intended to be applied particularly to robots in factories.

[0641] System Program

[0642] First, the user inputs the task they want to solve in natural language and registers it in the system. For example, a task might be, "Please suggest a way to resolve a bottleneck on the production line."

[0643] The server analyzes the input task information and extracts important keywords and concepts. Based on the extracted keywords, it selects the optimal domain model. For example, keywords such as "production line," "bottleneck," "solution," "method," and "proposal" are extracted.

[0644] Next, the server allocates the necessary computing resources to the selected domain model. In this case, CPUs and GPUs are used as computing resources, and specific examples include NVIDIA GPUs and Intel CPUs.

[0645] Using the allocated resources, each domain model performs tasks and generates solutions, such as optimizing a manufacturing process or analyzing the causes of bottlenecks.

[0646] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer, which is provided to the user in the form of a report. This process utilizes natural language generation technology and a generative AI model.

[0647] Users evaluate the answers they receive and provide feedback to the system, which then uses that feedback to improve the task-solving process and, if necessary, retrain the domain model.

[0648] Hardware and software used

[0649] Hardware: Factory robots, dedicated GPU servers, user devices (PCs, smartphones, etc.)

[0650] Software: Natural language processing libraries (NLTK, SpaCy), AI model selection libraries (custom libraries), resource allocation libraries, natural language generation models (GPT-3, etc.)

[0651] Adding specific examples

[0652] As a concrete example, consider the case where a user inputs "Please suggest a method to resolve bottlenecks in the production line." The system analyzes this task, extracts keywords such as "production line," "bottleneck," "resolve," "method," and "proposal," and based on these, selects the most appropriate domain-specific model (e.g., production engineering model, physical simulation model). Each model independently proposes a solution, and finally provides an integrated report.

[0653] An example of a prompt to be input to a generative AI model: "Please suggest a way to resolve a bottleneck on the production line."

[0654] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0655] Step 1:

[0656] Users input the task they want to solve in natural language from their own terminal and register it in the system.

[0657] Input: User task input (e.g., "Please suggest a way to resolve the bottleneck on the production line.")

[0658] Output: Task information is converted to JSON format and sent to the server.

[0659] Specific operation: When a user enters the task details in a dedicated application and presses the send button, the task information is converted into JSON format and sent to the server via the Internet.

[0660] Step 2:

[0661] The server analyzes the received task information and extracts important keywords and concepts.

[0662] Input: Task information sent to the server (JSON format)

[0663] Output: Extracted keywords and concepts (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[0664] How it works: The server parses the task information in JSON format and uses a natural language processing library (such as NLTK or SpaCy) to extract relevant keywords. During this process, it uses a specific algorithm to determine the frequency and importance of the keywords.

[0665] Step 3:

[0666] The server selects the most suitable specialized domain model based on the extracted keywords.

[0667] Input: Extracted keywords (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[0668] Output: List of selected domain-specific models (e.g., industrial engineering models, physical simulation models)

[0669] Specific operation: Based on the extracted keywords, the server selects an appropriate domain-specific model using a model selection library. This library recommends the best model based on past data and experience.

[0670] Step 4:

[0671] The server allocates the necessary computational resources to each selected domain model.

[0672] Input: List of selected domain models

[0673] Output: Allocated computing resources (e.g., 2 GPUs, 4 CPU cores, etc.)

[0674] How it works: The server uses a resource allocation library to allocate optimal resources to each model. For example, a physics simulation model requires a large number of GPUs, so it calculates and allocates appropriate resources.

[0675] Step 5:

[0676] The server uses the allocated resources to have each domain model perform a task and generate an answer.

[0677] Input: Allocated computing resources, selected domain model

[0678] Output: The answer generated by each model (e.g., "How to optimize a manufacturing process?")

[0679] How it works: The server runs each model and makes full use of its resources to generate the information needed to solve the task. Specifically, the production engineering model generates proposals for improving the efficiency of the manufacturing process, and the physical simulation model generates simulation results for eliminating bottlenecks.

[0680] Step 6:

[0681] The server collects the solutions generated from each model and combines them to generate a single unified solution.

[0682] Input: Answers generated by each model

[0683] Output: Unified answer (e.g. unified report)

[0684] How it works: The server uses a natural language generation model to integrate the answers from each model. During this process, it evaluates the contribution of each model and prioritizes the most important information. It then uses a generative AI model (e.g., GPT-3) to compile the results into a readable report.

[0685] Step 7:

[0686] The server sends the integrated answer to the user's terminal.

[0687] Input: Consolidated Answer

[0688] Output: The answer displayed on the user's terminal

[0689] Specific operation: The server converts the integrated answers into JSON format or another appropriate data format and sends them to the user's device. The user can then review the answers and use them to improve the manufacturing process.

[0690] Step 8:

[0691] Users rate the answers they receive and enter their feedback into the system.

[0692] Input: User rating and feedback (e.g., "Satisfied with the analysis results" or "More information needed")

[0693] Output: The feedback information is sent to the server.

[0694] Specific operation: The user uses a dedicated interface to input evaluations of answers and requests for additions, and sends them to the server.

[0695] Step 9:

[0696] The server uses the feedback it receives to improve the task-solving process for future tasks, including retraining the domain expertise model if necessary.

[0697] Input: User feedback

[0698] Output: Improved model and system settings

[0699] What happens: The server analyzes the feedback, identifies areas for improvement in the model and process, and, if necessary, retrains the domain-specific model to reflect it in the next task-solving process.

[0700] 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.

[0701] Understood. Below is the "Form for carrying out the invention".

[0702] ---

[0703] This invention is a system that utilizes a domain model and an emotion engine to efficiently allocate computational resources to obtain highly accurate answers for tasks registered by users. The processing of the system program is described in detail below.

[0704] System Overview

[0705] This system consists of a series of processes: a user registers the task they wish to solve, selects an appropriate domain model and emotion engine, allocates the necessary computing resources, generates an answer to the task, and provides the user with the integrated answer.

[0706] Task registration phase

[0707] 1. User operations

[0708] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[0709] An emotion engine analyzes emotion information from user input to measure, for example, urgency or anxiety.

[0710] 2. Terminal Processing

[0711] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[0712] Task analysis and model selection phase

[0713] 3. Server Processing

[0714] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[0715] Emotional data is also taken into consideration and the task is determined to be urgent.

[0716] 4. Server Processing

[0717] The server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the emotion data and analysis results. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[0718] Resource Allocation Phase

[0719] 5. Server Processing

[0720] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. If the emotional data indicates a high level of urgency, more resources are allocated. For example, a medical model may be allocated twice as many GPUs as usual.

[0721] 6. Server Processing

[0722] Using the allocated resources, prepare each model to execute the task. Provide each model with detailed task information and emotion information, and start the calculation.

[0723] Answer generation phase

[0724] 7. Server Processing

[0725] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data for a new drug and generates a result.

[0726] 8. Server Processing

[0727] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[0728] 9. Server Processing

[0729] The integrated answers are converted into an appropriate format and sent to the user's device. Emotional data is also reflected, and explanations that take the user's feelings into consideration are added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[0730] Feedback and Retraining Phase

[0731] 10. User Operations

[0732] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[0733] 11. Terminal Processing

[0734] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[0735] 12. Server Processing

[0736] The server receives the feedback and evaluates the performance of each specialized model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data. It also takes into account emotional data, reflecting emotional fluctuations due to the feedback.

[0737] In this way, the system of the present invention appropriately allocates the user's tasks to the domain model and emotion engine, and generates answers efficiently and accurately, thereby contributing to the user's problem solving.

[0738] The processing flow will be explained below.

[0739] I understand. I will explain the process in detail in the following format.

[0740] ---

[0741] Step 1:

[0742] User operations

[0743] The user enters the task they want to solve in natural language on their device and registers it in the system. For example, they might enter "I want to perform ingredient analysis for new drug development" and press the send button. At this time, the emotion engine also analyzes emotional information from the user's facial expressions and voice.

[0744] Step 2:

[0745] Terminal handling

[0746] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[0747] Step 3:

[0748] Server Processing

[0749] The server analyzes the received task information and emotional data. It uses a natural language processing model to understand the content of the task and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "ingredient analysis." It also determines the urgency and importance of the task from the emotional data.

[0750] Step 4:

[0751] Server Processing

[0752] The server selects the most appropriate specialized model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the analysis results and emotion data. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug. Furthermore, based on the emotion data, it creates a plan to allocate more resources to high-priority tasks.

[0753] Step 5:

[0754] Server Processing

[0755] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. Based on the emotion data, if the level of urgency is high, the server will increase resource allocation. For example, a medical model will be allocated twice as many GPUs as usual.

[0756] Step 6:

[0757] Server Processing

[0758] The server checks the resource allocation for each model and sends task details to each model, which prepares to execute the task and starts the calculation process.

[0759] Step 7:

[0760] Server Processing

[0761] Each specialized domain model generates a solution for the task it receives. For example, a chemistry model calculates component analysis data for a new drug and sends the results to a server.

[0762] Step 8:

[0763] Server Processing

[0764] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles component analysis data from chemical models, human body effect data from medical models, and drug efficacy data from pharmaceutical models into a single report format.

[0765] Step 9:

[0766] Server Processing

[0767] The integrated answer is converted into an appropriate format for presentation to the user and sent to the user's device. Emotional data is also reflected, and an explanation that takes the user's emotions into consideration is added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[0768] Step 10:

[0769] User operations

[0770] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information," along with emotional data.

[0771] Step 11:

[0772] Terminal handling

[0773] The user's device converts the feedback and emotion data into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis results\",\"emotion\":\"satisfied\"}\".

[0774] Step 12:

[0775] Server Processing

[0776] The server receives the feedback and emotion data and evaluates the performance of each domain-specific model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data to reflect changes in emotion due to feedback.

[0777] ---

[0778] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[0779] Example 2

[0780] 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."

[0781] Conventional problem-solving systems have difficulty efficiently analyzing user tasks and providing highly accurate solutions. Furthermore, they often fail to adequately consider the user's emotional information, resulting in a poor user experience. Furthermore, they fail to select an expert domain model or optimally allocate computing resources, resulting in the waste of computing resources. The purpose of this invention is to solve these problems.

[0782] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for analyzing task information and extracting appropriate keywords and concepts, means for selecting a specialized domain model based on the task and emotion information, and means for allocating necessary computational resources to the selected specialized domain model. This enables the generation of highly accurate answers to tasks. Furthermore, by taking the user's emotion information into consideration, a more satisfying user experience can be provided. Furthermore, efficient allocation of computational resources improves overall performance and reduces waste.

[0783] A "user" is an individual or group that uses the system to register tasks and receive answers.

[0784] A "task" refers to a specific problem or requirement that a user wants to solve.

[0785] "Emotion information" refers to data obtained by analyzing the user's emotional state at the time of task input.

[0786] A "domain model" is a computational model that includes knowledge and algorithms in a specific field.

[0787] "Computational resources" refers to the hardware and software resources (e.g., CPU, GPU, memory, etc.) required to run a domain-specific model.

[0788] "Answer" refers to the result or conclusion that a domain model generates for a task.

[0789] "Feedback" refers to the act of a user providing the system with an evaluation or opinion on the answer they received.

[0790] "Retraining" refers to the process of updating the domain-specific model based on feedback to improve accuracy the next time the task is performed.

[0791] The "server" refers to a central computing device that analyzes task information and emotional information, selects specialized domain models, allocates computing resources, and generates and integrates answers.

[0792] "Terminal" refers to a device (e.g., PC, smartphone, tablet, etc.) that a user uses to input tasks and receive answers.

[0793] "Data format" refers to a standardized structure (e.g., JSON, XML, etc.) for storing and transmitting information.

[0794] This invention is a system that utilizes domain expertise models and emotion engines to efficiently allocate computational resources to tasks registered by users and obtain highly accurate answers. The program for this system includes a series of processes from task registration to answer generation, feedback collection, and relearning.

[0795] In this system, a user first inputs the task they want to solve in natural language using a device (e.g., PC, smartphone, tablet, etc.). For example, they might input, "I would like you to analyze the ingredients for new drug development." The emotion engine then analyzes the user's input and extracts emotional information (e.g., urgency and anxiety). The device then converts this task information and emotional information into an appropriate data format (e.g., JSON format) and sends it to the server.

[0796] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis." It also takes into account emotional information to determine the urgency of the task. Based on this, the server selects the domain model (e.g., medical model, chemistry model, pharmaceutical model) that is best suited to the task.

[0797] Next, the server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected domain-specific model. If the emotional information indicates a high level of urgency, more computing resources are allocated. For example, a medical model is allocated twice as many GPUs as usual. Using the allocated resources, each model generates an answer to the task.

[0798] The server collects and integrates the generated answers from each specialized model. During integration, the contribution of each model is evaluated and a unified answer is generated. For example, the component analysis data from the chemical model and the human body effect data from the medical model are combined and integrated into a report format. The server then converts the integrated answer into an appropriate format and sends it to the user's device. The server also reflects the user's emotional information and adds an explanation that takes the user's feelings into consideration. For example, a report titled "Component Analysis Report of a New Drug" could be generated, with a comment added such as "We determined that an emergency response was required, so we conducted the analysis quickly."

[0799] Furthermore, the user evaluates the answers they receive and inputs their feedback into the system. For example, they may enter feedback such as "I'm satisfied with the analysis results" or "I need more information." The device converts the feedback into an appropriate format and sends it to the server. The server receives the feedback and evaluates the performance of each specialized domain model. If necessary, retraining is performed to improve accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data.

[0800] Through this series of processes, the system uses the domain model and emotion engine to process user tasks efficiently and with high accuracy, significantly contributing to solving user problems. Furthermore, by taking emotion information into consideration, the system can increase user satisfaction.

[0801] Examples of specific examples and prompts

[0802] Specific examples

[0803] The user types into the terminal, "I would like to request a component analysis for new drug development."

[0804] The server selects a specialized area model related to new drug development, determines that the task is of high urgency, and allocates more computing resources to it.

[0805] Each specialized model starts its analysis and sends the results to the server.

[0806] The integrated answer is provided to the user, along with a comment such as, "We determined that an emergency response was required, so we conducted a rapid analysis."

[0807] Prompt Sentence Examples

[0808] Input: "Please provide detailed ingredient analysis for my drug development project. It's very important."

[0809] The above is a specific embodiment for carrying out the present invention.

[0810] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0811] System processing steps

[0812] Step 1: Task Input and Sentiment Analysis

[0813] The user inputs the task they want to solve using the terminal in natural language.

[0814] Input: "I would like to request a component analysis for new drug development."

[0815] How it works: The device uses an emotion engine to extract emotion information (e.g., urgency) from the user's input.

[0816] Step 2: Convert to data format and send

[0817] The device converts the input task information and emotion information into an appropriate data format (e.g., JSON format) and sends it to the server.

[0818] Input: User task information and emotion information

[0819] Output: JSON data "{\"task\":\"I would like you to analyze ingredients for new drug development\",\"emotion\":\"high urgency\"}"

[0820] What it does: Serializes input data into JSON format and sends it to the server.

[0821] Step 3: Task analysis and keyword extraction

[0822] The server analyzes the received task information and extracts keywords and concepts.

[0823] Input: JSON data

[0824] Output: Keywords "new drug development" and "component analysis"

[0825] How it works: Extracts keywords using natural language processing algorithms.

[0826] Step 4: Selecting a domain model

[0827] The server selects an appropriate domain model based on task information and emotion information.

[0828] Input: Keywords, emotion information

[0829] Output: Selected domain models (e.g., medical model, chemical model)

[0830] Action: Searches for and selects a corresponding model from the database.

[0831] Step 5: Allocate Computational Resources

[0832] The server allocates the necessary computing resources to the selected domain model.

[0833] Input: Selected domain model, emotion information

[0834] Output: Allocated computing resources (e.g. GPU, CPU specs)

[0835] How it works: A resource management system is used to allocate optimal resources to each model.

[0836] Step 6: Prepare and provide information about the model

[0837] The server prepares each model to run using the allocated resources and provides task and emotion information.

[0838] Input: Liquidation resources, task information, emotion information

[0839] Output: Ready to run

[0840] How it works: Loads the necessary libraries and datasets, and feeds the model with task information.

[0841] Step 7: Generate the answer

[0842] The server uses each domain model to generate an answer to the task.

[0843] Input: Task information provided to the model

[0844] Output: Answer data for each model

[0845] Action: The model performs a calculation and generates an answer (e.g., compositional analysis data).

[0846] Step 8: Consolidating the Answers

[0847] The server collects and integrates the answers generated by each model.

[0848] Input: Answer data for each model

[0849] Output: Consolidated answer

[0850] How it works: Evaluate the contribution of each answer and generate a single unified answer.

[0851] Step 9: Provide the answer

[0852] The server converts the integrated answers into an appropriate format and sends them to the user's device.

[0853] Input: Integrated answers, emotional information

[0854] Output: Answer data to the user

[0855] What it does: Converts the answer data into a user-friendly format (e.g., report format) and adds sentiment-sensitive comments.

[0856] Step 10: Gather feedback

[0857] The user rates the answers they receive and provides feedback.

[0858] Input: Rating and comments on the answer

[0859] Output: Feedback data

[0860] How it works: Use the input form to enter your rating and comments and submit your feedback.

[0861] Step 11: Submit your feedback

[0862] The device converts the feedback into an appropriate format and sends it to the server.

[0863] Input: User feedback

[0864] Output: Feedback data in JSON format

[0865] What it does: Serializes the feedback into JSON format and sends it to the server.

[0866] Step 12: Retraining

[0867] The server evaluates the performance of each specialized domain model based on the feedback and performs re-learning.

[0868] Input: Feedback data

[0869] Output: Updated model

[0870] How it works: Analyzes feedback and applies a re-training algorithm to improve the model.

[0871] The above is the specific program processing flow of this system.

[0872] (Application example 2)

[0873] 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."

[0874] In modern content delivery services, it is extremely important to provide users with the content they need quickly and accurately. However, conventional systems do not dynamically allocate resources based on the user's emotional state or the urgency of the task, limiting the improvement of the user experience. Furthermore, they lack a mechanism for effectively utilizing user feedback to improve accuracy in subsequent task executions. As a result, user satisfaction declines, leading to a diminishing value of the service.

[0875] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0876] In this invention, the server includes: means for analyzing a user's emotional information, determining the urgency based on the information, and adjusting the allocation of computational resources; means for the user to register a task they wish to solve; means for selecting a specialized domain model appropriate for the task; means for allocating the necessary computational resources to the selected model; means for the model to execute the task using the allocated resources; means for integrating the execution results and providing them to the user; and means for collecting user feedback and reflecting it in subsequent processes. This enables dynamic resource allocation according to the user's emotional state and the urgency of the task, thereby enabling the rapid and accurate provision of content that meets the user's needs. Furthermore, by effectively utilizing user feedback to retrain the model, accuracy can be improved the next time the task is executed, thereby increasing user satisfaction.

[0877] A "task" refers to a specific request or purpose that a user requests the system to solve.

[0878] A "specialized domain model" is a machine learning model or algorithm built based on knowledge and data in a specific field or area.

[0879] "Emotional information" is emotional data such as emotional state and urgency that is analyzed from the user's input and behavior.

[0880] "Computational resources" refers to the computational resources such as CPU, GPU, and memory required to execute a task.

[0881] "Resource allocation" refers to the process of allocating the necessary computational resources to the selected model.

[0882] "Model execution" refers to the execution of calculations and analysis for a task by a selected domain-specific model using allocated computational resources.

[0883] "Execution result integration" refers to the compilation of results generated by multiple disciplinary models into a single unified answer or report.

[0884] "Feedback" refers to information indicating the user's evaluation and satisfaction with the answer they received.

[0885] "Retraining" is a learning process that improves the performance and accuracy of a domain-specific model based on collected feedback and newly acquired data.

[0886] "Urgency" is an indicator of how quickly a user's task should be resolved.

[0887] This invention is a system for quickly and accurately providing users with content they desire through a content distribution service. The system analyzes users' emotional information, determines the urgency of tasks based on that information, and dynamically adjusts the allocation of computing resources to provide personalized content tailored to the user's needs.

[0888] System Overview

[0889] This system utilizes domain expertise models and emotion engines to efficiently allocate the necessary computing resources to tasks registered by users, and provides highly accurate answers and content to users.

[0890] Hardware and software used

[0891] Hardware: Uses high-speed CPU, GPU and memory.

[0892] Software: Python 3.x, SentimentEngine (for sentiment analysis), and domain-specific models (news model, music model, video model, etc.).

[0893] System processing flow

[0894] 1. User Input and Sentiment Analysis

[0895] The user inputs the task they want to solve in natural language from their device. For example, they might input something like, "I want to know the latest technology news." The server receives this input and uses an emotion engine to analyze the emotional information and measure the urgency and anxiety of the task. The emotional information is treated as data that indicates, for example, the "high urgency" of the task.

[0896] 2. Selecting a Specialty Model and Allocating Resources

[0897] The server extracts necessary keywords and concepts based on the analyzed task information and emotion information, and selects the appropriate domain model. For example, for the keyword "technology news," a news model is selected. Then, the necessary computing resources are allocated based on the emotion data (urgency). If the urgency is high, more resources than usual are allocated.

[0898] 3. Execute the task

[0899] Using the allocated resources, selected domain-specific models perform tasks, such as a technology news model that analyzes the latest news and generates relevant articles.

[0900] 4. Consolidating and Presenting Results

[0901] The server collects and integrates the results generated by each model. This allows it to evaluate the contribution of each model and generate the optimal answer. The integrated answer is converted into an appropriate format and provided to the user along with a comment that takes into account the user's emotional information. For example, it may include a message such as, "Please check the technology news below. Due to its urgency, we analyzed it quickly."

[0902] 5. Feedback and Retraining

[0903] Users provide feedback on the answers they provide, and the server receives this feedback, evaluates the performance of each domain-specific model, and retrains it as needed, improving the model's accuracy the next time the task is performed.

[0904] Specific examples

[0905] For example, if a user requests "I want to know the latest technology news," the following prompt might be generated:

[0906] What's the latest technology news? It's very urgent right now.

[0907] Based on this prompt, the server performs a series of processes, including sentiment analysis, selection of domain expertise models, allocation of computing resources, task execution, integration and provision of results, and feedback collection and retraining, thereby quickly providing users with personalized, up-to-date technology news.

[0908] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0909] Step 1:

[0910] The device accepts the user's natural language input. The user types "I want to know the latest technology news" into the device. The device converts this input data into an appropriate data format (e.g., JSON) to send it to the server.

[0911] Input: User's natural language input

[0912] Data processing: converting natural language into data format

[0913] Output: JSON format data

[0914] Step 2:

[0915] The server receives the user input sent from the device and performs emotion analysis using an emotion engine. The emotion engine analyzes "urgency" and "anxiety" and extracts emotion information such as "high urgency."

[0916] Input: User-supplied data in JSON format

[0917] Data Computing: Sentiment Analysis

[0918] Output: Emotional information (e.g. "High Urgency")

[0919] Step 3:

[0920] The server extracts necessary keywords and concepts from the analyzed emotion information and user input. The keyword "technology news" is identified.

[0921] Input: User input and emotional information

[0922] Data processing: Keyword extraction

[0923] Output: The keyword "technology news"

[0924] Step 4:

[0925] The server selects an appropriate domain model (e.g., a news model) based on keywords and emotion information. At the same time, it dynamically allocates the necessary computing resources according to the emotion information, particularly the urgency. If the urgency is high, more CPU and GPU resources than usual are allocated.

[0926] Input: Extracted keywords and sentiment information

[0927] Data calculation: model selection, computational resource allocation

[0928] Output: Selected model and allocated resources

[0929] Step 5:

[0930] The server uses the allocated resources to run the selected news model. The model analyzes the task (e.g., get the latest technology news) and generates recommended news articles.

[0931] Input: Selected model and allocated resources

[0932] Data calculation: Model execution, recommended content generation

[0933] Output: Generated content (e.g., news article)

[0934] Step 6:

[0935] The server collects and integrates the generated content. If there are results from multiple domain-specific models, it evaluates each and generates a single integrated answer. At this time, it adds comments that take into account the user's emotional information.

[0936] Input: Multiple Content

[0937] Data processing: integrating results and adding comments

[0938] Output: Consolidated content and comments

[0939] Step 7:

[0940] The server transmits the integrated content to the terminal, which displays the content on the user's screen and makes it available to the user.

[0941] Input: Integrated content and comments

[0942] Data output: Send to terminal

[0943] Output: The content that is displayed on the user's screen

[0944] Step 8:

[0945] The user inputs feedback on the provided content, and the terminal converts this feedback into an appropriate data format for transmission to the server.

[0946] Input: User feedback

[0947] Data processing: Convert feedback into data format

[0948] Output: Feedback data in JSON format

[0949] Step 9:

[0950] The server analyzes the feedback received from the device and uses it to evaluate the performance of each specialized domain model and retrain it, and if necessary, retrains the model to improve its accuracy the next time the task is performed.

[0951] Input: Feedback data

[0952] Data calculation: performance evaluation, re-learning

[0953] Output: Improved model

[0954] 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.

[0955] 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.

[0956] 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.

[0957] [Third embodiment]

[0958] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0959] 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.

[0960] 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).

[0961] 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.

[0962] 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.

[0963] 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).

[0964] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0965] 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.

[0966] 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.

[0967] 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.

[0968] 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.

[0969] 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."

[0970] Understood. Below is the "Form for carrying out the invention".

[0971] ---

[0972] The present invention is a system for obtaining highly accurate solutions by selecting a domain model for a task registered by a user and efficiently allocating computational resources. The processing of the system program is described in detail below.

[0973] System Overview

[0974] This system consists of a series of processes in which a user registers a task they wish to solve, selects an appropriate domain model, allocates the necessary computational resources, generates an answer to the task, and provides the user with the integrated answer.

[0975] Task registration phase

[0976] 1. User operations

[0977] Users input the task they want to solve in natural language from their own device and register it in the system.

[0978] As a specific example, when a user requests a component analysis of a new drug, the user inputs "I would like to perform component analysis in new drug development."

[0979] 2. Terminal Processing

[0980] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server.

[0981] Task analysis and model selection phase

[0982] 3. Server Processing

[0983] The server analyzes the received task information and extracts necessary keywords and concepts. For example, keywords such as "new drug development" and "component analysis" are identified.

[0984] 4. Server Processing

[0985] Based on the analysis results, the server selects which specialized model (medical model, chemistry model, pharmaceutical model, etc.) is most appropriate.

[0986] As a specific example, we determine that specialized models in medicine, chemistry, and pharmacy are required for tasks related to new drug development.

[0987] Resource Allocation Phase

[0988] 5. Server Processing

[0989] The server allocates the necessary computing resources (GPU, CPU, memory, etc.) to each selected specialized domain model.

[0990] As a specific example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[0991] 6. Server Processing

[0992] With the allocated resources, each model is prepared to perform its task.

[0993] Answer generation phase

[0994] 7. Server Processing

[0995] Each domain model generates a solution for the task it receives. For example, a chemistry model analyzes the ingredients of a new drug and generates a result.

[0996] 8. Server Processing

[0997] The server collects and integrates the solutions generated by each model, evaluates the contribution of each model, and generates a unified solution.

[0998] As a specific example, component analysis data obtained from a chemical model, data on effects on the human body obtained from a medical model, and efficacy data obtained from a pharmaceutical model are integrated and compiled into a single report format.

[0999] 9. Server Processing

[1000] The integrated answers are converted into an appropriate format and sent to the user's terminal.

[1001] Feedback and Retraining Phase

[1002] 10. User Operations

[1003] The user evaluates the answers received and enters feedback into the system.

[1004] For example, feedback such as "I'm satisfied with the analysis results" or "I need more information" can be entered.

[1005] 11. Terminal Processing

[1006] The user's rating is converted into an appropriate format and sent to the server.

[1007] 12. Server Processing

[1008] The server receives feedback and evaluates the performance of each domain-specific model, retraining it as needed to improve its accuracy the next time the task is performed.

[1009] In this way, the system of the present invention contributes to solving the user's problem by appropriately allocating the user's tasks to the specialized domain model and generating an answer efficiently and accurately.

[1010] The processing flow will be explained below.

[1011] Understood. Below, I will explain the process step by step.

[1012] ---

[1013] Step 1:

[1014] User operations

[1015] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[1016] Step 2:

[1017] Terminal handling

[1018] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\"}".

[1019] Step 3:

[1020] Server Processing

[1021] The server analyzes the received task information. It uses a natural language processing model to understand the content of the task and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[1022] Step 4:

[1023] Server Processing

[1024] Based on the analysis results, the server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.). For example, it may determine that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[1025] Step 5:

[1026] Server Processing

[1027] The server allocates the necessary computing resources (e.g., GPUs, CPUs, memory, etc.) to each selected specialty model. For example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[1028] Step 6:

[1029] Server Processing

[1030] The server prepares each model to execute the task using the allocated resources, provides each model with detailed task information, and starts the calculation.

[1031] Step 7:

[1032] Server Processing

[1033] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data of a new drug and generates a result.

[1034] Step 8:

[1035] Server Processing

[1036] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[1037] Step 9:

[1038] Server Processing

[1039] The integrated answer is converted into an appropriate format and sent to the user's terminal. For example, a report is generated as a "New Drug Ingredient Analysis Report" and sent to the user.

[1040] Step 10:

[1041] User operations

[1042] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[1043] Step 11:

[1044] Terminal handling

[1045] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[1046] Step 12:

[1047] Server Processing

[1048] The server receives feedback and evaluates the performance of each domain model. If necessary, it retrains the model to improve its accuracy the next time it performs the task. For example, a medical model can be retrained using new clinical data.

[1049] ---

[1050] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[1051] Example 1

[1052] 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."

[1053] For complex tasks that users want to solve, there is a need to select the optimal domain model, appropriately allocate the necessary computational resources, and generate solutions effectively and efficiently. However, current technology does not fully automate the processes of task analysis, model selection, and resource management, placing a heavy burden on users. In addition, the feedback process to improve the accuracy of execution results is not fully integrated. Therefore, a system is needed that reduces the time and effort required for users to solve tasks and improve the accuracy of results.

[1054] 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.

[1055] In this invention, the server includes a means for analyzing task information and extracting related keywords and concepts, a means for selecting a specialized domain model appropriate for the task, and a means for allocating the necessary computational resources to each selected specialized domain model. This enables the selection of an optimal model for the task registered by the user and the efficient allocation of resources. This ultimately results in the rapid generation of highly accurate answers and reduces the burden on the user. Furthermore, by collecting feedback from the user on the answer results and reflecting this feedback in subsequent processes, the accuracy and efficiency of the entire system can be improved.

[1056] The "means for users to register tasks they wish to solve" is an interface that allows users to input specific problems or issues they wish to solve in natural language and register them in the system.

[1057] "Means for analyzing task information and extracting related keywords and concepts" refers to a function for automatically analyzing task information registered by the user and extracting important keywords and related concepts from it.

[1058] The "means for selecting a specialized domain model appropriate for the task" refers to an algorithm or process for selecting the specialized domain model (e.g., medical model, chemical model, pharmaceutical model, etc.) that is optimal for the task based on the extracted keywords and concepts.

[1059] The "means for allocating the necessary computing resources to each selected specialized domain model" is a mechanism for dynamically allocating the necessary computing resources (GPU, CPU, memory, etc.) to efficiently operate the specialized domain model.

[1060] The "means by which each specialized domain model generates an answer for a received task" refers to the process by which a specialized domain model generates an answer or result based on a given task using specialized knowledge or algorithms.

[1061] "Means for collecting and integrating execution results from each specialized model" refers to an integration process that brings together the answers and results generated by each specialized model to derive the optimal answer overall.

[1062] "Means for converting the integrated execution results into an appropriate format and providing them to the user" refers to a function that converts the integrated answers and results into a format that is easy for the user to understand (for example, a report format or graph) and provides them to the user.

[1063] "Means for collecting feedback from users and reflecting it in each process of the system" refers to the process of collecting evaluations and opinions on the answers and results provided by users and reflecting them in improving the system and increasing accuracy the next time the task is performed.

[1064] The present invention provides a system that selects an optimal domain model for a task that a user wants to solve, allocates the necessary computational resources, generates a highly accurate solution based on the model, and provides the solution to the user. Specific embodiments of the system are described in detail below.

[1065] This system mainly consists of three entities: users, terminals, and servers.

[1066] First, the user inputs the task they want to solve in natural language on their own device. For example, they might input, "I want to perform ingredient analysis for new drug development." By registering such a task, they can request the system to solve the problem.

[1067] Next, the device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server, which then transmits the task information throughout the system.

[1068] The server first analyzes the received task information and extracts important keywords and related concepts. For example, keywords such as "new drug development" and "ingredient analysis" are extracted. This analysis is generally performed using a natural language processing (NLP) engine. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK, spaCy).

[1069] Based on the analysis results, the server selects the domain-specific model appropriate for the task. For example, the task of ingredient analysis in new drug development requires a medical model, a chemical model, and a pharmaceutical model. To select these domain-specific models, the server uses Bayesian inference and rule-based systems.

[1070] Next, the necessary computing resources are allocated to each selected domain-specific model. The server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. For example, two GPUs are allocated to a medical model, and three GPUs to a chemistry model.

[1071] Using the allocated resources, each specialized model performs a task and generates an answer based on its own expertise. For example, a chemistry model analyzes input ingredient information and evaluates chemical properties.

[1072] The server collects and integrates the results returned by each model. The integration uses an algorithm that evaluates the contribution of each model to generate the most appropriate answer. For example, it combines component analysis data from a chemical model, data on effects on the human body from a medical model, and drug efficacy data from a pharmaceutical model.

[1073] The integrated answers are converted into an appropriate data format and sent to the user's device. The user then checks and evaluates the answers on their own device. The evaluation is then entered into the system as feedback. For example, the user can send feedback such as "I'm satisfied with the analysis results" or "Additional information is needed."

[1074] The collected feedback is analyzed by the server and used to improve the system's performance, specifically by retraining each domain-specific model to improve its accuracy when performing subsequent tasks.

[1075] Through the above process, the system of the present invention can provide highly accurate and efficient solutions to tasks input by the user, thereby enabling the user to solve problems quickly and reliably.

[1076] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1077] Step 1: The user enters the task they want to solve in natural language into the terminal and registers it in the system. For example, they might enter, "I want to perform ingredient analysis in new drug development." This is the input data. The terminal receives this data and displays it on the screen.

[1078] Step 2: The terminal converts the input task information into an appropriate data format (for example, JSON format). Specifically, the terminal converts the input text into a JSON object (for example, "task": "I would like to perform component analysis in new drug development"). This becomes the output data. This converted data is used in the next step.

[1079] Step 3: The terminal sends the converted data to the server. Specifically, the terminal uses an HTTP request to send JSON formatted data to the server. The input is JSON data, and the output is data transmission to the server.

[1080] Step 4: The server analyzes the received task information. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK or spaCy) to extract important keywords (e.g., "new drug development" and "ingredient analysis"). The input is the received JSON data, and the output is the extracted keywords.

[1081] Step 5: The server selects appropriate domain models based on the analysis results. Specifically, the server uses Bayesian inference and rule-based systems to select medical, chemical, and pharmaceutical models. The input is the extracted keywords, and the output is a list of selected models.

[1082] Step 6: The server allocates the necessary computing resources for each selected domain model. Specifically, the server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. The input is a list of selected models, and the output is the details of the resource allocation. For example, two GPUs are allocated to the medical model, and three GPUs are allocated to the chemistry model.

[1083] Step 7: The server prepares each model to run by adding a task to a queue for each model and launching a script to initialize the execution environment. The input is the resource allocation details, and the output is the ready state of each model.

[1084] Step 8: Each domain model generates a solution for the received task. Specifically, the chemical model analyzes the input ingredient information and evaluates chemical attributes. The input is the readiness status and resources of each model, and the output is the solution generated by each model.

[1085] Step 9: The server collects and integrates the answers from each model. Specifically, it aggregates the data returned from each model and generates a unified answer using a contribution evaluation algorithm. The input is the answer from each model, and the output is the integrated answer.

[1086] Step 10: The server converts the consolidated answer into an appropriate format and sends it to the user's device. Specifically, it converts the consolidated answer data into JSON or report format and returns it to the user using an HTTP response. The input is the consolidated answer, and the output is the answer returned to the user's device.

[1087] Step 11: The user evaluates the answer they received and enters their feedback into the system. For example, they might enter something like "I'm satisfied with the analysis results" or "I need more information." This is the input of feedback data.

[1088] Step 12: The device converts the user's rating into an appropriate format and sends it to the server. Specifically, the device converts the feedback into JSON format and sends it to the server via an HTTP request. The input is the feedback data, and the output is the data transmission to the server.

[1089] Step 13: The server receives the feedback, evaluates the performance of each domain model, and performs retraining. Specifically, it uses the feedback data to update the model's training dataset and retrains it. The input is the received feedback data, and the output is the updated model.

[1090] (Application example 1)

[1091] 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."

[1092] Modern factories have many tasks, and their efficient execution directly leads to improved productivity. However, optimal resource allocation and appropriate utilization of specialized knowledge for each task is difficult. Dynamic improvements based on feedback are also required, making it difficult to address these challenges with traditional static models. Therefore, a new system is needed that can efficiently optimize tasks and improve the performance of factory production lines.

[1093] 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.

[1094] In this invention, the server includes means for a user to register a task to be solved, means for selecting a specialized domain model suitable for the task, means for allocating necessary computational resources to the selected model, means for the model to execute the task using the allocated resources, means for integrating the execution results and providing them to the user, means for collecting feedback from the user and reflecting it in subsequent processes, and means for generating a plan for having robots in the factory execute the task and appropriately allocating resources, thereby enabling efficient optimization of factory tasks and dynamic improvement.

[1095] "User" refers to a person or organization that uses the system.

[1096] A "task" refers to a specific problem or task that a user wants the system to solve.

[1097] "Domain models" refer to machine learning or artificial intelligence algorithms or software that are specialized for a particular field or problem.

[1098] "Computational resources" refers to the computer hardware and software resources required to perform a task, including CPU, GPU, memory, etc.

[1099] "Execution" refers to the process by which a selected domain expertise model processes a task using the allocated computing resources.

[1100] "Synthesis" refers to the process of combining the results of multiple disciplinary models into a single conclusion or answer.

[1101] "Feedback" refers to the evaluations and opinions provided by users, and the system is improved based on that information.

[1102] "Factory robots" refer to autonomous or semi-autonomous machines that perform tasks in a factory.

[1103] "Generating a plan" refers to the process of automatically creating optimal execution procedures for a task.

[1104] "Appropriate resource allocation" refers to efficiently allocating the necessary computing resources and hardware to each task or model.

[1105] The present invention is a system that allows users to register a task they wish to solve, selects an appropriate domain model, efficiently allocates the necessary computational resources, and generates a highly accurate solution. This system is intended to be applied particularly to robots in factories.

[1106] System Program

[1107] First, the user inputs the task they want to solve in natural language and registers it in the system. For example, a task might be, "Please suggest a way to resolve a bottleneck on the production line."

[1108] The server analyzes the input task information and extracts important keywords and concepts. Based on the extracted keywords, it selects the optimal domain model. For example, keywords such as "production line," "bottleneck," "solution," "method," and "proposal" are extracted.

[1109] Next, the server allocates the necessary computing resources to the selected domain model. In this case, CPUs and GPUs are used as computing resources, and specific examples include NVIDIA GPUs and Intel CPUs.

[1110] Using the allocated resources, each domain model performs tasks and generates solutions, such as optimizing a manufacturing process or analyzing the causes of bottlenecks.

[1111] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer, which is provided to the user in the form of a report. This process utilizes natural language generation technology and a generative AI model.

[1112] Users evaluate the answers they receive and provide feedback to the system, which then uses that feedback to improve the task-solving process and, if necessary, retrain the domain model.

[1113] Hardware and software used

[1114] Hardware: Factory robots, dedicated GPU servers, user devices (PCs, smartphones, etc.)

[1115] Software: Natural language processing libraries (NLTK, SpaCy), AI model selection libraries (custom libraries), resource allocation libraries, natural language generation models (GPT-3, etc.)

[1116] Adding specific examples

[1117] As a concrete example, consider the case where a user inputs "Please suggest a method to resolve bottlenecks in the production line." The system analyzes this task, extracts keywords such as "production line," "bottleneck," "resolve," "method," and "proposal," and based on these, selects the most appropriate domain-specific model (e.g., production engineering model, physical simulation model). Each model independently proposes a solution, and finally provides an integrated report.

[1118] An example of a prompt to be input to a generative AI model: "Please suggest a way to resolve a bottleneck on the production line."

[1119] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1120] Step 1:

[1121] Users input the task they want to solve in natural language from their own terminal and register it in the system.

[1122] Input: User task input (e.g., "Please suggest a way to resolve the bottleneck on the production line.")

[1123] Output: Task information is converted to JSON format and sent to the server.

[1124] Specific operation: When a user enters the task details in a dedicated application and presses the send button, the task information is converted into JSON format and sent to the server via the Internet.

[1125] Step 2:

[1126] The server analyzes the received task information and extracts important keywords and concepts.

[1127] Input: Task information sent to the server (JSON format)

[1128] Output: Extracted keywords and concepts (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[1129] How it works: The server parses the task information in JSON format and uses a natural language processing library (such as NLTK or SpaCy) to extract relevant keywords. During this process, it uses a specific algorithm to determine the frequency and importance of the keywords.

[1130] Step 3:

[1131] The server selects the most suitable specialized domain model based on the extracted keywords.

[1132] Input: Extracted keywords (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[1133] Output: List of selected domain-specific models (e.g., industrial engineering models, physical simulation models)

[1134] Specific operation: Based on the extracted keywords, the server selects an appropriate domain-specific model using a model selection library. This library recommends the best model based on past data and experience.

[1135] Step 4:

[1136] The server allocates the necessary computational resources to each selected domain model.

[1137] Input: List of selected domain models

[1138] Output: Allocated computing resources (e.g., 2 GPUs, 4 CPU cores, etc.)

[1139] How it works: The server uses a resource allocation library to allocate optimal resources to each model. For example, a physics simulation model requires a large number of GPUs, so it calculates and allocates appropriate resources.

[1140] Step 5:

[1141] The server uses the allocated resources to have each domain model perform a task and generate an answer.

[1142] Input: Allocated computing resources, selected domain model

[1143] Output: The answer generated by each model (e.g., "How to optimize a manufacturing process?")

[1144] How it works: The server runs each model and makes full use of its resources to generate the information needed to solve the task. Specifically, the production engineering model generates proposals for improving the efficiency of the manufacturing process, and the physical simulation model generates simulation results for eliminating bottlenecks.

[1145] Step 6:

[1146] The server collects the solutions generated from each model and combines them to generate a single unified solution.

[1147] Input: Answers generated by each model

[1148] Output: Unified answer (e.g. unified report)

[1149] How it works: The server uses a natural language generation model to integrate the answers from each model. During this process, it evaluates the contribution of each model and prioritizes the most important information. It then uses a generative AI model (e.g., GPT-3) to compile the results into a readable report.

[1150] Step 7:

[1151] The server sends the integrated answer to the user's terminal.

[1152] Input: Consolidated Answer

[1153] Output: The answer displayed on the user's terminal

[1154] Specific operation: The server converts the integrated answers into JSON format or another appropriate data format and sends them to the user's device. The user can then review the answers and use them to improve the manufacturing process.

[1155] Step 8:

[1156] Users rate the answers they receive and enter their feedback into the system.

[1157] Input: User rating and feedback (e.g., "Satisfied with the analysis results" or "More information needed")

[1158] Output: The feedback information is sent to the server.

[1159] Specific operation: The user uses a dedicated interface to input evaluations of answers and requests for additions, and sends them to the server.

[1160] Step 9:

[1161] The server uses the feedback it receives to improve the task-solving process for future tasks, including retraining the domain expertise model if necessary.

[1162] Input: User feedback

[1163] Output: Improved model and system settings

[1164] What happens: The server analyzes the feedback, identifies areas for improvement in the model and process, and, if necessary, retrains the domain-specific model to reflect it in the next task-solving process.

[1165] 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.

[1166] Understood. Below is the "Form for carrying out the invention".

[1167] ---

[1168] This invention is a system that utilizes a domain model and an emotion engine to efficiently allocate computational resources to obtain highly accurate answers for tasks registered by users. The processing of the system program is described in detail below.

[1169] System Overview

[1170] This system consists of a series of processes: a user registers the task they wish to solve, selects an appropriate domain model and emotion engine, allocates the necessary computing resources, generates an answer to the task, and provides the user with the integrated answer.

[1171] Task registration phase

[1172] 1. User operations

[1173] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[1174] An emotion engine analyzes emotion information from user input to measure, for example, urgency or anxiety.

[1175] 2. Terminal Processing

[1176] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[1177] Task analysis and model selection phase

[1178] 3. Server Processing

[1179] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[1180] Emotional data is also taken into consideration and the task is determined to be urgent.

[1181] 4. Server Processing

[1182] The server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the emotion data and analysis results. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[1183] Resource Allocation Phase

[1184] 5. Server Processing

[1185] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. If the emotional data indicates a high level of urgency, more resources are allocated. For example, a medical model may be allocated twice as many GPUs as usual.

[1186] 6. Server Processing

[1187] Using the allocated resources, prepare each model to execute the task. Provide each model with detailed task information and emotion information, and start the calculation.

[1188] Answer generation phase

[1189] 7. Server Processing

[1190] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data for a new drug and generates a result.

[1191] 8. Server Processing

[1192] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[1193] 9. Server Processing

[1194] The integrated answers are converted into an appropriate format and sent to the user's device. Emotional data is also reflected, and explanations that take the user's feelings into consideration are added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[1195] Feedback and Retraining Phase

[1196] 10. User Operations

[1197] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[1198] 11. Terminal Processing

[1199] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[1200] 12. Server Processing

[1201] The server receives the feedback and evaluates the performance of each specialized model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data. It also takes into account emotional data, reflecting emotional fluctuations due to the feedback.

[1202] In this way, the system of the present invention appropriately allocates the user's tasks to the domain model and emotion engine, and generates answers efficiently and accurately, thereby contributing to the user's problem solving.

[1203] The processing flow will be explained below.

[1204] I understand. I will explain the process in detail in the following format.

[1205] ---

[1206] Step 1:

[1207] User operations

[1208] The user enters the task they want to solve in natural language on their device and registers it in the system. For example, they might enter "I want to perform ingredient analysis for new drug development" and press the send button. At this time, the emotion engine also analyzes emotional information from the user's facial expressions and voice.

[1209] Step 2:

[1210] Terminal handling

[1211] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[1212] Step 3:

[1213] Server Processing

[1214] The server analyzes the received task information and emotional data. It uses a natural language processing model to understand the content of the task and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "ingredient analysis." It also determines the urgency and importance of the task from the emotional data.

[1215] Step 4:

[1216] Server Processing

[1217] The server selects the most appropriate specialized model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the analysis results and emotion data. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug. Furthermore, based on the emotion data, it creates a plan to allocate more resources to high-priority tasks.

[1218] Step 5:

[1219] Server Processing

[1220] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. Based on the emotion data, if the level of urgency is high, the server will increase resource allocation. For example, a medical model will be allocated twice as many GPUs as usual.

[1221] Step 6:

[1222] Server Processing

[1223] The server checks the resource allocation for each model and sends task details to each model, which prepares to execute the task and starts the calculation process.

[1224] Step 7:

[1225] Server Processing

[1226] Each specialized domain model generates a solution for the task it receives. For example, a chemistry model calculates component analysis data for a new drug and sends the results to a server.

[1227] Step 8:

[1228] Server Processing

[1229] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles component analysis data from chemical models, human body effect data from medical models, and drug efficacy data from pharmaceutical models into a single report format.

[1230] Step 9:

[1231] Server Processing

[1232] The integrated answer is converted into an appropriate format for presentation to the user and sent to the user's device. Emotional data is also reflected, and an explanation that takes the user's emotions into consideration is added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[1233] Step 10:

[1234] User operations

[1235] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information," along with emotional data.

[1236] Step 11:

[1237] Terminal handling

[1238] The user's device converts the feedback and emotion data into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis results\",\"emotion\":\"satisfied\"}\".

[1239] Step 12:

[1240] Server Processing

[1241] The server receives the feedback and emotion data and evaluates the performance of each domain-specific model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data to reflect changes in emotion due to feedback.

[1242] ---

[1243] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[1244] Example 2

[1245] 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."

[1246] Conventional problem-solving systems have difficulty efficiently analyzing user tasks and providing highly accurate solutions. Furthermore, they often fail to adequately consider the user's emotional information, resulting in a poor user experience. Furthermore, they fail to select an expert domain model or optimally allocate computing resources, resulting in the waste of computing resources. The purpose of this invention is to solve these problems.

[1247] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for analyzing task information and extracting appropriate keywords and concepts, means for selecting a specialized domain model based on the task and emotion information, and means for allocating necessary computational resources to the selected specialized domain model. This enables the generation of highly accurate answers to tasks. Furthermore, by taking the user's emotion information into consideration, a more satisfying user experience can be provided. Furthermore, efficient allocation of computational resources improves overall performance and reduces waste.

[1248] A "user" is an individual or group that uses the system to register tasks and receive answers.

[1249] A "task" refers to a specific problem or requirement that a user wants to solve.

[1250] "Emotion information" refers to data obtained by analyzing the user's emotional state at the time of task input.

[1251] A "domain model" is a computational model that includes knowledge and algorithms in a specific field.

[1252] "Computational resources" refers to the hardware and software resources (e.g., CPU, GPU, memory, etc.) required to run a domain-specific model.

[1253] "Answer" refers to the result or conclusion that a domain model generates for a task.

[1254] "Feedback" refers to the act of a user providing the system with an evaluation or opinion on the answer they received.

[1255] "Retraining" refers to the process of updating the domain-specific model based on feedback to improve accuracy the next time the task is performed.

[1256] The "server" refers to a central computing device that analyzes task information and emotional information, selects specialized domain models, allocates computing resources, and generates and integrates answers.

[1257] "Terminal" refers to a device (e.g., PC, smartphone, tablet, etc.) that a user uses to input tasks and receive answers.

[1258] "Data format" refers to a standardized structure (e.g., JSON, XML, etc.) for storing and transmitting information.

[1259] This invention is a system that utilizes domain expertise models and emotion engines to efficiently allocate computational resources to tasks registered by users and obtain highly accurate answers. The program for this system includes a series of processes from task registration to answer generation, feedback collection, and relearning.

[1260] In this system, a user first inputs the task they want to solve in natural language using a device (e.g., PC, smartphone, tablet, etc.). For example, they might input, "I would like you to analyze the ingredients for new drug development." The emotion engine then analyzes the user's input and extracts emotional information (e.g., urgency and anxiety). The device then converts this task information and emotional information into an appropriate data format (e.g., JSON format) and sends it to the server.

[1261] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis." It also takes into account emotional information to determine the urgency of the task. Based on this, the server selects the domain model (e.g., medical model, chemistry model, pharmaceutical model) that is best suited to the task.

[1262] Next, the server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected domain-specific model. If the emotional information indicates a high level of urgency, more computing resources are allocated. For example, a medical model is allocated twice as many GPUs as usual. Using the allocated resources, each model generates an answer to the task.

[1263] The server collects and integrates the generated answers from each specialized model. During integration, the contribution of each model is evaluated and a unified answer is generated. For example, the component analysis data from the chemical model and the human body effect data from the medical model are combined and integrated into a report format. The server then converts the integrated answer into an appropriate format and sends it to the user's device. The server also reflects the user's emotional information and adds an explanation that takes the user's feelings into consideration. For example, a report titled "Component Analysis Report of a New Drug" could be generated, with a comment added such as "We determined that an emergency response was required, so we conducted the analysis quickly."

[1264] Furthermore, the user evaluates the answers they receive and inputs their feedback into the system. For example, they may enter feedback such as "I'm satisfied with the analysis results" or "I need more information." The device converts the feedback into an appropriate format and sends it to the server. The server receives the feedback and evaluates the performance of each specialized domain model. If necessary, retraining is performed to improve accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data.

[1265] Through this series of processes, the system uses the domain model and emotion engine to process user tasks efficiently and with high accuracy, significantly contributing to solving user problems. Furthermore, by taking emotion information into consideration, the system can increase user satisfaction.

[1266] Examples of specific examples and prompts

[1267] Specific examples

[1268] The user types into the terminal, "I would like to request a component analysis for new drug development."

[1269] The server selects a specialized area model related to new drug development, determines that the task is of high urgency, and allocates more computing resources to it.

[1270] Each specialized model starts its analysis and sends the results to the server.

[1271] The integrated answer is provided to the user, along with a comment such as, "We determined that an emergency response was required, so we conducted a rapid analysis."

[1272] Prompt Sentence Examples

[1273] Input: "Please provide detailed ingredient analysis for my drug development project. It's very important."

[1274] The above is a specific embodiment for carrying out the present invention.

[1275] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1276] System processing steps

[1277] Step 1: Task Input and Sentiment Analysis

[1278] The user inputs the task they want to solve using the terminal in natural language.

[1279] Input: "I would like to request a component analysis for new drug development."

[1280] How it works: The device uses an emotion engine to extract emotion information (e.g., urgency) from the user's input.

[1281] Step 2: Convert to data format and send

[1282] The device converts the input task information and emotion information into an appropriate data format (e.g., JSON format) and sends it to the server.

[1283] Input: User task information and emotion information

[1284] Output: JSON data "{\"task\":\"I would like you to analyze ingredients for new drug development\",\"emotion\":\"high urgency\"}"

[1285] What it does: Serializes input data into JSON format and sends it to the server.

[1286] Step 3: Task analysis and keyword extraction

[1287] The server analyzes the received task information and extracts keywords and concepts.

[1288] Input: JSON data

[1289] Output: Keywords "new drug development" and "component analysis"

[1290] How it works: Extracts keywords using natural language processing algorithms.

[1291] Step 4: Selecting a domain model

[1292] The server selects an appropriate domain model based on task information and emotion information.

[1293] Input: Keywords, emotion information

[1294] Output: Selected domain models (e.g., medical model, chemical model)

[1295] Action: Searches for and selects a corresponding model from the database.

[1296] Step 5: Allocate Computational Resources

[1297] The server allocates the necessary computing resources to the selected domain model.

[1298] Input: Selected domain model, emotion information

[1299] Output: Allocated computing resources (e.g. GPU, CPU specs)

[1300] How it works: A resource management system is used to allocate optimal resources to each model.

[1301] Step 6: Prepare and provide information about the model

[1302] The server prepares each model to run using the allocated resources and provides task and emotion information.

[1303] Input: Liquidation resources, task information, emotion information

[1304] Output: Ready to run

[1305] How it works: Loads the necessary libraries and datasets, and feeds the model with task information.

[1306] Step 7: Generate the answer

[1307] The server uses each domain model to generate an answer to the task.

[1308] Input: Task information provided to the model

[1309] Output: Answer data for each model

[1310] Action: The model performs a calculation and generates an answer (e.g., compositional analysis data).

[1311] Step 8: Consolidating the Answers

[1312] The server collects and integrates the answers generated by each model.

[1313] Input: Answer data for each model

[1314] Output: Consolidated answer

[1315] How it works: Evaluate the contribution of each answer and generate a single unified answer.

[1316] Step 9: Provide the answer

[1317] The server converts the integrated answers into an appropriate format and sends them to the user's device.

[1318] Input: Integrated answers, emotional information

[1319] Output: Answer data to the user

[1320] What it does: Converts the answer data into a user-friendly format (e.g., report format) and adds sentiment-sensitive comments.

[1321] Step 10: Gather feedback

[1322] The user rates the answers they receive and provides feedback.

[1323] Input: Rating and comments on the answer

[1324] Output: Feedback data

[1325] How it works: Use the input form to enter your rating and comments and submit your feedback.

[1326] Step 11: Submit your feedback

[1327] The device converts the feedback into an appropriate format and sends it to the server.

[1328] Input: User feedback

[1329] Output: Feedback data in JSON format

[1330] What it does: Serializes the feedback into JSON format and sends it to the server.

[1331] Step 12: Retraining

[1332] The server evaluates the performance of each specialized domain model based on the feedback and performs re-learning.

[1333] Input: Feedback data

[1334] Output: Updated model

[1335] How it works: Analyzes feedback and applies a re-training algorithm to improve the model.

[1336] The above is the specific program processing flow of this system.

[1337] (Application example 2)

[1338] 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."

[1339] In modern content delivery services, it is extremely important to provide users with the content they need quickly and accurately. However, conventional systems do not dynamically allocate resources based on the user's emotional state or the urgency of the task, limiting the improvement of the user experience. Furthermore, they lack a mechanism for effectively utilizing user feedback to improve accuracy in subsequent task executions. As a result, user satisfaction declines, leading to a diminishing value of the service.

[1340] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1341] In this invention, the server includes: means for analyzing a user's emotional information, determining the urgency based on the information, and adjusting the allocation of computational resources; means for the user to register a task they wish to solve; means for selecting a specialized domain model appropriate for the task; means for allocating the necessary computational resources to the selected model; means for the model to execute the task using the allocated resources; means for integrating the execution results and providing them to the user; and means for collecting user feedback and reflecting it in subsequent processes. This enables dynamic resource allocation according to the user's emotional state and the urgency of the task, thereby enabling the rapid and accurate provision of content that meets the user's needs. Furthermore, by effectively utilizing user feedback to retrain the model, accuracy can be improved the next time the task is executed, thereby increasing user satisfaction.

[1342] A "task" refers to a specific request or purpose that a user requests the system to solve.

[1343] A "specialized domain model" is a machine learning model or algorithm built based on knowledge and data in a specific field or area.

[1344] "Emotional information" is emotional data such as emotional state and urgency that is analyzed from the user's input and behavior.

[1345] "Computational resources" refers to the computational resources such as CPU, GPU, and memory required to execute a task.

[1346] "Resource allocation" refers to the process of allocating the necessary computational resources to the selected model.

[1347] "Model execution" refers to the execution of calculations and analysis for a task by a selected domain-specific model using allocated computational resources.

[1348] "Execution result integration" refers to the compilation of results generated by multiple disciplinary models into a single unified answer or report.

[1349] "Feedback" refers to information indicating the user's evaluation and satisfaction with the answer they received.

[1350] "Retraining" is a learning process that improves the performance and accuracy of a domain-specific model based on collected feedback and newly acquired data.

[1351] "Urgency" is an indicator of how quickly a user's task should be resolved.

[1352] This invention is a system for quickly and accurately providing users with content they desire through a content distribution service. The system analyzes users' emotional information, determines the urgency of tasks based on that information, and dynamically adjusts the allocation of computing resources to provide personalized content tailored to the user's needs.

[1353] System Overview

[1354] This system utilizes domain expertise models and emotion engines to efficiently allocate the necessary computing resources to tasks registered by users, and provides highly accurate answers and content to users.

[1355] Hardware and software used

[1356] Hardware: Uses high-speed CPU, GPU and memory.

[1357] Software: Python 3.x, SentimentEngine (for sentiment analysis), and domain-specific models (news model, music model, video model, etc.).

[1358] System processing flow

[1359] 1. User Input and Sentiment Analysis

[1360] The user inputs the task they want to solve in natural language from their device. For example, they might input something like, "I want to know the latest technology news." The server receives this input and uses an emotion engine to analyze the emotional information and measure the urgency and anxiety of the task. The emotional information is treated as data that indicates, for example, the "high urgency" of the task.

[1361] 2. Selecting a Specialty Model and Allocating Resources

[1362] The server extracts necessary keywords and concepts based on the analyzed task information and emotion information, and selects the appropriate domain model. For example, for the keyword "technology news," a news model is selected. Then, the necessary computing resources are allocated based on the emotion data (urgency). If the urgency is high, more resources than usual are allocated.

[1363] 3. Execute the task

[1364] Using the allocated resources, selected domain-specific models perform tasks, such as a technology news model that analyzes the latest news and generates relevant articles.

[1365] 4. Consolidating and Presenting Results

[1366] The server collects and integrates the results generated by each model. This allows it to evaluate the contribution of each model and generate the optimal answer. The integrated answer is converted into an appropriate format and provided to the user along with a comment that takes into account the user's emotional information. For example, it may include a message such as, "Please check the technology news below. Due to its urgency, we analyzed it quickly."

[1367] 5. Feedback and Retraining

[1368] Users provide feedback on the answers they provide, and the server receives this feedback, evaluates the performance of each domain-specific model, and retrains it as needed, improving the model's accuracy the next time the task is performed.

[1369] Specific examples

[1370] For example, if a user requests "I want to know the latest technology news," the following prompt might be generated:

[1371] What's the latest technology news? It's very urgent right now.

[1372] Based on this prompt, the server performs a series of processes, including sentiment analysis, selection of domain expertise models, allocation of computing resources, task execution, integration and provision of results, and feedback collection and retraining, thereby quickly providing users with personalized, up-to-date technology news.

[1373] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1374] Step 1:

[1375] The device accepts the user's natural language input. The user types "I want to know the latest technology news" into the device. The device converts this input data into an appropriate data format (e.g., JSON) to send it to the server.

[1376] Input: User's natural language input

[1377] Data processing: converting natural language into data format

[1378] Output: JSON format data

[1379] Step 2:

[1380] The server receives the user input sent from the device and performs emotion analysis using an emotion engine. The emotion engine analyzes "urgency" and "anxiety" and extracts emotion information such as "high urgency."

[1381] Input: User-supplied data in JSON format

[1382] Data Computing: Sentiment Analysis

[1383] Output: Emotional information (e.g. "High Urgency")

[1384] Step 3:

[1385] The server extracts necessary keywords and concepts from the analyzed emotion information and user input. The keyword "technology news" is identified.

[1386] Input: User input and emotional information

[1387] Data processing: Keyword extraction

[1388] Output: The keyword "technology news"

[1389] Step 4:

[1390] The server selects an appropriate domain model (e.g., a news model) based on keywords and emotion information. At the same time, it dynamically allocates the necessary computing resources according to the emotion information, particularly the urgency. If the urgency is high, more CPU and GPU resources than usual are allocated.

[1391] Input: Extracted keywords and sentiment information

[1392] Data calculation: model selection, computational resource allocation

[1393] Output: Selected model and allocated resources

[1394] Step 5:

[1395] The server uses the allocated resources to run the selected news model. The model analyzes the task (e.g., get the latest technology news) and generates recommended news articles.

[1396] Input: Selected model and allocated resources

[1397] Data calculation: Model execution, recommended content generation

[1398] Output: Generated content (e.g., news article)

[1399] Step 6:

[1400] The server collects and integrates the generated content. If there are results from multiple domain-specific models, it evaluates each and generates a single integrated answer. At this time, it adds comments that take into account the user's emotional information.

[1401] Input: Multiple Content

[1402] Data processing: integrating results and adding comments

[1403] Output: Consolidated content and comments

[1404] Step 7:

[1405] The server transmits the integrated content to the terminal, which displays the content on the user's screen and makes it available to the user.

[1406] Input: Integrated content and comments

[1407] Data output: Send to terminal

[1408] Output: The content that is displayed on the user's screen

[1409] Step 8:

[1410] The user inputs feedback on the provided content, and the terminal converts this feedback into an appropriate data format for transmission to the server.

[1411] Input: User feedback

[1412] Data processing: Convert feedback into data format

[1413] Output: Feedback data in JSON format

[1414] Step 9:

[1415] The server analyzes the feedback received from the device and uses it to evaluate the performance of each specialized domain model and retrain it, and if necessary, retrains the model to improve its accuracy the next time the task is performed.

[1416] Input: Feedback data

[1417] Data calculation: performance evaluation, re-learning

[1418] Output: Improved model

[1419] 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.

[1420] 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.

[1421] 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.

[1422] [Fourth embodiment]

[1423] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1424] 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.

[1425] 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).

[1426] 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.

[1427] 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.

[1428] 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).

[1429] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1430] 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.

[1431] 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.

[1432] 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.

[1433] 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.

[1434] 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.

[1435] 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."

[1436] Understood. Below is the "Form for carrying out the invention".

[1437] ---

[1438] The present invention is a system for obtaining highly accurate solutions by selecting a domain model for a task registered by a user and efficiently allocating computational resources. The processing of the system program is described in detail below.

[1439] System Overview

[1440] This system consists of a series of processes in which a user registers a task they wish to solve, selects an appropriate domain model, allocates the necessary computational resources, generates an answer to the task, and provides the user with the integrated answer.

[1441] Task registration phase

[1442] 1. User operations

[1443] Users input the task they want to solve in natural language from their own device and register it in the system.

[1444] As a specific example, when a user requests a component analysis of a new drug, the user inputs "I would like to perform component analysis in new drug development."

[1445] 2. Terminal Processing

[1446] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server.

[1447] Task analysis and model selection phase

[1448] 3. Server Processing

[1449] The server analyzes the received task information and extracts necessary keywords and concepts. For example, keywords such as "new drug development" and "component analysis" are identified.

[1450] 4. Server Processing

[1451] Based on the analysis results, the server selects which specialized model (medical model, chemistry model, pharmaceutical model, etc.) is most appropriate.

[1452] As a specific example, we determine that specialized models in medicine, chemistry, and pharmacy are required for tasks related to new drug development.

[1453] Resource Allocation Phase

[1454] 5. Server Processing

[1455] The server allocates the necessary computing resources (GPU, CPU, memory, etc.) to each selected specialized domain model.

[1456] As a specific example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[1457] 6. Server Processing

[1458] With the allocated resources, each model is prepared to perform its task.

[1459] Answer generation phase

[1460] 7. Server Processing

[1461] Each domain model generates a solution for the task it receives. For example, a chemistry model analyzes the ingredients of a new drug and generates a result.

[1462] 8. Server Processing

[1463] The server collects and integrates the solutions generated by each model, evaluates the contribution of each model, and generates a unified solution.

[1464] As a specific example, component analysis data obtained from a chemical model, data on effects on the human body obtained from a medical model, and efficacy data obtained from a pharmaceutical model are integrated and compiled into a single report format.

[1465] 9. Server Processing

[1466] The integrated answers are converted into an appropriate format and sent to the user's terminal.

[1467] Feedback and Retraining Phase

[1468] 10. User Operations

[1469] The user evaluates the answers received and enters feedback into the system.

[1470] For example, feedback such as "I'm satisfied with the analysis results" or "I need more information" can be entered.

[1471] 11. Terminal Processing

[1472] The user's rating is converted into an appropriate format and sent to the server.

[1473] 12. Server Processing

[1474] The server receives feedback and evaluates the performance of each domain-specific model, retraining it as needed to improve its accuracy the next time the task is performed.

[1475] In this way, the system of the present invention contributes to solving the user's problem by appropriately allocating the user's tasks to the specialized domain model and generating an answer efficiently and accurately.

[1476] The processing flow will be explained below.

[1477] Understood. Below, I will explain the process step by step.

[1478] ---

[1479] Step 1:

[1480] User operations

[1481] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[1482] Step 2:

[1483] Terminal handling

[1484] The user's device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis in new drug development\"}".

[1485] Step 3:

[1486] Server Processing

[1487] The server analyzes the received task information. It uses a natural language processing model to understand the task content and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[1488] Step 4:

[1489] Server Processing

[1490] Based on the analysis results, the server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.). For example, it may determine that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[1491] Step 5:

[1492] Server Processing

[1493] The server allocates the necessary computing resources (e.g., GPUs, CPUs, memory, etc.) to each selected specialty model. For example, two GPUs are allocated to the medical model, three GPUs to the chemistry model, and one GPU to the pharmacology model.

[1494] Step 6:

[1495] Server Processing

[1496] The server prepares each model to execute the task using the allocated resources, provides each model with detailed task information, and starts the calculation.

[1497] Step 7:

[1498] Server Processing

[1499] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data of a new drug and generates a result.

[1500] Step 8:

[1501] Server Processing

[1502] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[1503] Step 9:

[1504] Server Processing

[1505] The integrated answer is converted into an appropriate format and sent to the user's terminal. For example, a report is generated as a "New Drug Ingredient Analysis Report" and sent to the user.

[1506] Step 10:

[1507] User operations

[1508] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[1509] Step 11:

[1510] Terminal handling

[1511] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[1512] Step 12:

[1513] Server Processing

[1514] The server receives feedback and evaluates the performance of each domain model. If necessary, it retrains the model to improve its accuracy the next time it performs the task. For example, a medical model can be retrained using new clinical data.

[1515] ---

[1516] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[1517] Example 1

[1518] 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."

[1519] For complex tasks that users want to solve, there is a need to select the optimal domain model, appropriately allocate the necessary computational resources, and generate solutions effectively and efficiently. However, current technology does not fully automate the processes of task analysis, model selection, and resource management, placing a heavy burden on users. In addition, the feedback process to improve the accuracy of execution results is not fully integrated. Therefore, a system is needed that reduces the time and effort required for users to solve tasks and improve the accuracy of results.

[1520] 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.

[1521] In this invention, the server includes a means for analyzing task information and extracting related keywords and concepts, a means for selecting a specialized domain model appropriate for the task, and a means for allocating the necessary computational resources to each selected specialized domain model. This enables the selection of an optimal model for the task registered by the user and the efficient allocation of resources. This ultimately results in the rapid generation of highly accurate answers and reduces the burden on the user. Furthermore, by collecting feedback from the user on the answer results and reflecting this feedback in subsequent processes, the accuracy and efficiency of the entire system can be improved.

[1522] The "means for users to register tasks they wish to solve" is an interface that allows users to input specific problems or issues they wish to solve in natural language and register them in the system.

[1523] "Means for analyzing task information and extracting related keywords and concepts" refers to a function for automatically analyzing task information registered by the user and extracting important keywords and related concepts from it.

[1524] The "means for selecting a specialized domain model appropriate for the task" refers to an algorithm or process for selecting the specialized domain model (e.g., medical model, chemical model, pharmaceutical model, etc.) that is optimal for the task based on the extracted keywords and concepts.

[1525] The "means for allocating the necessary computing resources to each selected specialized domain model" is a mechanism for dynamically allocating the necessary computing resources (GPU, CPU, memory, etc.) to efficiently operate the specialized domain model.

[1526] The "means by which each specialized domain model generates an answer for a received task" refers to the process by which a specialized domain model generates an answer or result based on a given task using specialized knowledge or algorithms.

[1527] "Means for collecting and integrating execution results from each specialized model" refers to an integration process that brings together the answers and results generated by each specialized model to derive the optimal answer overall.

[1528] "Means for converting the integrated execution results into an appropriate format and providing them to the user" refers to a function that converts the integrated answers and results into a format that is easy for the user to understand (for example, a report format or graph) and provides them to the user.

[1529] "Means for collecting feedback from users and reflecting it in each process of the system" refers to the process of collecting evaluations and opinions on the answers and results provided by users and reflecting them in improving the system and increasing accuracy the next time the task is performed.

[1530] The present invention provides a system that selects an optimal domain model for a task that a user wants to solve, allocates the necessary computational resources, generates a highly accurate solution based on the model, and provides the solution to the user. Specific embodiments of the system are described in detail below.

[1531] This system mainly consists of three entities: users, terminals, and servers.

[1532] First, the user inputs the task they want to solve in natural language on their own device. For example, they might input, "I want to perform ingredient analysis for new drug development." By registering such a task, they can request the system to solve the problem.

[1533] Next, the device converts the input task information into an appropriate data format (e.g., JSON) and sends it to the server, which then transmits the task information throughout the system.

[1534] The server first analyzes the received task information and extracts important keywords and related concepts. For example, keywords such as "new drug development" and "ingredient analysis" are extracted. This analysis is generally performed using a natural language processing (NLP) engine. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK, spaCy).

[1535] Based on the analysis results, the server selects the domain-specific model appropriate for the task. For example, the task of ingredient analysis in new drug development requires a medical model, a chemical model, and a pharmaceutical model. To select these domain-specific models, the server uses Bayesian inference and rule-based systems.

[1536] Next, the necessary computing resources are allocated to each selected domain-specific model. The server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. For example, two GPUs are allocated to a medical model, and three GPUs to a chemistry model.

[1537] Using the allocated resources, each specialized model performs a task and generates an answer based on its own expertise. For example, a chemistry model analyzes input ingredient information and evaluates chemical properties.

[1538] The server collects and integrates the results returned by each model. The integration uses an algorithm that evaluates the contribution of each model to generate the most appropriate answer. For example, it combines component analysis data from a chemical model, data on effects on the human body from a medical model, and drug efficacy data from a pharmaceutical model.

[1539] The integrated answers are converted into an appropriate data format and sent to the user's device. The user then checks and evaluates the answers on their own device. The evaluation is then entered into the system as feedback. For example, the user can send feedback such as "I'm satisfied with the analysis results" or "Additional information is needed."

[1540] The collected feedback is analyzed by the server and used to improve the system's performance, specifically by retraining each domain-specific model to improve its accuracy when performing subsequent tasks.

[1541] Through the above process, the system of the present invention can provide highly accurate and efficient solutions to tasks input by the user, thereby enabling the user to solve problems quickly and reliably.

[1542] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1543] Step 1: The user enters the task they want to solve in natural language into the terminal and registers it in the system. For example, they might enter, "I want to perform ingredient analysis in new drug development." This is the input data. The terminal receives this data and displays it on the screen.

[1544] Step 2: The terminal converts the input task information into an appropriate data format (for example, JSON format). Specifically, the terminal converts the input text into a JSON object (for example, "task": "I would like to perform component analysis in new drug development"). This becomes the output data. This converted data is used in the next step.

[1545] Step 3: The terminal sends the converted data to the server. Specifically, the terminal uses an HTTP request to send JSON formatted data to the server. The input is JSON data, and the output is data transmission to the server.

[1546] Step 4: The server analyzes the received task information. Specifically, the server performs text analysis using a Python NLP library (e.g., NLTK or spaCy) to extract important keywords (e.g., "new drug development" and "ingredient analysis"). The input is the received JSON data, and the output is the extracted keywords.

[1547] Step 5: The server selects appropriate domain models based on the analysis results. Specifically, the server uses Bayesian inference and rule-based systems to select medical, chemical, and pharmaceutical models. The input is the extracted keywords, and the output is a list of selected models.

[1548] Step 6: The server allocates the necessary computing resources for each selected domain model. Specifically, the server uses a resource allocator in the cloud environment to dynamically allocate the GPUs and CPUs required for each model. The input is a list of selected models, and the output is the details of the resource allocation. For example, two GPUs are allocated to the medical model, and three GPUs are allocated to the chemistry model.

[1549] Step 7: The server prepares each model to run by adding a task to a queue for each model and launching a script to initialize the execution environment. The input is the resource allocation details, and the output is the ready state of each model.

[1550] Step 8: Each domain model generates a solution for the received task. Specifically, the chemical model analyzes the input ingredient information and evaluates chemical attributes. The input is the readiness status and resources of each model, and the output is the solution generated by each model.

[1551] Step 9: The server collects and integrates the answers from each model. Specifically, it aggregates the data returned from each model and generates a unified answer using a contribution evaluation algorithm. The input is the answer from each model, and the output is the integrated answer.

[1552] Step 10: The server converts the consolidated answer into an appropriate format and sends it to the user's device. Specifically, it converts the consolidated answer data into JSON or report format and returns it to the user using an HTTP response. The input is the consolidated answer, and the output is the answer returned to the user's device.

[1553] Step 11: The user evaluates the answer they received and enters their feedback into the system. For example, they might enter something like "I'm satisfied with the analysis results" or "I need more information." This is the input of feedback data.

[1554] Step 12: The device converts the user's rating into an appropriate format and sends it to the server. Specifically, the device converts the feedback into JSON format and sends it to the server via an HTTP request. The input is the feedback data, and the output is the data transmission to the server.

[1555] Step 13: The server receives the feedback, evaluates the performance of each domain model, and performs retraining. Specifically, it uses the feedback data to update the model's training dataset and retrains it. The input is the received feedback data, and the output is the updated model.

[1556] (Application example 1)

[1557] 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."

[1558] Modern factories have many tasks, and their efficient execution directly leads to improved productivity. However, optimal resource allocation and appropriate utilization of specialized knowledge for each task is difficult. Dynamic improvements based on feedback are also required, making it difficult to address these challenges with traditional static models. Therefore, a new system is needed that can efficiently optimize tasks and improve the performance of factory production lines.

[1559] 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.

[1560] In this invention, the server includes means for a user to register a task to be solved, means for selecting a specialized domain model suitable for the task, means for allocating necessary computational resources to the selected model, means for the model to execute the task using the allocated resources, means for integrating the execution results and providing them to the user, means for collecting feedback from the user and reflecting it in subsequent processes, and means for generating a plan for having robots in the factory execute the task and appropriately allocating resources, thereby enabling efficient optimization of factory tasks and dynamic improvement.

[1561] "User" refers to a person or organization that uses the system.

[1562] A "task" refers to a specific problem or task that a user wants the system to solve.

[1563] "Domain models" refer to machine learning or artificial intelligence algorithms or software that are specialized for a particular field or problem.

[1564] "Computational resources" refers to the computer hardware and software resources required to perform a task, including CPU, GPU, memory, etc.

[1565] "Execution" refers to the process by which a selected domain expertise model processes a task using the allocated computing resources.

[1566] "Synthesis" refers to the process of combining the results of multiple disciplinary models into a single conclusion or answer.

[1567] "Feedback" refers to the evaluations and opinions provided by users, and the system is improved based on that information.

[1568] "Factory robots" refer to autonomous or semi-autonomous machines that perform tasks in a factory.

[1569] "Generating a plan" refers to the process of automatically creating optimal execution procedures for a task.

[1570] "Appropriate resource allocation" refers to efficiently allocating the necessary computing resources and hardware to each task or model.

[1571] The present invention is a system that allows users to register a task they wish to solve, selects an appropriate domain model, efficiently allocates the necessary computational resources, and generates a highly accurate solution. This system is intended to be applied particularly to robots in factories.

[1572] System Program

[1573] First, the user inputs the task they want to solve in natural language and registers it in the system. For example, a task might be, "Please suggest a way to resolve a bottleneck on the production line."

[1574] The server analyzes the input task information and extracts important keywords and concepts. Based on the extracted keywords, it selects the optimal domain model. For example, keywords such as "production line," "bottleneck," "solution," "method," and "proposal" are extracted.

[1575] Next, the server allocates the necessary computing resources to the selected domain model. In this case, CPUs and GPUs are used as computing resources, and specific examples include NVIDIA GPUs and Intel CPUs.

[1576] Using the allocated resources, each domain model performs tasks and generates solutions, such as optimizing a manufacturing process or analyzing the causes of bottlenecks.

[1577] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer, which is provided to the user in the form of a report. This process utilizes natural language generation technology and a generative AI model.

[1578] Users evaluate the answers they receive and provide feedback to the system, which then uses that feedback to improve the task-solving process and, if necessary, retrain the domain model.

[1579] Hardware and software used

[1580] Hardware: Factory robots, dedicated GPU servers, user devices (PCs, smartphones, etc.)

[1581] Software: Natural language processing libraries (NLTK, SpaCy), AI model selection libraries (custom libraries), resource allocation libraries, natural language generation models (GPT-3, etc.)

[1582] Adding specific examples

[1583] As a concrete example, consider the case where a user inputs "Please suggest a method to resolve bottlenecks in the production line." The system analyzes this task, extracts keywords such as "production line," "bottleneck," "resolve," "method," and "proposal," and based on these, selects the most appropriate domain-specific model (e.g., production engineering model, physical simulation model). Each model independently proposes a solution, and finally provides an integrated report.

[1584] An example of a prompt to be input to a generative AI model: "Please suggest a way to resolve a bottleneck on the production line."

[1585] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1586] Step 1:

[1587] Users input the task they want to solve in natural language from their own terminal and register it in the system.

[1588] Input: User task input (e.g., "Please suggest a way to resolve the bottleneck on the production line.")

[1589] Output: Task information is converted to JSON format and sent to the server.

[1590] Specific operation: When a user enters the task details in a dedicated application and presses the send button, the task information is converted into JSON format and sent to the server via the Internet.

[1591] Step 2:

[1592] The server analyzes the received task information and extracts important keywords and concepts.

[1593] Input: Task information sent to the server (JSON format)

[1594] Output: Extracted keywords and concepts (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[1595] How it works: The server parses the JSON-formatted task information and uses a natural language processing library (such as NLTK or SpaCy) to extract relevant keywords. During this process, it uses a specific algorithm to determine the frequency and importance of the keywords.

[1596] Step 3:

[1597] The server selects the most suitable specialized domain model based on the extracted keywords.

[1598] Input: Extracted keywords (e.g., "production line," "bottleneck," "solution," "method," "proposal")

[1599] Output: List of selected domain-specific models (e.g., industrial engineering models, physical simulation models)

[1600] Specific operation: Based on the extracted keywords, the server selects an appropriate domain-specific model using a model selection library. This library recommends the best model based on past data and experience.

[1601] Step 4:

[1602] The server allocates the necessary computational resources to each selected domain model.

[1603] Input: List of selected domain models

[1604] Output: Allocated computing resources (e.g., 2 GPUs, 4 CPU cores, etc.)

[1605] How it works: The server uses a resource allocation library to allocate optimal resources to each model. For example, a physics simulation model requires a large number of GPUs, so it calculates and allocates appropriate resources.

[1606] Step 5:

[1607] The server uses the allocated resources to have each domain model perform a task and generate an answer.

[1608] Input: Allocated computing resources, selected domain model

[1609] Output: The answer generated by each model (e.g., "How to optimize a manufacturing process?")

[1610] How it works: The server runs each model and makes full use of its resources to generate the information needed to solve the task. Specifically, the production engineering model generates proposals for improving the efficiency of the manufacturing process, and the physical simulation model generates simulation results for eliminating bottlenecks.

[1611] Step 6:

[1612] The server collects the solutions generated from each model and combines them to generate a single unified solution.

[1613] Input: Answers generated by each model

[1614] Output: Unified answer (e.g. unified report)

[1615] How it works: The server uses a natural language generation model to integrate the answers from each model. During this process, it evaluates the contribution of each model and prioritizes the most important information. It then uses a generative AI model (e.g., GPT-3) to compile the results into a readable report.

[1616] Step 7:

[1617] The server sends the integrated answer to the user's terminal.

[1618] Input: Consolidated Answer

[1619] Output: The answer displayed on the user's terminal

[1620] Specific operation: The server converts the integrated answers into JSON format or another appropriate data format and sends them to the user's device. The user can then review the answers and use them to improve the manufacturing process.

[1621] Step 8:

[1622] Users rate the answers they receive and enter their feedback into the system.

[1623] Input: User rating and feedback (e.g., "Satisfied with the analysis results" or "More information needed")

[1624] Output: The feedback information is sent to the server.

[1625] Specific operation: The user uses a dedicated interface to input evaluations of answers and requests for additions, and sends them to the server.

[1626] Step 9:

[1627] The server uses the feedback it receives to improve the task-solving process for future tasks, including retraining the domain expertise model if necessary.

[1628] Input: User feedback

[1629] Output: Improved model and system settings

[1630] What happens: The server analyzes the feedback, identifies areas for improvement in the model and process, and, if necessary, retrains the domain model to reflect this in the next task-solving process.

[1631] 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.

[1632] Understood. Below is the "Form for carrying out the invention".

[1633] ---

[1634] This invention is a system that utilizes a domain model and an emotion engine to efficiently allocate computational resources to obtain highly accurate answers for tasks registered by users. The processing of the system program is described in detail below.

[1635] System Overview

[1636] This system consists of a series of processes: a user registers the task they wish to solve, selects an appropriate domain model and emotion engine, allocates the necessary computing resources, generates an answer to the task, and provides the user with the integrated answer.

[1637] Task registration phase

[1638] 1. User operations

[1639] The user inputs the task they want to solve in natural language from their own device and registers it in the system. For example, they might input "I want to perform ingredient analysis for new drug development" and press the send button.

[1640] An emotion engine analyzes emotion information from user input to measure, for example, urgency or anxiety.

[1641] 2. Terminal Processing

[1642] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[1643] Task analysis and model selection phase

[1644] 3. Server Processing

[1645] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis."

[1646] Emotional data is also taken into consideration and the task is determined to be urgent.

[1647] 4. Server Processing

[1648] The server selects the most appropriate specialized domain model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the emotion data and analysis results. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug.

[1649] Resource Allocation Phase

[1650] 5. Server Processing

[1651] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. If the emotional data indicates a high level of urgency, more resources are allocated. For example, a medical model may be allocated twice as many GPUs as usual.

[1652] 6. Server Processing

[1653] Using the allocated resources, prepare each model to execute the task. Provide each model with detailed task information and emotion information, and start the calculation.

[1654] Answer generation phase

[1655] 7. Server Processing

[1656] Each specialized model generates a solution for the task it receives. For example, a chemistry model calculates the component analysis data for a new drug and generates a result.

[1657] 8. Server Processing

[1658] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles the component analysis data from chemical models and the human body effect data from medical models into a single report format.

[1659] 9. Server Processing

[1660] The integrated answers are converted into an appropriate format and sent to the user's device. Emotional data is also reflected, and explanations that take the user's feelings into consideration are added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[1661] Feedback and Retraining Phase

[1662] 10. User Operations

[1663] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information."

[1664] 11. Terminal Processing

[1665] The user's device converts the feedback into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis result\"}" and sends it.

[1666] 12. Server Processing

[1667] The server receives the feedback and evaluates the performance of each specialized model. It retrains the model as needed to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data. It also takes into account emotional data, reflecting emotional fluctuations due to the feedback.

[1668] In this way, the system of the present invention appropriately allocates the user's tasks to the domain model and emotion engine, and generates answers efficiently and accurately, thereby contributing to the user's problem solving.

[1669] The processing flow will be explained below.

[1670] I understand. I will explain the process in detail in the following format.

[1671] ---

[1672] Step 1:

[1673] User operations

[1674] The user enters the task they want to solve in natural language on their device and registers it in the system. For example, they might enter "I want to perform ingredient analysis for new drug development" and press the send button. At this time, the emotion engine also analyzes emotional information from the user's facial expressions and voice.

[1675] Step 2:

[1676] Terminal handling

[1677] The user's device converts the input task information and emotion data into an appropriate data format (e.g., JSON) and sends it to the server. For example, the input data is converted into the format "{\"task\":\"I want to perform ingredient analysis for new drug development\",\"emotion\":\"high urgency\"}\".

[1678] Step 3:

[1679] Server Processing

[1680] The server analyzes the received task information and emotional data. It uses a natural language processing model to understand the content of the task and extract necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "ingredient analysis." It also determines the urgency and importance of the task from the emotional data.

[1681] Step 4:

[1682] Server Processing

[1683] The server selects the most appropriate specialized model for the task (e.g., medical model, chemistry model, pharmaceutical model, etc.) based on the analysis results and emotion data. For example, it determines that a medical model, chemistry model, and pharmaceutical model are required for the task of developing a new drug. Furthermore, based on the emotion data, it creates a plan to allocate more resources to high-priority tasks.

[1684] Step 5:

[1685] Server Processing

[1686] The server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected specialized model. Based on the emotion data, if the level of urgency is high, the server will increase resource allocation. For example, a medical model will be allocated twice as many GPUs as usual.

[1687] Step 6:

[1688] Server Processing

[1689] The server checks the resource allocation for each model and sends task details to each model, which prepares to execute the task and starts the calculation process.

[1690] Step 7:

[1691] Server Processing

[1692] Each specialized domain model generates a solution for the task it receives. For example, a chemistry model calculates component analysis data for a new drug and sends the results to a server.

[1693] Step 8:

[1694] Server Processing

[1695] The server collects and integrates the answers generated by each model, evaluates the contribution of each model, and generates a unified answer. For example, it compiles component analysis data from chemical models, human body effect data from medical models, and drug efficacy data from pharmaceutical models into a single report format.

[1696] Step 9:

[1697] Server Processing

[1698] The integrated answer is converted into an appropriate format for presentation to the user and sent to the user's device. Emotional data is also reflected, and an explanation that takes the user's emotions into consideration is added. For example, when generating a report titled "Ingredient Analysis Report of a New Drug" and sending it to the user, a comment such as "We determined that an emergency response was necessary, so we performed the analysis quickly" is added.

[1699] Step 10:

[1700] User operations

[1701] The user evaluates the answer they received and enters their feedback into the system, such as "I'm satisfied with the analysis results" or "I need more information," as well as emotional data.

[1702] Step 11:

[1703] Terminal handling

[1704] The user's device converts the feedback and emotion data into an appropriate format and sends it to the server. For example, it converts it into the format "{\"feedback\":\"Satisfied with the analysis results\",\"emotion\":\"satisfied\"}\".

[1705] Step 12:

[1706] Server Processing

[1707] The server receives the feedback and emotion data and evaluates the performance of each domain-specific model. If necessary, it retrains the model to improve its accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data to reflect changes in emotion due to feedback.

[1708] ---

[1709] The above are the details of the specific processing steps of the system. Through this process, users can obtain professional and accurate answers, and the overall performance of the system can be improved.

[1710] Example 2

[1711] 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."

[1712] Conventional problem-solving systems have difficulty efficiently analyzing user tasks and providing highly accurate solutions. Furthermore, they often fail to adequately consider the user's emotional information, resulting in a poor user experience. Furthermore, they fail to select an expert domain model or optimally allocate computing resources, resulting in the waste of computing resources. The purpose of this invention is to solve these problems.

[1713] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for analyzing task information and extracting appropriate keywords and concepts, means for selecting a specialized domain model based on the task and emotion information, and means for allocating necessary computational resources to the selected specialized domain model. This enables the generation of highly accurate answers to tasks. Furthermore, by taking the user's emotion information into consideration, a more satisfying user experience can be provided. Furthermore, efficient allocation of computational resources improves overall performance and reduces waste.

[1714] A "user" is an individual or group that uses the system to register tasks and receive answers.

[1715] A "task" refers to a specific problem or requirement that a user wants to solve.

[1716] "Emotion information" refers to data obtained by analyzing the user's emotional state at the time of task input.

[1717] A "domain model" is a computational model that includes knowledge and algorithms in a specific field.

[1718] "Computational resources" refers to the hardware and software resources (e.g., CPU, GPU, memory, etc.) required to run a domain-specific model.

[1719] "Answer" refers to the result or conclusion that a domain model generates for a task.

[1720] "Feedback" refers to the act of a user providing the system with an evaluation or opinion on the answer they received.

[1721] "Retraining" refers to the process of updating the domain-specific model based on feedback to improve accuracy the next time the task is performed.

[1722] The "server" refers to a central computing device that analyzes task information and emotional information, selects specialized domain models, allocates computing resources, and generates and integrates answers.

[1723] "Terminal" refers to a device (e.g., PC, smartphone, tablet, etc.) that a user uses to input tasks and receive answers.

[1724] "Data format" refers to a standardized structure (e.g., JSON, XML, etc.) for storing and transmitting information.

[1725] This invention is a system that utilizes domain expertise models and emotion engines to efficiently allocate computational resources to tasks registered by users and obtain highly accurate answers. The program for this system includes a series of processes from task registration to answer generation, feedback collection, and relearning.

[1726] In this system, a user first inputs the task they want to solve in natural language using a device (e.g., PC, smartphone, tablet, etc.). For example, they might input, "I would like you to analyze the ingredients for new drug development." The emotion engine then analyzes the user's input and extracts emotional information (e.g., urgency and anxiety). The device then converts this task information and emotional information into an appropriate data format (e.g., JSON format) and sends it to the server.

[1727] The server analyzes the received task information and extracts necessary keywords and concepts. For example, it identifies keywords such as "new drug development" and "component analysis." It also takes into account emotional information to determine the urgency of the task. Based on this, the server selects the domain model (e.g., medical model, chemistry model, pharmaceutical model) that is best suited to the task.

[1728] Next, the server allocates the necessary computing resources (e.g., GPU, CPU, memory, etc.) to each selected domain-specific model. If the emotional information indicates a high level of urgency, more computing resources are allocated. For example, a medical model is allocated twice as many GPUs as usual. Using the allocated resources, each model generates an answer to the task.

[1729] The server collects and integrates the generated answers from each specialized model. During integration, the contribution of each model is evaluated and a unified answer is generated. For example, the component analysis data from the chemical model and the human body effect data from the medical model are combined and integrated into a report format. The server then converts the integrated answer into an appropriate format and sends it to the user's device. The server also reflects the user's emotional information and adds an explanation that takes the user's feelings into consideration. For example, a report titled "Component Analysis Report of a New Drug" could be generated, with a comment added such as "We determined that an emergency response was required, so we conducted the analysis quickly."

[1730] Furthermore, the user evaluates the answers they receive and inputs their feedback into the system. For example, they may enter feedback such as "I'm satisfied with the analysis results" or "I need more information." The device converts the feedback into an appropriate format and sends it to the server. The server receives the feedback and evaluates the performance of each specialized domain model. If necessary, retraining is performed to improve accuracy the next time the task is performed. For example, a medical model can be retrained using new clinical data.

[1731] Through this series of processes, the system uses the domain model and emotion engine to process user tasks efficiently and with high accuracy, significantly contributing to solving user problems. Furthermore, by taking emotion information into consideration, the system can increase user satisfaction.

[1732] Examples of concrete examples and prompts

[1733] Specific examples

[1734] The user types into the terminal, "I would like to request a component analysis for new drug development."

[1735] The server selects a specialized area model related to new drug development, determines that the task is of high urgency, and allocates more computing resources to it.

[1736] Each specialized model starts its analysis and sends the results to the server.

[1737] The integrated answer is provided to the user, along with a comment such as, "We determined that an emergency response was required, so we conducted a rapid analysis."

[1738] Prompt Sentence Examples

[1739] Input: "Please provide detailed ingredient analysis for my drug development project. It's very important."

[1740] The above is a specific embodiment for carrying out the present invention.

[1741] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1742] System processing steps

[1743] Step 1: Task Input and Sentiment Analysis

[1744] The user inputs the task they want to solve using the terminal in natural language.

[1745] Input: "I would like to request a component analysis for new drug development."

[1746] How it works: The device uses an emotion engine to extract emotion information (e.g., urgency) from the user's input.

[1747] Step 2: Convert to data format and send

[1748] The device converts the input task information and emotion information into an appropriate data format (e.g., JSON format) and sends it to the server.

[1749] Input: User task information and emotion information

[1750] Output: JSON data "{\"task\":\"I would like you to analyze ingredients for new drug development\",\"emotion\":\"high urgency\"}"

[1751] What it does: Serializes input data into JSON format and sends it to the server.

[1752] Step 3: Task analysis and keyword extraction

[1753] The server analyzes the received task information and extracts keywords and concepts.

[1754] Input: JSON data

[1755] Output: Keywords "new drug development" and "component analysis"

[1756] How it works: Extracts keywords using natural language processing algorithms.

[1757] Step 4: Selecting a domain model

[1758] The server selects an appropriate domain model based on task information and emotion information.

[1759] Input: Keywords, emotion information

[1760] Output: Selected domain models (e.g., medical model, chemical model)

[1761] Action: Searches for and selects a corresponding model from the database.

[1762] Step 5: Allocate Computational Resources

[1763] The server allocates the necessary computing resources to the selected domain model.

[1764] Input: Selected domain model, emotion information

[1765] Output: Allocated computing resources (e.g. GPU, CPU specs)

[1766] How it works: A resource management system is used to allocate optimal resources to each model.

[1767] Step 6: Prepare and provide information about the model

[1768] The server prepares each model to run using the allocated resources and provides task and emotion information.

[1769] Input: Liquidation resources, task information, emotion information

[1770] Output: Ready to run

[1771] How it works: Loads the necessary libraries and datasets, and feeds the model with task information.

[1772] Step 7: Generate the answer

[1773] The server uses each domain model to generate an answer to the task.

[1774] Input: Task information provided to the model

[1775] Output: Answer data for each model

[1776] Action: The model performs a calculation and generates an answer (e.g., compositional analysis data).

[1777] Step 8: Consolidating the Answers

[1778] The server collects and integrates the answers generated by each model.

[1779] Input: Answer data for each model

[1780] Output: Consolidated answer

[1781] How it works: Evaluate the contribution of each answer and generate a single unified answer.

[1782] Step 9: Provide the answer

[1783] The server converts the integrated answers into an appropriate format and sends them to the user's device.

[1784] Input: Integrated answers, emotional information

[1785] Output: Answer data to the user

[1786] What it does: Converts the answer data into a user-friendly format (e.g., report format) and adds sentiment-sensitive comments.

[1787] Step 10: Gather feedback

[1788] The user rates the answers they receive and provides feedback.

[1789] Input: Rating and comments on the answer

[1790] Output: Feedback data

[1791] How it works: Use the input form to enter your rating and comments and submit your feedback.

[1792] Step 11: Submit your feedback

[1793] The device converts the feedback into an appropriate format and sends it to the server.

[1794] Input: User feedback

[1795] Output: Feedback data in JSON format

[1796] What it does: Serializes the feedback into JSON format and sends it to the server.

[1797] Step 12: Retraining

[1798] The server evaluates the performance of each specialized domain model based on the feedback and performs re-learning.

[1799] Input: Feedback data

[1800] Output: Updated model

[1801] How it works: Analyzes feedback and applies a re-training algorithm to improve the model.

[1802] The above is the specific program processing flow of this system.

[1803] (Application example 2)

[1804] 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."

[1805] In modern content delivery services, it is extremely important to provide users with the content they need quickly and accurately. However, conventional systems do not dynamically allocate resources based on the user's emotional state or the urgency of the task, limiting the improvement of the user experience. Furthermore, they lack a mechanism for effectively utilizing user feedback to improve accuracy in subsequent task executions. As a result, user satisfaction declines, leading to a diminishing value of the service.

[1806] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1807] In this invention, the server includes: means for analyzing a user's emotional information, determining the urgency based on the information, and adjusting the allocation of computational resources; means for the user to register a task they wish to solve; means for selecting a specialized domain model appropriate for the task; means for allocating the necessary computational resources to the selected model; means for the model to execute the task using the allocated resources; means for integrating and providing the execution results to the user; and means for collecting user feedback and reflecting it in subsequent processes. This enables dynamic resource allocation according to the user's emotional state and the urgency of the task, thereby enabling the rapid and accurate provision of content that meets the user's needs. Furthermore, by effectively utilizing user feedback to retrain the model, accuracy can be improved the next time the task is executed, thereby increasing user satisfaction.

[1808] A "task" refers to a specific request or purpose that a user requests the system to solve.

[1809] A "specialized domain model" is a machine learning model or algorithm built based on knowledge and data in a specific field or area.

[1810] "Emotional information" is emotional data such as emotional state and urgency that is analyzed from the user's input and behavior.

[1811] "Computational resources" refers to the computational resources such as CPU, GPU, and memory required to execute a task.

[1812] "Resource allocation" refers to the process of allocating the necessary computational resources to the selected model.

[1813] "Model execution" refers to the execution of calculations and analysis for a task by a selected domain-specific model using allocated computational resources.

[1814] "Execution result integration" refers to the compilation of results generated by multiple disciplinary models into a single unified answer or report.

[1815] "Feedback" refers to information indicating the user's evaluation and satisfaction with the answer they received.

[1816] "Retraining" is a learning process that improves the performance and accuracy of a domain-specific model based on collected feedback and newly acquired data.

[1817] "Urgency" is an indicator of how quickly a user's task should be resolved.

[1818] This invention is a system for quickly and accurately providing users with content they desire through a content distribution service. The system analyzes users' emotional information, determines the urgency of tasks based on that information, and dynamically adjusts the allocation of computing resources to provide personalized content tailored to the user's needs.

[1819] System Overview

[1820] This system utilizes domain expertise models and emotion engines to efficiently allocate the necessary computing resources to tasks registered by users, and provides highly accurate answers and content to users.

[1821] Hardware and software used

[1822] Hardware: Uses high-speed CPU, GPU and memory.

[1823] Software: Python 3.x, SentimentEngine (for sentiment analysis), and domain-specific models (news model, music model, video model, etc.).

[1824] System processing flow

[1825] 1. User Input and Sentiment Analysis

[1826] The user inputs the task they want to solve in natural language from their device. For example, they might input something like, "I want to know the latest technology news." The server receives this input and uses an emotion engine to analyze the emotional information and measure the urgency and anxiety of the task. The emotional information is treated as data that indicates, for example, the "high urgency" of the task.

[1827] 2. Selecting a Specialty Model and Allocating Resources

[1828] The server extracts necessary keywords and concepts based on the analyzed task information and emotion information, and selects the appropriate domain model. For example, for the keyword "technology news," a news model is selected. Then, the necessary computing resources are allocated based on the emotion data (urgency). If the urgency is high, more resources than usual are allocated.

[1829] 3. Execute the task

[1830] Using the allocated resources, selected domain-specific models perform tasks, such as a technology news model that analyzes the latest news and generates relevant articles.

[1831] 4. Consolidating and Presenting Results

[1832] The server collects and integrates the results generated by each model. This allows it to evaluate the contribution of each model and generate the optimal answer. The integrated answer is converted into an appropriate format and provided to the user along with a comment that takes into account the user's emotional information. For example, it may include a message such as, "Please check the technology news below. Due to its urgency, we analyzed it quickly."

[1833] 5. Feedback and Retraining

[1834] Users provide feedback on the answers they provide, and the server receives this feedback, evaluates the performance of each domain-specific model, and retrains it as needed, improving the model's accuracy the next time the task is performed.

[1835] Specific examples

[1836] For example, if a user requests "I want to know the latest technology news," the following prompt might be generated:

[1837] What's the latest technology news? It's very urgent right now.

[1838] Based on this prompt, the server performs a series of processes, including sentiment analysis, selection of domain expertise models, allocation of computing resources, task execution, integration and provision of results, and feedback collection and retraining, thereby quickly providing users with personalized, up-to-date technology news.

[1839] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1840] Step 1:

[1841] The device accepts the user's natural language input. The user types "I want to know the latest technology news" into the device. The device converts this input data into an appropriate data format (e.g., JSON) to send it to the server.

[1842] Input: User's natural language input

[1843] Data processing: converting natural language into data format

[1844] Output: JSON format data

[1845] Step 2:

[1846] The server receives the user input sent from the device and performs emotion analysis using an emotion engine. The emotion engine analyzes "urgency" and "anxiety" and extracts emotion information such as "high urgency."

[1847] Input: User-supplied data in JSON format

[1848] Data Computing: Sentiment Analysis

[1849] Output: Emotional information (e.g. "High Urgency")

[1850] Step 3:

[1851] The server extracts necessary keywords and concepts from the analyzed emotion information and user input. The keyword "technology news" is identified.

[1852] Input: User input and emotional information

[1853] Data processing: Keyword extraction

[1854] Output: The keyword "technology news"

[1855] Step 4:

[1856] The server selects an appropriate domain model (e.g., a news model) based on keywords and emotion information. At the same time, it dynamically allocates the necessary computing resources according to the emotion information, particularly the urgency. If the urgency is high, more CPU and GPU resources than usual are allocated.

[1857] Input: Extracted keywords and sentiment information

[1858] Data calculation: model selection, computational resource allocation

[1859] Output: Selected model and allocated resources

[1860] Step 5:

[1861] The server uses the allocated resources to run the selected news model. The model analyzes the task (e.g., get the latest technology news) and generates recommended news articles.

[1862] Input: Selected model and allocated resources

[1863] Data calculation: Model execution, recommended content generation

[1864] Output: Generated content (e.g., news article)

[1865] Step 6:

[1866] The server collects and integrates the generated content. If there are results from multiple domain-specific models, it evaluates each and generates a single integrated answer. At this time, it adds comments that take into account the user's emotional information.

[1867] Input: Multiple Content

[1868] Data processing: integrating results and adding comments

[1869] Output: Consolidated content and comments

[1870] Step 7:

[1871] The server transmits the integrated content to the terminal, which displays the content on the user's screen and makes it available to the user.

[1872] Input: Integrated content and comments

[1873] Data output: Send to terminal

[1874] Output: The content that is displayed on the user's screen

[1875] Step 8:

[1876] The user inputs feedback on the provided content, and the terminal converts this feedback into an appropriate data format for transmission to the server.

[1877] Input: User feedback

[1878] Data processing: Convert feedback into data format

[1879] Output: Feedback data in JSON format

[1880] Step 9:

[1881] The server analyzes the feedback received from the device and uses it to evaluate the performance of each specialized domain model and retrain it, and if necessary, retrains the model to improve its accuracy the next time the task is performed.

[1882] Input: Feedback data

[1883] Data calculation: performance evaluation, re-learning

[1884] Output: Improved model

[1885] 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.

[1886] 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.

[1887] 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 robot 414.

[1888] 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.

[1889] 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.

[1890] 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.

[1891] 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).

[1892] 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.

[1893] 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."

[1894] 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.

[1895] 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).

[1896] 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.

[1897] 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.

[1898] 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.

[1899] 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.

[1900] 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.

[1901] 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.

[1902] 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.

[1903] 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.

[1904] 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.

[1905] 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.

[1906] The following is further disclosed regarding the above embodiment.

[1907] Understood. Below are propose...

Claims

1. A means for users to register tasks they wish to solve; A means for selecting a domain expertise model appropriate for the task; A means for allocating necessary computational resources to the selected model; a means by which the model executes tasks using the allocated resources; and a means for integrating the execution results and providing them to the user; A means of collecting user feedback and incorporating it into future processes. A system including:

2. The system of claim 1 further comprising means for analyzing a task and evaluating the contributions of multiple domain-specific models.

3. The system of claim 1 further comprising means for retraining the domain expertise model based on the feedback.

Images