System

Abstract

A system is provided.SOLUTION: The system includes a fact check means for checking accuracy of the generated information, a legal check means for evaluating legal conformity to the generated information, an answer check means for evaluating whether the generated information is an optimal answer, an integrated evaluation means for integrating evaluation results to generate an overall evaluation and an explanation, a certification means for performing quality certification on a product that has obtained a high evaluation, and a relearning means for causing the generation model to relearn the evaluation results.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] There are three main concerns regarding information generated by generative AI models. First, it is necessary to verify whether the generated information is factual. Second, it is necessary to confirm whether the generated information infringes the rights of others. Third, it is necessary to verify whether the generated information is the optimal answer. To address these concerns, a system is needed that guarantees that the generated information is accurate, legally sound, and the optimal answer. [Means for solving the problem]

[0005] The present invention provides a system for evaluating the accuracy, legal compliance, and optimality of generated information. Specifically, it includes a fact-checking means, a legal check means, an answer check means, an integrated evaluation means, a certification means, and a re-learning means. This system comprehensively evaluates the output of a generative AI model to ensure its quality. The fact-checking means compares the generated information with a fact-checking database and verifies the facts using an external fact-checking API. The legal check means analyzes the generated information and checks for legal issues such as copyright infringement, privacy violations, and defamation. The answer check means evaluates whether the generated information is the optimal answer and provides feedback. The integrated evaluation means combines the evaluation results of each module to generate a final evaluation and commentary. Highly rated products are certified as having quality, and the evaluation results are retrained into the generative model to improve future evaluation accuracy.

[0006] A "generative AI model" is an artificial intelligence model that generates new output from blank input data for tasks such as natural language processing and image generation.

[0007] A "fact-checking tool" is a method or device for verifying whether generated information is factual.

[0008] "Legal check means" refers to a method or device for verifying whether the generated information is legally compliant.

[0009] An "answer check means" is a method or device for evaluating whether the generated information is the optimal answer.

[0010] An "integrated evaluation means" is a method or device that synthesizes the evaluation results from each module to generate a final evaluation and commentary.

[0011] "Qualification means" refers to a method or device that certifies the quality of a highly rated product.

[0012] "Relearning means" refers to a method or device for relearning a generative AI model based on evaluation results. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0014] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

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

[0016] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0017] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0018] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0019] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0020] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0021] [First embodiment]

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

[0023] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0024] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0025] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0026] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0027] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0028] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

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

[0030] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0031] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0032] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0033] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0034] The present invention is a system for evaluating the accuracy, legality, and appropriateness of generated information. The system is composed of the following main modules: a fact-checking module, a legal check module, an answer-checking module, an integrated evaluation module, a certification module, and a re-learning module.

[0035] Fact Checking Module

[0036] server:

[0037] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[0038] Legal Check Module

[0039] Device:

[0040] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[0041] Answer Check Module

[0042] User:

[0043] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[0044] Integrated Evaluation Module

[0045] server:

[0046] The evaluation results of each module are combined to generate a final evaluation and commentary. For example, the results of fact-checking, legal checks, and answer checks are combined to ensure that the generated information is factual, legally compliant, and the most appropriate answer.

[0047] Certification Module

[0048] server:

[0049] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[0050] Retraining Module

[0051] server:

[0052] The evaluation results are retrained into the generative AI model, thereby improving the future evaluation accuracy of the generative AI model.

[0053] This system comprehensively evaluates the output of the generative AI model, ensuring that the generated result is accurate, legally sound, and the most appropriate answer. The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[0054] The processing flow will be explained below.

[0055] Step 1:

[0056] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[0057] Step 2:

[0058] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[0059] Step 3:

[0060] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[0061] Step 4:

[0062] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[0063] Step 5:

[0064] The server sends the fact-checked product to the legal check module, which analyzes the product.

[0065] Step 6:

[0066] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0067] Step 7:

[0068] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[0069] Step 8:

[0070] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[0071] Step 9:

[0072] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[0073] Step 10:

[0074] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[0075] Step 11:

[0076] The server combines the results of the fact checks, legal checks, and answer checks to generate a final rating and commentary.

[0077] Step 12:

[0078] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[0079] Step 13:

[0080] The server retrains the generative AI model with the evaluation results, thereby improving the future evaluation accuracy of the generative AI model.

[0081] This detailed processing flow creates a system that ensures that the output of the generative AI model is accurate, legally sound, and the optimal answer.

[0082] Example 1

[0083] 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."

[0084] There is a need to comprehensively evaluate the accuracy, legal compliance, and optimality of products generated by generative AI models and ensure their quality in order to improve the reliability of the products. However, in conventional systems, these evaluations are often performed individually, making comprehensive evaluation difficult and resulting in a lack of reliability in the quality of the products. In addition, there is a lack of an interface that allows users to input prompts into the generative AI model and directly check the evaluation results.

[0085] 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.

[0086] In this invention, the server includes a fact-checking means for checking the accuracy of the generated information, a legal evaluation means for evaluating the legal compliance of the generated information, a response evaluation means for evaluating whether the generated information is the optimal answer, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an authentication means for certifying the quality of highly rated products, a re-learning means for relearning the evaluation results to the generative model, a user interface means for inputting prompt sentences to the generative AI model, and an evaluation result display means for displaying various evaluation results for the product. This makes it possible to comprehensively evaluate the accuracy, legal compliance, and optimality of the product and provide a highly reliable product.

[0087] A "generative AI model" is an artificial intelligence algorithm for generating text in natural language based on a prompt entered by a user.

[0088] A "fact-checking tool" is a device or program that cross-checks the generated information with a fact-checking database or external fact-checking API to verify its accuracy.

[0089] A "legal check means" is a device or program that analyzes information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation in order to assess whether the generated information is legally problematic.

[0090] An "answer checker" is a device or program that evaluates content based on specialized knowledge or additional research to assess whether the information generated is the best answer.

[0091] The "integrated evaluation means" is a device or program that synthesizes the evaluation results of the fact-checking means, legal check means, and answer check means to generate a final evaluation and commentary of the product.

[0092] A "certification means" is a device or program that certifies the quality of products that receive high ratings and issues an "AI quality certification certificate."

[0093] A "relearning means" is a device or program that retrains the generative AI model based on the evaluation results, thereby improving future evaluation accuracy.

[0094] The "user interface means" is an interface through which the user inputs a prompt statement to the generated AI model and the system starts the analysis process.

[0095] The "evaluation result display means" is a device or program for displaying various evaluation results for the product to the user.

[0096] This invention is an information generation system that utilizes a generative AI model, and is a system that evaluates the accuracy, legal compliance, and optimality of the generated information and guarantees its quality. Below, we will explain how to specifically implement this system.

[0097] Fact Checking Module

[0098] server:

[0099] The server receives the product of the generative AI model and performs fact-checking using the fact-checking module. Specifically, it compares the generated information with a fact-checking database and verifies the accuracy of the information using an external fact-checking API. For example, if the product generated is "The 2022 Nobel Peace Prize winner is Mr. A," the server queries the official Nobel Peace Prize database for that information.

[0100] Legal Check Module

[0101] Device:

[0102] The terminal analyzes the generated information sent from the server and performs a legal evaluation using the legal check module. Specifically, the generated information is analyzed and checked for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, if a generated information is "The CEO of Company X was arrested for illegal activity," the terminal checks the content against the law to see if there are any problems.

[0103] Answer Check Module

[0104] User:

[0105] The user receives the product sent from the server and evaluates it in the answer check module. Specifically, the user evaluates whether the generated information is the best answer based on their expert knowledge and additional research. For example, the user evaluates the product "The appropriate daily caffeine intake is XX mg" and provides feedback on the results.

[0106] Integrated Evaluation Module

[0107] server:

[0108] The server combines the evaluation results of the fact-checking module, legal check module, and answer check module to generate a final evaluation in the integrated evaluation module. Specifically, it aggregates the evaluation results of each module and checks whether the generated information is accurate, legally compliant, and the most appropriate answer.

[0109] Certification Module

[0110] server:

[0111] The server certifies the products that receive high ratings. Specifically, it issues an "AI Quality Certificate" based on the overall evaluation results, guaranteeing the reliability of the generated information.

[0112] Retraining Module

[0113] server:

[0114] The server updates the generative AI model using a re-learning module based on all evaluation results. Specifically, the server re-learns the generative AI model based on the evaluation results to improve the evaluation accuracy of future products.

[0115] User Interface Means

[0116] User:

[0117] The user inputs a prompt sentence to the generative AI model. Specifically, the system starts the analysis process through an interface for inputting the prompt sentence. The input prompt sentence is sent to the server.

[0118] Evaluation result display means

[0119] server:

[0120] The server displays various evaluation results for the product, including fact checks, legal checks, answer checks, and integrated evaluations, to allow users to visually check the quality of the product.

[0121] Examples of prompt statements

[0122] "Who will win the Nobel Peace Prize in 2022?"

[0123] "Please fact-check the reports that the CEO of Company X was arrested for illegal activity."

[0124] "What is an appropriate amount of caffeine to consume per day?"

[0125] This makes it possible to comprehensively evaluate the reliability of the products generated by the generative AI model and provide high-quality information.

[0126] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0127] Step 1:

[0128] User:

[0129] The user inputs a prompt to the generative AI model. The input prompt triggers the system to start the analysis process. For example, the prompt might be, "Please tell me who the Nobel Peace Prize winners are in 2022." The input data is the prompt, which becomes the input to the next step.

[0130] Step 2:

[0131] server:

[0132] The server inputs the prompt received from the user into a generative AI model. The generative AI model (for example, a natural language processing algorithm) generates text based on the prompt. The generated text is returned to the server as a product. For example, the product output is "The 2022 Nobel Peace Prize winner is Mr. A." Here, the input is the prompt, and the output is the product.

[0133] Step 3:

[0134] server:

[0135] The server receives the product and calls the fact-checking module. The fact-checking module analyzes the product and compares it with a fact-checking database. If necessary, it uses an external fact-checking API to verify the accuracy of the information. For example, it checks whether "Mr. A" is a 2022 Nobel Peace Prize winner. The input is the product, and the output is the result of the fact-checking.

[0136] Step 4:

[0137] Device:

[0138] The terminal receives the product sent from the server and calls the legal check module. The legal check module analyzes the product and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, the legal check module checks whether the product "The CEO of Company X was arrested for illegal activity" is legally compliant. The input is the product, and the output is the result of the legal evaluation.

[0139] Step 5:

[0140] User:

[0141] The user receives the product sent from the server and evaluates it using the answer check module. Based on their expertise and additional research, the user checks whether the product is the optimal answer and provides feedback. For example, the user evaluates whether the product "The appropriate daily caffeine intake is XX mg" is appropriate. The input is the product, and the output is the result of the optimality evaluation.

[0142] Step 6:

[0143] server:

[0144] The server aggregates the evaluation results of each module and calls the integrated evaluation module. The integrated evaluation module combines the results of fact checks, legal checks, and answer checks to generate a final evaluation. For example, it outputs an overall evaluation such as "The information is factual, legally compliant, and the best answer." The input is the evaluation results, and the output is the final evaluation.

[0145] Step 7:

[0146] server:

[0147] The server calls the certification module for the products that receive high evaluations. The certification module issues an "AI quality certificate" based on the overall evaluation results, guaranteeing the reliability of the products. The input is the final evaluation, and the output is the certificate.

[0148] Step 8:

[0149] server:

[0150] The server calls the re-learning module based on all the evaluation results. The re-learning module retrains the generative AI model with the evaluation results to improve the evaluation accuracy of future products. The input is the evaluation results, and the output is the updated generative AI model.

[0151] (Application example 1)

[0152] 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."

[0153] Conventional content distribution systems lack effective means for evaluating the accuracy, legal compliance, and appropriateness of the generated information, resulting in a high risk of distributing inaccurate or legally questionable information. Furthermore, the lack of comprehensive evaluation and quality certification to ensure reliable content often leaves users unsure about the reliability of the information provided. As a result, the reliability of content distribution services declines, leading to problems of reduced user satisfaction.

[0154] 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.

[0155] In this invention, the server includes a fact-checking means, a legal checking means, an answer checking means, an integrated evaluation means that integrates the evaluation results to generate an overall evaluation and commentary, a certification means that certifies the quality of highly rated products, a re-learning means that re-learns the evaluation results into a generative model, a distribution means that evaluates the accuracy, legal compliance, and optimality of the information to be distributed and distributes the quality-certified information, and an information assurance means that guarantees the reliability of the distributed information based on the evaluation results. This makes it possible to evaluate the accuracy, legal compliance, and optimality of the content to be distributed in advance, and to provide high-quality information to users.

[0156] "Generated information" refers to text data, image data, audio data, etc. generated by a natural language processing model or other information generation means.

[0157] "Fact-checking methods" are methods of fact-checking using fact-checking databases or external fact-checking APIs to verify the accuracy of the generated information.

[0158] "Legal check means" is a means for evaluating the legal compliance of generated information (copyright infringement, violation of privacy rights, defamation, etc.).

[0159] The "answer check means" is a means for evaluating whether the generated information is the optimal answer.

[0160] An "integrated evaluation method" is a method that integrates the evaluation results of fact-checking methods, legal check methods, and answer check methods to generate a comprehensive evaluation and commentary.

[0161] "Certification means" refers to a means for certifying the quality of highly rated products.

[0162] The "relearning means" is a means for improving the evaluation accuracy of a generative model by having the generative model relearn the evaluation results.

[0163] The "distribution means" is a means for distributing information whose quality has been certified based on the evaluation results to users.

[0164] "Information assurance means" is a means for assuring the reliability of distributed information based on the evaluation results.

[0165] This invention is a system that evaluates the accuracy, legal compliance, and optimality of generated information, distributes quality-certified information, and guarantees its reliability.

[0166] System configuration

[0167] The server includes the following means:

[0168] Fact-checking tools

[0169] Legal check methods

[0170] Answer check method

[0171] Integrated Assessment Instrument

[0172] Certification means

[0173] Re-learning methods

[0174] Delivery Method

[0175] Information Assurance Measures

[0176] System Operation

[0177] 1. Fact-checking methods:

[0178] The server compares the information generated by the generative AI model with a fact-checking database. It uses an external fact-checking API to verify the facts. For example, if the generative AI outputs "The winner of the 2022 award is ____," it compares that information with the database and uses an external API to verify the facts.

[0179] 2. Legal check methods:

[0180] The server analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, defamation, etc. For example, if the generating AI outputs "The CEO of a certain company was arrested for legal issues," the content is analyzed to check for legal issues.

[0181] 3. Answer check method:

[0182] The user receives the output sent by the server, evaluates its suitability based on their expertise and additional research, and provides feedback. For example, they evaluate the output, saying, "The appropriate daily caffeine intake is XX mg."

[0183] 4. Integrated assessment instruments:

[0184] The server combines the results of fact-checking, legal checking, and answer checking to generate an overall rating and commentary, for example, combining the rating results from each module to ensure the generated information is accurate, legally compliant, and the best answer.

[0185] 5. Certification means:

[0186] The server certifies the quality of the products that receive high evaluations, issuing a "Generation AI Quality Certificate" and providing evidence of the reliability of the products.

[0187] 6. Re-learning methods:

[0188] The server retrains the generative AI model with the evaluation results, thereby improving the evaluation accuracy of the generative AI model.

[0189] 7. Delivery Method:

[0190] The server delivers high-quality information that has been evaluated to users, such as articles and news that have been evaluated.

[0191] 8. Information Assurance Measures:

[0192] The server guarantees the authenticity of the information delivered, allowing users to feel confident about the accuracy and reliability of the information provided.

[0193] Specific examples

[0194] For example, you might input the following prompt into a generative AI model:

[0195] "Please tell me more about the 2022 Nobel Peace Prize winners."

[0196] The generative AI model generates information such as "The 2022 Nobel Peace Prize winner is XX, and his / her achievements are △△." This information is evaluated through fact-checking means, legal checking means, and answer checking means, and is then comprehensively evaluated by comprehensive evaluation means. If it is subsequently evaluated highly, quality is certified by certification means, and the information is distributed to users via distribution means. The reliability of the distributed information is guaranteed by information assurance means.

[0197] As a result, the reliability and quality of the information being distributed are ensured, and content distribution that users can use with peace of mind is realized.

[0198] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0199] Step 1:

[0200] The server inputs a prompt into the generative AI model and receives the generated information. Text data is generated based on this prompt, and the generated information is then evaluated. For example, the server inputs a prompt such as, "Please explain in detail the winners of the 2022 Nobel Peace Prize."

[0201] Step 2:

[0202] The server passes the generated information to the fact-checking means, which checks it against a fact-checking database. It uses an external fact-checking API to check the facts and retrieves the results from the database. For example, it checks whether the generated information, "The 2022 Nobel Peace Prize winner is ____," matches the actual name of the winner.

[0203] Step 3:

[0204] The server receives the fact-check results and then passes the generated information to the legal check means. The legal check means analyzes the generated information and detects legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, it checks whether the generated information "The CEO of a company was arrested for legal issues" constitutes defamation.

[0205] Step 4:

[0206] The server receives the results of the legal check and passes the generated information to the answer check means. The user receives the generated information sent from the server, evaluates its appropriateness based on their expertise and additional research, and provides feedback. For example, they evaluate whether the generated information, "The appropriate daily caffeine intake is XX mg," is medically appropriate.

[0207] Step 5:

[0208] The server combines the results of the fact checks, legal checks, and answer checks to generate an overall rating and commentary. The results of each check are averaged to calculate an overall rating score. For example, if the checks each score 95, 100, and 90, the average of each score would be 95.

[0209] Step 6:

[0210] The server certifies the quality of the products that receive high evaluations. It issues a "Generation AI Quality Certificate" through the certification method, providing evidence of the reliability of the product. For example, a certificate is issued if the overall evaluation score is 90 points or higher.

[0211] Step 7:

[0212] The server distributes certified, high-quality information to users through distribution methods. Articles and news that have been rated and certified are delivered to users. For example, rated news articles are distributed on a news app.

[0213] Step 8:

[0214] The server guarantees the reliability of the delivered information using information assurance measures. Users can use information with guaranteed reliability with peace of mind. The reliability is visually indicated by displaying an information authentication badge, etc.

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

[0216] The present invention combines a system that evaluates the accuracy, legality, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: a fact-checking module, a legal check module, an answer check module, an integrated evaluation module, a certification module, a re-learning module, and an emotion engine.

[0217] Fact Checking Module

[0218] server:

[0219] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[0220] Legal Check Module

[0221] Device:

[0222] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[0223] Answer Check Module

[0224] User:

[0225] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[0226] Emotion Engine

[0227] server:

[0228] An emotion engine is used to recognize the user's emotions, which analyzes how the user feels about the product.

[0229] Examples:

[0230] When the generation AI outputs an "opinion about a particular policy," the emotion engine analyzes whether the user has any dissatisfaction or doubts about the output and sends the results to the server.

[0231] Integrated Evaluation Module

[0232] server:

[0233] The evaluation results of each module are combined to generate a final rating and commentary. Fact checks, legal checks, answer checks, and sentiment data from the sentiment engine are combined to ensure the generated information is factual, legally compliant, and the most appropriate answer.

[0234] Certification Module

[0235] server:

[0236] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[0237] Retraining Module

[0238] server:

[0239] The generative AI model is retrained using the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0240] This system not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[0241] The processing flow will be explained below.

[0242] Step 1:

[0243] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[0244] Step 2:

[0245] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[0246] Step 3:

[0247] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[0248] Step 4:

[0249] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[0250] Step 5:

[0251] The server sends the fact-checked product to the legal check module, which analyzes the product.

[0252] Step 6:

[0253] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0254] Step 7:

[0255] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[0256] Step 8:

[0257] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[0258] Step 9:

[0259] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[0260] Step 10:

[0261] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[0262] Step 11:

[0263] The server acquires the user's emotion data using an emotion engine to recognize the user's emotion, analyzes this data, and evaluates the emotion of the product.

[0264] Step 12:

[0265] The server combines the results from the fact check, legal check, answer check, and sentiment engine to generate a final rating and commentary. The server ensures that the generated rating is factual, legally compliant, and the best answer.

[0266] Step 13:

[0267] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[0268] Step 14:

[0269] The server retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0270] This detailed processing flow not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.

[0271] Example 2

[0272] 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."

[0273] It is difficult to evaluate whether the information generated by generative AI models is factually accurate, legally compliant, or the most appropriate answer. It is also important to consider the user's emotions when making an evaluation, and a system that can achieve this in an integrated manner is needed.

[0274] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a fact confirmation means for checking the accuracy of the generated information, a legal review means for evaluating the legal compliance of the generated information, an answer evaluation means for evaluating whether the generated information is an optimal answer, a sentiment analysis means for recognizing and evaluating the user's sentiment, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an accreditation means for certifying the quality of highly rated products, and a re-learning means for relearning the evaluation results to the generative model. This enables the generated information to be based on facts, legally compliant, and an optimal answer, while also being evaluated taking the user's sentiment into consideration.

[0275] "Fact-checking means" refers to a means of verifying the accuracy of generated information using a database or external API.

[0276] "Legal review measures" are measures to evaluate whether the generated information is legally compliant and check for legal issues such as infringement of intellectual property rights, infringement of the right to protect personal information, and defamation.

[0277] The "answer evaluation means" is a means for evaluating whether the generated information is the optimal answer.

[0278] The "emotion analysis means" is a means for recognizing the user's emotions and reflecting them in the evaluation of the generated information.

[0279] An "integrated evaluation means" is a means for integrating the results from each evaluation means to generate an overall evaluation and commentary.

[0280] "Certification means" is a means of certifying the quality of products that have received high praise.

[0281] The "relearning means" is a means for feeding back the evaluation results to the generative model to perform re-learning, thereby improving future evaluation accuracy.

[0282] A "generative model" is an artificial intelligence model that generates information based on prompts from users.

[0283] The present invention combines an emotion engine that recognizes user emotions in a system that evaluates the accuracy, legal compliance, and appropriateness of generated information. The system includes a fact-checking means, a legal check means, an answer check means, a sentiment analysis means, an integrated evaluation means, a certification means, and a re-learning means.

[0284] Hardware and software used

[0285] Hardware

[0286] Server: central system for data processing and evaluation

[0287] Terminal: User input and display

[0288] User Device: Input Device for Emotion Analysis

[0289] software

[0290] Generative AI models (e.g., open-source generative models)

[0291] Fact-checking databases and external fact-checking APIs

[0292] Legal databases and legal document analysis tools

[0293] Sentiment Analysis Engine

[0294] Program processing

[0295] server

[0296] When a user sends a prompt, the server uses a generative AI model to generate output. For example, if the prompt is "When AI describes the future of driving, please generate output taking into consideration technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," the generative AI generates the corresponding text. The generated information is then fact-checked using a fact-checking method, which checks it against a fact-checking database and external fact-checking APIs (e.g., FactCheck.org and Snopes).

[0297] Terminal

[0298] The device receives the generated text from the server and analyzes the generated information using legal checks. For example, if the generated text contains a statement such as "The CEO of Company X was arrested for illegal conduct," the device analyzes the content and checks for legal issues such as intellectual property infringement, privacy infringement, and defamation. This check involves referencing legal databases (e.g., LexisNexis and Westlaw).

[0299] User

[0300] After receiving a product that has passed the legal check, the user evaluates whether the content is the optimal answer using the answer check tool. For example, for a product that says "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their expertise and additional research and provides feedback.

[0301] server

[0302] The emotion analysis means analyzes the user's feedback and emotions toward the product. For example, if the user feels that the information about the product is insufficient, the emotion engine recognizes that emotion from the user's text, facial expressions, and voice, and sends the results to the server. Next, the evaluation results of each module are integrated by the integrated evaluation means, and products that receive high ratings are issued a "quality certificate" by the certification means. The evaluation results and emotion analysis results are then fed back into the generative model, and the model is retrained by the retraining means.

[0303] Specific examples

[0304] As a concrete example, consider a generative AI model that generates output on the topic "The Future of AI-Driven Driving." The prompt in this case would be:

[0305] "When describing the future of AI driving, please generate output taking into account the technological trends announced in 2022, the associated legal issues, and how the general public will feel about them."

[0306] Based on this, a system is provided in which each means works together, the generated information is factual, legally compliant, provides the optimal answer, and is evaluated taking into account the user's emotions.

[0307] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0308] Step 1:

[0309] The user inputs a prompt, for example, "When AI describes the future of car driving, please generate an output taking into consideration the technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," and sends it to the server. Based on this input, the generative AI model generates an appropriate output.

[0310] Step 2:

[0311] The server uses a generative AI model to generate output based on the prompt. The generated text is analyzed and compared with a fact-checking database and an external fact-checking API. For example, if a product is generated such as "The 2022 Nobel Peace Prize winner is so-and-so," the database and API are used to fact-check the output. The output is the fact-checked text.

[0312] Step 3:

[0313] The server sends the fact-checked product to the terminal. The terminal then performs a legal review of the received product. Specifically, it analyzes the text in the product to determine whether it contains legal issues such as intellectual property infringement, personal information protection infringement, or defamation. For example, it analyzes a statement such as "The CEO of Company X was arrested for illegal activity" and checks whether there are any problems. The output is a product that is legally sound.

[0314] Step 4:

[0315] The terminal sends the legally reviewed product to the user, who evaluates whether the received product is the optimal answer. For example, for a product such as "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their own expertise and additional research and provides feedback. The output is the user's feedback and the evaluation result.

[0316] Step 5:

[0317] The server analyzes the feedback and evaluation results received from the user using emotion analysis means. It recognizes emotions from the user's text, facial expressions, and voice, and sends the evaluation results to the server. For example, if the user feels that "this information is insufficient," the emotion data is analyzed and sent to the server. The output is the result of the user's emotion analysis.

[0318] Step 6:

[0319] The server integrates the evaluation results obtained from the fact-checking means, legal checking means, answer checking means, and sentiment analysis means in an integrated evaluation means. For example, the server aggregates the evaluation results of each module to generate an overall score, and then generates a final evaluation and commentary. The output is the overall evaluation result and commentary.

[0320] Step 7:

[0321] If the overall evaluation result is high, the server issues a "quality certificate" using the certification means. This provides evidence of the reliability of the product. The quality certificate is assigned to the product. The output is the product with the issued quality certificate.

[0322] Step 8:

[0323] The server feeds back the overall evaluation results and sentiment analysis results to the generative model using a retraining method. This improves the future evaluation accuracy of the generative AI model. For example, the evaluation results are added to the dataset and the model is retrained. The output is a retrained generative AI model.

[0324] (Application example 2)

[0325] 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."

[0326] While existing technologies exist for individually evaluating the accuracy, legal compliance, and optimality of products generated by generative AI models, these evaluations do not take user emotions into account, making it difficult to improve the user experience or provide appropriate feedback based on emotions.The present invention aims to combine user emotions with the evaluation of generative AI models to produce products that are more realistic and reliable, and to provide optimized results that correspond to the user's emotions.

[0327] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: a fact-checking means for checking the accuracy of the generated information; a legal-checking means for evaluating the legal compliance of the generated information; an answer-checking means for evaluating whether the generated information is the optimal answer; a sentiment-analysis means for making an evaluation based on the user's emotions using an emotion engine that analyzes the user's emotions; an integrated-evaluation means for integrating the evaluation results to generate an overall evaluation and commentary; an accreditation means for certifying the quality of highly rated products; and a re-learning means for relearning the evaluation results to the generative model. This makes it possible for the generated information and warnings to be provided in a form that is accurate, legally sound, and optimized according to the user's emotions.

[0328] A "fact-checking tool" is a module that checks the accuracy of the generated information against a fact-checking database and uses an external fact-checking API to verify the facts.

[0329] The "legal check means" is a module that analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0330] The "answer check means" is a module for evaluating whether the generated information is the optimal answer.

[0331] The "emotion analysis means" is an emotion engine that analyzes the user's emotions and reevaluates the generated information based on those emotions.

[0332] The "integrated evaluation means" is a module that integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to generate a comprehensive evaluation and commentary.

[0333] The "certification means" is a module that certifies the quality of products that have received high evaluations.

[0334] The "relearning means" is a module that re-learns the evaluation results into the generative model, thereby improving the evaluation accuracy of the model.

[0335] This invention combines a system that evaluates the accuracy, legal compliance, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: fact-checking means, legal checking means, answer checking means, emotion analysis means, integrated evaluation means, certification means, and relearning means.

[0336] In this embodiment, the server performs the following process.

[0337] First, the fact-checking method verifies the accuracy of the information generated by the generative AI model. Specifically, it compares the generated information with a fact-checking database and, if necessary, uses an external fact-checking API to verify the facts. This ensures the accuracy of the output.

[0338] Next, the legal check means analyzes the generated information and evaluates its legal compliance. This means checks whether the generated information contains any legal issues such as copyright infringement, violation of privacy rights, or defamation. This ensures that the generated information is legally free.

[0339] The answer checker evaluates whether the generated information is the best answer. Based on the user's expertise and additional research, it verifies whether the generated answer is the best answer for the given situation. This ensures that the generated content is appropriate.

[0340] Furthermore, in this embodiment, an emotion engine is used as an emotion analysis means. The emotion engine analyzes the user's emotions in real time and provides information and feedback based on the analysis. It evaluates how the user feels about the generated information and adjusts the content and presentation method of the information accordingly.

[0341] The integrated evaluation method combines the results of each evaluation to generate a final evaluation and commentary. It integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to comprehensively judge the reliability and appropriateness of the generated product.

[0342] The certification method will certify the quality of products that receive high evaluations. Products that receive quality certification will be given an "AI Quality Certificate," which can be provided to external parties as evidence of the reliability of the information.

[0343] Finally, the retraining means retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0344] Specific examples

[0345] For example, if a user receives an alert that reads "Critical security breach detected" and feels "anxious," the emotion engine analyzes that emotion in real time. Based on the analysis results, the alert content is provided with information that is adjusted to prevent excessive tension. This emotion analysis allows the user to receive appropriate feedback, enabling them to respond more calmly.

[0346] Prompt Sentence Examples

[0347] An example of a prompt to be input to the generative AI model is as follows:

[0348] "Ask the Generative AI to create an incident report. Please create an incident report with the following information:

[0349] 1. Details of the incident and its scope.

[0350] 2. Include fact-checked data.

[0351] 3. Use legally acceptable language.

[0352] 4. Customize alert content based on user emotions.

[0353] As described above, the present invention is a system that can provide more realistic and reliable information by combining the user's emotions with the evaluation of generated information.

[0354] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0355] Step 1:

[0356] Receiving generated information

[0357] The server receives information generated by the generative AI model. The input is the generated information (e.g., incident report), and the output is the content of the information itself. This information is passed to various subsequent check modules.

[0358] Step 2:

[0359] Fact Check

[0360] The server passes the received information to the fact-checking means to check its accuracy. Specifically, it compares the input information with a fact-checking database and uses an external fact-checking API to check the facts. The output is a result of whether the information is true or not.

[0361] Step 3:

[0362] Legal Check

[0363] Based on the fact-checking results, the server passes the information to the legal checker. The input is the fact-checked information, which is analyzed to evaluate its legal compliance. Specifically, it checks for legal issues such as copyright infringement, violation of privacy rights, and defamation, and outputs a result indicating whether or not there is a legal problem.

[0364] Step 4:

[0365] Answer Check

[0366] The server passes the information to the answer checker based on the results of the legal check. The input is the legally checked information, and the server evaluates whether this information is the best answer. The output is an evaluation of the information's suitability based on the user's expertise and additional research.

[0367] Step 5:

[0368] Emotion analysis

[0369] The server passes the information to the emotion analysis means based on the answer check results. The input is the answer-checked information and the user's emotion data, which the emotion engine analyzes to understand the user's emotional state. The output is an analysis result based on the user's emotion.

[0370] Step 6:

[0371] Integrated Evaluation

[0372] The server integrates all the evaluation results based on the results of sentiment analysis. The inputs are the results of fact-checking, legal checks, answer checks, and sentiment analysis, and combines them to generate a final evaluation and commentary. The output is an overall evaluation result based on the information's reliability, legal relevance, appropriateness, and user sentiment.

[0373] Step 7:

[0374] Certification

[0375] The server certifies the quality of information that receives a high rating based on the results of the integrated evaluation. The input is the results of the integrated evaluation, and the output is an "AI Quality Certificate."

[0376] Step 8:

[0377] Relearn

[0378] The server retrains the generative AI model based on the evaluation results and sentiment analysis results based on the certification results. The inputs are the evaluation results and sentiment analysis results, and the generative AI model is updated based on these, improving the model's evaluation accuracy as output.

[0379] The above are the specific processing steps of the program for the system that realizes the application example.

[0380] 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.

[0381] 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.

[0382] 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.

[0383] [Second embodiment]

[0384] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0385] 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.

[0386] 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).

[0387] 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.

[0388] 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.

[0389] 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).

[0390] 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.

[0391] 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.

[0392] 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.

[0393] 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.

[0394] 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.

[0395] 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."

[0396] The present invention is a system for evaluating the accuracy, legality, and appropriateness of generated information. The system is composed of the following main modules: a fact-checking module, a legal check module, an answer-checking module, an integrated evaluation module, a certification module, and a re-learning module.

[0397] Fact Checking Module

[0398] server:

[0399] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[0400] Legal Check Module

[0401] Device:

[0402] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[0403] Answer Check Module

[0404] User:

[0405] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[0406] Integrated Evaluation Module

[0407] server:

[0408] The evaluation results of each module are combined to generate a final evaluation and commentary. For example, the results of fact-checking, legal checks, and answer checks are combined to ensure that the generated information is factual, legally compliant, and the most appropriate answer.

[0409] Certification Module

[0410] server:

[0411] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[0412] Retraining Module

[0413] server:

[0414] The evaluation results are retrained into the generative AI model, thereby improving the future evaluation accuracy of the generative AI model.

[0415] This system comprehensively evaluates the output of the generative AI model, ensuring that the generated result is accurate, legally sound, and the most appropriate answer. The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[0416] The processing flow will be explained below.

[0417] Step 1:

[0418] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[0419] Step 2:

[0420] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[0421] Step 3:

[0422] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[0423] Step 4:

[0424] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[0425] Step 5:

[0426] The server sends the fact-checked product to the legal check module, which analyzes the product.

[0427] Step 6:

[0428] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0429] Step 7:

[0430] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[0431] Step 8:

[0432] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[0433] Step 9:

[0434] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[0435] Step 10:

[0436] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[0437] Step 11:

[0438] The server combines the results of the fact checks, legal checks, and answer checks to generate a final rating and commentary.

[0439] Step 12:

[0440] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[0441] Step 13:

[0442] The server retrains the generative AI model with the evaluation results, thereby improving the future evaluation accuracy of the generative AI model.

[0443] This detailed processing flow creates a system that ensures that the output of the generative AI model is accurate, legally sound, and the optimal answer.

[0444] Example 1

[0445] 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."

[0446] There is a need to comprehensively evaluate the accuracy, legal compliance, and optimality of products generated by generative AI models and ensure their quality in order to improve the reliability of the products. However, in conventional systems, these evaluations are often performed individually, making comprehensive evaluation difficult and resulting in a lack of reliability in the quality of the products. In addition, there is a lack of an interface that allows users to input prompts into the generative AI model and directly check the evaluation results.

[0447] 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.

[0448] In this invention, the server includes a fact-checking means for checking the accuracy of the generated information, a legal evaluation means for evaluating the legal compliance of the generated information, a response evaluation means for evaluating whether the generated information is the optimal answer, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an authentication means for certifying the quality of highly rated products, a re-learning means for relearning the evaluation results to the generative model, a user interface means for inputting prompt sentences to the generative AI model, and an evaluation result display means for displaying various evaluation results for the product. This makes it possible to comprehensively evaluate the accuracy, legal compliance, and optimality of the product and provide a highly reliable product.

[0449] A "generative AI model" is an artificial intelligence algorithm for generating text in natural language based on a prompt entered by a user.

[0450] A "fact-checking tool" is a device or program that cross-checks the generated information with a fact-checking database or external fact-checking API to verify its accuracy.

[0451] A "legal check means" is a device or program that analyzes information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation in order to assess whether the generated information is legally problematic.

[0452] An "answer checker" is a device or program that evaluates content based on specialized knowledge or additional research to assess whether the information generated is the best answer.

[0453] The "integrated evaluation means" is a device or program that synthesizes the evaluation results of the fact-checking means, legal check means, and answer check means to generate a final evaluation and commentary of the product.

[0454] A "certification means" is a device or program that certifies the quality of products that receive high ratings and issues an "AI quality certification certificate."

[0455] A "relearning means" is a device or program that retrains the generative AI model based on the evaluation results, thereby improving future evaluation accuracy.

[0456] The "user interface means" is an interface through which the user inputs a prompt statement to the generated AI model and the system starts the analysis process.

[0457] The "evaluation result display means" is a device or program for displaying various evaluation results for the product to the user.

[0458] This invention is an information generation system that utilizes a generative AI model, and is a system that evaluates the accuracy, legal compliance, and optimality of the generated information and guarantees its quality. Below, we will explain how to specifically implement this system.

[0459] Fact Checking Module

[0460] server:

[0461] The server receives the product of the generative AI model and performs fact-checking using the fact-checking module. Specifically, it compares the generated information with a fact-checking database and verifies the accuracy of the information using an external fact-checking API. For example, if the product generated is "The 2022 Nobel Peace Prize winner is Mr. A," the server queries the official Nobel Peace Prize database for that information.

[0462] Legal Check Module

[0463] Device:

[0464] The terminal analyzes the generated information sent from the server and performs a legal evaluation using the legal check module. Specifically, the generated information is analyzed and checked for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, if a generated information is "The CEO of Company X was arrested for illegal activity," the terminal checks the content against the law to see if there are any problems.

[0465] Answer Check Module

[0466] User:

[0467] The user receives the product sent from the server and evaluates it in the answer check module. Specifically, the user evaluates whether the generated information is the best answer based on their expert knowledge and additional research. For example, the user evaluates the product "The appropriate daily caffeine intake is XX mg" and provides feedback on the results.

[0468] Integrated Evaluation Module

[0469] server:

[0470] The server combines the evaluation results of the fact-checking module, legal check module, and answer check module to generate a final evaluation in the integrated evaluation module. Specifically, it aggregates the evaluation results of each module and checks whether the generated information is accurate, legally compliant, and the most appropriate answer.

[0471] Certification Module

[0472] server:

[0473] The server certifies the products that receive high ratings. Specifically, it issues an "AI Quality Certificate" based on the overall evaluation results, guaranteeing the reliability of the generated information.

[0474] Retraining Module

[0475] server:

[0476] The server updates the generative AI model using a re-learning module based on all evaluation results. Specifically, the server re-learns the generative AI model based on the evaluation results to improve the evaluation accuracy of future products.

[0477] User Interface Means

[0478] User:

[0479] The user inputs a prompt sentence to the generative AI model. Specifically, the system starts the analysis process through an interface for inputting the prompt sentence. The input prompt sentence is sent to the server.

[0480] Evaluation result display means

[0481] server:

[0482] The server displays various evaluation results for the product, including fact checks, legal checks, answer checks, and integrated evaluations, to allow users to visually check the quality of the product.

[0483] Examples of prompt statements

[0484] "Who will win the Nobel Peace Prize in 2022?"

[0485] "Please fact-check the reports that the CEO of Company X was arrested for illegal activity."

[0486] "What is an appropriate amount of caffeine to consume per day?"

[0487] This makes it possible to comprehensively evaluate the reliability of the products generated by the generative AI model and provide high-quality information.

[0488] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0489] Step 1:

[0490] User:

[0491] The user inputs a prompt to the generative AI model. The input prompt triggers the system to start the analysis process. For example, the prompt might be, "Please tell me who the Nobel Peace Prize winners are in 2022." The input data is the prompt, which becomes the input to the next step.

[0492] Step 2:

[0493] server:

[0494] The server inputs the prompt received from the user into a generative AI model. The generative AI model (for example, a natural language processing algorithm) generates text based on the prompt. The generated text is returned to the server as a product. For example, the product output is "The 2022 Nobel Peace Prize winner is Mr. A." Here, the input is the prompt, and the output is the product.

[0495] Step 3:

[0496] server:

[0497] The server receives the product and calls the fact-checking module. The fact-checking module analyzes the product and compares it with a fact-checking database. If necessary, it uses an external fact-checking API to verify the accuracy of the information. For example, it checks whether "Mr. A" is a 2022 Nobel Peace Prize winner. The input is the product, and the output is the result of the fact-checking.

[0498] Step 4:

[0499] Device:

[0500] The terminal receives the product sent from the server and calls the legal check module. The legal check module analyzes the product and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, the legal check module checks whether the product "The CEO of Company X was arrested for illegal activity" is legally compliant. The input is the product, and the output is the result of the legal evaluation.

[0501] Step 5:

[0502] User:

[0503] The user receives the product sent from the server and evaluates it using the answer check module. Based on their expertise and additional research, the user checks whether the product is the optimal answer and provides feedback. For example, the user evaluates whether the product "The appropriate daily caffeine intake is XX mg" is appropriate. The input is the product, and the output is the result of the optimality evaluation.

[0504] Step 6:

[0505] server:

[0506] The server aggregates the evaluation results of each module and calls the integrated evaluation module. The integrated evaluation module combines the results of fact checks, legal checks, and answer checks to generate a final evaluation. For example, it outputs an overall evaluation such as "The information is factual, legally compliant, and the best answer." The input is the evaluation results, and the output is the final evaluation.

[0507] Step 7:

[0508] server:

[0509] The server calls the certification module for the products that receive high evaluations. The certification module issues an "AI quality certificate" based on the overall evaluation results, guaranteeing the reliability of the products. The input is the final evaluation, and the output is the certificate.

[0510] Step 8:

[0511] server:

[0512] The server calls the re-learning module based on all the evaluation results. The re-learning module retrains the generative AI model with the evaluation results to improve the evaluation accuracy of future products. The input is the evaluation results, and the output is the updated generative AI model.

[0513] (Application example 1)

[0514] 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."

[0515] Conventional content distribution systems lack effective means for evaluating the accuracy, legal compliance, and appropriateness of the generated information, resulting in a high risk of distributing inaccurate or legally questionable information. Furthermore, the lack of comprehensive evaluation and quality certification to ensure reliable content often leaves users unsure about the reliability of the information provided. As a result, the reliability of content distribution services declines, leading to problems of reduced user satisfaction.

[0516] 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.

[0517] In this invention, the server includes a fact-checking means, a legal checking means, an answer checking means, an integrated evaluation means that integrates the evaluation results to generate an overall evaluation and commentary, a certification means that certifies the quality of highly rated products, a re-learning means that re-learns the evaluation results into a generative model, a distribution means that evaluates the accuracy, legal compliance, and optimality of the information to be distributed and distributes the quality-certified information, and an information assurance means that guarantees the reliability of the distributed information based on the evaluation results. This makes it possible to evaluate the accuracy, legal compliance, and optimality of the content to be distributed in advance, and to provide high-quality information to users.

[0518] "Generated information" refers to text data, image data, audio data, etc. generated by a natural language processing model or other information generation means.

[0519] "Fact-checking methods" are methods of fact-checking using fact-checking databases or external fact-checking APIs to verify the accuracy of the generated information.

[0520] "Legal check means" is a means for evaluating the legal compliance of generated information (copyright infringement, violation of privacy rights, defamation, etc.).

[0521] The "answer check means" is a means for evaluating whether the generated information is the optimal answer.

[0522] An "integrated evaluation method" is a method that integrates the evaluation results of fact-checking methods, legal check methods, and answer check methods to generate a comprehensive evaluation and commentary.

[0523] "Certification means" refers to a means for certifying the quality of highly rated products.

[0524] The "relearning means" is a means for improving the evaluation accuracy of a generative model by having the generative model relearn the evaluation results.

[0525] The "distribution means" is a means for distributing information whose quality has been certified based on the evaluation results to users.

[0526] "Information assurance means" is a means for assuring the reliability of distributed information based on the evaluation results.

[0527] This invention is a system that evaluates the accuracy, legal compliance, and optimality of generated information, distributes quality-certified information, and guarantees its reliability.

[0528] System configuration

[0529] The server includes the following means:

[0530] Fact-checking tools

[0531] Legal check methods

[0532] Answer check method

[0533] Integrated Assessment Instrument

[0534] Certification means

[0535] Re-learning methods

[0536] Delivery Method

[0537] Information Assurance Measures

[0538] System Operation

[0539] 1. Fact-checking methods:

[0540] The server compares the information generated by the generative AI model with a fact-checking database. It uses an external fact-checking API to verify the facts. For example, if the generative AI outputs "The winner of the 2022 award is ____," it compares that information with the database and uses an external API to verify the facts.

[0541] 2. Legal check methods:

[0542] The server analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, defamation, etc. For example, if the generating AI outputs "The CEO of a certain company was arrested for legal issues," the content is analyzed to check for legal issues.

[0543] 3. Answer check method:

[0544] The user receives the output sent by the server, evaluates its suitability based on their expertise and additional research, and provides feedback. For example, they evaluate the output, saying, "The appropriate daily caffeine intake is XX mg."

[0545] 4. Integrated assessment instruments:

[0546] The server combines the results of fact-checking, legal checking, and answer checking to generate an overall rating and commentary, for example, combining the rating results from each module to ensure the generated information is accurate, legally compliant, and the best answer.

[0547] 5. Certification means:

[0548] The server certifies the quality of the products that receive high evaluations, issuing a "Generation AI Quality Certificate" and providing evidence of the reliability of the products.

[0549] 6. Re-learning methods:

[0550] The server retrains the generative AI model with the evaluation results, thereby improving the evaluation accuracy of the generative AI model.

[0551] 7. Delivery Method:

[0552] The server delivers high-quality information that has been evaluated to users, such as articles and news that have been evaluated.

[0553] 8. Information Assurance Measures:

[0554] The server guarantees the authenticity of the information delivered, allowing users to feel confident about the accuracy and reliability of the information provided.

[0555] Specific examples

[0556] For example, you might input the following prompt into a generative AI model:

[0557] "Please tell me more about the 2022 Nobel Peace Prize winners."

[0558] The generative AI model generates information such as "The 2022 Nobel Peace Prize winner is XX, and his / her achievements are △△." This information is evaluated through fact-checking means, legal checking means, and answer checking means, and is then comprehensively evaluated by comprehensive evaluation means. If it is subsequently evaluated highly, quality is certified by certification means, and the information is distributed to users via distribution means. The reliability of the distributed information is guaranteed by information assurance means.

[0559] As a result, the reliability and quality of the information being distributed are ensured, and content distribution that users can use with peace of mind is realized.

[0560] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0561] Step 1:

[0562] The server inputs a prompt into the generative AI model and receives the generated information. Text data is generated based on this prompt, and the generated information is then evaluated. For example, the server inputs a prompt such as, "Please explain in detail the winners of the 2022 Nobel Peace Prize."

[0563] Step 2:

[0564] The server passes the generated information to the fact-checking means, which checks it against a fact-checking database. It uses an external fact-checking API to check the facts and retrieves the results from the database. For example, it checks whether the generated information, "The 2022 Nobel Peace Prize winner is ____," matches the actual name of the winner.

[0565] Step 3:

[0566] The server receives the fact-check results and then passes the generated information to the legal check means. The legal check means analyzes the generated information and detects legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, it checks whether the generated information "The CEO of a company was arrested for legal issues" constitutes defamation.

[0567] Step 4:

[0568] The server receives the results of the legal check and passes the generated information to the answer check means. The user receives the generated information sent from the server, evaluates its appropriateness based on their expertise and additional research, and provides feedback. For example, they evaluate whether the generated information, "The appropriate daily caffeine intake is XX mg," is medically appropriate.

[0569] Step 5:

[0570] The server combines the results of the fact checks, legal checks, and answer checks to generate an overall rating and commentary. The results of each check are averaged to calculate an overall rating score. For example, if the checks each score 95, 100, and 90, the average of each score would be 95.

[0571] Step 6:

[0572] The server certifies the quality of the products that receive high evaluations. It issues a "Generation AI Quality Certificate" through the certification method, providing evidence of the reliability of the product. For example, a certificate is issued if the overall evaluation score is 90 points or higher.

[0573] Step 7:

[0574] The server distributes certified, high-quality information to users through distribution methods. Articles and news that have been rated and certified are delivered to users. For example, rated news articles are distributed on a news app.

[0575] Step 8:

[0576] The server guarantees the reliability of the delivered information using information assurance measures. Users can use information with guaranteed reliability with peace of mind. The reliability is visually indicated by displaying an information authentication badge, etc.

[0577] 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.

[0578] The present invention combines a system that evaluates the accuracy, legality, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: a fact-checking module, a legal check module, an answer check module, an integrated evaluation module, a certification module, a re-learning module, and an emotion engine.

[0579] Fact Checking Module

[0580] server:

[0581] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[0582] Legal Check Module

[0583] Device:

[0584] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[0585] Answer Check Module

[0586] User:

[0587] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[0588] Emotion Engine

[0589] server:

[0590] An emotion engine is used to recognize the user's emotions, which analyzes how the user feels about the product.

[0591] Examples:

[0592] When the generation AI outputs an "opinion about a particular policy," the emotion engine analyzes whether the user has any dissatisfaction or doubts about the output and sends the results to the server.

[0593] Integrated Evaluation Module

[0594] server:

[0595] The evaluation results of each module are combined to generate a final rating and commentary. Fact checks, legal checks, answer checks, and sentiment data from the sentiment engine are combined to ensure the generated information is factual, legally compliant, and the most appropriate answer.

[0596] Certification Module

[0597] server:

[0598] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[0599] Retraining Module

[0600] server:

[0601] The generative AI model is retrained using the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0602] This system not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[0603] The processing flow will be explained below.

[0604] Step 1:

[0605] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[0606] Step 2:

[0607] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[0608] Step 3:

[0609] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[0610] Step 4:

[0611] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[0612] Step 5:

[0613] The server sends the fact-checked product to the legal check module, which analyzes the product.

[0614] Step 6:

[0615] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0616] Step 7:

[0617] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[0618] Step 8:

[0619] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[0620] Step 9:

[0621] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[0622] Step 10:

[0623] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[0624] Step 11:

[0625] The server acquires the user's emotion data using an emotion engine to recognize the user's emotion, analyzes this data, and evaluates the emotion of the product.

[0626] Step 12:

[0627] The server combines the results from the fact check, legal check, answer check, and sentiment engine to generate a final rating and commentary. The server ensures that the generated rating is factual, legally compliant, and the best answer.

[0628] Step 13:

[0629] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[0630] Step 14:

[0631] The server retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0632] This detailed processing flow not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.

[0633] Example 2

[0634] 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."

[0635] It is difficult to evaluate whether the information generated by generative AI models is factually accurate, legally compliant, or the most appropriate answer. It is also important to consider the user's emotions when making an evaluation, and a system that can achieve this in an integrated manner is needed.

[0636] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a fact confirmation means for checking the accuracy of the generated information, a legal review means for evaluating the legal compliance of the generated information, an answer evaluation means for evaluating whether the generated information is an optimal answer, a sentiment analysis means for recognizing and evaluating the user's sentiment, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an accreditation means for certifying the quality of highly rated products, and a re-learning means for relearning the evaluation results to the generative model. This enables the generated information to be based on facts, legally compliant, and an optimal answer, while also being evaluated taking the user's sentiment into consideration.

[0637] "Fact-checking means" refers to a means of verifying the accuracy of generated information using a database or external API.

[0638] "Legal review measures" are measures to evaluate whether the generated information is legally compliant and check for legal issues such as infringement of intellectual property rights, infringement of the right to protect personal information, and defamation.

[0639] The "answer evaluation means" is a means for evaluating whether the generated information is the optimal answer.

[0640] The "emotion analysis means" is a means for recognizing the user's emotions and reflecting them in the evaluation of the generated information.

[0641] An "integrated evaluation means" is a means for integrating the results from each evaluation means to generate an overall evaluation and commentary.

[0642] "Certification means" is a means of certifying the quality of products that have received high praise.

[0643] The "relearning means" is a means for feeding back the evaluation results to the generative model to perform re-learning, thereby improving future evaluation accuracy.

[0644] A "generative model" is an artificial intelligence model that generates information based on prompts from users.

[0645] The present invention combines an emotion engine that recognizes user emotions in a system that evaluates the accuracy, legal compliance, and appropriateness of generated information. The system includes a fact-checking means, a legal check means, an answer check means, a sentiment analysis means, an integrated evaluation means, a certification means, and a re-learning means.

[0646] Hardware and software used

[0647] Hardware

[0648] Server: central system for data processing and evaluation

[0649] Terminal: User input and display

[0650] User Device: Input Device for Emotion Analysis

[0651] software

[0652] Generative AI models (e.g., open-source generative models)

[0653] Fact-checking databases and external fact-checking APIs

[0654] Legal databases and legal document analysis tools

[0655] Sentiment Analysis Engine

[0656] Program processing

[0657] server

[0658] When a user sends a prompt, the server uses a generative AI model to generate output. For example, if the prompt is "When AI describes the future of driving, please generate output taking into consideration technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," the generative AI generates the corresponding text. The generated information is then fact-checked using a fact-checking method, which checks it against a fact-checking database and external fact-checking APIs (e.g., FactCheck.org and Snopes).

[0659] Terminal

[0660] The device receives the generated text from the server and analyzes the generated information using legal checks. For example, if the generated text contains a statement such as "The CEO of Company X was arrested for illegal conduct," the device analyzes the content and checks for legal issues such as intellectual property infringement, privacy infringement, and defamation. This check involves referencing legal databases (e.g., LexisNexis and Westlaw).

[0661] User

[0662] After receiving a product that has passed the legal check, the user evaluates whether the content is the optimal answer using the answer check tool. For example, for a product that says "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their expertise and additional research and provides feedback.

[0663] server

[0664] The emotion analysis means analyzes the user's feedback and emotions toward the product. For example, if the user feels that the information about the product is insufficient, the emotion engine recognizes that emotion from the user's text, facial expressions, and voice, and sends the results to the server. Next, the evaluation results of each module are integrated by the integrated evaluation means, and products that receive high ratings are issued a "quality certificate" by the certification means. The evaluation results and emotion analysis results are then fed back into the generative model, and the model is retrained by the retraining means.

[0665] Specific examples

[0666] As a concrete example, consider a generative AI model that generates output on the topic "The Future of AI-Driven Driving." The prompt in this case would be:

[0667] "When describing the future of AI driving, please generate output taking into account the technological trends announced in 2022, the associated legal issues, and how the general public will feel about them."

[0668] Based on this, a system is provided in which each means works together, the generated information is factual, legally compliant, provides the optimal answer, and is evaluated taking into account the user's emotions.

[0669] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0670] Step 1:

[0671] The user inputs a prompt, for example, "When AI describes the future of car driving, please generate an output taking into consideration the technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," and sends it to the server. Based on this input, the generative AI model generates an appropriate output.

[0672] Step 2:

[0673] The server uses a generative AI model to generate output based on the prompt. The generated text is analyzed and compared with a fact-checking database and an external fact-checking API. For example, if a product is generated such as "The 2022 Nobel Peace Prize winner is so-and-so," the database and API are used to fact-check the output. The output is the fact-checked text.

[0674] Step 3:

[0675] The server sends the fact-checked product to the terminal. The terminal then performs a legal review of the received product. Specifically, it analyzes the text in the product to determine whether it contains legal issues such as intellectual property infringement, personal information protection infringement, or defamation. For example, it analyzes a statement such as "The CEO of Company X was arrested for illegal activity" and checks whether there are any problems. The output is a product that is legally sound.

[0676] Step 4:

[0677] The terminal sends the legally reviewed product to the user, who evaluates whether the received product is the optimal answer. For example, for a product such as "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their own expertise and additional research and provides feedback. The output is the user's feedback and the evaluation result.

[0678] Step 5:

[0679] The server analyzes the feedback and evaluation results received from the user using emotion analysis means. It recognizes emotions from the user's text, facial expressions, and voice, and sends the evaluation results to the server. For example, if the user feels that "this information is insufficient," the emotion data is analyzed and sent to the server. The output is the result of the user's emotion analysis.

[0680] Step 6:

[0681] The server integrates the evaluation results obtained from the fact-checking means, legal checking means, answer checking means, and sentiment analysis means in an integrated evaluation means. For example, the server aggregates the evaluation results of each module to generate an overall score, and then generates a final evaluation and commentary. The output is the overall evaluation result and commentary.

[0682] Step 7:

[0683] If the overall evaluation result is high, the server issues a "quality certificate" using the certification means. This provides evidence of the reliability of the product. The quality certificate is assigned to the product. The output is the product with the issued quality certificate.

[0684] Step 8:

[0685] The server feeds back the overall evaluation results and sentiment analysis results to the generative model using a retraining method. This improves the future evaluation accuracy of the generative AI model. For example, the evaluation results are added to the dataset and the model is retrained. The output is a retrained generative AI model.

[0686] (Application example 2)

[0687] 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."

[0688] While existing technologies exist for individually evaluating the accuracy, legal compliance, and optimality of products generated by generative AI models, these evaluations do not take user emotions into account, making it difficult to improve the user experience or provide appropriate feedback based on emotions.The present invention aims to combine user emotions with the evaluation of generative AI models to produce products that are more realistic and reliable, and to provide optimized results that correspond to the user's emotions.

[0689] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: a fact-checking means for checking the accuracy of the generated information; a legal-checking means for evaluating the legal compliance of the generated information; an answer-checking means for evaluating whether the generated information is the optimal answer; a sentiment-analysis means for making an evaluation based on the user's emotions using an emotion engine that analyzes the user's emotions; an integrated-evaluation means for integrating the evaluation results to generate an overall evaluation and commentary; an accreditation means for certifying the quality of highly rated products; and a re-learning means for relearning the evaluation results to the generative model. This makes it possible for the generated information and warnings to be provided in a form that is accurate, legally sound, and optimized according to the user's emotions.

[0690] A "fact-checking tool" is a module that checks the accuracy of the generated information against a fact-checking database and uses an external fact-checking API to verify the facts.

[0691] The "legal check means" is a module that analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0692] The "answer check means" is a module for evaluating whether the generated information is the optimal answer.

[0693] The "emotion analysis means" is an emotion engine that analyzes the user's emotions and reevaluates the generated information based on those emotions.

[0694] The "integrated evaluation means" is a module that integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to generate a comprehensive evaluation and commentary.

[0695] The "certification means" is a module that certifies the quality of products that have received high evaluations.

[0696] The "relearning means" is a module that re-learns the evaluation results into the generative model, thereby improving the evaluation accuracy of the model.

[0697] This invention combines a system that evaluates the accuracy, legal compliance, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: fact-checking means, legal checking means, answer checking means, emotion analysis means, integrated evaluation means, certification means, and relearning means.

[0698] In this embodiment, the server performs the following process.

[0699] First, the fact-checking method verifies the accuracy of the information generated by the generative AI model. Specifically, it compares the generated information with a fact-checking database and, if necessary, uses an external fact-checking API to verify the facts. This ensures the accuracy of the output.

[0700] Next, the legal check means analyzes the generated information and evaluates its legal compliance. This means checks whether the generated information contains any legal issues such as copyright infringement, violation of privacy rights, or defamation. This ensures that the generated information is legally free.

[0701] The answer checker evaluates whether the generated information is the best answer. Based on the user's expertise and additional research, it verifies whether the generated answer is the best answer for the given situation. This ensures that the generated content is appropriate.

[0702] Furthermore, in this embodiment, an emotion engine is used as an emotion analysis means. The emotion engine analyzes the user's emotions in real time and provides information and feedback based on the analysis. It evaluates how the user feels about the generated information and adjusts the content and presentation method of the information accordingly.

[0703] The integrated evaluation method combines the results of each evaluation to generate a final evaluation and commentary. It integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to comprehensively judge the reliability and appropriateness of the generated product.

[0704] The certification method will certify the quality of products that receive high evaluations. Products that receive quality certification will be given an "AI Quality Certificate," which can be provided to external parties as evidence of the reliability of the information.

[0705] Finally, the retraining means retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0706] Specific examples

[0707] For example, if a user receives an alert that reads "Critical security breach detected" and feels "anxious," the emotion engine analyzes that emotion in real time. Based on the analysis results, the alert content is provided with information that is adjusted to prevent excessive tension. This emotion analysis allows the user to receive appropriate feedback, enabling them to respond more calmly.

[0708] Prompt Sentence Examples

[0709] An example of a prompt to be input to the generative AI model is as follows:

[0710] "Ask the Generative AI to create an incident report. Please create an incident report with the following information:

[0711] 1. Details of the incident and its scope.

[0712] 2. Include fact-checked data.

[0713] 3. Use legally acceptable language.

[0714] 4. Customize alert content based on user emotions.

[0715] As described above, the present invention is a system that can provide more realistic and reliable information by combining the user's emotions with the evaluation of generated information.

[0716] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0717] Step 1:

[0718] Receiving generated information

[0719] The server receives information generated by the generative AI model. The input is the generated information (e.g., incident report), and the output is the content of the information itself. This information is passed to various subsequent check modules.

[0720] Step 2:

[0721] Fact Check

[0722] The server passes the received information to the fact-checking means to check its accuracy. Specifically, it compares the input information with a fact-checking database and uses an external fact-checking API to check the facts. The output is a result of whether the information is true or not.

[0723] Step 3:

[0724] Legal Check

[0725] Based on the fact-checking results, the server passes the information to the legal checker. The input is the fact-checked information, which is analyzed to evaluate its legal compliance. Specifically, it checks for legal issues such as copyright infringement, violation of privacy rights, and defamation, and outputs a result indicating whether or not there is a legal problem.

[0726] Step 4:

[0727] Answer Check

[0728] The server passes the information to the answer checker based on the results of the legal check. The input is the legally checked information, and the server evaluates whether this information is the best answer. The output is an evaluation of the information's suitability based on the user's expertise and additional research.

[0729] Step 5:

[0730] Emotion analysis

[0731] The server passes the information to the emotion analysis means based on the answer check results. The input is the answer-checked information and the user's emotion data, which the emotion engine analyzes to understand the user's emotional state. The output is an analysis result based on the user's emotion.

[0732] Step 6:

[0733] Integrated Evaluation

[0734] The server integrates all the evaluation results based on the results of sentiment analysis. The inputs are the results of fact-checking, legal checks, answer checks, and sentiment analysis, and combines them to generate a final evaluation and commentary. The output is an overall evaluation result based on the information's reliability, legal relevance, appropriateness, and user sentiment.

[0735] Step 7:

[0736] Certification

[0737] The server certifies the quality of information that receives a high rating based on the results of the integrated evaluation. The input is the results of the integrated evaluation, and the output is an "AI Quality Certificate."

[0738] Step 8:

[0739] Relearn

[0740] The server retrains the generative AI model based on the evaluation results and sentiment analysis results based on the certification results. The inputs are the evaluation results and sentiment analysis results, and the generative AI model is updated based on these, improving the model's evaluation accuracy as output.

[0741] The above are the specific processing steps of the program for the system that realizes the application example.

[0742] 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.

[0743] 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.

[0744] 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.

[0745] [Third embodiment]

[0746] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0747] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0748] 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).

[0749] 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.

[0750] 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.

[0751] 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).

[0752] 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.

[0753] 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.

[0754] 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.

[0755] 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.

[0756] 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.

[0757] 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."

[0758] The present invention is a system for evaluating the accuracy, legality, and appropriateness of generated information. The system is composed of the following main modules: a fact-checking module, a legal check module, an answer-checking module, an integrated evaluation module, a certification module, and a re-learning module.

[0759] Fact Checking Module

[0760] server:

[0761] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[0762] Legal Check Module

[0763] Device:

[0764] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[0765] Answer Check Module

[0766] User:

[0767] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[0768] Integrated Evaluation Module

[0769] server:

[0770] The evaluation results of each module are combined to generate a final evaluation and commentary. For example, the results of fact-checking, legal checks, and answer checks are combined to ensure that the generated information is factual, legally compliant, and the most appropriate answer.

[0771] Certification Module

[0772] server:

[0773] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[0774] Retraining Module

[0775] server:

[0776] The evaluation results are retrained into the generative AI model, thereby improving the future evaluation accuracy of the generative AI model.

[0777] This system comprehensively evaluates the output of the generative AI model, ensuring that the generated result is accurate, legally sound, and the most appropriate answer. The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[0778] The processing flow will be explained below.

[0779] Step 1:

[0780] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[0781] Step 2:

[0782] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[0783] Step 3:

[0784] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[0785] Step 4:

[0786] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[0787] Step 5:

[0788] The server sends the fact-checked product to the legal check module, which analyzes the product.

[0789] Step 6:

[0790] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0791] Step 7:

[0792] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[0793] Step 8:

[0794] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[0795] Step 9:

[0796] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[0797] Step 10:

[0798] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[0799] Step 11:

[0800] The server combines the results of the fact checks, legal checks, and answer checks to generate a final rating and commentary.

[0801] Step 12:

[0802] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[0803] Step 13:

[0804] The server retrains the generative AI model with the evaluation results, thereby improving the future evaluation accuracy of the generative AI model.

[0805] This detailed processing flow creates a system that ensures that the output of the generative AI model is accurate, legally sound, and the optimal answer.

[0806] Example 1

[0807] 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."

[0808] There is a need to comprehensively evaluate the accuracy, legal compliance, and optimality of products generated by generative AI models and ensure their quality in order to improve the reliability of the products. However, in conventional systems, these evaluations are often performed individually, making comprehensive evaluation difficult and resulting in a lack of reliability in the quality of the products. In addition, there is a lack of an interface that allows users to input prompts into the generative AI model and directly check the evaluation results.

[0809] 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.

[0810] In this invention, the server includes a fact-checking means for checking the accuracy of the generated information, a legal evaluation means for evaluating the legal compliance of the generated information, a response evaluation means for evaluating whether the generated information is the optimal answer, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an authentication means for certifying the quality of highly rated products, a re-learning means for relearning the evaluation results to the generative model, a user interface means for inputting prompt sentences to the generative AI model, and an evaluation result display means for displaying various evaluation results for the product. This makes it possible to comprehensively evaluate the accuracy, legal compliance, and optimality of the product and provide a highly reliable product.

[0811] A "generative AI model" is an artificial intelligence algorithm for generating text in natural language based on a prompt entered by a user.

[0812] A "fact-checking tool" is a device or program that cross-checks the generated information with a fact-checking database or external fact-checking API to verify its accuracy.

[0813] A "legal check means" is a device or program that analyzes information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation in order to assess whether the generated information is legally problematic.

[0814] An "answer checker" is a device or program that evaluates content based on specialized knowledge or additional research to assess whether the information generated is the best answer.

[0815] The "integrated evaluation means" is a device or program that synthesizes the evaluation results of the fact-checking means, legal check means, and answer check means to generate a final evaluation and commentary of the product.

[0816] A "certification means" is a device or program that certifies the quality of products that receive high ratings and issues an "AI quality certification certificate."

[0817] A "relearning means" is a device or program that retrains the generative AI model based on the evaluation results, thereby improving future evaluation accuracy.

[0818] The "user interface means" is an interface through which the user inputs a prompt statement to the generated AI model and the system starts the analysis process.

[0819] The "evaluation result display means" is a device or program for displaying various evaluation results for the product to the user.

[0820] This invention is an information generation system that utilizes a generative AI model, and is a system that evaluates the accuracy, legal compliance, and optimality of the generated information and guarantees its quality. Below, we will explain how to specifically implement this system.

[0821] Fact Checking Module

[0822] server:

[0823] The server receives the product of the generative AI model and performs fact-checking using the fact-checking module. Specifically, it compares the generated information with a fact-checking database and verifies the accuracy of the information using an external fact-checking API. For example, if the product generated is "The 2022 Nobel Peace Prize winner is Mr. A," the server queries the official Nobel Peace Prize database for that information.

[0824] Legal Check Module

[0825] Device:

[0826] The terminal analyzes the generated information sent from the server and performs a legal evaluation using the legal check module. Specifically, the generated information is analyzed and checked for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, if a generated information is "The CEO of Company X was arrested for illegal activity," the terminal checks the content against the law to see if there are any problems.

[0827] Answer Check Module

[0828] User:

[0829] The user receives the product sent from the server and evaluates it in the answer check module. Specifically, the user evaluates whether the generated information is the best answer based on their expert knowledge and additional research. For example, the user evaluates the product "The appropriate daily caffeine intake is XX mg" and provides feedback on the results.

[0830] Integrated Evaluation Module

[0831] server:

[0832] The server combines the evaluation results of the fact-checking module, legal check module, and answer check module to generate a final evaluation in the integrated evaluation module. Specifically, it aggregates the evaluation results of each module and checks whether the generated information is accurate, legally compliant, and the most appropriate answer.

[0833] Certification Module

[0834] server:

[0835] The server certifies the products that receive high ratings. Specifically, it issues an "AI Quality Certificate" based on the overall evaluation results, guaranteeing the reliability of the generated information.

[0836] Retraining Module

[0837] server:

[0838] The server updates the generative AI model using a re-learning module based on all evaluation results. Specifically, the server re-learns the generative AI model based on the evaluation results to improve the evaluation accuracy of future products.

[0839] User Interface Means

[0840] User:

[0841] The user inputs a prompt sentence to the generative AI model. Specifically, the system starts the analysis process through an interface for inputting the prompt sentence. The input prompt sentence is sent to the server.

[0842] Evaluation result display means

[0843] server:

[0844] The server displays various evaluation results for the product, including fact checks, legal checks, answer checks, and integrated evaluations, to allow users to visually check the quality of the product.

[0845] Examples of prompt statements

[0846] "Who will win the Nobel Peace Prize in 2022?"

[0847] "Please fact-check the reports that the CEO of Company X was arrested for illegal activity."

[0848] "What is an appropriate amount of caffeine to consume per day?"

[0849] This makes it possible to comprehensively evaluate the reliability of the products generated by the generative AI model and provide high-quality information.

[0850] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0851] Step 1:

[0852] User:

[0853] The user inputs a prompt to the generative AI model. The input prompt triggers the system to start the analysis process. For example, the prompt might be, "Please tell me who the Nobel Peace Prize winners are in 2022." The input data is the prompt, which becomes the input to the next step.

[0854] Step 2:

[0855] server:

[0856] The server inputs the prompt received from the user into a generative AI model. The generative AI model (for example, a natural language processing algorithm) generates text based on the prompt. The generated text is returned to the server as a product. For example, the product output is "The 2022 Nobel Peace Prize winner is Mr. A." Here, the input is the prompt, and the output is the product.

[0857] Step 3:

[0858] server:

[0859] The server receives the product and calls the fact-checking module. The fact-checking module analyzes the product and compares it with a fact-checking database. If necessary, it uses an external fact-checking API to verify the accuracy of the information. For example, it checks whether "Mr. A" is a 2022 Nobel Peace Prize winner. The input is the product, and the output is the result of the fact-checking.

[0860] Step 4:

[0861] Device:

[0862] The terminal receives the product sent from the server and calls the legal check module. The legal check module analyzes the product and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, the legal check module checks whether the product "The CEO of Company X was arrested for illegal activity" is legally compliant. The input is the product, and the output is the result of the legal evaluation.

[0863] Step 5:

[0864] User:

[0865] The user receives the product sent from the server and evaluates it using the answer check module. Based on their expertise and additional research, the user checks whether the product is the optimal answer and provides feedback. For example, the user evaluates whether the product "The appropriate daily caffeine intake is XX mg" is appropriate. The input is the product, and the output is the result of the optimality evaluation.

[0866] Step 6:

[0867] server:

[0868] The server aggregates the evaluation results of each module and calls the integrated evaluation module. The integrated evaluation module combines the results of fact checks, legal checks, and answer checks to generate a final evaluation. For example, it outputs an overall evaluation such as "The information is factual, legally compliant, and the best answer." The input is the evaluation results, and the output is the final evaluation.

[0869] Step 7:

[0870] server:

[0871] The server calls the certification module for the products that receive high evaluations. The certification module issues an "AI quality certificate" based on the overall evaluation results, guaranteeing the reliability of the products. The input is the final evaluation, and the output is the certificate.

[0872] Step 8:

[0873] server:

[0874] The server calls the re-learning module based on all the evaluation results. The re-learning module retrains the generative AI model with the evaluation results to improve the evaluation accuracy of future products. The input is the evaluation results, and the output is the updated generative AI model.

[0875] (Application example 1)

[0876] 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."

[0877] Conventional content distribution systems lack effective means for evaluating the accuracy, legal compliance, and appropriateness of the generated information, resulting in a high risk of distributing inaccurate or legally questionable information. Furthermore, the lack of comprehensive evaluation and quality certification to ensure reliable content often leaves users unsure about the reliability of the information provided. As a result, the reliability of content distribution services declines, leading to problems of reduced user satisfaction.

[0878] 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.

[0879] In this invention, the server includes a fact-checking means, a legal checking means, an answer checking means, an integrated evaluation means that integrates the evaluation results to generate an overall evaluation and commentary, a certification means that certifies the quality of highly rated products, a re-learning means that re-learns the evaluation results into a generative model, a distribution means that evaluates the accuracy, legal compliance, and optimality of the information to be distributed and distributes the quality-certified information, and an information assurance means that guarantees the reliability of the distributed information based on the evaluation results. This makes it possible to evaluate the accuracy, legal compliance, and optimality of the content to be distributed in advance, and to provide high-quality information to users.

[0880] "Generated information" refers to text data, image data, audio data, etc. generated by a natural language processing model or other information generation means.

[0881] "Fact-checking methods" are methods of fact-checking using fact-checking databases or external fact-checking APIs to verify the accuracy of the generated information.

[0882] "Legal check means" is a means for evaluating the legal compliance of generated information (copyright infringement, violation of privacy rights, defamation, etc.).

[0883] The "answer check means" is a means for evaluating whether the generated information is the optimal answer.

[0884] An "integrated evaluation method" is a method that integrates the evaluation results of fact-checking methods, legal check methods, and answer check methods to generate a comprehensive evaluation and commentary.

[0885] "Certification means" refers to a means for certifying the quality of highly rated products.

[0886] The "relearning means" is a means for improving the evaluation accuracy of a generative model by having the generative model relearn the evaluation results.

[0887] The "distribution means" is a means for distributing information whose quality has been certified based on the evaluation results to users.

[0888] "Information assurance means" is a means for assuring the reliability of distributed information based on the evaluation results.

[0889] This invention is a system that evaluates the accuracy, legal compliance, and optimality of generated information, distributes quality-certified information, and guarantees its reliability.

[0890] System configuration

[0891] The server includes the following means:

[0892] Fact-checking tools

[0893] Legal check methods

[0894] Answer check method

[0895] Integrated Assessment Instrument

[0896] Certification means

[0897] Re-learning methods

[0898] Delivery Method

[0899] Information Assurance Measures

[0900] System Operation

[0901] 1. Fact-checking methods:

[0902] The server compares the information generated by the generative AI model with a fact-checking database. It uses an external fact-checking API to verify the facts. For example, if the generative AI outputs "The winner of the 2022 award is ____," it compares that information with the database and uses an external API to verify the facts.

[0903] 2. Legal check methods:

[0904] The server analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, defamation, etc. For example, if the generating AI outputs "The CEO of a certain company was arrested for legal issues," the content is analyzed to check for legal issues.

[0905] 3. Answer check method:

[0906] The user receives the output sent by the server, evaluates its suitability based on their expertise and additional research, and provides feedback. For example, they evaluate the output, saying, "The appropriate daily caffeine intake is XX mg."

[0907] 4. Integrated assessment instruments:

[0908] The server combines the results of fact-checking, legal checking, and answer checking to generate an overall rating and commentary, for example, combining the rating results from each module to ensure the generated information is accurate, legally compliant, and the best answer.

[0909] 5. Certification means:

[0910] The server certifies the quality of the products that receive high evaluations, issuing a "Generation AI Quality Certificate" and providing evidence of the reliability of the products.

[0911] 6. Re-learning methods:

[0912] The server retrains the generative AI model with the evaluation results, thereby improving the evaluation accuracy of the generative AI model.

[0913] 7. Delivery Method:

[0914] The server delivers high-quality information that has been evaluated to users, such as articles and news that have been evaluated.

[0915] 8. Information Assurance Measures:

[0916] The server guarantees the authenticity of the information delivered, allowing users to feel confident about the accuracy and reliability of the information provided.

[0917] Specific examples

[0918] For example, you might input the following prompt into a generative AI model:

[0919] "Please tell me more about the 2022 Nobel Peace Prize winners."

[0920] The generative AI model generates information such as "The 2022 Nobel Peace Prize winner is XX, and his / her achievements are △△." This information is evaluated through fact-checking means, legal checking means, and answer checking means, and is then comprehensively evaluated by comprehensive evaluation means. If it is subsequently evaluated highly, quality is certified by certification means, and the information is distributed to users via distribution means. The reliability of the distributed information is guaranteed by information assurance means.

[0921] As a result, the reliability and quality of the information being distributed are ensured, and content distribution that users can use with peace of mind is realized.

[0922] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0923] Step 1:

[0924] The server inputs a prompt into the generative AI model and receives the generated information. Text data is generated based on this prompt, and the generated information is then evaluated. For example, the server inputs a prompt such as, "Please explain in detail the winners of the 2022 Nobel Peace Prize."

[0925] Step 2:

[0926] The server passes the generated information to the fact-checking means, which checks it against a fact-checking database. It uses an external fact-checking API to check the facts and retrieves the results from the database. For example, it checks whether the generated information, "The 2022 Nobel Peace Prize winner is ____," matches the actual name of the winner.

[0927] Step 3:

[0928] The server receives the fact-check results and then passes the generated information to the legal check means. The legal check means analyzes the generated information and detects legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, it checks whether the generated information "The CEO of a company was arrested for legal issues" constitutes defamation.

[0929] Step 4:

[0930] The server receives the results of the legal check and passes the generated information to the answer check means. The user receives the generated information sent from the server, evaluates its appropriateness based on their expertise and additional research, and provides feedback. For example, they evaluate whether the generated information, "The appropriate daily caffeine intake is XX mg," is medically appropriate.

[0931] Step 5:

[0932] The server combines the results of the fact checks, legal checks, and answer checks to generate an overall rating and commentary. The results of each check are averaged to calculate an overall rating score. For example, if the checks each score 95, 100, and 90, the average of each score would be 95.

[0933] Step 6:

[0934] The server certifies the quality of the products that receive high evaluations. It issues a "Generation AI Quality Certificate" through the certification method, providing evidence of the reliability of the product. For example, a certificate is issued if the overall evaluation score is 90 points or higher.

[0935] Step 7:

[0936] The server distributes certified, high-quality information to users through distribution methods. Articles and news that have been rated and certified are delivered to users. For example, rated news articles are distributed on a news app.

[0937] Step 8:

[0938] The server guarantees the reliability of the delivered information using information assurance measures. Users can use information with guaranteed reliability with peace of mind. The reliability is visually indicated by displaying an information authentication badge, etc.

[0939] 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.

[0940] The present invention combines a system that evaluates the accuracy, legality, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: a fact-checking module, a legal check module, an answer check module, an integrated evaluation module, a certification module, a re-learning module, and an emotion engine.

[0941] Fact Checking Module

[0942] server:

[0943] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[0944] Legal Check Module

[0945] Device:

[0946] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[0947] Answer Check Module

[0948] User:

[0949] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[0950] Emotion Engine

[0951] server:

[0952] An emotion engine is used to recognize the user's emotions, which analyzes how the user feels about the product.

[0953] Examples:

[0954] When the generation AI outputs an "opinion about a particular policy," the emotion engine analyzes whether the user has any dissatisfaction or doubts about the output and sends the results to the server.

[0955] Integrated Evaluation Module

[0956] server:

[0957] The evaluation results of each module are combined to generate a final rating and commentary. Fact checks, legal checks, answer checks, and sentiment data from the sentiment engine are combined to ensure the generated information is factual, legally compliant, and the most appropriate answer.

[0958] Certification Module

[0959] server:

[0960] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[0961] Retraining Module

[0962] server:

[0963] The generative AI model is retrained using the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0964] This system not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[0965] The processing flow will be explained below.

[0966] Step 1:

[0967] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[0968] Step 2:

[0969] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[0970] Step 3:

[0971] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[0972] Step 4:

[0973] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[0974] Step 5:

[0975] The server sends the fact-checked product to the legal check module, which analyzes the product.

[0976] Step 6:

[0977] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[0978] Step 7:

[0979] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[0980] Step 8:

[0981] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[0982] Step 9:

[0983] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[0984] Step 10:

[0985] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[0986] Step 11:

[0987] The server acquires the user's emotion data using an emotion engine to recognize the user's emotion, analyzes this data, and evaluates the emotion of the product.

[0988] Step 12:

[0989] The server combines the results from the fact check, legal check, answer check, and sentiment engine to generate a final rating and commentary. The server ensures that the generated rating is factual, legally compliant, and the best answer.

[0990] Step 13:

[0991] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[0992] Step 14:

[0993] The server retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[0994] This detailed processing flow not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.

[0995] Example 2

[0996] 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."

[0997] It is difficult to evaluate whether the information generated by generative AI models is factually accurate, legally compliant, or the most appropriate answer. It is also important to consider the user's emotions when making an evaluation, and a system that can achieve this in an integrated manner is needed.

[0998] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a fact confirmation means for checking the accuracy of the generated information, a legal review means for evaluating the legal compliance of the generated information, an answer evaluation means for evaluating whether the generated information is an optimal answer, a sentiment analysis means for recognizing and evaluating the user's sentiment, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an accreditation means for certifying the quality of highly rated products, and a re-learning means for relearning the evaluation results to the generative model. This enables the generated information to be based on facts, legally compliant, and an optimal answer, while also being evaluated taking the user's sentiment into consideration.

[0999] "Fact-checking means" refers to a means of verifying the accuracy of generated information using a database or external API.

[1000] "Legal review measures" are measures to evaluate whether the generated information is legally compliant and check for legal issues such as infringement of intellectual property rights, infringement of the right to protect personal information, and defamation.

[1001] The "answer evaluation means" is a means for evaluating whether the generated information is the optimal answer.

[1002] The "emotion analysis means" is a means for recognizing the user's emotions and reflecting them in the evaluation of the generated information.

[1003] An "integrated evaluation means" is a means for integrating the results from each evaluation means to generate an overall evaluation and commentary.

[1004] "Certification means" is a means of certifying the quality of products that have received high praise.

[1005] The "relearning means" is a means for feeding back the evaluation results to the generative model to perform re-learning, thereby improving future evaluation accuracy.

[1006] A "generative model" is an artificial intelligence model that generates information based on prompts from users.

[1007] The present invention combines an emotion engine that recognizes user emotions in a system that evaluates the accuracy, legal compliance, and appropriateness of generated information. The system includes a fact-checking means, a legal check means, an answer check means, a sentiment analysis means, an integrated evaluation means, a certification means, and a re-learning means.

[1008] Hardware and software used

[1009] Hardware

[1010] Server: central system for data processing and evaluation

[1011] Terminal: User input and display

[1012] User Device: Input Device for Emotion Analysis

[1013] software

[1014] Generative AI models (e.g., open-source generative models)

[1015] Fact-checking databases and external fact-checking APIs

[1016] Legal databases and legal document analysis tools

[1017] Sentiment Analysis Engine

[1018] Program processing

[1019] server

[1020] When a user sends a prompt, the server uses a generative AI model to generate output. For example, if the prompt is "When AI describes the future of driving, please generate output taking into consideration technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," the generative AI generates the corresponding text. The generated information is then fact-checked using a fact-checking method, which checks it against a fact-checking database and external fact-checking APIs (e.g., FactCheck.org and Snopes).

[1021] Terminal

[1022] The device receives the generated text from the server and analyzes the generated information using legal checks. For example, if the generated text contains a statement such as "The CEO of Company X was arrested for illegal conduct," the device analyzes the content and checks for legal issues such as intellectual property infringement, privacy infringement, and defamation. This check involves referencing legal databases (e.g., LexisNexis and Westlaw).

[1023] User

[1024] After receiving a product that has passed the legal check, the user evaluates whether the content is the optimal answer using the answer check tool. For example, for a product that says "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their expertise and additional research and provides feedback.

[1025] server

[1026] The emotion analysis means analyzes the user's feedback and emotions toward the product. For example, if the user feels that the information about the product is insufficient, the emotion engine recognizes that emotion from the user's text, facial expressions, and voice, and sends the results to the server. Next, the evaluation results of each module are integrated by the integrated evaluation means, and products that receive high ratings are issued a "quality certificate" by the certification means. The evaluation results and emotion analysis results are then fed back into the generative model, and the model is retrained by the retraining means.

[1027] Specific examples

[1028] As a concrete example, consider a generative AI model that generates output on the topic "The Future of AI-Driven Driving." The prompt in this case would be:

[1029] "When describing the future of AI driving, please generate output taking into account the technological trends announced in 2022, the associated legal issues, and how the general public will feel about them."

[1030] Based on this, a system is provided in which each means works together, the generated information is factual, legally compliant, provides the optimal answer, and is evaluated taking into account the user's emotions.

[1031] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1032] Step 1:

[1033] The user inputs a prompt, for example, "When AI describes the future of car driving, please generate an output taking into consideration the technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," and sends it to the server. Based on this input, the generative AI model generates an appropriate output.

[1034] Step 2:

[1035] The server uses a generative AI model to generate output based on the prompt. The generated text is analyzed and compared with a fact-checking database and an external fact-checking API. For example, if a product is generated such as "The 2022 Nobel Peace Prize winner is so-and-so," the database and API are used to fact-check the output. The output is the fact-checked text.

[1036] Step 3:

[1037] The server sends the fact-checked product to the terminal. The terminal then performs a legal review of the received product. Specifically, it analyzes the text in the product to determine whether it contains legal issues such as intellectual property infringement, personal information protection infringement, or defamation. For example, it analyzes a statement such as "The CEO of Company X was arrested for illegal activity" and checks whether there are any problems. The output is a product that is legally sound.

[1038] Step 4:

[1039] The terminal sends the legally reviewed product to the user, who evaluates whether the received product is the optimal answer. For example, for a product such as "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their own expertise and additional research and provides feedback. The output is the user's feedback and the evaluation result.

[1040] Step 5:

[1041] The server analyzes the feedback and evaluation results received from the user using emotion analysis means. It recognizes emotions from the user's text, facial expressions, and voice, and sends the evaluation results to the server. For example, if the user feels that "this information is insufficient," the emotion data is analyzed and sent to the server. The output is the result of the user's emotion analysis.

[1042] Step 6:

[1043] The server integrates the evaluation results obtained from the fact-checking means, legal checking means, answer checking means, and sentiment analysis means in an integrated evaluation means. For example, the server aggregates the evaluation results of each module to generate an overall score, and then generates a final evaluation and commentary. The output is the overall evaluation result and commentary.

[1044] Step 7:

[1045] If the overall evaluation result is high, the server issues a "quality certificate" using the certification means. This provides evidence of the reliability of the product. The quality certificate is assigned to the product. The output is the product with the issued quality certificate.

[1046] Step 8:

[1047] The server feeds back the overall evaluation results and sentiment analysis results to the generative model using a retraining method. This improves the future evaluation accuracy of the generative AI model. For example, the evaluation results are added to the dataset and the model is retrained. The output is a retrained generative AI model.

[1048] (Application example 2)

[1049] 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."

[1050] While existing technologies exist for individually evaluating the accuracy, legal compliance, and optimality of products generated by generative AI models, these evaluations do not take user emotions into account, making it difficult to improve the user experience or provide appropriate feedback based on emotions.The present invention aims to combine user emotions with the evaluation of generative AI models to produce products that are more realistic and reliable, and to provide optimized results that correspond to the user's emotions.

[1051] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: a fact-checking means for checking the accuracy of the generated information; a legal-checking means for evaluating the legal compliance of the generated information; an answer-checking means for evaluating whether the generated information is the optimal answer; a sentiment-analysis means for making an evaluation based on the user's emotions using an emotion engine that analyzes the user's emotions; an integrated-evaluation means for integrating the evaluation results to generate an overall evaluation and commentary; an accreditation means for certifying the quality of highly rated products; and a re-learning means for relearning the evaluation results to the generative model. This makes it possible for the generated information and warnings to be provided in a form that is accurate, legally sound, and optimized according to the user's emotions.

[1052] A "fact-checking tool" is a module that checks the accuracy of the generated information against a fact-checking database and uses an external fact-checking API to verify the facts.

[1053] The "legal check means" is a module that analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1054] The "answer check means" is a module for evaluating whether the generated information is the optimal answer.

[1055] The "emotion analysis means" is an emotion engine that analyzes the user's emotions and reevaluates the generated information based on those emotions.

[1056] The "integrated evaluation means" is a module that integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to generate a comprehensive evaluation and commentary.

[1057] The "certification means" is a module that certifies the quality of products that have received high evaluations.

[1058] The "relearning means" is a module that re-learns the evaluation results into the generative model, thereby improving the evaluation accuracy of the model.

[1059] This invention combines a system that evaluates the accuracy, legal compliance, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: fact-checking means, legal checking means, answer checking means, emotion analysis means, integrated evaluation means, certification means, and relearning means.

[1060] In this embodiment, the server performs the following process.

[1061] First, the fact-checking method verifies the accuracy of the information generated by the generative AI model. Specifically, it compares the generated information with a fact-checking database and, if necessary, uses an external fact-checking API to verify the facts. This ensures the accuracy of the output.

[1062] Next, the legal check means analyzes the generated information and evaluates its legal compliance. This means checks whether the generated information contains any legal issues such as copyright infringement, violation of privacy rights, or defamation. This ensures that the generated information is legally free.

[1063] The answer checker evaluates whether the generated information is the best answer. Based on the user's expertise and additional research, it verifies whether the generated answer is the best answer for the given situation. This ensures that the generated content is appropriate.

[1064] Furthermore, in this embodiment, an emotion engine is used as an emotion analysis means. The emotion engine analyzes the user's emotions in real time and provides information and feedback based on the analysis. It evaluates how the user feels about the generated information and adjusts the content and presentation method of the information accordingly.

[1065] The integrated evaluation method combines the results of each evaluation to generate a final evaluation and commentary. It integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to comprehensively judge the reliability and appropriateness of the generated product.

[1066] The certification method will certify the quality of products that receive high evaluations. Products that receive quality certification will be given an "AI Quality Certificate," which can be provided to external parties as evidence of the reliability of the information.

[1067] Finally, the retraining means retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[1068] Specific examples

[1069] For example, if a user receives an alert that reads "Critical security breach detected" and feels "anxious," the emotion engine analyzes that emotion in real time. Based on the analysis results, the alert content is provided with information that is adjusted to prevent excessive tension. This emotion analysis allows the user to receive appropriate feedback, enabling them to respond more calmly.

[1070] Prompt Sentence Examples

[1071] An example of a prompt to be input to the generative AI model is as follows:

[1072] "Ask the Generative AI to create an incident report. Please create an incident report with the following information:

[1073] 1. Details of the incident and its scope.

[1074] 2. Include fact-checked data.

[1075] 3. Use legally acceptable language.

[1076] 4. Customize alert content based on user emotions.

[1077] As described above, the present invention is a system that can provide more realistic and reliable information by combining the user's emotions with the evaluation of generated information.

[1078] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1079] Step 1:

[1080] Receiving generated information

[1081] The server receives information generated by the generative AI model. The input is the generated information (e.g., incident report), and the output is the content of the information itself. This information is passed to various subsequent check modules.

[1082] Step 2:

[1083] Fact Check

[1084] The server passes the received information to the fact-checking means to check its accuracy. Specifically, it compares the input information with a fact-checking database and uses an external fact-checking API to check the facts. The output is a result of whether the information is true or not.

[1085] Step 3:

[1086] Legal Check

[1087] Based on the fact-checking results, the server passes the information to the legal checker. The input is the fact-checked information, which is analyzed to evaluate its legal compliance. Specifically, it checks for legal issues such as copyright infringement, violation of privacy rights, and defamation, and outputs a result indicating whether or not there is a legal problem.

[1088] Step 4:

[1089] Answer Check

[1090] The server passes the information to the answer checker based on the results of the legal check. The input is the legally checked information, and the server evaluates whether this information is the best answer. The output is an evaluation of the information's suitability based on the user's expertise and additional research.

[1091] Step 5:

[1092] Emotion analysis

[1093] The server passes the information to the emotion analysis means based on the answer check results. The input is the answer-checked information and the user's emotion data, which the emotion engine analyzes to understand the user's emotional state. The output is an analysis result based on the user's emotion.

[1094] Step 6:

[1095] Integrated Evaluation

[1096] The server integrates all the evaluation results based on the results of sentiment analysis. The inputs are the results of fact-checking, legal checks, answer checks, and sentiment analysis, and combines them to generate a final evaluation and commentary. The output is an overall evaluation result based on the information's reliability, legal relevance, appropriateness, and user sentiment.

[1097] Step 7:

[1098] Certification

[1099] The server certifies the quality of information that receives a high rating based on the results of the integrated evaluation. The input is the results of the integrated evaluation, and the output is an "AI Quality Certificate."

[1100] Step 8:

[1101] Relearn

[1102] The server retrains the generative AI model based on the evaluation results and sentiment analysis results based on the certification results. The inputs are the evaluation results and sentiment analysis results, and the generative AI model is updated based on these, improving the model's evaluation accuracy as output.

[1103] The above are the specific processing steps of the program for the system that realizes the application example.

[1104] 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.

[1105] 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.

[1106] 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.

[1107] [Fourth embodiment]

[1108] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1109] 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.

[1110] 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).

[1111] 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.

[1112] 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.

[1113] 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).

[1114] 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.

[1115] 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.

[1116] 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.

[1117] 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.

[1118] 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.

[1119] 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.

[1120] 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."

[1121] The present invention is a system for evaluating the accuracy, legality, and appropriateness of generated information. The system is composed of the following main modules: a fact-checking module, a legal check module, an answer-checking module, an integrated evaluation module, a certification module, and a re-learning module.

[1122] Fact Checking Module

[1123] server:

[1124] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[1125] Legal Check Module

[1126] Device:

[1127] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[1128] Answer Check Module

[1129] User:

[1130] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[1131] Integrated Evaluation Module

[1132] server:

[1133] The evaluation results of each module are combined to generate a final evaluation and commentary. For example, the results of fact-checking, legal checks, and answer checks are combined to ensure that the generated information is factual, legally compliant, and the most appropriate answer.

[1134] Certification Module

[1135] server:

[1136] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[1137] Retraining Module

[1138] server:

[1139] The evaluation results are retrained into the generative AI model, thereby improving the future evaluation accuracy of the generative AI model.

[1140] This system comprehensively evaluates the output of the generative AI model, ensuring that the generated result is accurate, legally sound, and the most appropriate answer. The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[1141] The processing flow will be explained below.

[1142] Step 1:

[1143] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[1144] Step 2:

[1145] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[1146] Step 3:

[1147] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[1148] Step 4:

[1149] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[1150] Step 5:

[1151] The server sends the fact-checked product to the legal check module, which analyzes the product.

[1152] Step 6:

[1153] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1154] Step 7:

[1155] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[1156] Step 8:

[1157] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[1158] Step 9:

[1159] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[1160] Step 10:

[1161] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[1162] Step 11:

[1163] The server combines the results of the fact checks, legal checks, and answer checks to generate a final rating and commentary.

[1164] Step 12:

[1165] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[1166] Step 13:

[1167] The server retrains the generative AI model with the evaluation results, thereby improving the future evaluation accuracy of the generative AI model.

[1168] This detailed processing flow creates a system that ensures that the output of the generative AI model is accurate, legally sound, and the optimal answer.

[1169] Example 1

[1170] 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."

[1171] There is a need to comprehensively evaluate the accuracy, legal compliance, and optimality of products generated by generative AI models and ensure their quality in order to improve the reliability of the products. However, in conventional systems, these evaluations are often performed individually, making comprehensive evaluation difficult and resulting in a lack of reliability in the quality of the products. In addition, there is a lack of an interface that allows users to input prompts into the generative AI model and directly check the evaluation results.

[1172] 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.

[1173] In this invention, the server includes a fact-checking means for checking the accuracy of the generated information, a legal evaluation means for evaluating the legal compliance of the generated information, a response evaluation means for evaluating whether the generated information is the optimal answer, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an authentication means for certifying the quality of highly rated products, a re-learning means for relearning the evaluation results to the generative model, a user interface means for inputting prompt sentences to the generative AI model, and an evaluation result display means for displaying various evaluation results for the product. This makes it possible to comprehensively evaluate the accuracy, legal compliance, and optimality of the product and provide a highly reliable product.

[1174] A "generative AI model" is an artificial intelligence algorithm for generating text in natural language based on a prompt entered by a user.

[1175] A "fact-checking tool" is a device or program that cross-checks the generated information with a fact-checking database or external fact-checking API to verify its accuracy.

[1176] A "legal check means" is a device or program that analyzes information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation in order to assess whether the generated information is legally problematic.

[1177] An "answer checker" is a device or program that evaluates content based on specialized knowledge or additional research to assess whether the information generated is the best answer.

[1178] The "integrated evaluation means" is a device or program that synthesizes the evaluation results of the fact-checking means, legal check means, and answer check means to generate a final evaluation and commentary of the product.

[1179] A "certification means" is a device or program that certifies the quality of products that receive high ratings and issues an "AI quality certification certificate."

[1180] A "relearning means" is a device or program that retrains the generative AI model based on the evaluation results, thereby improving future evaluation accuracy.

[1181] The "user interface means" is an interface through which the user inputs a prompt statement to the generated AI model and the system starts the analysis process.

[1182] The "evaluation result display means" is a device or program for displaying various evaluation results for the product to the user.

[1183] This invention is an information generation system that utilizes a generative AI model, and is a system that evaluates the accuracy, legal compliance, and optimality of the generated information and guarantees its quality. Below, we will explain how to specifically implement this system.

[1184] Fact Checking Module

[1185] server:

[1186] The server receives the product of the generative AI model and performs fact-checking using the fact-checking module. Specifically, it compares the generated information with a fact-checking database and verifies the accuracy of the information using an external fact-checking API. For example, if the product generated is "The 2022 Nobel Peace Prize winner is Mr. A," the server queries the official Nobel Peace Prize database for that information.

[1187] Legal Check Module

[1188] Device:

[1189] The terminal analyzes the generated information sent from the server and performs a legal evaluation using the legal check module. Specifically, the generated information is analyzed and checked for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, if a generated information is "The CEO of Company X was arrested for illegal activity," the terminal checks the content against the law to see if there are any problems.

[1190] Answer Check Module

[1191] User:

[1192] The user receives the product sent from the server and evaluates it in the answer check module. Specifically, the user evaluates whether the generated information is the best answer based on their expert knowledge and additional research. For example, the user evaluates the product "The appropriate daily caffeine intake is XX mg" and provides feedback on the results.

[1193] Integrated Evaluation Module

[1194] server:

[1195] The server combines the evaluation results of the fact-checking module, legal check module, and answer check module to generate a final evaluation in the integrated evaluation module. Specifically, it aggregates the evaluation results of each module and checks whether the generated information is accurate, legally compliant, and the most appropriate answer.

[1196] Certification Module

[1197] server:

[1198] The server certifies the products that receive high ratings. Specifically, it issues an "AI Quality Certificate" based on the overall evaluation results, guaranteeing the reliability of the generated information.

[1199] Retraining Module

[1200] server:

[1201] The server updates the generative AI model using a re-learning module based on all evaluation results. Specifically, the server re-learns the generative AI model based on the evaluation results to improve the evaluation accuracy of future products.

[1202] User Interface Means

[1203] User:

[1204] The user inputs a prompt sentence to the generative AI model. Specifically, the system starts the analysis process through an interface for inputting the prompt sentence. The input prompt sentence is sent to the server.

[1205] Evaluation result display means

[1206] server:

[1207] The server displays various evaluation results for the product, including fact checks, legal checks, answer checks, and integrated evaluations, to allow users to visually check the quality of the product.

[1208] Examples of prompt statements

[1209] "Who will win the Nobel Peace Prize in 2022?"

[1210] "Please fact-check the reports that the CEO of Company X was arrested for illegal activity."

[1211] "What is an appropriate amount of caffeine to consume per day?"

[1212] This makes it possible to comprehensively evaluate the reliability of the products generated by the generative AI model and provide high-quality information.

[1213] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1214] Step 1:

[1215] User:

[1216] The user inputs a prompt to the generative AI model. The input prompt triggers the system to start the analysis process. For example, the prompt might be, "Please tell me who the Nobel Peace Prize winners are in 2022." The input data is the prompt, which becomes the input to the next step.

[1217] Step 2:

[1218] server:

[1219] The server inputs the prompt received from the user into a generative AI model. The generative AI model (for example, a natural language processing algorithm) generates text based on the prompt. The generated text is returned to the server as a product. For example, the product output is "The 2022 Nobel Peace Prize winner is Mr. A." Here, the input is the prompt, and the output is the product.

[1220] Step 3:

[1221] server:

[1222] The server receives the product and calls the fact-checking module. The fact-checking module analyzes the product and compares it with a fact-checking database. If necessary, it uses an external fact-checking API to verify the accuracy of the information. For example, it checks whether "Mr. A" is a 2022 Nobel Peace Prize winner. The input is the product, and the output is the result of the fact-checking.

[1223] Step 4:

[1224] Device:

[1225] The terminal receives the product sent from the server and calls the legal check module. The legal check module analyzes the product and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, the legal check module checks whether the product "The CEO of Company X was arrested for illegal activity" is legally compliant. The input is the product, and the output is the result of the legal evaluation.

[1226] Step 5:

[1227] User:

[1228] The user receives the product sent from the server and evaluates it using the answer check module. Based on their expertise and additional research, the user checks whether the product is the optimal answer and provides feedback. For example, the user evaluates whether the product "The appropriate daily caffeine intake is XX mg" is appropriate. The input is the product, and the output is the result of the optimality evaluation.

[1229] Step 6:

[1230] server:

[1231] The server aggregates the evaluation results of each module and calls the integrated evaluation module. The integrated evaluation module combines the results of fact checks, legal checks, and answer checks to generate a final evaluation. For example, it outputs an overall evaluation such as "The information is factual, legally compliant, and the best answer." The input is the evaluation results, and the output is the final evaluation.

[1232] Step 7:

[1233] server:

[1234] The server calls the certification module for the products that receive high evaluations. The certification module issues an "AI quality certificate" based on the overall evaluation results, guaranteeing the reliability of the products. The input is the final evaluation, and the output is the certificate.

[1235] Step 8:

[1236] server:

[1237] The server calls the re-learning module based on all the evaluation results. The re-learning module retrains the generative AI model with the evaluation results to improve the evaluation accuracy of future products. The input is the evaluation results, and the output is the updated generative AI model.

[1238] (Application example 1)

[1239] 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."

[1240] Conventional content distribution systems lack effective means for evaluating the accuracy, legal compliance, and appropriateness of the generated information, resulting in a high risk of distributing inaccurate or legally questionable information. Furthermore, the lack of comprehensive evaluation and quality certification to ensure reliable content often leaves users unsure about the reliability of the information provided. As a result, the reliability of content distribution services declines, leading to problems of reduced user satisfaction.

[1241] 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.

[1242] In this invention, the server includes a fact-checking means, a legal checking means, an answer checking means, an integrated evaluation means that integrates the evaluation results to generate an overall evaluation and commentary, a certification means that certifies the quality of highly rated products, a re-learning means that re-learns the evaluation results into a generative model, a distribution means that evaluates the accuracy, legal compliance, and optimality of the information to be distributed and distributes the quality-certified information, and an information assurance means that guarantees the reliability of the distributed information based on the evaluation results. This makes it possible to evaluate the accuracy, legal compliance, and optimality of the content to be distributed in advance, and to provide high-quality information to users.

[1243] "Generated information" refers to text data, image data, audio data, etc. generated by a natural language processing model or other information generation means.

[1244] "Fact-checking methods" are methods of fact-checking using fact-checking databases or external fact-checking APIs to verify the accuracy of the generated information.

[1245] "Legal check means" is a means for evaluating the legal compliance of generated information (copyright infringement, violation of privacy rights, defamation, etc.).

[1246] The "answer check means" is a means for evaluating whether the generated information is the optimal answer.

[1247] An "integrated evaluation method" is a method that integrates the evaluation results of fact-checking methods, legal check methods, and answer check methods to generate a comprehensive evaluation and commentary.

[1248] "Certification means" refers to a means for certifying the quality of highly rated products.

[1249] The "relearning means" is a means for improving the evaluation accuracy of a generative model by having the generative model relearn the evaluation results.

[1250] The "distribution means" is a means for distributing information whose quality has been certified based on the evaluation results to users.

[1251] "Information assurance means" is a means for assuring the reliability of distributed information based on the evaluation results.

[1252] This invention is a system that evaluates the accuracy, legal compliance, and optimality of generated information, distributes quality-certified information, and guarantees its reliability.

[1253] System configuration

[1254] The server includes the following means:

[1255] Fact-checking tools

[1256] Legal check methods

[1257] Answer check method

[1258] Integrated Assessment Instrument

[1259] Certification means

[1260] Re-learning methods

[1261] Delivery Method

[1262] Information Assurance Measures

[1263] System Operation

[1264] 1. Fact-checking methods:

[1265] The server compares the information generated by the generative AI model with a fact-checking database. It uses an external fact-checking API to verify the facts. For example, if the generative AI outputs "The winner of the 2022 award is ____," it compares that information with the database and uses an external API to verify the facts.

[1266] 2. Legal check methods:

[1267] The server analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, defamation, etc. For example, if the generating AI outputs "The CEO of a certain company was arrested for legal issues," the content is analyzed to check for legal issues.

[1268] 3. Answer check method:

[1269] The user receives the output sent by the server, evaluates its suitability based on their expertise and additional research, and provides feedback. For example, they evaluate the output, saying, "The appropriate daily caffeine intake is XX mg."

[1270] 4. Integrated assessment instruments:

[1271] The server combines the results of fact-checking, legal checking, and answer checking to generate an overall rating and commentary, for example, combining the rating results from each module to ensure the generated information is accurate, legally compliant, and the best answer.

[1272] 5. Certification means:

[1273] The server certifies the quality of the products that receive high evaluations, issuing a "Generation AI Quality Certificate" and providing evidence of the reliability of the products.

[1274] 6. Re-learning methods:

[1275] The server retrains the generative AI model with the evaluation results, thereby improving the evaluation accuracy of the generative AI model.

[1276] 7. Delivery Method:

[1277] The server delivers high-quality information that has been evaluated to users, such as articles and news that have been evaluated.

[1278] 8. Information Assurance Measures:

[1279] The server guarantees the authenticity of the information delivered, allowing users to feel confident about the accuracy and reliability of the information provided.

[1280] Specific examples

[1281] For example, you might input the following prompt into a generative AI model:

[1282] "Please tell me more about the 2022 Nobel Peace Prize winners."

[1283] The generative AI model generates information such as "The 2022 Nobel Peace Prize winner is XX, and his / her achievements are △△." This information is evaluated through fact-checking means, legal checking means, and answer checking means, and is then comprehensively evaluated by comprehensive evaluation means. If it is subsequently evaluated highly, quality is certified by certification means, and the information is distributed to users via distribution means. The reliability of the distributed information is guaranteed by information assurance means.

[1284] As a result, the reliability and quality of the information being distributed are ensured, and content distribution that users can use with peace of mind is realized.

[1285] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1286] Step 1:

[1287] The server inputs a prompt into the generative AI model and receives the generated information. Text data is generated based on this prompt, and the generated information is then evaluated. For example, the server inputs a prompt such as, "Please explain in detail the winners of the 2022 Nobel Peace Prize."

[1288] Step 2:

[1289] The server passes the generated information to the fact-checking means, which checks it against a fact-checking database. It uses an external fact-checking API to check the facts and retrieves the results from the database. For example, it checks whether the generated information, "The 2022 Nobel Peace Prize winner is ____," matches the actual name of the winner.

[1290] Step 3:

[1291] The server receives the fact-check results and then passes the generated information to the legal check means. The legal check means analyzes the generated information and detects legal issues such as copyright infringement, violation of privacy rights, and defamation. For example, it checks whether the generated information "The CEO of a company was arrested for legal issues" constitutes defamation.

[1292] Step 4:

[1293] The server receives the results of the legal check and passes the generated information to the answer check means. The user receives the generated information sent from the server, evaluates its appropriateness based on their expertise and additional research, and provides feedback. For example, they evaluate whether the generated information, "The appropriate daily caffeine intake is XX mg," is medically appropriate.

[1294] Step 5:

[1295] The server combines the results of the fact checks, legal checks, and answer checks to generate an overall rating and commentary. The results of each check are averaged to calculate an overall rating score. For example, if the checks each score 95, 100, and 90, the average of each score would be 95.

[1296] Step 6:

[1297] The server certifies the quality of the products that receive high evaluations. It issues a "Generation AI Quality Certificate" through the certification method, providing evidence of the reliability of the product. For example, a certificate is issued if the overall evaluation score is 90 points or higher.

[1298] Step 7:

[1299] The server distributes certified, high-quality information to users through distribution methods. Articles and news that have been rated and certified are delivered to users. For example, rated news articles are distributed on a news app.

[1300] Step 8:

[1301] The server guarantees the reliability of the delivered information using information assurance measures. Users can use information with guaranteed reliability with peace of mind. The reliability is visually indicated by displaying an information authentication badge, etc.

[1302] 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.

[1303] The present invention combines a system that evaluates the accuracy, legality, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: a fact-checking module, a legal check module, an answer check module, an integrated evaluation module, a certification module, a re-learning module, and an emotion engine.

[1304] Fact Checking Module

[1305] server:

[1306] The server receives the output of the generative AI model and compares it with a fact-checking database. For example, if the generative AI outputs "The 2022 Nobel Peace Prize winner is so-and-so," the server checks that information against the fact-checking database. If necessary, it uses an external fact-checking API to check the facts and compiles the results.

[1307] Legal Check Module

[1308] Device:

[1309] The generated output sent from the server is analyzed and a legal check is performed. For example, if the generating AI outputs "The CEO of Company X was arrested for illegal activity," the device analyzes the content and checks whether it contains any legal issues such as copyright infringement, violation of privacy rights, or defamation.

[1310] Answer Check Module

[1311] User:

[1312] The generated results are received from the server and evaluated to determine whether they are the optimal answer. For example, if the generated AI outputs "The appropriate daily caffeine intake is XX mg," the user evaluates the results based on their expertise and additional research and provides feedback.

[1313] Emotion Engine

[1314] server:

[1315] An emotion engine is used to recognize the user's emotions, which analyzes how the user feels about the product.

[1316] Examples:

[1317] When the generation AI outputs an "opinion about a particular policy," the emotion engine analyzes whether the user has any dissatisfaction or doubts about the output and sends the results to the server.

[1318] Integrated Evaluation Module

[1319] server:

[1320] The evaluation results of each module are combined to generate a final rating and commentary. Fact checks, legal checks, answer checks, and sentiment data from the sentiment engine are combined to ensure the generated information is factual, legally compliant, and the most appropriate answer.

[1321] Certification Module

[1322] server:

[1323] Products that receive high ratings will be issued an "AI Quality Certificate," which will provide evidence of the reliability of the product.

[1324] Retraining Module

[1325] server:

[1326] The generative AI model is retrained using the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[1327] This system not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.The specific processing steps will be explained in detail later, but here we have described the basic functions and roles of each module.

[1328] The processing flow will be explained below.

[1329] Step 1:

[1330] The server receives the output from the generative AI model and begins preparing to send this output to the next checking process.

[1331] Step 2:

[1332] The server sends the product to a fact-checking module, which checks the product's content against a fact-checking database.

[1333] Step 3:

[1334] The server uses external fact-checking APIs as needed, and sends the resulting information to the external APIs to perform additional fact-checking.

[1335] Step 4:

[1336] The server aggregates the results from the fact-check database and the fact-check API. The server prepares the fact-check results for further processing.

[1337] Step 5:

[1338] The server sends the fact-checked product to the legal check module, which analyzes the product.

[1339] Step 6:

[1340] The terminal analyzes the content of the generated data and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1341] Step 7:

[1342] The terminal sends the results of the legal check to the server, which then begins preparations for the next step.

[1343] Step 8:

[1344] The server sends the legally checked product to the answer check module, which evaluates whether the product is the best answer.

[1345] Step 9:

[1346] The user receives the product and rates it for being the best answer, providing feedback based on their own expertise and additional research.

[1347] Step 10:

[1348] The user sends the feedback to the server, which starts preparing to aggregate the evaluation results of each module.

[1349] Step 11:

[1350] The server acquires the user's emotion data using an emotion engine to recognize the user's emotion, analyzes this data, and evaluates the emotion of the product.

[1351] Step 12:

[1352] The server combines the results from the fact check, legal check, answer check, and sentiment engine to generate a final rating and commentary. The server ensures that the generated rating is factual, legally compliant, and the best answer.

[1353] Step 13:

[1354] The server issues an "AI Quality Certificate" to highly rated products, providing evidence of the reliability of the products.

[1355] Step 14:

[1356] The server retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[1357] This detailed processing flow not only ensures that the output of the generative AI model is accurate, legally sound, and optimal, but also allows for evaluation based on the user's emotions.

[1358] Example 2

[1359] 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."

[1360] It is difficult to evaluate whether the information generated by generative AI models is factually accurate, legally compliant, or the most appropriate answer. It is also important to consider the user's emotions when making an evaluation, and a system that can achieve this in an integrated manner is needed.

[1361] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a fact confirmation means for checking the accuracy of the generated information, a legal review means for evaluating the legal compliance of the generated information, an answer evaluation means for evaluating whether the generated information is an optimal answer, a sentiment analysis means for recognizing and evaluating the user's sentiment, an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary, an accreditation means for certifying the quality of highly rated products, and a re-learning means for relearning the evaluation results to the generative model. This enables the generated information to be based on facts, legally compliant, and an optimal answer, while also being evaluated taking the user's sentiment into consideration.

[1362] "Fact-checking means" refers to a means of verifying the accuracy of generated information using a database or external API.

[1363] "Legal review measures" are measures to evaluate whether the generated information is legally compliant and check for legal issues such as infringement of intellectual property rights, infringement of the right to protect personal information, and defamation.

[1364] The "answer evaluation means" is a means for evaluating whether the generated information is the optimal answer.

[1365] The "emotion analysis means" is a means for recognizing the user's emotions and reflecting them in the evaluation of the generated information.

[1366] An "integrated evaluation means" is a means for integrating the results from each evaluation means to generate an overall evaluation and commentary.

[1367] "Certification means" is a means of certifying the quality of products that have received high praise.

[1368] The "relearning means" is a means for feeding back the evaluation results to the generative model to perform re-learning, thereby improving future evaluation accuracy.

[1369] A "generative model" is an artificial intelligence model that generates information based on prompts from users.

[1370] The present invention combines an emotion engine that recognizes user emotions in a system that evaluates the accuracy, legal compliance, and appropriateness of generated information. The system includes a fact-checking means, a legal check means, an answer check means, a sentiment analysis means, an integrated evaluation means, a certification means, and a re-learning means.

[1371] Hardware and software used

[1372] Hardware

[1373] Server: central system for data processing and evaluation

[1374] Terminal: User input and display

[1375] User Device: Input Device for Emotion Analysis

[1376] software

[1377] Generative AI models (e.g., open-source generative models)

[1378] Fact-checking databases and external fact-checking APIs

[1379] Legal databases and legal document analysis tools

[1380] Sentiment Analysis Engine

[1381] Program processing

[1382] server

[1383] When a user sends a prompt, the server uses a generative AI model to generate output. For example, if the prompt is "When AI describes the future of driving, please generate output taking into consideration technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," the generative AI generates the corresponding text. The generated information is then fact-checked using a fact-checking method, which checks it against a fact-checking database and external fact-checking APIs (e.g., FactCheck.org and Snopes).

[1384] Terminal

[1385] The device receives the generated text from the server and analyzes the generated information using legal checks. For example, if the generated text contains a statement such as "The CEO of Company X was arrested for illegal conduct," the device analyzes the content and checks for legal issues such as intellectual property infringement, privacy infringement, and defamation. This check involves referencing legal databases (e.g., LexisNexis and Westlaw).

[1386] User

[1387] After receiving a product that has passed the legal check, the user evaluates whether the content is the optimal answer using the answer check tool. For example, for a product that says "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their expertise and additional research and provides feedback.

[1388] server

[1389] The emotion analysis means analyzes the user's feedback and emotions toward the product. For example, if the user feels that the information about the product is insufficient, the emotion engine recognizes that emotion from the user's text, facial expressions, and voice, and sends the results to the server. Next, the evaluation results of each module are integrated by the integrated evaluation means, and products that receive high ratings are issued a "quality certificate" by the certification means. The evaluation results and emotion analysis results are then fed back into the generative model, and the model is retrained by the retraining means.

[1390] Specific examples

[1391] As a concrete example, consider a generative AI model that generates output on the topic "The Future of AI-Driven Driving." The prompt in this case would be:

[1392] "When describing the future of AI driving, please generate output taking into account the technological trends announced in 2022, the associated legal issues, and how the general public will feel about them."

[1393] Based on this, a system is provided in which each means works together, the generated information is factual, legally compliant, provides the optimal answer, and is evaluated taking into account the user's emotions.

[1394] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1395] Step 1:

[1396] The user inputs a prompt, for example, "When AI describes the future of car driving, please generate an output taking into consideration the technological trends announced in 2022, related legal issues, and how ordinary users would feel about them," and sends it to the server. Based on this input, the generative AI model generates an appropriate output.

[1397] Step 2:

[1398] The server uses a generative AI model to generate output based on the prompt. The generated text is analyzed and compared with a fact-checking database and an external fact-checking API. For example, if a product is generated such as "The 2022 Nobel Peace Prize winner is so-and-so," the database and API are used to fact-check the output. The output is the fact-checked text.

[1399] Step 3:

[1400] The server sends the fact-checked product to the terminal. The terminal then performs a legal review of the received product. Specifically, it analyzes the text in the product to determine whether it contains legal issues such as intellectual property infringement, personal information protection infringement, or defamation. For example, it analyzes a statement such as "The CEO of Company X was arrested for illegal activity" and checks whether there are any problems. The output is a product that is legally sound.

[1401] Step 4:

[1402] The terminal sends the legally reviewed product to the user, who evaluates whether the received product is the optimal answer. For example, for a product such as "The appropriate daily caffeine intake is XX mg," the user evaluates it based on their own expertise and additional research and provides feedback. The output is the user's feedback and the evaluation result.

[1403] Step 5:

[1404] The server analyzes the feedback and evaluation results received from the user using emotion analysis means. It recognizes emotions from the user's text, facial expressions, and voice, and sends the evaluation results to the server. For example, if the user feels that "this information is insufficient," the emotion data is analyzed and sent to the server. The output is the result of the user's emotion analysis.

[1405] Step 6:

[1406] The server integrates the evaluation results obtained from the fact-checking means, legal checking means, answer checking means, and sentiment analysis means in an integrated evaluation means. For example, the server aggregates the evaluation results of each module to generate an overall score, and then generates a final evaluation and commentary. The output is the overall evaluation result and commentary.

[1407] Step 7:

[1408] If the overall evaluation result is high, the server issues a "quality certificate" using the certification means. This provides evidence of the reliability of the product. The quality certificate is assigned to the product. The output is the product with the issued quality certificate.

[1409] Step 8:

[1410] The server feeds back the overall evaluation results and sentiment analysis results to the generative model using a retraining method. This improves the future evaluation accuracy of the generative AI model. For example, the evaluation results are added to the dataset and the model is retrained. The output is a retrained generative AI model.

[1411] (Application example 2)

[1412] 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."

[1413] While existing technologies exist for individually evaluating the accuracy, legal compliance, and optimality of products generated by generative AI models, these evaluations do not take user emotions into account, making it difficult to improve the user experience or provide appropriate feedback based on emotions.The present invention aims to combine user emotions with the evaluation of generative AI models to produce products that are more realistic and reliable, and to provide optimized results that correspond to the user's emotions.

[1414] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: a fact-checking means for checking the accuracy of the generated information; a legal-checking means for evaluating the legal compliance of the generated information; an answer-checking means for evaluating whether the generated information is the optimal answer; a sentiment-analysis means for making an evaluation based on the user's emotions using an emotion engine that analyzes the user's emotions; an integrated-evaluation means for integrating the evaluation results to generate an overall evaluation and commentary; an accreditation means for certifying the quality of highly rated products; and a re-learning means for relearning the evaluation results to the generative model. This makes it possible for the generated information and warnings to be provided in a form that is accurate, legally sound, and optimized according to the user's emotions.

[1415] A "fact-checking tool" is a module that checks the accuracy of the generated information against a fact-checking database and uses an external fact-checking API to verify the facts.

[1416] The "legal check means" is a module that analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1417] The "answer check means" is a module for evaluating whether the generated information is the optimal answer.

[1418] The "emotion analysis means" is an emotion engine that analyzes the user's emotions and reevaluates the generated information based on those emotions.

[1419] The "integrated evaluation means" is a module that integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to generate a comprehensive evaluation and commentary.

[1420] The "certification means" is a module that certifies the quality of products that have received high evaluations.

[1421] The "relearning means" is a module that re-learns the evaluation results into the generative model, thereby improving the evaluation accuracy of the model.

[1422] This invention combines a system that evaluates the accuracy, legal compliance, and appropriateness of generated information with an emotion engine that recognizes user emotions. The system consists of the following main modules: fact-checking means, legal checking means, answer checking means, emotion analysis means, integrated evaluation means, certification means, and relearning means.

[1423] In this embodiment, the server performs the following process.

[1424] First, the fact-checking method verifies the accuracy of the information generated by the generative AI model. Specifically, it compares the generated information with a fact-checking database and, if necessary, uses an external fact-checking API to verify the facts. This ensures the accuracy of the output.

[1425] Next, the legal check means analyzes the generated information and evaluates its legal compliance. This means checks whether the generated information contains any legal issues such as copyright infringement, violation of privacy rights, or defamation. This ensures that the generated information is legally free.

[1426] The answer checker evaluates whether the generated information is the best answer. Based on the user's expertise and additional research, it verifies whether the generated answer is the best answer for the given situation. This ensures that the generated content is appropriate.

[1427] Furthermore, in this embodiment, an emotion engine is used as an emotion analysis means. The emotion engine analyzes the user's emotions in real time and provides information and feedback based on the analysis. It evaluates how the user feels about the generated information and adjusts the content and presentation method of the information accordingly.

[1428] The integrated evaluation method combines the results of each evaluation to generate a final evaluation and commentary. It integrates the results of fact-checking, legal checks, answer checks, and sentiment analysis to comprehensively judge the reliability and appropriateness of the generated product.

[1429] The certification method will certify the quality of products that receive high evaluations. Products that receive quality certification will be given an "AI Quality Certificate," which can be provided to external parties as evidence of the reliability of the information.

[1430] Finally, the retraining means retrains the generative AI model with the evaluation results and sentiment analysis results, thereby improving the future evaluation accuracy of the generative AI model.

[1431] Specific examples

[1432] For example, if a user receives an alert that reads "Critical security breach detected" and feels "anxious," the emotion engine analyzes that emotion in real time. Based on the analysis results, the alert content is provided with information that is adjusted to prevent excessive tension. This emotion analysis allows the user to receive appropriate feedback, enabling them to respond more calmly.

[1433] Prompt Sentence Examples

[1434] An example of a prompt to be input to the generative AI model is as follows:

[1435] "Ask the Generative AI to create an incident report. Please create an incident report with the following information:

[1436] 1. Details of the incident and its scope.

[1437] 2. Include fact-checked data.

[1438] 3. Use legally acceptable language.

[1439] 4. Customize alert content based on user emotions.

[1440] As described above, the present invention is a system that can provide more realistic and reliable information by combining the user's emotions with the evaluation of generated information.

[1441] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1442] Step 1:

[1443] Receiving generated information

[1444] The server receives information generated by the generative AI model. The input is the generated information (e.g., incident report), and the output is the content of the information itself. This information is passed to various subsequent check modules.

[1445] Step 2:

[1446] Fact Check

[1447] The server passes the received information to the fact-checking means to check its accuracy. Specifically, it compares the input information with a fact-checking database and uses an external fact-checking API to check the facts. The output is a result of whether the information is true or not.

[1448] Step 3:

[1449] Legal Check

[1450] Based on the fact-checking results, the server passes the information to the legal checker. The input is the fact-checked information, which is analyzed to evaluate its legal compliance. Specifically, it checks for legal issues such as copyright infringement, violation of privacy rights, and defamation, and outputs a result indicating whether or not there is a legal problem.

[1451] Step 4:

[1452] Answer Check

[1453] The server passes the information to the answer checker based on the results of the legal check. The input is the legally checked information, and the server evaluates whether this information is the best answer. The output is an evaluation of the information's suitability based on the user's expertise and additional research.

[1454] Step 5:

[1455] Emotion analysis

[1456] The server passes the information to the emotion analysis means based on the answer check results. The input is the answer-checked information and the user's emotion data, which the emotion engine analyzes to understand the user's emotional state. The output is an analysis result based on the user's emotion.

[1457] Step 6:

[1458] Integrated Evaluation

[1459] The server integrates all the evaluation results based on the results of sentiment analysis. The inputs are the results of fact-checking, legal checks, answer checks, and sentiment analysis, and combines them to generate a final evaluation and commentary. The output is an overall evaluation result based on the information's reliability, legal relevance, appropriateness, and user sentiment.

[1460] Step 7:

[1461] Certification

[1462] The server certifies the quality of information that receives a high rating based on the results of the integrated evaluation. The input is the results of the integrated evaluation, and the output is an "AI Quality Certificate."

[1463] Step 8:

[1464] Relearn

[1465] The server retrains the generative AI model based on the evaluation results and sentiment analysis results based on the certification results. The inputs are the evaluation results and sentiment analysis results, and the generative AI model is updated based on these, improving the model's evaluation accuracy as output.

[1466] The above are the specific processing steps of the program for the system that realizes the application example.

[1467] 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.

[1468] 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.

[1469] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1470] 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.

[1471] 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.

[1472] 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.

[1473] 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).

[1474] 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.

[1475] 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."

[1476] 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.

[1477] 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).

[1478] 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.

[1479] 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.

[1480] 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.

[1481] 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.

[1482] 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.

[1483] 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.

[1484] 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.

[1485] 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.

[1486] 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.

[1487] 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.

[1488] The following is further disclosed regarding the above embodiment.

[1489] (Claim 1)

[1490] fact-checking measures to check the accuracy of the information generated;

[1491] a legal check means for assessing the legal compliance of the generated information;

[1492] an answer check means for evaluating whether the generated information is the optimal answer;

[1493] an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary;

[1494] A certification method for certifying the quality of highly rated products;

[1495] The system includes a re-learning means for re-learning the evaluation results into a generative model.

[1496] (Claim 2)

[1497] 2. The system of claim 1, wherein the fact-checking means compares the generated information with a fact-checking database and performs fact-checking using an external fact-checking API.

[1498] (Claim 3)

[1499] 2. The system according to claim 1, wherein the legal check means analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1500] "Example 1"

[1501] (Claim 1)

[1502] fact-checking means to check the accuracy of the information generated;

[1503] a legal assessment tool for assessing the legal compliance of the generated information;

[1504] a response evaluation means for evaluating whether the generated information is an optimal answer;

[1505] an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary;

[1506] A certification method to certify the quality of highly rated products;

[1507] a re-learning means for re-learning the evaluation results into a generative model;

[1508] A user interface means for inputting a prompt sentence to the generative AI model;

[1509] A system including an evaluation result display means for displaying various evaluation results for a product.

[1510] (Claim 2)

[1511] 2. The system of claim 1, wherein the fact-checking means checks the generated information against a fact-checking database and performs fact-checking using an external fact-checking API.

[1512] (Claim 3)

[1513] 2. The system of claim 1, wherein the legal evaluation means analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1514] "Application Example 1"

[1515] (Claim 1)

[1516] fact-checking measures to check the accuracy of the information generated;

[1517] a legal check means for assessing the legal compliance of the generated information;

[1518] an answer check means for evaluating whether the generated information is the optimal answer;

[1519] an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary;

[1520] A certification method for certifying the quality of highly rated products;

[1521] a re-learning means for re-learning the evaluation results into a generative model;

[1522] a distribution means for evaluating the accuracy, legal compliance, and appropriateness of the information to be distributed and distributing quality-certified information;

[1523] Information assurance measures to guarantee the reliability of distributed information based on evaluation results

[1524] A system including:

[1525] (Claim 2)

[1526] 2. The system of claim 1, wherein the fact-checking means compares the generated information with a fact-checking database and performs fact-checking using an external fact-checking API.

[1527] (Claim 3)

[1528] 2. The system according to claim 1, wherein the legal check means analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

[1529] "Example 2: Combining Emotion Engines"

[1530] (Claim 1)

[1531] fact-checking means to check the accuracy of the information generated;

[1532] a legal review instrument to assess the legal compliance of the generated information;

[1533] an answer evaluation means for evaluating whether the generated information is an optimal answer;

[1534] emotion analysis means for recognizing and evaluating the emotions of a user;

[1535] an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary;

[1536] A certification method for certifying the quality of highly rated products;

[1537] The system includes a re-learning means for re-learning the evaluation results into a generative model.

[1538] (Claim 2)

[1539] 2. The system of claim 1, wherein the fact-checking means checks the generated information against a fact-checking database and performs fact-checking using an external fact-checking API.

[1540] (Claim 3)

[1541] 2. The system according to claim 1, wherein the legal review means analyzes the generated information and checks for legal issues such as infringement of intellectual property rights, infringement of personal information protection rights, and defamation.

[1542] "Application example 2 when combining emotion engines"

[1543] (Claim 1)

[1544] fact-checking measures to check the accuracy of the information generated;

[1545] a legal check means for assessing the legal compliance of the generated information;

[1546] an answer check means for evaluating whether the generated information is the optimal answer;

[1547] emotion analysis means for performing evaluation based on the user's emotion using an emotion engine that analyzes the user's emotion;

[1548] an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary;

[1549] A certification method for certifying the quality of highly rated products;

[1550] and a re-learning means for re-learning the evaluation results into the generative model.

[1551] (Claim 2)

[1552] 2. The system of claim 1, wherein the fact-checking means compares the generated information with a fact-checking database and performs fact-checking using an external fact-checking API.

[1553] (Claim 3)

[1554] 2. The system according to claim 1, wherein the legal check means analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation. [Explanation of symbols]

[1555] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. fact-checking measures to check the accuracy of the information produced; a legal check means for assessing the legal compliance of the generated information; an answer check means for evaluating whether the generated information is the optimal answer; an integrated evaluation means for integrating the evaluation results to generate an overall evaluation and commentary; A certification method for certifying the quality of highly rated products; The system includes a re-learning means for re-learning the evaluation results into a generative model.

2. 2. The system of claim 1, wherein the fact-checking means compares the generated information with a fact-checking database and performs fact-checking using an external fact-checking API.

3. 2. The system according to claim 1, wherein the legal check means analyzes the generated information and checks for legal issues such as copyright infringement, violation of privacy rights, and defamation.

Images