System

Abstract

A system is provided.SOLUTION: A system comprising: means for a user to perform a login operation; means for passing received textual data to a generative AI model; means for matching candidates extracted by the generative AI model against a database; means for generating and presenting a list of indicated points to the user; means for the user to select or input correction content; and means for storing final text.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] Despite its convenience, currently used text generation AI is often limited to specific occupations, and there is a risk that the generated text may contain errors or suspicious points. In particular, incorrect information generation (hallucination) can cause reliability issues, preventing widespread use. The purpose of this invention is to expand the scope of use of text generation AI and realize highly reliable text generation. [Means for solving the problem]

[0005] To solve the above problems, the present invention provides the following means: a system including a means for a user to perform a login operation, a means for passing received text data to a generative AI model, a means for comparing candidates extracted by the generative AI model with a database, a means for generating a list of issues to present to the user, a means for the user to select or input corrections, and a means for saving the final text. Furthermore, by adding to the system a means for using RAG technology to compare candidates extracted by the generative AI model and a means for generating final text that reflects the corrections selected or input by the user, reliable text generation becomes available to a wide range of enterprise users.

[0006] A "user" is an entity that accesses the system and performs login operations, text input, and modifications.

[0007] A "login operation" is a procedure in which a user authenticates to the system using an ID and password.

[0008] A "generative AI model" is a machine learning model that analyzes received text data and extracts potential errors or suspicious points.

[0009] A "database" is a collection of information, including public or closed data, used for collation.

[0010] "RAG technology" stands for Retrieval-Augmented Generation, and is a technology that compares candidates extracted by a generative AI model with external data.

[0011] The "list of points of concern" is a list of items that point out errors or suspicious points based on the results of comparing the generated AI model with the database.

[0012] "Corrections" are changes or corrections that the user makes to the text based on the presented list of suggestions.

[0013] "Saving" refers to recording and storing the final text data after the user has completed the modifications. [Brief explanation of the drawings]

[0014] [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

[0015] 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.

[0016] First, the terms used in the following description will be explained.

[0017] 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).

[0018] 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.

[0019] 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.

[0020] 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.

[0021] 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."

[0022] [First embodiment]

[0023] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0024] 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.

[0025] 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).

[0026] 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.

[0027] 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.

[0028] 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.

[0029] 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.

[0030] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0031] 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.

[0032] 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.

[0033] 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.

[0034] 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."

[0035] This invention is a text validation UI system that points out errors and suspicious points in text created by users of a text generation service and generates accurate and reliable text. Specific embodiments of this system are described below.

[0036] Initial Setup

[0037] 1. The user logs in

[0038] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[0039] 2. The server authenticates the user

[0040] The server authenticates the user by checking the ID and password against a database. If authentication is successful, the user session is started and the user information is loaded into the system.

[0041] Text Input

[0042] 1. The user enters text

[0043] Users enter the text of business email drafts and documents into the device's interface.

[0044] 2. The device sends the entered text to the server

[0045] When the user clicks the input button, the terminal sends the entered text data to the server in an appropriate format, such as JSON format.

[0046] Text analytics

[0047] 1. The server passes the text to the generative AI model

[0048] The server passes the received text data to an existing generative AI model, which then receives the input text and begins analyzing it.

[0049] 2. A generative AI model analyzes the text

[0050] The generative AI model analyzes the content of the text and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[0051] Validation

[0052] 1. The server collates data using RAG technology

[0053] The server uses RAG technology to compare candidates extracted by the generative AI model with public and closed databases, and evaluates the reliability of the candidates.

[0054] 2. The server generates a list of issues

[0055] Based on the matching results, the server generates a list of issues, such as "misspelled company name" or "incorrect date format."

[0056] feedback

[0057] 1. The server sends the list of issues to the user's device.

[0058] The list of issues is sent to the user's device in an appropriate format, where the user can view it.

[0059] 2. The user confirms the issue

[0060] The user reviews the presented list of issues and carefully examines the points that need to be corrected.

[0061] 3. User makes corrections

[0062] Users can make corrections based on the suggestions, either manually or by simply clicking on the suggested corrections.

[0063] Final confirmation and saving

[0064] 1. The user checks the final text

[0065] The user checks the final text after correction and confirms the content.

[0066] 2. The user clicks the save button

[0067] When the user clicks the save button, the final edited text is sent to the server.

[0068] 3. The server saves the final text

[0069] The server stores the final text in a database and associates it with the user's account.

[0070] Specific examples

[0071] For example, suppose a user is composing a business email containing information about a new product. First, the user logs in to the system and enters the email content. The entered text is sent to the server and analyzed by the generative AI model. The user then receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the user saves the email with the corrections completed, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the diverse needs of enterprise users.

[0072] The processing flow will be explained below.

[0073] Step 1:

[0074] The user performs a login operation.

[0075] The user accesses the login screen of the text validation UI from their device, enters their ID and password, and clicks the login button to send the entered information to the server.

[0076] Step 2:

[0077] The server authenticates the user

[0078] The server checks the received ID and password against the authentication information in the database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[0079] Step 3:

[0080] The user enters text

[0081] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[0082] Step 4:

[0083] The device sends the entered text to the server

[0084] The device sends the text data entered by the user to the server in JSON format or an appropriate data format, including the session ID.

[0085] Step 5:

[0086] The server passes the text to the generative AI model

[0087] The server processes the received text data and sends a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[0088] Step 6:

[0089] Generative AI models parse text

[0090] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, and suspicious points, using natural language processing (NLP) techniques.

[0091] Step 7:

[0092] The server collates data using RAG technology

[0093] The server receives the candidates extracted by the generative AI model and compares them with public and closed databases using RAG technology, thereby evaluating the reliability of each candidate.

[0094] Step 8:

[0095] The server generates a list of issues

[0096] The server generates a list of issues based on the results of the match, including specific corrections such as grammar errors and incorrect data.

[0097] Step 9:

[0098] The server sends the list of issues to the user's device.

[0099] The server sends the list of issues to the user's device in JSON format, etc., which the user can view on the interface.

[0100] Step 10:

[0101] The user confirms the points

[0102] The user can check the list of issues on their device and carefully examine the details of each correction.

[0103] Step 11:

[0104] User makes corrections

[0105] The user can then make the necessary corrections based on the list, either by manually entering the corrections or by clicking on the suggested corrections provided to automatically incorporate them.

[0106] Step 12:

[0107] User confirms final text

[0108] After completing the corrections, the user checks the final text to ensure there are no problems with the content.

[0109] Step 13:

[0110] The user clicks the save button

[0111] When the user clicks the save button, the final text is sent to the server.

[0112] Step 14:

[0113] The server saves the final text

[0114] The server stores the final text in a database and associates it with the user's account, completing the process.

[0115] Example 1

[0116] 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."

[0117] Conventional text generation systems have difficulty efficiently detecting errors and inconsistencies and providing appropriate feedback to users. Furthermore, they lack an interface that allows users to review corrections and quickly and accurately correct the text. This can lead to problems such as a decrease in the reliability and quality of the generated text.

[0118] 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.

[0119] In this invention, the server includes a means for user authentication, a means for passing received text data to the generative AI model, and a means for comparing candidates extracted by the generative AI model with a database using RAG technology, which enables reliable evaluation of errors and suspicious points in the text detected by the generative AI model and provides accurate feedback and suggested corrections to the user.

[0120] "User authentication" is the process in which the server checks the authentication information (such as ID and password) entered by the user against a database to verify whether the user is legitimate.

[0121] A "terminal" is a device, such as a computer or smartphone, that a user uses to enter text or perform login operations.

[0122] "Text data" refers to character string information such as sentences or words entered by the user.

[0123] A "generative AI model" is an algorithm that uses natural language processing technology to analyze input text and extract grammatical errors, typos, and suspicious points.

[0124] "RAG technology" stands for Retrieval-Augmented Generation, and is a technology that compares generated candidates with existing databases to evaluate their reliability.

[0125] A "database" is a structured collection of data that is accessed by a server and stores user authentication information and data for verification.

[0126] The "list of points of concern" is a list of grammatical errors, typos, suspicious points, and other areas that need to be corrected, generated as a result of analysis by the generative AI model.

[0127] "Saving" is the process by which the server stores the text, once the user has completed the modifications, in a database for later access.

[0128] "Corrections" refer to corrections suggested by the generative AI model or corrections manually entered by the user.

[0129] This invention provides a system for pointing out errors and suspicious points in text created by a user using a text generation service, and generating accurate and reliable text. Specific embodiments of this system are described below.

[0130] Initial Setup

[0131] A user opens a browser on their device (e.g., a PC or smartphone) and accesses the text validation UI. They enter their ID and password on the login screen and perform the login operation. The server checks the ID and password entered by the user against the database to confirm whether they are a legitimate user. If authentication is successful, the server starts a user session, loads related user information, and provides the user with the dashboard or main operation screen.

[0132] Text Input

[0133] For example, when composing a business email containing new product information, a user can enter text into the device interface or copy and paste it from another source. When the user clicks the "Done" button, the device converts the entered text data into JSON format and sends it to the server via the API.

[0134] Text analytics

[0135] The server then passes the received text data to a generative AI model (e.g., GPT-3 or BERT). This generative AI model receives the input text and analyzes it to identify grammatical errors, typos, and other suspicious features. Specifically, it uses natural language processing techniques to tokenize each sentence and extract anomalous patterns.

[0136] Validation

[0137] The server uses Retrieval-Augmented Generation (RAG) technology to compare the suspicious points and candidates extracted by the generative AI model with public databases and internal closed data. This comparison evaluates the reliability and accuracy of the candidates. Based on the comparison results, the server creates a list of issues to be addressed. Examples of issues on this list include "misspellings in company names" and "date format errors."

[0138] feedback

[0139] The server converts the list of issues into an appropriate format (e.g., JSON) and sends it to the user's device. The user's device displays the received data on an interface. The user reviews the list of issues and carefully examines which parts are incorrect and how they should be corrected. The user corrects the text based on the issues, either manually or by applying the system's suggested corrections with one click.

[0140] Final confirmation and saving

[0141] The user then finalizes the edited text to ensure it is correct. Once the user clicks the save button, the final edited text is sent to the server, which stores the edited text in a database and associates it with the user's account.

[0142] Specific examples

[0143] For example, a user creates a business email containing new product information and enters the content into the system. The entered text is sent to the server and analyzed by the generative AI model. The user receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the corrected email is saved, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the needs of various enterprise users.

[0144] Prompt Sentence Examples

[0145] The following prompts can be fed to the generative AI model:

[0146] Enter the paragraph "Please check this document." Point out the errors in the text output by the generative AI model and provide the best possible corrections.

[0147] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0148] Step 1:

[0149] The user performs a login operation.

[0150] Input: The user enters their ID and password from the device.

[0151] Data processing: The terminal converts the entered authentication information into a standardized format (e.g., JSON).

[0152] Output: The device sends the authentication information to the server.

[0153] What happens: The user opens a browser, accesses the login screen, and enters the required credentials.

[0154] Step 2:

[0155] The server authenticates the user

[0156] Input: The authentication information sent from the device.

[0157] Data processing: The server queries the database for authentication information and verifies whether the user's ID and password are correct.

[0158] Output: Prepares the authentication result (success or failure).

[0159] Specific operation: The server references the database and starts a user session based on the authentication result.

[0160] Step 3:

[0161] The user enters text

[0162] Input: A user types text for a business email or document.

[0163] Data processing: The entered text is temporarily stored on the terminal as string data.

[0164] Output: The input text data.

[0165] What happens: The user types text directly into the device interface or copies and pastes it from another source.

[0166] Step 4:

[0167] The device sends the entered text to the server

[0168] Input: The user completes text input.

[0169] Data processing: The terminal converts the input text into JSON format.

[0170] Output: The converted text data is sent to the server.

[0171] Specific operation: When the user clicks the "Complete input" button, the terminal sends the text data to the server.

[0172] Step 5:

[0173] The server passes the text to the generative AI model

[0174] Input: Text data sent from the terminal.

[0175] Data processing: The server formats the received text data into a format suitable for the generative AI model.

[0176] Output: The formatted text data is fed into a generative AI model.

[0177] How it works: The server passes the text data to a generative AI model (e.g., GPT-3 or BERT).

[0178] Step 6:

[0179] Generative AI models parse text

[0180] Input: Text data passed from the server.

[0181] Data processing: The generative AI model analyzes the text and extracts grammatical errors, typos, and suspicious points.

[0182] Output: A list of candidates as the analysis result.

[0183] How it works: The generative AI model uses natural language processing techniques to analyze text for tokens and extract anomalous patterns.

[0184] Step 7:

[0185] The server collates data using RAG technology

[0186] Input: A list of candidates extracted by a generative AI model.

[0187] Data processing: The server uses RAG technology to check the candidate list against the database.

[0188] Output: A list of candidates with a confidence rating.

[0189] Specific operation: The server checks candidates against public databases and closed data.

[0190] Step 8:

[0191] The server generates a list of issues

[0192] Input: A list of candidates with confidence ratings.

[0193] Data processing: The server generates a list of findings.

[0194] Output: A list of issues.

[0195] Specific operation: Based on the matching results, the server creates a list of issues such as "misspelled company name" or "date format error."

[0196] Step 9:

[0197] The server sends the list of issues to the user's device.

[0198] Input: The generated issue list.

[0199] Data processing: The server converts the list of findings into an appropriate format (e.g., JSON).

[0200] Output: Send the converted issue list to the user's terminal.

[0201] Specific operation: The server sends a list of issues to the user's device and displays it on the user's interface.

[0202] Step 10:

[0203] The user confirms the points

[0204] Input: A list of issues displayed on the terminal.

[0205] Data processing: The user confirms the information pointed out.

[0206] Output: The confirmed issue.

[0207] Specific action: The user looks at the list of issues on the interface and identifies the problem.

[0208] Step 11:

[0209] User makes corrections

[0210] Input: The confirmed issue.

[0211] Data processing: The user corrects the text based on the suggestions.

[0212] Output: The corrected text.

[0213] What happens: The user manually corrects the text or clicks to apply the suggested correction.

[0214] Step 12:

[0215] User confirms final text

[0216] Input: The corrected text.

[0217] Data processing: The user makes a final check of the revised text.

[0218] Output: The final text that was checked.

[0219] Specific operation: The user reviews the revised text in detail and performs a final check.

[0220] Step 13:

[0221] The user clicks the save button

[0222] Input: Final text as verified.

[0223] Data processing: The user clicks the save button.

[0224] Output: The final text is sent to the server.

[0225] Specific behavior: When the user clicks the save button, the device sends the final text to the server.

[0226] Step 14:

[0227] The server saves the final text

[0228] Input: The final text sent from the terminal.

[0229] Data processing: The server stores the final text in a database.

[0230] Output: The final text saved.

[0231] What it does: The server stores the final text it receives in a database and associates it with the user's account.

[0232] (Application example 1)

[0233] 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."

[0234] When creating business texts or security documents, it is difficult for users to quickly and accurately identify grammatical errors, typographical errors, and even security risk-related errors to create reliable final text. In addition, real-time analysis and correction suggestions are required, while ensuring compliance with industry regulations and internal rules. An efficient system to solve these challenges is needed.

[0235] 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.

[0236] In this invention, the server includes means for a user to perform authentication operations, means for passing received text data to a generative AI model, means for comparing candidates extracted by the generative AI model with a data storage location, means for generating a list of points to be pointed out and presenting it to the user, means for the user to select or input corrections, means for saving the final text, means for analyzing voice-input text in real time and pointing out errors and security risks, means for displaying correction suggestions in real time, and means for checking compliance with industry regulations and internal rules.This enables users to check and correct text errors and security risks in real time and efficiently create reliable final text.

[0237] The "means by which a user performs authentication operations" is a function for verifying the identity of a user based on the authentication information entered by the user.

[0238] "Means for passing received text data to a generative AI model" refers to a function that transfers text data obtained from a user to a generative AI model that analyzes the data and suggests corrections.

[0239] "Means for comparing candidates extracted by the generative AI model with data storage locations" refers to a function that compares and verifies candidates for errors and corrections in text extracted by the generative AI model with an existing database.

[0240] "Means for generating a list of points of concern and presenting it to the user" refers to a function that creates a list of errors and suspicious points in the text based on the generative AI model and the matching results, and displays it to the user.

[0241] The "means for the user to select or input corrections" refers to an interface or function that allows the user to select or newly input corrections to the pointed out points in the text.

[0242] The "means for saving the final text" is a function for saving the final text after correction in a database or a specific storage location.

[0243] "Means for analyzing voice-input text in real time and pointing out errors and security risks" is a function that instantly analyzes voice-input text, detects grammatical errors and security risks, and points them out in real time.

[0244] "Means for displaying suggested revisions in real time" refers to a function that instantly presents users with suggested revisions based on the analysis results.

[0245] The "means for checking compliance with industry regulations and internal rules" is a function for checking whether the generated text complies with the relevant industry regulations and internal rules.

[0246] The "Secure Text Guard" system allows users to create text using voice input or text input, and detects grammatical errors, typos, suspicious points, and security risks in the text in real time and makes correction suggestions. Specific embodiments of the system are described below.

[0247] Hardware Configuration

[0248] 1. Smart Glasses

[0249] A device for voice input and text display.

[0250] It has a voice recognition function that converts voice input data into text.

[0251] 2. Server

[0252] A central server responsible for user authentication, text analysis, generating correction suggestions, and data storage.

[0253] Integrate with databases of data locations and industry regulations.

[0254] Software Configuration

[0255] 1. Generative AI Models

[0256] An AI model that analyzes user-entered text to detect grammatical errors, typos, and security risks.

[0257] 2. NLP technology

[0258] As a natural language processing technology, it is used to analyze input text and extract suspicious points.

[0259] 3. RAG technology

[0260] A technology that compares the results of text analysis with a database and evaluates their reliability.

[0261] System operation flow

[0262] 1. Authentication Operation

[0263] The user puts on the smart glasses and enters authentication information (user ID and password) using voice or touch.

[0264] The server checks this authentication information against a database to authenticate the user.

[0265] 2. Text Input

[0266] After successful authentication, the user can then use voice input or direct text input.

[0267] The entered text data is sent to the server in real time.

[0268] 3. Text analysis and suggested revisions

[0269] The server passes the received text to the generative AI model and begins analysis.

[0270] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[0271] Meanwhile, RAG technology is used to compare the analysis results with the data storage location and evaluate the reliability of the analysis.

[0272] 4. Corrections and Final Text Generation

[0273] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[0274] The user selects or inputs the corrections based on this list.

[0275] The server applies the modifications and generates the final text.

[0276] 5. Save

[0277] The final text, after corrections are complete, is stored in a data repository by the server.

[0278] Specific examples

[0279] For example, a user can use smart glasses to voice-input "important meeting materials." The system converts this into text in real time and begins analysis. If grammatical errors or risks of leaking confidential information are detected, they are immediately presented to the user and suggested corrections are displayed. The user can select or enter the suggested corrections, and the final text is generated.

[0280] Example prompt sentence:

[0281] "Analyzes the entered text and points out errors and suspicious points such as grammar mistakes, typos, and risk of leaking confidential information."

[0282] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0283] Step 1:

[0284] The user puts on the smart glasses and enters authentication information (user ID and password) by voice or touch.

[0285] Input: User ID, Password

[0286] Output: Authentication request

[0287] Step 2:

[0288] The server receives the authentication information and authenticates the user against a database.

[0289] Input:Authentication request

[0290] Data processing and calculation: Verification with authentication information in the database

[0291] Output: Authentication result (success / failure)

[0292] Step 3:

[0293] After successful authentication, the user can enter voice or text.

[0294] Input: Audio or text data

[0295] Output: Input text data

[0296] Step 4:

[0297] The entered text data is sent to the server in real time.

[0298] Input: Input text data

[0299] Output: Text data sent to the server

[0300] Step 5:

[0301] The server passes the received text data to the generative AI model and begins analysis.

[0302] Input: Text data sent to the server

[0303] Data processing and computation: Text analysis with generative AI models

[0304] Output: Analysis results (potential errors or suspicious points)

[0305] Step 6:

[0306] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[0307] Input: Analysis results (candidates for errors or suspicious points)

[0308] Data processing and calculation: Additional detection and correction suggestion generation using NLP technology

[0309] Output: List of suggested fixes

[0310] Step 7:

[0311] Using RAG technology, the analysis results are compared with the data storage location to evaluate the reliability of the analysis.

[0312] Input: list of suggested revisions

[0313] Data processing and calculation: Verification with data storage location

[0314] Output: Reliability evaluation results

[0315] Step 8:

[0316] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[0317] Input: Reliability evaluation results

[0318] Output: A list of issues and suggested fixes displayed on the user's screen

[0319] Step 9:

[0320] The user selects or enters corrections based on this list.

[0321] Input: List of issues and suggested fixes

[0322] Output: Modifications selected or entered

[0323] Step 10:

[0324] The server applies the modifications and generates the final text.

[0325] Input: The selected or entered correction

[0326] Data processing and calculation: Reflection of corrections

[0327] Output: Final text

[0328] Step 11:

[0329] The final text, after corrections are complete, is stored in a data repository by the server.

[0330] Input: Final text

[0331] Output: Saved text data

[0332] 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.

[0333] This invention is a text validation UI system that identifies errors and suspicious points in text created by users of a text generation service, recognizes the user's emotions, and generates more accurate and reliable text. A specific embodiment of this system is described below.

[0334] Initial Setup

[0335] 1. The user logs in

[0336] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[0337] 2. The server authenticates the user

[0338] The server authenticates the user by checking the ID and password entered by the user against the authentication information in the database. If authentication is successful, a user session is started, a session ID is generated, and returned to the user. At the same time, the user information is loaded into the system.

[0339] Text Input

[0340] 1. The user enters text

[0341] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[0342] 2. The device sends the entered text to the server

[0343] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[0344] Text analytics

[0345] 1. The server passes the text to the generative AI model

[0346] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[0347] 2. A generative AI model analyzes the text

[0348] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[0349] emotion recognition

[0350] 1. Emotion engine recognizes user emotions

[0351] The emotion engine recognizes users' emotions in real time as they type, and this emotional information is reflected in the tone of the text and suggested corrections.

[0352] 2. The emotion engine uses emotional information for analysis

[0353] The emotion engine analyzes the user's emotions as they type and combines them with the results of a generative AI model, which adjusts the tone of the text.

[0354] Validation

[0355] 1. The server collates data using RAG technology

[0356] The server receives the candidates extracted by the generative AI model and compares them with public and closed databases using RAG technology, thereby evaluating the reliability of each candidate.

[0357] 2. The server generates a list of issues

[0358] Based on the matching results, the server generates a list of issues to be fixed, including specific corrections such as grammar errors and incorrect data.

[0359] feedback

[0360] 1. The server sends the list of issues to the user's device.

[0361] The list of issues is sent to the user's terminal in an appropriate format, and the user can view it on the interface.

[0362] 2. The user confirms the issue

[0363] Users can review the list of issues and review the details of each correction on their device, and the emotion engine will provide feedback tailored to the user's emotions.

[0364] 3. User makes corrections

[0365] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[0366] Final confirmation and saving

[0367] 1. The user checks the final text

[0368] The user checks the final text after corrections and confirms that there are no problems with the content.

[0369] 2. The user clicks the save button

[0370] When the user clicks the save button, the final text is sent to the server.

[0371] 3. The server saves the final text

[0372] The server stores the final text in a database and associates it with the user's account.

[0373] Specific examples

[0374] For example, imagine a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, it provides optimal feedback based on the user's emotions. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[0375] The processing flow will be explained below.

[0376] Step 1:

[0377] The user performs a login operation.

[0378] The user accesses the login screen of the text validation UI from the terminal, enters their ID and password, and clicks the login button to send the entered information to the server.

[0379] Step 2:

[0380] The server authenticates the user

[0381] The server checks the received ID and password against the authentication information in its database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[0382] Step 3:

[0383] The user enters text

[0384] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[0385] Step 4:

[0386] The device sends the entered text to the server

[0387] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[0388] Step 5:

[0389] The server passes the text to the generative AI model

[0390] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[0391] Step 6:

[0392] Generative AI models parse text

[0393] The generative AI model analyzes the text data, extracts candidates for grammatical errors, typos, and suspicious points, and returns the results to the server. This analysis uses natural language processing (NLP) technology.

[0394] Step 7:

[0395] Emotion engine recognizes user emotions

[0396] The emotion engine collects and analyzes real-time emotion data while the user is entering text, thereby identifying the user's emotional state.

[0397] Step 8:

[0398] The emotion engine uses emotional information for analysis

[0399] The emotion engine combines the results returned by the generative AI model with the user's emotional information, allowing the tone of the text and suggested revisions to be adjusted based on the user's emotions.

[0400] Step 9:

[0401] The server collates data using RAG technology

[0402] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[0403] Step 10:

[0404] The server generates a list of issues

[0405] Based on the results of the match, the server generates a list of issues, including specific corrections such as "grammatical errors" or "incorrect data."

[0406] Step 11:

[0407] The server sends the list of issues to the user's device.

[0408] The list of issues is sent to the user's device in an appropriate format, where the user can view it in the interface.

[0409] Step 12:

[0410] The user confirms the points

[0411] Users can view a list of issues and review the details of each correction on their device, and the emotion engine provides feedback tailored to the user's emotions.

[0412] Step 13:

[0413] User makes corrections

[0414] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[0415] Step 14:

[0416] User confirms final text

[0417] The user checks the final text after corrections and confirms that there are no problems with the content.

[0418] Step 15:

[0419] The user clicks the save button

[0420] When the user clicks the save button, the final text is sent to the server.

[0421] Step 16:

[0422] The server saves the final text

[0423] The server stores the final text in a database and associates it with the user's account, completing the process.

[0424] Example 2

[0425] 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."

[0426] Modern text generation services require a method to effectively identify errors and suspicious points in user-generated text. However, current systems are unable to provide feedback that takes into account the user's emotions, and the reliability of the generated feedback is not sufficiently guaranteed. As a result, users spend a lot of time correcting the text, and the quality of the final text is difficult to guarantee.

[0427] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0428] In this invention, the server includes an emotion recognition means for recognizing a user's emotions in real time, a means for adjusting the analysis results of the generative AI model based on the emotion information, and a means for reflecting emotion-based feedback in the list of points to be pointed out. This makes it possible to provide feedback that takes the user's emotions into consideration and improve the reliability and quality of the text.

[0429] "Means for users to log in" refers to the means by which users access the system and perform authentication using their own account information.

[0430] "Means for passing received text data to the generative AI model" refers to the means by which the server receives text data entered by the user and sends it to the generative AI model for analysis.

[0431] "Means for comparing candidates extracted by the generative AI model with a database" refers to a means for comparing grammatical errors, typos, and suspicious points pointed out by the generative AI model with information in a database and evaluating their reliability.

[0432] "Means for generating a list of issues and presenting it to the user" refers to a means for compiling correction suggestions obtained from a generative AI model or database matching into a list and displaying it to the user.

[0433] The "means for the user to select or input correction content" refers to the means by which the user selects the necessary correction from the presented list of issues or inputs the correction themselves.

[0434] The "means for saving the final text" is a means for saving the text after the user has completed the corrections in a database and associating it with the user's account information.

[0435] The "emotion recognition means for recognizing a user's emotions in real time" is a means for recognizing and analyzing the emotions of a user while the user is inputting text.

[0436] "Means for adjusting the analysis results of a generative AI model based on emotional information" refers to a means for optimizing the analysis results of a generative AI model by taking into account the user's emotions and providing more appropriate feedback.

[0437] The "means for reflecting emotion-based feedback in the points to be pointed out list" is a means for adjusting the feedback content displayed in the points to be pointed out list based on the user's emotion information.

[0438] This invention is a system that identifies errors and suspicious points in text created by users of a text generation service, and further recognizes the user's emotions to generate more accurate and reliable text. Specific components and operations of this system will be described below.

[0439] System Configuration

[0440] The system consists of the following main components:

[0441] 1. User Device

[0442] A device that allows a user to input text and receive feedback, such as a PC, smartphone, or tablet.

[0443] The input interface is built with front-end frameworks such as React or Vue.js.

[0444] 2. Server

[0445] This is a central processing unit that receives and analyzes text data entered by the user and executes the generative AI model. The server uses a cloud server such as AWS (Amazon Web Services) or Google Cloud Platform.

[0446] 3. Database

[0447] This is data storage for saving authentication information, text data, analysis results, etc. SQL databases and NoSQL databases are used.

[0448] 4. Generative AI Models

[0449] It is an artificial intelligence model used to analyze text for grammatical errors, typos, and suspicious points. Specifically, it uses large-scale language models such as GPT-3.

[0450] 5. Emotion Recognition Engine

[0451] It is an engine for recognizing user emotions in real time and adjusting the tone of the text. TensorFlow and OpenCV are often used.

[0452] 6. RAG technology

[0453] This technology evaluates the reliability of the findings extracted by the generative AI model by comparing them with public databases and closed data.

[0454] Operation explanation

[0455] Below, we will explain the specific steps involved in the operation of the system and how to implement them.

[0456] 1. The user logs in

[0457] Users use the login interface on their devices to enter their ID and password, which is then encrypted by SSL / TLS and sent to the server.

[0458] 2. The server authenticates the user

[0459] The server authenticates the user by checking the received ID and password against the authentication information in the database. If authentication is successful, a session ID is generated and returned to the user.

[0460] 3. The user enters text

[0461] The user enters the content of a business email or document into the device interface, and when the input is complete, clicks the send button.

[0462] 4. The device sends the entered text to the server

[0463] The entered text data is sent to the server in JSON format, along with the session ID.

[0464] 5. The server passes the text to the generative AI model

[0465] The received text data is passed to the generative AI model, and a request is sent to analyze it for grammatical errors, typos, and suspicious points.

[0466] 6. Generative AI models analyze text

[0467] A generative AI model analyzes text data and extracts insights using natural language processing (NLP) techniques.

[0468] 7. Emotion recognition engine recognizes user emotions

[0469] Recognize and analyze the user's emotions in real time while they are entering text.

[0470] 8. Adjust the analysis results of generative AI models based on emotional information

[0471] Based on the results of the emotion recognition engine, the analysis results are optimized according to the user's emotions.

[0472] 9. The server collates the data using RAG technology

[0473] The findings extracted by the generative AI model are verified using RAG technology to assess their reliability, for example by comparing them with public databases or closed data.

[0474] 10. The server generates a list of issues and sends it to the user's device.

[0475] A list of the verified issues is generated and transmitted to the user terminal based on this information.

[0476] 11. The user checks the issues and makes corrections

[0477] The user can review the list of issues and manually correct each one, or click on the suggested fixes to have them applied automatically.

[0478] 12. Check and save the final text

[0479] The user confirms the final text after correction and clicks the save button, and the server saves this final text in its database and associates it with the user's account.

[0480] Specific examples

[0481] For example, consider a case where a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, optimal feedback based on the user's emotions is provided. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[0482] Prompt Sentence Examples

[0483] Please review this email and let us know about any grammatical or typo errors or anything suspicious. Also, please adjust the tone of the email to a more calm one.

[0484] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0485] Step 1:

[0486] The user performs a login operation.

[0487] Input: User ID, Password

[0488] Operation: The user accesses the login screen of the text validation UI from the terminal and enters their ID and password.

[0489] Output: Login information is sent to the server.

[0490] Step 2:

[0491] The server authenticates the user

[0492] Input: Login information (user ID, password)

[0493] How it works: The server checks the received ID and password against the information in its database.

[0494] Output: If authentication is successful, a session ID is generated and returned to the user. At the same time, the user information is loaded into memory or cache.

[0495] Step 3:

[0496] The user enters text

[0497] Input: Text of business emails and documents

[0498] Action: A user enters text into the interface on their device and clicks the send button.

[0499] Output: The input text data is stored in the terminal.

[0500] Step 4:

[0501] The device sends the entered text to the server

[0502] Input: Text data, session ID

[0503] What happens: The device sends text data in JSON format to the server, which is sent as an HTTP POST request.

[0504] Output: The server receives the text data and the session ID.

[0505] Step 5:

[0506] The server passes the text to the generative AI model

[0507] Input: Text data

[0508] How it works: The server formats the text data and sends a request to the generative AI model endpoint.

[0509] Output: The generative AI model receives the text data.

[0510] Step 6:

[0511] Generative AI models parse text

[0512] Input: Text data

[0513] How it works: Generative AI models use natural language processing (NLP) techniques to analyze text data for grammatical errors, typos, and other suspicious content.

[0514] Output: A list of issues is generated and returned to the server.

[0515] Step 7:

[0516] Emotion recognition engine recognizes user emotions

[0517] Input: User input status, facial expression data, etc.

[0518] How it works: The emotion recognition engine recognizes the user's emotions in real time, analyzing images and input speed.

[0519] Output: User emotion data is generated and sent to the server.

[0520] Step 8:

[0521] Adjusting the analysis results of generative AI models based on emotional information

[0522] Input: List of points to be pointed out, user's emotional data

[0523] How it works: The server optimizes the analysis results of the generative AI model based on the results of the emotion recognition engine, adjusting the tone of the feedback, etc.

[0524] Output: A reconciled list of issues is generated.

[0525] Step 9:

[0526] The server collates data using RAG technology

[0527] Input: Reconciled issue list

[0528] How it works: The server uses RAG technology to match items in the findings list against public and closed databases.

[0529] Output: The reliability evaluation result is generated.

[0530] Step 10:

[0531] The server generates a list of issues and sends it to the user's device.

[0532] Input: Reliability evaluation results, adjustment results

[0533] Operation: The server generates a final list of issues and sends it to the user's terminal.

[0534] Output: The final issue list is displayed on the user's terminal.

[0535] Step 11:

[0536] The user checks the issues and makes corrections

[0537] Input: Final list of issues

[0538] How it works: The user reviews the list of issues on their device and either manually fixes each one or clicks on the suggested fixes to automatically apply them.

[0539] Output: The modified text data is saved to the terminal.

[0540] Step 12:

[0541] User confirms and saves the final text

[0542] Input: Modified text data

[0543] Action: The user reviews the final text after corrections and clicks the save button.

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

[0545] Step 13:

[0546] The server saves the final text

[0547] Input: Final text

[0548] How it works: The server stores the final text in a database and associates it with the user's account.

[0549] Output: A confirmation message is displayed to the user indicating that the save was successful.

[0550] (Application example 2)

[0551] 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."

[0552] Conventional text generation services have limited functionality for pointing out errors or suspicious points in text entered by users, making it difficult to provide feedback that takes emotions into account. Furthermore, even in multilingual translation, they have been unable to detect grammatical errors, typos, and suspicious points, or provide translations with a natural tone that reflects the user's emotions. Therefore, the objective of this invention is to provide an advanced text validation system that not only points out errors and suspicious points in text, but also includes multilingual translation and user sentiment analysis.

[0553] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0554] In this invention, the server includes a means for translating text entered by a user into a different language, a means for detecting grammatical errors, typos, and suspicious points in the translated text, and a means for analyzing the user's emotions and adjusting the tone of the text, thereby enabling the generation of grammatically correct, natural-looking text that reflects the user's emotions even in multiple languages.

[0555] "User" is a person who uses the system to input text and perform translation and validation.

[0556] "Login operation" refers to the authentication procedure required when a user accesses a system.

[0557] A "generative AI model" refers to an artificial intelligence model that analyzes received text data and extracts grammatical errors, typos, and suspicious points.

[0558] A "database" is a data storage device that stores information used to verify the reliability of candidates extracted by a generative AI model.

[0559] "Translation facility" refers to a facility that converts user-entered text into a different language.

[0560] "Grammar error detector" refers to a function that detects grammatical errors present in the translated text.

[0561] "Sentiment analysis means" refers to the function that analyzes the user's emotions when typing and adjusts the tone of the text.

[0562] The "list of points of criticism" is a list of information summarizing errors and corrections detected by the generative AI model, grammar error detection means, and sentiment analysis means.

[0563] The "means for selecting correction content" refers to a function that allows the user to select or newly input the content to be corrected based on the list of pointed out points.

[0564] "Final text" refers to text data that has been corrected by the user, finally confirmed, and saved.

[0565] "RAG technology" refers to a technology that compares candidates extracted by a generative AI model with existing data and evaluates their reliability.

[0566] MODE FOR CARRYING OUT THE INVENTION

[0567] System Program

[0568] The system that realizes this application example includes the following programs that perform a series of functions: user authentication, text input, text processing, sentiment analysis, grammar error detection, translation, presenting a list of corrections to the user, and saving the corrections.

[0569] Program processing

[0570] The server provides a means for users to log in when they access the system. The login operation uses a user authentication system using Django to verify the ID and password.

[0571] When the user enters text, the terminal sends the entered text data to the backend, which uses React for the frontend, making the user interface intuitive and responsive.

[0572] The server passes the received text data to a generative AI model, which can use BERT, for example, to detect grammatical errors, typos, and other suspicious aspects of the text. The output from the generative AI model is then sent back to the server.

[0573] The server then compares the candidates extracted by the generative AI model against a database of existing data and uses RAG technology to assess the reliability of the candidates.

[0574] Additionally, the server provides a means to analyze the user's emotions, using the Google Cloud Natural Language API to calculate an emotion score in real time as the user types.

[0575] Based on the results of the sentiment analysis, the server adjusts the tone of the text and performs the translation function, which uses the Google Cloud Translation API to translate the text entered by the user into different languages.

[0576] Based on the results of these processes, the server generates a list of issues and presents it to the user. The user can check the list on their device and carefully examine the details of each correction. If there are grammatical errors or misleading expressions, the user can correct them or select the correction suggested by the system.

[0577] Finally, the final text that the user has confirmed is stored on the server, and this final text is highly reliable and in line with the user's sentiment and intent.

[0578] Hardware and software used

[0579] Hardware: Smartphones, computers, and cloud-based server environments.

[0580] software:

[0581] Frontend: React

[0582] Backend: Django

[0583] Translation API: Google Cloud Translation API

[0584] Sentiment analysis: Google Cloud Natural Language API

[0585] Grammar Error Detection: BERT (TensorFlow / PyTorch)

[0586] Specific examples

[0587] For example, a user might enter the text "Good morning, how can I help you today?" in Japanese. The application translates this text into English and displays "Good morning, how can I help you today?". At the same time, it performs sentiment analysis to ensure the user maintains a friendly tone. The generative AI model detects grammatical errors, typos, and other suspicious points and presents a list of issues to the user.

[0588] Prompt example

[0589] User Input: "Good morning, how are you doing today?"

[0590] -> Translate to English

[0591] -> Detect grammar issues

[0592] -> Analyze sentiment score

[0593] In this way, the system provides grammatically accurate text generation across multiple languages ​​and feedback that reflects the user's sentiment, making user communication more efficient and reliable.

[0594] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0595] Step 1:

[0596] The server provides a login screen for the user to log in. The user enters their ID and password and clicks the login button. The server receives this input data and checks it against the authentication information in the database. If authentication is successful, a session ID is generated and the user information is loaded into the system.

[0597] Input: User ID and password

[0598] Output: Session ID and user information

[0599] Step 2:

[0600] The terminal displays a text input interface to the authenticated user. The user enters text and clicks the send button to send the input text to the server. The terminal then sends the entered text data to the server in JSON format, including the session ID.

[0601] Input: Text entered by the user

[0602] Output: JSON format text data and session ID

[0603] Step 3:

[0604] The server passes the received text data to a generative AI model. The generative AI model (e.g., BERT) analyzes the text data and extracts grammatical errors, typos, and suspicious points. The server then sends a request to the generative AI model's endpoint to obtain the analysis results.

[0605] Input: Received JSON format text data

[0606] Output: Analysis results from the generative AI model

[0607] Step 4:

[0608] The server compares the candidates extracted by the generative AI model with a database, using RAG technology to compare the candidates with public and closed databases and evaluate their reliability.

[0609] Input: Analysis results from generative AI model

[0610] Output: Matching results and confidence scores for each candidate

[0611] Step 5:

[0612] The server passes the received text data to the emotion engine to analyze the user's emotions. The emotion engine calculates the user's emotion score in real time using Google Cloud Natural Language API and other tools.

[0613] Input: Received text data

[0614] Output: Sentiment score

[0615] Step 6:

[0616] The server adjusts the tone of the text based on the results of sentiment analysis, and uses the Google Cloud Translation API to translate the text into different languages. Finally, it generates the evaluated translation results and adjusted text.

[0617] Input: Sentiment scores and received text data

[0618] Output: Adjusted translated text

[0619] Step 7:

[0620] The server generates and presents a list of issues to the user, including grammar, typos, suspicious points, and feedback based on sentiment scores, and sends this information to the front-end in an appropriate format.

[0621] Input: Evaluated translated text, sentiment score, and analysis results

[0622] Output: List of issues

[0623] Step 8:

[0624] The user can view a list of issues on their device, review the details of each correction, and make any necessary changes. They can either manually correct the text or click on the system's suggested corrections to automatically apply them.

[0625] Input: List of issues

[0626] Output: User's corrected text

[0627] Step 9:

[0628] The user finally checks the corrected text and clicks the save button. The server saves the final text in its database and associates it with the user's account.

[0629] Input: Final text as modified by the user

[0630] Output: Final text stored in the database

[0631] 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.

[0632] 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.

[0633] 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.

[0634] [Second embodiment]

[0635] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0636] 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.

[0637] 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).

[0638] 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.

[0639] 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.

[0640] 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).

[0641] 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.

[0642] 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.

[0643] 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.

[0644] 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.

[0645] In the smart glasses 214, 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.

[0646] 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."

[0647] This invention is a text validation UI system that points out errors and suspicious points in text created by users of a text generation service and generates accurate and reliable text. Specific embodiments of this system are described below.

[0648] Initial Setup

[0649] 1. The user logs in

[0650] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[0651] 2. The server authenticates the user

[0652] The server authenticates the user by checking the ID and password against a database. If authentication is successful, the user session is started and the user information is loaded into the system.

[0653] Text Input

[0654] 1. The user enters text

[0655] Users enter the text of business email drafts and documents into the device's interface.

[0656] 2. The device sends the entered text to the server

[0657] When the user clicks the input button, the terminal sends the entered text data to the server in an appropriate format, such as JSON format.

[0658] Text analytics

[0659] 1. The server passes the text to the generative AI model

[0660] The server passes the received text data to an existing generative AI model, which then receives the input text and begins analyzing it.

[0661] 2. A generative AI model analyzes the text

[0662] The generative AI model analyzes the content of the text and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[0663] Validation

[0664] 1. The server collates data using RAG technology

[0665] The server uses RAG technology to compare candidates extracted by the generative AI model with public and closed databases, and evaluates the reliability of the candidates.

[0666] 2. The server generates a list of issues

[0667] Based on the matching results, the server generates a list of issues, such as "misspelled company name" or "incorrect date format."

[0668] feedback

[0669] 1. The server sends the list of issues to the user's device.

[0670] The list of issues is sent to the user's device in an appropriate format, where the user can view it.

[0671] 2. The user confirms the issue

[0672] The user reviews the presented list of issues and carefully examines the points that need to be corrected.

[0673] 3. User makes corrections

[0674] Users can make corrections based on the suggestions, either manually or by simply clicking on the suggested corrections.

[0675] Final confirmation and saving

[0676] 1. The user checks the final text

[0677] The user checks the final text after correction and confirms the content.

[0678] 2. The user clicks the save button

[0679] When the user clicks the save button, the final edited text is sent to the server.

[0680] 3. The server saves the final text

[0681] The server stores the final text in a database and associates it with the user's account.

[0682] Specific examples

[0683] For example, suppose a user is composing a business email containing information about a new product. First, the user logs in to the system and enters the email content. The entered text is sent to the server and analyzed by the generative AI model. The user then receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the user saves the email with the corrections completed, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the diverse needs of enterprise users.

[0684] The processing flow will be explained below.

[0685] Step 1:

[0686] The user performs a login operation.

[0687] The user accesses the login screen of the text validation UI from their device, enters their ID and password, and clicks the login button to send the entered information to the server.

[0688] Step 2:

[0689] The server authenticates the user

[0690] The server checks the received ID and password against the authentication information in the database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[0691] Step 3:

[0692] The user enters text

[0693] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[0694] Step 4:

[0695] The device sends the entered text to the server

[0696] The device sends the text data entered by the user to the server in JSON format or an appropriate data format, including the session ID.

[0697] Step 5:

[0698] The server passes the text to the generative AI model

[0699] The server processes the received text data and sends a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[0700] Step 6:

[0701] Generative AI models parse text

[0702] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, and suspicious points, using natural language processing (NLP) techniques.

[0703] Step 7:

[0704] The server collates data using RAG technology

[0705] The server receives the candidates extracted by the generative AI model and compares them with public and closed databases using RAG technology, thereby evaluating the reliability of each candidate.

[0706] Step 8:

[0707] The server generates a list of issues

[0708] The server generates a list of issues based on the results of the match, including specific corrections such as grammar errors and incorrect data.

[0709] Step 9:

[0710] The server sends the list of issues to the user's device.

[0711] The server sends the list of issues to the user's device in JSON format, etc., which the user can view on the interface.

[0712] Step 10:

[0713] The user confirms the points

[0714] The user can check the list of issues on their device and carefully examine the details of each correction.

[0715] Step 11:

[0716] User makes corrections

[0717] The user can then make the necessary corrections based on the list, either by manually entering the corrections or by clicking on the suggested corrections provided to automatically incorporate them.

[0718] Step 12:

[0719] User confirms final text

[0720] After completing the corrections, the user checks the final text to ensure there are no problems with the content.

[0721] Step 13:

[0722] The user clicks the save button

[0723] When the user clicks the save button, the final text is sent to the server.

[0724] Step 14:

[0725] The server saves the final text

[0726] The server stores the final text in a database and associates it with the user's account, completing the process.

[0727] Example 1

[0728] 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."

[0729] Conventional text generation systems have difficulty efficiently detecting errors and inconsistencies and providing appropriate feedback to users. Furthermore, they lack an interface that allows users to review corrections and quickly and accurately correct the text. This can lead to problems such as a decrease in the reliability and quality of the generated text.

[0730] 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.

[0731] In this invention, the server includes a means for user authentication, a means for passing received text data to the generative AI model, and a means for comparing candidates extracted by the generative AI model with a database using RAG technology, which enables reliable evaluation of errors and suspicious points in the text detected by the generative AI model and provides accurate feedback and suggested corrections to the user.

[0732] "User authentication" is the process in which the server checks the authentication information (such as ID and password) entered by the user against a database to verify whether the user is legitimate.

[0733] A "terminal" is a device, such as a computer or smartphone, that a user uses to enter text or perform login operations.

[0734] "Text data" refers to character string information such as sentences or words entered by the user.

[0735] A "generative AI model" is an algorithm that uses natural language processing technology to analyze input text and extract grammatical errors, typos, and suspicious points.

[0736] "RAG technology" stands for Retrieval-Augmented Generation, and is a technology that compares generated candidates with existing databases to evaluate their reliability.

[0737] A "database" is a structured collection of data that is accessed by a server and stores user authentication information and data for verification.

[0738] The "list of points of concern" is a list of grammatical errors, typos, suspicious points, and other areas that need to be corrected, generated as a result of analysis by the generative AI model.

[0739] "Saving" is the process by which the server stores the text, once the user has completed the modifications, in a database for later access.

[0740] "Corrections" refer to corrections suggested by the generative AI model or corrections manually entered by the user.

[0741] This invention provides a system for pointing out errors and suspicious points in text created by a user using a text generation service, and generating accurate and reliable text. Specific embodiments of this system are described below.

[0742] Initial Setup

[0743] A user opens a browser on their device (e.g., a PC or smartphone) and accesses the text validation UI. They enter their ID and password on the login screen and perform the login operation. The server checks the ID and password entered by the user against the database to confirm whether they are a legitimate user. If authentication is successful, the server starts a user session, loads related user information, and provides the user with the dashboard or main operation screen.

[0744] Text Input

[0745] For example, when composing a business email containing new product information, a user can enter text into the device interface or copy and paste it from another source. When the user clicks the "Done" button, the device converts the entered text data into JSON format and sends it to the server via the API.

[0746] Text analytics

[0747] The server then passes the received text data to a generative AI model (e.g., GPT-3 or BERT). This generative AI model receives the input text and analyzes it to identify grammatical errors, typos, and other suspicious features. Specifically, it uses natural language processing techniques to tokenize each sentence and extract anomalous patterns.

[0748] Validation

[0749] The server uses Retrieval-Augmented Generation (RAG) technology to compare the suspicious points and candidates extracted by the generative AI model with public databases and internal closed data. This comparison evaluates the reliability and accuracy of the candidates. Based on the comparison results, the server creates a list of issues to be addressed. Examples of issues on this list include "misspellings in company names" and "date format errors."

[0750] feedback

[0751] The server converts the list of issues into an appropriate format (e.g., JSON) and sends it to the user's device. The user's device displays the received data on an interface. The user reviews the list of issues and carefully examines which parts are incorrect and how they should be corrected. The user corrects the text based on the issues, either manually or by applying the system's suggested corrections with one click.

[0752] Final confirmation and saving

[0753] The user then finalizes the edited text to ensure it is correct. Once the user clicks the save button, the final edited text is sent to the server, which stores the edited text in a database and associates it with the user's account.

[0754] Specific examples

[0755] For example, a user creates a business email containing new product information and enters the content into the system. The entered text is sent to the server and analyzed by the generative AI model. The user receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the corrected email is saved, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the needs of various enterprise users.

[0756] Prompt Sentence Examples

[0757] The following prompts can be fed to the generative AI model:

[0758] Enter the paragraph "Please check this document." Point out the errors in the text output by the generative AI model and provide the best possible corrections.

[0759] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0760] Step 1:

[0761] The user performs a login operation.

[0762] Input: The user enters their ID and password from the device.

[0763] Data processing: The terminal converts the entered authentication information into a standardized format (e.g., JSON).

[0764] Output: The device sends the authentication information to the server.

[0765] What happens: The user opens a browser, accesses the login screen, and enters the required credentials.

[0766] Step 2:

[0767] The server authenticates the user

[0768] Input: The authentication information sent from the device.

[0769] Data processing: The server queries the database for authentication information and verifies whether the user's ID and password are correct.

[0770] Output: Prepares the authentication result (success or failure).

[0771] Specific operation: The server references the database and starts a user session based on the authentication result.

[0772] Step 3:

[0773] The user enters text

[0774] Input: A user types text for a business email or document.

[0775] Data processing: The entered text is temporarily stored on the terminal as string data.

[0776] Output: The input text data.

[0777] What happens: The user types text directly into the device interface or copies and pastes it from another source.

[0778] Step 4:

[0779] The device sends the entered text to the server

[0780] Input: The user completes text input.

[0781] Data processing: The terminal converts the input text into JSON format.

[0782] Output: The converted text data is sent to the server.

[0783] Specific operation: When the user clicks the "Complete input" button, the terminal sends the text data to the server.

[0784] Step 5:

[0785] The server passes the text to the generative AI model

[0786] Input: Text data sent from the terminal.

[0787] Data processing: The server formats the received text data into a format suitable for the generative AI model.

[0788] Output: The formatted text data is fed into a generative AI model.

[0789] How it works: The server passes the text data to a generative AI model (e.g., GPT-3 or BERT).

[0790] Step 6:

[0791] Generative AI models parse text

[0792] Input: Text data passed from the server.

[0793] Data processing: The generative AI model analyzes the text and extracts grammatical errors, typos, and suspicious points.

[0794] Output: A list of candidates as the analysis result.

[0795] How it works: The generative AI model uses natural language processing techniques to analyze text for tokens and extract anomalous patterns.

[0796] Step 7:

[0797] The server collates data using RAG technology

[0798] Input: A list of candidates extracted by a generative AI model.

[0799] Data processing: The server uses RAG technology to check the candidate list against the database.

[0800] Output: A list of candidates with a confidence rating.

[0801] Specific operation: The server checks candidates against public databases and closed data.

[0802] Step 8:

[0803] The server generates a list of issues

[0804] Input: A list of candidates with confidence ratings.

[0805] Data processing: The server generates a list of findings.

[0806] Output: A list of issues.

[0807] Specific operation: Based on the matching results, the server creates a list of issues such as "misspelled company name" or "date format error."

[0808] Step 9:

[0809] The server sends the list of issues to the user's device.

[0810] Input: The generated issue list.

[0811] Data processing: The server converts the list of findings into an appropriate format (e.g., JSON).

[0812] Output: Send the converted issue list to the user's terminal.

[0813] Specific operation: The server sends a list of issues to the user's device and displays it on the user's interface.

[0814] Step 10:

[0815] The user confirms the points

[0816] Input: A list of issues displayed on the terminal.

[0817] Data processing: The user confirms the information pointed out.

[0818] Output: The confirmed issue.

[0819] Specific action: The user looks at the list of issues on the interface and identifies the problem.

[0820] Step 11:

[0821] User makes corrections

[0822] Input: The confirmed issue.

[0823] Data processing: The user corrects the text based on the suggestions.

[0824] Output: The corrected text.

[0825] What happens: The user manually corrects the text or clicks to apply the suggested correction.

[0826] Step 12:

[0827] User confirms final text

[0828] Input: The corrected text.

[0829] Data processing: The user makes a final check of the revised text.

[0830] Output: The final text that was checked.

[0831] Specific operation: The user reviews the revised text in detail and performs a final check.

[0832] Step 13:

[0833] The user clicks the save button

[0834] Input: Final text as verified.

[0835] Data processing: The user clicks the save button.

[0836] Output: The final text is sent to the server.

[0837] Specific behavior: When the user clicks the save button, the device sends the final text to the server.

[0838] Step 14:

[0839] The server saves the final text

[0840] Input: The final text sent from the terminal.

[0841] Data processing: The server stores the final text in a database.

[0842] Output: The final text saved.

[0843] What it does: The server stores the final text it receives in a database and associates it with the user's account.

[0844] (Application example 1)

[0845] 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."

[0846] When creating business texts or security documents, it is difficult for users to quickly and accurately identify grammatical errors, typographical errors, and even security risk-related errors to create reliable final text. In addition, real-time analysis and correction suggestions are required, while ensuring compliance with industry regulations and internal rules. An efficient system to solve these challenges is needed.

[0847] 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.

[0848] In this invention, the server includes means for a user to perform authentication operations, means for passing received text data to a generative AI model, means for comparing candidates extracted by the generative AI model with a data storage location, means for generating a list of points to be pointed out and presenting it to the user, means for the user to select or input corrections, means for saving the final text, means for analyzing voice-input text in real time and pointing out errors and security risks, means for displaying correction suggestions in real time, and means for checking compliance with industry regulations and internal rules.This enables users to check and correct text errors and security risks in real time and efficiently create reliable final text.

[0849] The "means by which a user performs authentication operations" is a function for verifying the identity of a user based on the authentication information entered by the user.

[0850] "Means for passing received text data to a generative AI model" refers to a function that transfers text data obtained from a user to a generative AI model that analyzes the data and suggests corrections.

[0851] "Means for comparing candidates extracted by the generative AI model with data storage locations" refers to a function that compares and verifies candidates for errors and corrections in text extracted by the generative AI model with an existing database.

[0852] "Means for generating a list of points of concern and presenting it to the user" refers to a function that creates a list of errors and suspicious points in the text based on the generative AI model and the matching results, and displays it to the user.

[0853] The "means for the user to select or input corrections" refers to an interface or function that allows the user to select or newly input corrections to the pointed out points in the text.

[0854] The "means for saving the final text" is a function for saving the final text after correction in a database or a specific storage location.

[0855] "Means for analyzing voice-input text in real time and pointing out errors and security risks" is a function that instantly analyzes voice-input text, detects grammatical errors and security risks, and points them out in real time.

[0856] "Means for displaying suggested revisions in real time" refers to a function that instantly presents users with suggested revisions based on the analysis results.

[0857] The "means for checking compliance with industry regulations and internal rules" is a function for checking whether the generated text complies with the relevant industry regulations and internal rules.

[0858] The "Secure Text Guard" system allows users to create text using voice input or text input, and detects grammatical errors, typos, suspicious points, and security risks in the text in real time and makes correction suggestions. Specific embodiments of the system are described below.

[0859] Hardware Configuration

[0860] 1. Smart Glasses

[0861] A device for voice input and text display.

[0862] It has a voice recognition function that converts voice input data into text.

[0863] 2. Server

[0864] A central server responsible for user authentication, text analysis, generating correction suggestions, and data storage.

[0865] Integrate with databases of data locations and industry regulations.

[0866] Software Configuration

[0867] 1. Generative AI Models

[0868] An AI model that analyzes user-entered text to detect grammatical errors, typos, and security risks.

[0869] 2. NLP technology

[0870] As a natural language processing technology, it is used to analyze input text and extract suspicious points.

[0871] 3. RAG technology

[0872] A technology that compares the results of text analysis with a database and evaluates their reliability.

[0873] System operation flow

[0874] 1. Authentication Operation

[0875] The user puts on the smart glasses and enters authentication information (user ID and password) using voice or touch.

[0876] The server checks this authentication information against a database to authenticate the user.

[0877] 2. Text Input

[0878] After successful authentication, the user can then use voice input or direct text input.

[0879] The entered text data is sent to the server in real time.

[0880] 3. Text analysis and suggested revisions

[0881] The server passes the received text to the generative AI model and begins analysis.

[0882] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[0883] Meanwhile, RAG technology is used to compare the analysis results with the data storage location and evaluate the reliability of the analysis.

[0884] 4. Corrections and Final Text Generation

[0885] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[0886] The user selects or inputs the corrections based on this list.

[0887] The server applies the modifications and generates the final text.

[0888] 5. Save

[0889] The final text, after corrections are complete, is stored in a data repository by the server.

[0890] Specific examples

[0891] For example, a user can use smart glasses to voice-input "important meeting materials." The system converts this into text in real time and begins analysis. If grammatical errors or risks of leaking confidential information are detected, they are immediately presented to the user and suggested corrections are displayed. The user can select or enter the suggested corrections, and the final text is generated.

[0892] Example prompt sentence:

[0893] "Analyzes the entered text and points out errors and suspicious points such as grammar mistakes, typos, and risk of leaking confidential information."

[0894] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0895] Step 1:

[0896] The user puts on the smart glasses and enters authentication information (user ID and password) by voice or touch.

[0897] Input: User ID, Password

[0898] Output: Authentication request

[0899] Step 2:

[0900] The server receives the authentication information and authenticates the user against a database.

[0901] Input:Authentication request

[0902] Data processing and calculation: Verification with authentication information in the database

[0903] Output: Authentication result (success / failure)

[0904] Step 3:

[0905] After successful authentication, the user can enter voice or text.

[0906] Input: Audio or text data

[0907] Output: Input text data

[0908] Step 4:

[0909] The entered text data is sent to the server in real time.

[0910] Input: Input text data

[0911] Output: Text data sent to the server

[0912] Step 5:

[0913] The server passes the received text data to the generative AI model and begins analysis.

[0914] Input: Text data sent to the server

[0915] Data processing and computation: Text analysis with generative AI models

[0916] Output: Analysis results (potential errors or suspicious points)

[0917] Step 6:

[0918] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[0919] Input: Analysis results (candidates for errors or suspicious points)

[0920] Data processing and calculation: Additional detection and correction suggestion generation using NLP technology

[0921] Output: List of suggested fixes

[0922] Step 7:

[0923] Using RAG technology, the analysis results are compared with the data storage location to evaluate the reliability of the analysis.

[0924] Input: list of suggested revisions

[0925] Data processing and calculation: Verification with data storage location

[0926] Output: Reliability evaluation results

[0927] Step 8:

[0928] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[0929] Input: Reliability evaluation results

[0930] Output: A list of issues and suggested fixes displayed on the user's screen

[0931] Step 9:

[0932] The user selects or enters corrections based on this list.

[0933] Input: List of issues and suggested fixes

[0934] Output: Modifications selected or entered

[0935] Step 10:

[0936] The server applies the modifications and generates the final text.

[0937] Input: The selected or entered correction

[0938] Data processing and calculation: Reflection of corrections

[0939] Output: Final text

[0940] Step 11:

[0941] The final text, after corrections are complete, is stored in a data repository by the server.

[0942] Input: Final text

[0943] Output: Saved text data

[0944] 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.

[0945] This invention is a text validation UI system that identifies errors and suspicious points in text created by users of a text generation service, recognizes the user's emotions, and generates more accurate and reliable text. A specific embodiment of this system is described below.

[0946] Initial Setup

[0947] 1. The user logs in

[0948] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[0949] 2. The server authenticates the user

[0950] The server authenticates the user by checking the ID and password entered by the user against the authentication information in the database. If authentication is successful, a user session is started, a session ID is generated, and returned to the user. At the same time, the user information is loaded into the system.

[0951] Text Input

[0952] 1. The user enters text

[0953] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[0954] 2. The device sends the entered text to the server

[0955] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[0956] Text analytics

[0957] 1. The server passes the text to the generative AI model

[0958] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[0959] 2. A generative AI model analyzes the text

[0960] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[0961] emotion recognition

[0962] 1. Emotion engine recognizes user emotions

[0963] The emotion engine recognizes users' emotions in real time as they type, and this emotional information is reflected in the tone of the text and suggested corrections.

[0964] 2. The emotion engine uses emotional information for analysis

[0965] The emotion engine analyzes the user's emotions as they type and combines them with the results of a generative AI model, which adjusts the tone of the text.

[0966] Validation

[0967] 1. The server collates data using RAG technology

[0968] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[0969] 2. The server generates a list of issues

[0970] Based on the matching results, the server generates a list of issues to be fixed, including specific corrections such as grammar errors and incorrect data.

[0971] feedback

[0972] 1. The server sends the list of issues to the user's device.

[0973] The list of issues is sent to the user's terminal in an appropriate format, and the user can view it on the interface.

[0974] 2. The user confirms the issue

[0975] Users can review the list of issues and review the details of each correction on their device, and the emotion engine will provide feedback tailored to the user's emotions.

[0976] 3. User makes corrections

[0977] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[0978] Final confirmation and saving

[0979] 1. The user checks the final text

[0980] The user checks the final text after corrections and confirms that there are no problems with the content.

[0981] 2. The user clicks the save button

[0982] When the user clicks the save button, the final text is sent to the server.

[0983] 3. The server saves the final text

[0984] The server stores the final text in a database and associates it with the user's account.

[0985] Specific examples

[0986] For example, imagine a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, it provides optimal feedback based on the user's emotions. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[0987] The processing flow will be explained below.

[0988] Step 1:

[0989] The user performs a login operation.

[0990] The user accesses the login screen of the text validation UI from the terminal, enters their ID and password, and clicks the login button to send the entered information to the server.

[0991] Step 2:

[0992] The server authenticates the user

[0993] The server checks the received ID and password against the authentication information in its database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[0994] Step 3:

[0995] The user enters text

[0996] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[0997] Step 4:

[0998] The device sends the entered text to the server

[0999] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[1000] Step 5:

[1001] The server passes the text to the generative AI model

[1002] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[1003] Step 6:

[1004] Generative AI models parse text

[1005] The generative AI model analyzes the text data, extracts candidates for grammatical errors, typos, and suspicious points, and returns the results to the server. This analysis uses natural language processing (NLP) technology.

[1006] Step 7:

[1007] Emotion engine recognizes user emotions

[1008] The emotion engine collects and analyzes real-time emotion data while the user is entering text, thereby identifying the user's emotional state.

[1009] Step 8:

[1010] The emotion engine uses emotional information for analysis

[1011] The emotion engine combines the results returned by the generative AI model with the user's emotional information, allowing the tone of the text and suggested revisions to be adjusted based on the user's emotions.

[1012] Step 9:

[1013] The server collates data using RAG technology

[1014] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[1015] Step 10:

[1016] The server generates a list of issues

[1017] Based on the results of the match, the server generates a list of issues, including specific corrections such as "grammatical errors" or "incorrect data."

[1018] Step 11:

[1019] The server sends the list of issues to the user's device.

[1020] The list of issues is sent to the user's device in an appropriate format, where the user can view it in the interface.

[1021] Step 12:

[1022] The user confirms the points

[1023] Users can view a list of issues and review the details of each correction on their device, and the emotion engine provides feedback tailored to the user's emotions.

[1024] Step 13:

[1025] User makes corrections

[1026] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[1027] Step 14:

[1028] User confirms final text

[1029] The user checks the final text after corrections and confirms that there are no problems with the content.

[1030] Step 15:

[1031] The user clicks the save button

[1032] When the user clicks the save button, the final text is sent to the server.

[1033] Step 16:

[1034] The server saves the final text

[1035] The server stores the final text in a database and associates it with the user's account, completing the process.

[1036] Example 2

[1037] 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."

[1038] Modern text generation services require a method to effectively identify errors and suspicious points in user-generated text. However, current systems are unable to provide feedback that takes into account the user's emotions, and the reliability of the generated feedback is not sufficiently guaranteed. As a result, users spend a lot of time correcting the text, and the quality of the final text is difficult to guarantee.

[1039] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1040] In this invention, the server includes an emotion recognition means for recognizing a user's emotions in real time, a means for adjusting the analysis results of the generative AI model based on the emotion information, and a means for reflecting emotion-based feedback in the list of points to be pointed out. This makes it possible to provide feedback that takes the user's emotions into consideration and improve the reliability and quality of the text.

[1041] "Means for users to log in" refers to the means by which users access the system and perform authentication using their own account information.

[1042] "Means for passing received text data to the generative AI model" refers to the means by which the server receives text data entered by the user and sends it to the generative AI model for analysis.

[1043] "Means for comparing candidates extracted by the generative AI model with a database" refers to a means for comparing grammatical errors, typos, and suspicious points pointed out by the generative AI model with information in a database and evaluating their reliability.

[1044] "Means for generating a list of issues and presenting it to the user" refers to a means for compiling correction suggestions obtained from a generative AI model or database matching into a list and displaying it to the user.

[1045] The "means for the user to select or input correction content" refers to the means by which the user selects the necessary correction from the presented list of issues or inputs the correction themselves.

[1046] The "means for saving the final text" is a means for saving the text after the user has completed the corrections in a database and associating it with the user's account information.

[1047] The "emotion recognition means for recognizing a user's emotions in real time" is a means for recognizing and analyzing the emotions of a user while the user is inputting text.

[1048] "Means for adjusting the analysis results of a generative AI model based on emotional information" refers to a means for optimizing the analysis results of a generative AI model by taking into account the user's emotions and providing more appropriate feedback.

[1049] The "means for reflecting emotion-based feedback in the points to be pointed out list" is a means for adjusting the feedback content displayed in the points to be pointed out list based on the user's emotion information.

[1050] This invention is a system that identifies errors and suspicious points in text created by users of a text generation service, and further recognizes the user's emotions to generate more accurate and reliable text. Specific components and operations of this system will be described below.

[1051] System Configuration

[1052] The system consists of the following main components:

[1053] 1. User Device

[1054] A device that allows a user to input text and receive feedback, such as a PC, smartphone, or tablet.

[1055] The input interface is built with front-end frameworks such as React or Vue.js.

[1056] 2. Server

[1057] This is a central processing unit that receives and analyzes text data entered by the user and executes the generative AI model. The server uses a cloud server such as AWS (Amazon Web Services) or Google Cloud Platform.

[1058] 3. Database

[1059] This is data storage for saving authentication information, text data, analysis results, etc. SQL databases and NoSQL databases are used.

[1060] 4. Generative AI Models

[1061] It is an artificial intelligence model used to analyze text for grammatical errors, typos, and suspicious points. Specifically, it uses large-scale language models such as GPT-3.

[1062] 5. Emotion Recognition Engine

[1063] It is an engine for recognizing user emotions in real time and adjusting the tone of the text. TensorFlow and OpenCV are often used.

[1064] 6. RAG technology

[1065] This technology evaluates the reliability of the findings extracted by the generative AI model by comparing them with public databases and closed data.

[1066] Operation explanation

[1067] Below, we will explain the specific steps involved in the operation of the system and how to implement them.

[1068] 1. The user logs in

[1069] Users use the login interface on their devices to enter their ID and password, which is then encrypted by SSL / TLS and sent to the server.

[1070] 2. The server authenticates the user

[1071] The server authenticates the user by checking the received ID and password against the authentication information in the database. If authentication is successful, a session ID is generated and returned to the user.

[1072] 3. The user enters text

[1073] The user enters the content of a business email or document into the device interface, and when the input is complete, clicks the send button.

[1074] 4. The device sends the entered text to the server

[1075] The entered text data is sent to the server in JSON format, along with the session ID.

[1076] 5. The server passes the text to the generative AI model

[1077] The received text data is passed to the generative AI model, and a request is sent to analyze it for grammatical errors, typos, and suspicious points.

[1078] 6. Generative AI models analyze text

[1079] A generative AI model analyzes text data and extracts insights using natural language processing (NLP) techniques.

[1080] 7. Emotion recognition engine recognizes user emotions

[1081] Recognize and analyze the user's emotions in real time while they are entering text.

[1082] 8. Adjust the analysis results of generative AI models based on emotional information

[1083] Based on the results of the emotion recognition engine, the analysis results are optimized according to the user's emotions.

[1084] 9. The server collates the data using RAG technology

[1085] The findings extracted by the generative AI model are verified using RAG technology to assess their reliability, for example by comparing them with public databases or closed data.

[1086] 10. The server generates a list of issues and sends it to the user's device.

[1087] A list of the verified issues is generated and transmitted to the user terminal based on this information.

[1088] 11. The user checks the issues and makes corrections

[1089] The user can review the list of issues and manually correct each one, or click on the suggested fixes to have them applied automatically.

[1090] 12. Check and save the final text

[1091] The user confirms the final text after correction and clicks the save button, and the server saves this final text in its database and associates it with the user's account.

[1092] Specific examples

[1093] For example, consider a case where a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, optimal feedback based on the user's emotions is provided. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[1094] Prompt Sentence Examples

[1095] Please review this email and let us know about any grammatical or typo errors or anything suspicious. Also, please adjust the tone of the email to a more calm one.

[1096] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1097] Step 1:

[1098] The user performs a login operation.

[1099] Input: User ID, Password

[1100] Operation: The user accesses the login screen of the text validation UI from the terminal and enters their ID and password.

[1101] Output: Login information is sent to the server.

[1102] Step 2:

[1103] The server authenticates the user

[1104] Input: Login information (user ID, password)

[1105] How it works: The server checks the received ID and password against the information in its database.

[1106] Output: If authentication is successful, a session ID is generated and returned to the user. At the same time, the user information is loaded into memory or cache.

[1107] Step 3:

[1108] The user enters text

[1109] Input: Text of business emails and documents

[1110] Action: A user enters text into the interface on their device and clicks the send button.

[1111] Output: The input text data is stored in the terminal.

[1112] Step 4:

[1113] The device sends the entered text to the server

[1114] Input: Text data, session ID

[1115] What happens: The device sends text data in JSON format to the server, which is sent as an HTTP POST request.

[1116] Output: The server receives the text data and the session ID.

[1117] Step 5:

[1118] The server passes the text to the generative AI model

[1119] Input: Text data

[1120] How it works: The server formats the text data and sends a request to the generative AI model endpoint.

[1121] Output: The generative AI model receives the text data.

[1122] Step 6:

[1123] Generative AI models parse text

[1124] Input: Text data

[1125] How it works: Generative AI models use natural language processing (NLP) techniques to analyze text data for grammatical errors, typos, and other suspicious content.

[1126] Output: A list of issues is generated and returned to the server.

[1127] Step 7:

[1128] Emotion recognition engine recognizes user emotions

[1129] Input: User input status, facial expression data, etc.

[1130] How it works: The emotion recognition engine recognizes the user's emotions in real time, analyzing images and input speed.

[1131] Output: User emotion data is generated and sent to the server.

[1132] Step 8:

[1133] Adjusting the analysis results of generative AI models based on emotional information

[1134] Input: List of points to be pointed out, user's emotional data

[1135] How it works: The server optimizes the analysis results of the generative AI model based on the results of the emotion recognition engine, adjusting the tone of the feedback, etc.

[1136] Output: A reconciled list of issues is generated.

[1137] Step 9:

[1138] The server collates data using RAG technology

[1139] Input: Reconciled issue list

[1140] How it works: The server uses RAG technology to match items in the findings list against public and closed databases.

[1141] Output: The reliability evaluation result is generated.

[1142] Step 10:

[1143] The server generates a list of issues and sends it to the user's device.

[1144] Input: Reliability evaluation results, adjustment results

[1145] Operation: The server generates a final list of issues and sends it to the user's terminal.

[1146] Output: The final issue list is displayed on the user's terminal.

[1147] Step 11:

[1148] The user checks the issues and makes corrections

[1149] Input: Final list of issues

[1150] How it works: The user reviews the list of issues on their device and either manually fixes each one or clicks on the suggested fixes to automatically apply them.

[1151] Output: The modified text data is saved to the terminal.

[1152] Step 12:

[1153] User confirms and saves the final text

[1154] Input: Modified text data

[1155] Action: The user reviews the final text after corrections and clicks the save button.

[1156] Output: The final text is sent to the server.

[1157] Step 13:

[1158] The server saves the final text

[1159] Input: Final text

[1160] How it works: The server stores the final text in a database and associates it with the user's account.

[1161] Output: A confirmation message is displayed to the user indicating that the save was successful.

[1162] (Application example 2)

[1163] 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."

[1164] Conventional text generation services have limited functionality for pointing out errors or suspicious points in text entered by users, making it difficult to provide feedback that takes emotions into account. Furthermore, even in multilingual translation, they have been unable to detect grammatical errors, typos, and suspicious points, or provide translations with a natural tone that reflects the user's emotions. Therefore, the objective of this invention is to provide an advanced text validation system that not only points out errors and suspicious points in text, but also includes multilingual translation and user sentiment analysis.

[1165] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1166] In this invention, the server includes a means for translating text entered by a user into a different language, a means for detecting grammatical errors, typos, and suspicious points in the translated text, and a means for analyzing the user's emotions and adjusting the tone of the text, thereby enabling the generation of grammatically correct, natural-looking text that reflects the user's emotions even in multiple languages.

[1167] "User" is a person who uses the system to input text and perform translation and validation.

[1168] "Login operation" refers to the authentication procedure required when a user accesses a system.

[1169] A "generative AI model" refers to an artificial intelligence model that analyzes received text data and extracts grammatical errors, typos, and suspicious points.

[1170] A "database" is a data storage device that stores information used to verify the reliability of candidates extracted by a generative AI model.

[1171] "Translation facility" refers to a facility that converts user-entered text into a different language.

[1172] "Grammar error detector" refers to a function that detects grammatical errors present in the translated text.

[1173] "Sentiment analysis means" refers to the function that analyzes the user's emotions when typing and adjusts the tone of the text.

[1174] The "list of points of criticism" is a list of information summarizing errors and corrections detected by the generative AI model, grammar error detection means, and sentiment analysis means.

[1175] The "means for selecting correction content" refers to a function that allows the user to select or newly input the content to be corrected based on the list of pointed out points.

[1176] "Final text" refers to text data that has been corrected by the user, finally confirmed, and saved.

[1177] "RAG technology" refers to a technology that compares candidates extracted by a generative AI model with existing data and evaluates their reliability.

[1178] MODE FOR CARRYING OUT THE INVENTION

[1179] System Program

[1180] The system that realizes this application example includes the following programs that perform a series of functions: user authentication, text input, text processing, sentiment analysis, grammar error detection, translation, presenting a list of corrections to the user, and saving the corrections.

[1181] Program processing

[1182] The server provides a means for users to log in when they access the system. The login operation uses a user authentication system using Django to verify the ID and password.

[1183] When the user enters text, the terminal sends the entered text data to the backend, which uses React for the frontend, making the user interface intuitive and responsive.

[1184] The server passes the received text data to a generative AI model, which can use BERT, for example, to detect grammatical errors, typos, and other suspicious aspects of the text. The output from the generative AI model is then sent back to the server.

[1185] The server then compares the candidates extracted by the generative AI model against a database of existing data and uses RAG technology to assess the reliability of the candidates.

[1186] Additionally, the server provides a means to analyze the user's emotions, using the Google Cloud Natural Language API to calculate an emotion score in real time as the user types.

[1187] Based on the results of the sentiment analysis, the server adjusts the tone of the text and performs the translation function, which uses the Google Cloud Translation API to translate the text entered by the user into different languages.

[1188] Based on the results of these processes, the server generates a list of issues and presents it to the user. The user can check the list on their device and carefully examine the details of each correction. If there are grammatical errors or misleading expressions, the user can correct them or select the correction suggested by the system.

[1189] Finally, the final text that the user has confirmed is stored on the server, and this final text is highly reliable and in line with the user's sentiment and intent.

[1190] Hardware and software used

[1191] Hardware: Smartphones, computers, and cloud-based server environments.

[1192] software:

[1193] Frontend: React

[1194] Backend: Django

[1195] Translation API: Google Cloud Translation API

[1196] Sentiment analysis: Google Cloud Natural Language API

[1197] Grammar Error Detection: BERT (TensorFlow / PyTorch)

[1198] Specific examples

[1199] For example, a user might enter the text "Good morning, how can I help you today?" in Japanese. The application translates this text into English and displays "Good morning, how can I help you today?". At the same time, it performs sentiment analysis to ensure the user maintains a friendly tone. The generative AI model detects grammatical errors, typos, and other suspicious points and presents a list of issues to the user.

[1200] Prompt example

[1201] User Input: "Good morning, how are you doing today?"

[1202] -> Translate to English

[1203] -> Detect grammar issues

[1204] -> Analyze sentiment score

[1205] In this way, the system provides grammatically accurate text generation across multiple languages ​​and feedback that reflects the user's sentiment, making user communication more efficient and reliable.

[1206] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1207] Step 1:

[1208] The server provides a login screen for the user to log in. The user enters their ID and password and clicks the login button. The server receives this input data and checks it against the authentication information in the database. If authentication is successful, a session ID is generated and the user information is loaded into the system.

[1209] Input: User ID and password

[1210] Output: Session ID and user information

[1211] Step 2:

[1212] The terminal displays a text input interface to the authenticated user. The user enters text and clicks the send button to send the input text to the server. The terminal then sends the entered text data to the server in JSON format, including the session ID.

[1213] Input: Text entered by the user

[1214] Output: JSON format text data and session ID

[1215] Step 3:

[1216] The server passes the received text data to a generative AI model. The generative AI model (e.g., BERT) analyzes the text data and extracts grammatical errors, typos, and suspicious points. The server then sends a request to the generative AI model's endpoint to obtain the analysis results.

[1217] Input: Received JSON format text data

[1218] Output: Analysis results from the generative AI model

[1219] Step 4:

[1220] The server compares the candidates extracted by the generative AI model with a database, using RAG technology to compare the candidates with public and closed databases and evaluate their reliability.

[1221] Input: Analysis results from generative AI model

[1222] Output: Matching results and confidence scores for each candidate

[1223] Step 5:

[1224] The server passes the received text data to the emotion engine to analyze the user's emotions. The emotion engine calculates the user's emotion score in real time using Google Cloud Natural Language API and other tools.

[1225] Input: Received text data

[1226] Output: Sentiment score

[1227] Step 6:

[1228] The server adjusts the tone of the text based on the results of sentiment analysis, and uses the Google Cloud Translation API to translate the text into different languages. Finally, it generates the evaluated translation results and adjusted text.

[1229] Input: Sentiment scores and received text data

[1230] Output: Adjusted translated text

[1231] Step 7:

[1232] The server generates and presents a list of issues to the user, including grammar, typos, suspicious points, and feedback based on sentiment scores, and sends this information to the front-end in an appropriate format.

[1233] Input: Evaluated translated text, sentiment score, and analysis results

[1234] Output: List of issues

[1235] Step 8:

[1236] The user can view a list of issues on their device, review the details of each correction, and make any necessary changes. They can either manually correct the text or click on the system's suggested corrections to automatically apply them.

[1237] Input: List of issues

[1238] Output: User's corrected text

[1239] Step 9:

[1240] The user finally checks the corrected text and clicks the save button. The server saves the final text in its database and associates it with the user's account.

[1241] Input: Final text as modified by the user

[1242] Output: Final text stored in the database

[1243] 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.

[1244] 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.

[1245] 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.

[1246] [Third embodiment]

[1247] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1248] 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.

[1249] 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).

[1250] 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.

[1251] 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.

[1252] 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).

[1253] 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.

[1254] 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.

[1255] 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.

[1256] 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.

[1257] 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.

[1258] 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."

[1259] This invention is a text validation UI system that points out errors and suspicious points in text created by users of a text generation service and generates accurate and reliable text. Specific embodiments of this system are described below.

[1260] Initial Setup

[1261] 1. The user logs in

[1262] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[1263] 2. The server authenticates the user

[1264] The server authenticates the user by checking the ID and password against a database. If authentication is successful, the user session is started and the user information is loaded into the system.

[1265] Text Input

[1266] 1. The user enters text

[1267] Users enter the text of business email drafts and documents into the device's interface.

[1268] 2. The device sends the entered text to the server

[1269] When the user clicks the input button, the terminal sends the entered text data to the server in an appropriate format, such as JSON format.

[1270] Text analytics

[1271] 1. The server passes the text to the generative AI model

[1272] The server passes the received text data to an existing generative AI model, which then receives the input text and begins analyzing it.

[1273] 2. A generative AI model analyzes the text

[1274] The generative AI model analyzes the content of the text and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[1275] Validation

[1276] 1. The server collates data using RAG technology

[1277] The server uses RAG technology to compare candidates extracted by the generative AI model with public and closed databases, and evaluates the reliability of the candidates.

[1278] 2. The server generates a list of issues

[1279] Based on the matching results, the server generates a list of issues, such as "misspelled company name" or "incorrect date format."

[1280] feedback

[1281] 1. The server sends the list of issues to the user's device.

[1282] The list of issues is sent to the user's device in an appropriate format, where the user can view it.

[1283] 2. The user confirms the issue

[1284] The user reviews the presented list of issues and carefully examines the points that need to be corrected.

[1285] 3. User makes corrections

[1286] Users can make corrections based on the suggestions, either manually or by simply clicking on the suggested corrections.

[1287] Final confirmation and saving

[1288] 1. The user checks the final text

[1289] The user checks the final text after correction and confirms the content.

[1290] 2. The user clicks the save button

[1291] When the user clicks the save button, the final edited text is sent to the server.

[1292] 3. The server saves the final text

[1293] The server stores the final text in a database and associates it with the user's account.

[1294] Specific examples

[1295] For example, suppose a user is composing a business email containing information about a new product. First, the user logs in to the system and enters the email content. The entered text is sent to the server and analyzed by the generative AI model. The user then receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the user saves the email with the corrections completed, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the diverse needs of enterprise users.

[1296] The processing flow will be explained below.

[1297] Step 1:

[1298] The user performs a login operation.

[1299] The user accesses the login screen of the text validation UI from their device, enters their ID and password, and clicks the login button to send the entered information to the server.

[1300] Step 2:

[1301] The server authenticates the user

[1302] The server checks the received ID and password against the authentication information in the database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[1303] Step 3:

[1304] The user enters text

[1305] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[1306] Step 4:

[1307] The device sends the entered text to the server

[1308] The device sends the text data entered by the user to the server in JSON format or an appropriate data format, including the session ID.

[1309] Step 5:

[1310] The server passes the text to the generative AI model

[1311] The server processes the received text data and sends a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[1312] Step 6:

[1313] Generative AI models parse text

[1314] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, and suspicious points, using natural language processing (NLP) techniques.

[1315] Step 7:

[1316] The server collates data using RAG technology

[1317] The server receives the candidates extracted by the generative AI model and compares them with public and closed databases using RAG technology, thereby evaluating the reliability of each candidate.

[1318] Step 8:

[1319] The server generates a list of issues

[1320] The server generates a list of issues based on the results of the match, including specific corrections such as grammar errors and incorrect data.

[1321] Step 9:

[1322] The server sends the list of issues to the user's device.

[1323] The server sends the list of issues to the user's device in JSON format, etc., which the user can view on the interface.

[1324] Step 10:

[1325] The user confirms the points

[1326] The user can check the list of issues on their device and carefully examine the details of each correction.

[1327] Step 11:

[1328] User makes corrections

[1329] The user can then make the necessary corrections based on the list, either by manually entering the corrections or by clicking on the suggested corrections provided to automatically incorporate them.

[1330] Step 12:

[1331] User confirms final text

[1332] After completing the corrections, the user checks the final text to ensure there are no problems with the content.

[1333] Step 13:

[1334] The user clicks the save button

[1335] When the user clicks the save button, the final text is sent to the server.

[1336] Step 14:

[1337] The server saves the final text

[1338] The server stores the final text in a database and associates it with the user's account, completing the process.

[1339] Example 1

[1340] 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."

[1341] Conventional text generation systems have difficulty efficiently detecting errors and inconsistencies and providing appropriate feedback to users. Furthermore, they lack an interface that allows users to review corrections and quickly and accurately correct the text. This can lead to problems such as a decrease in the reliability and quality of the generated text.

[1342] 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.

[1343] In this invention, the server includes a means for user authentication, a means for passing received text data to the generative AI model, and a means for comparing candidates extracted by the generative AI model with a database using RAG technology, which enables reliable evaluation of errors and suspicious points in the text detected by the generative AI model and provides accurate feedback and suggested corrections to the user.

[1344] "User authentication" is the process in which the server checks the authentication information (such as ID and password) entered by the user against a database to verify whether the user is legitimate.

[1345] A "terminal" is a device, such as a computer or smartphone, that a user uses to enter text or perform login operations.

[1346] "Text data" refers to character string information such as sentences or words entered by the user.

[1347] A "generative AI model" is an algorithm that uses natural language processing technology to analyze input text and extract grammatical errors, typos, and suspicious points.

[1348] "RAG technology" stands for Retrieval-Augmented Generation, and is a technology that compares generated candidates with existing databases to evaluate their reliability.

[1349] A "database" is a structured collection of data that is accessed by a server and stores user authentication information and data for verification.

[1350] The "list of points of concern" is a list of grammatical errors, typos, suspicious points, and other areas that need to be corrected, generated as a result of analysis by the generative AI model.

[1351] "Saving" is the process by which the server stores the text, once the user has completed the modifications, in a database for later access.

[1352] "Corrections" refer to corrections suggested by the generative AI model or corrections manually entered by the user.

[1353] This invention provides a system for pointing out errors and suspicious points in text created by a user using a text generation service, and generating accurate and reliable text. Specific embodiments of this system are described below.

[1354] Initial Setup

[1355] A user opens a browser on their device (e.g., a PC or smartphone) and accesses the text validation UI. They enter their ID and password on the login screen and perform the login operation. The server checks the ID and password entered by the user against the database to confirm whether they are a legitimate user. If authentication is successful, the server starts a user session, loads related user information, and provides the user with the dashboard or main operation screen.

[1356] Text Input

[1357] For example, when composing a business email containing new product information, a user can enter text into the device interface or copy and paste it from another source. When the user clicks the "Done" button, the device converts the entered text data into JSON format and sends it to the server via the API.

[1358] Text analytics

[1359] The server then passes the received text data to a generative AI model (e.g., GPT-3 or BERT). This generative AI model receives the input text and analyzes it to identify grammatical errors, typos, and other suspicious features. Specifically, it uses natural language processing techniques to tokenize each sentence and extract anomalous patterns.

[1360] Validation

[1361] The server uses Retrieval-Augmented Generation (RAG) technology to compare the suspicious points and candidates extracted by the generative AI model with public databases and internal closed data. This comparison evaluates the reliability and accuracy of the candidates. Based on the comparison results, the server creates a list of issues to be addressed. Examples of issues on this list include "misspellings in company names" and "date format errors."

[1362] feedback

[1363] The server converts the list of issues into an appropriate format (e.g., JSON) and sends it to the user's device. The user's device displays the received data on an interface. The user reviews the list of issues and carefully examines which parts are incorrect and how they should be corrected. The user corrects the text based on the issues, either manually or by applying the system's suggested corrections with one click.

[1364] Final confirmation and saving

[1365] The user then finalizes the edited text to ensure it is correct. Once the user clicks the save button, the final edited text is sent to the server, which stores the edited text in a database and associates it with the user's account.

[1366] Specific examples

[1367] For example, a user creates a business email containing new product information and enters the content into the system. The entered text is sent to the server and analyzed by the generative AI model. The user receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the corrected email is saved, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the needs of various enterprise users.

[1368] Prompt Sentence Examples

[1369] The following prompts can be fed to the generative AI model:

[1370] Enter the paragraph "Please check this document." Point out the errors in the text output by the generative AI model and provide the best possible corrections.

[1371] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1372] Step 1:

[1373] The user performs a login operation.

[1374] Input: The user enters their ID and password from the device.

[1375] Data processing: The terminal converts the entered authentication information into a standardized format (e.g., JSON).

[1376] Output: The device sends the authentication information to the server.

[1377] What happens: The user opens a browser, accesses the login screen, and enters the required credentials.

[1378] Step 2:

[1379] The server authenticates the user

[1380] Input: The authentication information sent from the device.

[1381] Data processing: The server queries the database for authentication information and verifies whether the user's ID and password are correct.

[1382] Output: Prepares the authentication result (success or failure).

[1383] Specific operation: The server references the database and starts a user session based on the authentication result.

[1384] Step 3:

[1385] The user enters text

[1386] Input: A user types text for a business email or document.

[1387] Data processing: The entered text is temporarily stored on the terminal as string data.

[1388] Output: The input text data.

[1389] What happens: The user types text directly into the device interface or copies and pastes it from another source.

[1390] Step 4:

[1391] The device sends the entered text to the server

[1392] Input: The user completes text input.

[1393] Data processing: The terminal converts the input text into JSON format.

[1394] Output: The converted text data is sent to the server.

[1395] Specific operation: When the user clicks the "Complete input" button, the terminal sends the text data to the server.

[1396] Step 5:

[1397] The server passes the text to the generative AI model

[1398] Input: Text data sent from the terminal.

[1399] Data processing: The server formats the received text data into a format suitable for the generative AI model.

[1400] Output: The formatted text data is fed into a generative AI model.

[1401] How it works: The server passes the text data to a generative AI model (e.g., GPT-3 or BERT).

[1402] Step 6:

[1403] Generative AI models parse text

[1404] Input: Text data passed from the server.

[1405] Data processing: The generative AI model analyzes the text and extracts grammatical errors, typos, and suspicious points.

[1406] Output: A list of candidates as the analysis result.

[1407] How it works: The generative AI model uses natural language processing techniques to analyze text for tokens and extract anomalous patterns.

[1408] Step 7:

[1409] The server collates data using RAG technology

[1410] Input: A list of candidates extracted by a generative AI model.

[1411] Data processing: The server uses RAG technology to check the candidate list against the database.

[1412] Output: A list of candidates with a confidence rating.

[1413] Specific operation: The server checks candidates against public databases and closed data.

[1414] Step 8:

[1415] The server generates a list of issues

[1416] Input: A list of candidates with confidence ratings.

[1417] Data processing: The server generates a list of findings.

[1418] Output: A list of issues.

[1419] Specific operation: Based on the matching results, the server creates a list of issues such as "misspelled company name" or "date format error."

[1420] Step 9:

[1421] The server sends the list of issues to the user's device.

[1422] Input: The generated issue list.

[1423] Data processing: The server converts the list of findings into an appropriate format (e.g., JSON).

[1424] Output: Send the converted issue list to the user's terminal.

[1425] Specific operation: The server sends a list of issues to the user's device and displays it on the user's interface.

[1426] Step 10:

[1427] The user confirms the points

[1428] Input: A list of issues displayed on the terminal.

[1429] Data processing: The user confirms the information pointed out.

[1430] Output: The confirmed issue.

[1431] Specific action: The user looks at the list of issues on the interface and identifies the problem.

[1432] Step 11:

[1433] User makes corrections

[1434] Input: The confirmed issue.

[1435] Data processing: The user corrects the text based on the suggestions.

[1436] Output: The corrected text.

[1437] What happens: The user manually corrects the text or clicks to apply the suggested correction.

[1438] Step 12:

[1439] User confirms final text

[1440] Input: The corrected text.

[1441] Data processing: The user makes a final check of the revised text.

[1442] Output: The final text that was checked.

[1443] Specific operation: The user reviews the revised text in detail and performs a final check.

[1444] Step 13:

[1445] The user clicks the save button

[1446] Input: Final text as verified.

[1447] Data processing: The user clicks the save button.

[1448] Output: The final text is sent to the server.

[1449] Specific behavior: When the user clicks the save button, the device sends the final text to the server.

[1450] Step 14:

[1451] The server saves the final text

[1452] Input: The final text sent from the terminal.

[1453] Data processing: The server stores the final text in a database.

[1454] Output: The final text saved.

[1455] What it does: The server stores the final text it receives in a database and associates it with the user's account.

[1456] (Application example 1)

[1457] 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."

[1458] When creating business texts or security documents, it is difficult for users to quickly and accurately identify grammatical errors, typographical errors, and even security risk-related errors to create reliable final text. In addition, real-time analysis and correction suggestions are required, while ensuring compliance with industry regulations and internal rules. An efficient system to solve these challenges is needed.

[1459] 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.

[1460] In this invention, the server includes means for a user to perform authentication operations, means for passing received text data to a generative AI model, means for comparing candidates extracted by the generative AI model with a data storage location, means for generating a list of points to be pointed out and presenting it to the user, means for the user to select or input corrections, means for saving the final text, means for analyzing voice-input text in real time and pointing out errors and security risks, means for displaying correction suggestions in real time, and means for checking compliance with industry regulations and internal rules.This enables users to check and correct text errors and security risks in real time and efficiently create reliable final text.

[1461] The "means by which a user performs authentication operations" is a function for verifying the identity of a user based on the authentication information entered by the user.

[1462] "Means for passing received text data to a generative AI model" refers to a function that transfers text data obtained from a user to a generative AI model that analyzes the data and suggests corrections.

[1463] "Means for comparing candidates extracted by the generative AI model with data storage locations" refers to a function that compares and verifies candidates for errors and corrections in text extracted by the generative AI model with an existing database.

[1464] "Means for generating a list of points of concern and presenting it to the user" refers to a function that creates a list of errors and suspicious points in the text based on the generative AI model and the matching results, and displays it to the user.

[1465] The "means for the user to select or input corrections" refers to an interface or function that allows the user to select or newly input corrections to the pointed out points in the text.

[1466] The "means for saving the final text" is a function for saving the final text after correction in a database or a specific storage location.

[1467] "Means for analyzing voice-input text in real time and pointing out errors and security risks" is a function that instantly analyzes voice-input text, detects grammatical errors and security risks, and points them out in real time.

[1468] "Means for displaying suggested revisions in real time" refers to a function that instantly presents users with suggested revisions based on the analysis results.

[1469] The "means for checking compliance with industry regulations and internal rules" is a function for checking whether the generated text complies with the relevant industry regulations and internal rules.

[1470] The "Secure Text Guard" system allows users to create text using voice input or text input, and detects grammatical errors, typos, suspicious points, and security risks in the text in real time and makes correction suggestions. Specific embodiments of the system are described below.

[1471] Hardware Configuration

[1472] 1. Smart Glasses

[1473] A device for voice input and text display.

[1474] It has a voice recognition function that converts voice input data into text.

[1475] 2. Server

[1476] A central server responsible for user authentication, text analysis, generating correction suggestions, and data storage.

[1477] Integrate with databases of data locations and industry regulations.

[1478] Software Configuration

[1479] 1. Generative AI Models

[1480] An AI model that analyzes user-entered text to detect grammatical errors, typos, and security risks.

[1481] 2. NLP technology

[1482] As a natural language processing technology, it is used to analyze input text and extract suspicious points.

[1483] 3. RAG technology

[1484] A technology that compares the results of text analysis with a database and evaluates their reliability.

[1485] System operation flow

[1486] 1. Authentication Operation

[1487] The user puts on the smart glasses and enters authentication information (user ID and password) using voice or touch.

[1488] The server checks this authentication information against a database to authenticate the user.

[1489] 2. Text Input

[1490] After successful authentication, the user can then use voice input or direct text input.

[1491] The entered text data is sent to the server in real time.

[1492] 3. Text analysis and suggested revisions

[1493] The server passes the received text to the generative AI model and begins analysis.

[1494] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[1495] Meanwhile, RAG technology is used to compare the analysis results with the data storage location and evaluate the reliability of the analysis.

[1496] 4. Corrections and Final Text Generation

[1497] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[1498] The user selects or inputs the corrections based on this list.

[1499] The server applies the modifications and generates the final text.

[1500] 5. Save

[1501] The final text, after corrections are complete, is stored in a data repository by the server.

[1502] Specific examples

[1503] For example, a user can use smart glasses to voice-input "important meeting materials." The system converts this into text in real time and begins analysis. If grammatical errors or risks of leaking confidential information are detected, they are immediately presented to the user and suggested corrections are displayed. The user can select or enter the suggested corrections, and the final text is generated.

[1504] Example prompt sentence:

[1505] "Analyzes the entered text and points out errors and suspicious points such as grammar mistakes, typos, and risk of leaking confidential information."

[1506] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1507] Step 1:

[1508] The user puts on the smart glasses and enters authentication information (user ID and password) by voice or touch.

[1509] Input: User ID, Password

[1510] Output: Authentication request

[1511] Step 2:

[1512] The server receives the authentication information and authenticates the user against a database.

[1513] Input:Authentication request

[1514] Data processing and calculation: Verification with authentication information in the database

[1515] Output: Authentication result (success / failure)

[1516] Step 3:

[1517] After successful authentication, the user can enter voice or text.

[1518] Input: Audio or text data

[1519] Output: Input text data

[1520] Step 4:

[1521] The entered text data is sent to the server in real time.

[1522] Input: Input text data

[1523] Output: Text data sent to the server

[1524] Step 5:

[1525] The server passes the received text data to the generative AI model and begins analysis.

[1526] Input: Text data sent to the server

[1527] Data processing and computation: Text analysis with generative AI models

[1528] Output: Analysis results (potential errors or suspicious points)

[1529] Step 6:

[1530] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[1531] Input: Analysis results (candidates for errors or suspicious points)

[1532] Data processing and calculation: Additional detection and correction suggestion generation using NLP technology

[1533] Output: List of suggested fixes

[1534] Step 7:

[1535] Using RAG technology, the analysis results are compared with the data storage location to evaluate the reliability of the analysis.

[1536] Input: list of suggested revisions

[1537] Data processing and calculation: Verification with data storage location

[1538] Output: Reliability evaluation results

[1539] Step 8:

[1540] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[1541] Input: Reliability evaluation results

[1542] Output: A list of issues and suggested fixes displayed on the user's screen

[1543] Step 9:

[1544] The user selects or enters corrections based on this list.

[1545] Input: List of issues and suggested fixes

[1546] Output: Modifications selected or entered

[1547] Step 10:

[1548] The server applies the modifications and generates the final text.

[1549] Input: The selected or entered correction

[1550] Data processing and calculation: Reflection of corrections

[1551] Output: Final text

[1552] Step 11:

[1553] The final text, after corrections are complete, is stored in a data repository by the server.

[1554] Input: Final text

[1555] Output: Saved text data

[1556] 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.

[1557] This invention is a text validation UI system that identifies errors and suspicious points in text created by users of a text generation service, recognizes the user's emotions, and generates more accurate and reliable text. A specific embodiment of this system is described below.

[1558] Initial Setup

[1559] 1. The user logs in

[1560] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[1561] 2. The server authenticates the user

[1562] The server authenticates the user by checking the ID and password entered by the user against the authentication information in the database. If authentication is successful, a user session is started, a session ID is generated, and returned to the user. At the same time, the user information is loaded into the system.

[1563] Text Input

[1564] 1. The user enters text

[1565] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[1566] 2. The device sends the entered text to the server

[1567] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[1568] Text analytics

[1569] 1. The server passes the text to the generative AI model

[1570] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[1571] 2. A generative AI model analyzes the text

[1572] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[1573] emotion recognition

[1574] 1. Emotion engine recognizes user emotions

[1575] The emotion engine recognizes users' emotions in real time as they type, and this emotional information is reflected in the tone of the text and suggested corrections.

[1576] 2. The emotion engine uses emotional information for analysis

[1577] The emotion engine analyzes the user's emotions as they type and combines them with the results of a generative AI model, which adjusts the tone of the text.

[1578] Validation

[1579] 1. The server collates data using RAG technology

[1580] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[1581] 2. The server generates a list of issues

[1582] Based on the matching results, the server generates a list of issues to be fixed, including specific corrections such as grammar errors and incorrect data.

[1583] feedback

[1584] 1. The server sends the list of issues to the user's device.

[1585] The list of issues is sent to the user's terminal in an appropriate format, and the user can view it on the interface.

[1586] 2. The user confirms the issue

[1587] Users can review the list of issues and review the details of each correction on their device, and the emotion engine will provide feedback tailored to the user's emotions.

[1588] 3. User makes corrections

[1589] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[1590] Final confirmation and saving

[1591] 1. The user checks the final text

[1592] The user checks the final text after corrections and confirms that there are no problems with the content.

[1593] 2. The user clicks the save button

[1594] When the user clicks the save button, the final text is sent to the server.

[1595] 3. The server saves the final text

[1596] The server stores the final text in a database and associates it with the user's account.

[1597] Specific examples

[1598] For example, imagine a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, it provides optimal feedback based on the user's emotions. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[1599] The processing flow will be explained below.

[1600] Step 1:

[1601] The user performs a login operation.

[1602] The user accesses the login screen of the text validation UI from the terminal, enters their ID and password, and clicks the login button to send the entered information to the server.

[1603] Step 2:

[1604] The server authenticates the user

[1605] The server checks the received ID and password against the authentication information in its database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[1606] Step 3:

[1607] The user enters text

[1608] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[1609] Step 4:

[1610] The device sends the entered text to the server

[1611] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[1612] Step 5:

[1613] The server passes the text to the generative AI model

[1614] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[1615] Step 6:

[1616] Generative AI models parse text

[1617] The generative AI model analyzes the text data, extracts candidates for grammatical errors, typos, and suspicious points, and returns the results to the server. This analysis uses natural language processing (NLP) technology.

[1618] Step 7:

[1619] Emotion engine recognizes user emotions

[1620] The emotion engine collects and analyzes real-time emotion data while the user is entering text, thereby identifying the user's emotional state.

[1621] Step 8:

[1622] The emotion engine uses emotional information for analysis

[1623] The emotion engine combines the results returned by the generative AI model with the user's emotional information, allowing the tone of the text and suggested revisions to be adjusted based on the user's emotions.

[1624] Step 9:

[1625] The server collates data using RAG technology

[1626] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[1627] Step 10:

[1628] The server generates a list of issues

[1629] Based on the results of the match, the server generates a list of issues, including specific corrections such as "grammatical errors" or "incorrect data."

[1630] Step 11:

[1631] The server sends the list of issues to the user's device.

[1632] The list of issues is sent to the user's device in an appropriate format, where the user can view it in the interface.

[1633] Step 12:

[1634] The user confirms the points

[1635] Users can view a list of issues and review the details of each correction on their device, and the emotion engine provides feedback tailored to the user's emotions.

[1636] Step 13:

[1637] User makes corrections

[1638] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[1639] Step 14:

[1640] User confirms final text

[1641] The user checks the final text after corrections and confirms that there are no problems with the content.

[1642] Step 15:

[1643] The user clicks the save button

[1644] When the user clicks the save button, the final text is sent to the server.

[1645] Step 16:

[1646] The server saves the final text

[1647] The server stores the final text in a database and associates it with the user's account, completing the process.

[1648] Example 2

[1649] 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."

[1650] Modern text generation services require a method to effectively identify errors and suspicious points in user-generated text. However, current systems are unable to provide feedback that takes into account the user's emotions, and the reliability of the generated feedback is not sufficiently guaranteed. As a result, users spend a lot of time correcting the text, and the quality of the final text is difficult to guarantee.

[1651] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1652] In this invention, the server includes an emotion recognition means for recognizing a user's emotions in real time, a means for adjusting the analysis results of the generative AI model based on the emotion information, and a means for reflecting emotion-based feedback in the list of points to be pointed out. This makes it possible to provide feedback that takes the user's emotions into consideration and improve the reliability and quality of the text.

[1653] "Means for users to log in" refers to the means by which users access the system and perform authentication using their own account information.

[1654] "Means for passing received text data to the generative AI model" refers to the means by which the server receives text data entered by the user and sends it to the generative AI model for analysis.

[1655] "Means for comparing candidates extracted by the generative AI model with a database" refers to a means for comparing grammatical errors, typos, and suspicious points pointed out by the generative AI model with information in a database and evaluating their reliability.

[1656] "Means for generating a list of issues and presenting it to the user" refers to a means for compiling correction suggestions obtained from a generative AI model or database matching into a list and displaying it to the user.

[1657] The "means for the user to select or input correction content" refers to the means by which the user selects the necessary correction from the presented list of issues or inputs the correction themselves.

[1658] The "means for saving the final text" is a means for saving the text after the user has completed the corrections in a database and associating it with the user's account information.

[1659] The "emotion recognition means for recognizing a user's emotions in real time" is a means for recognizing and analyzing the emotions of a user while the user is inputting text.

[1660] "Means for adjusting the analysis results of a generative AI model based on emotional information" refers to a means for optimizing the analysis results of a generative AI model by taking into account the user's emotions and providing more appropriate feedback.

[1661] The "means for reflecting emotion-based feedback in the points to be pointed out list" is a means for adjusting the feedback content displayed in the points to be pointed out list based on the user's emotion information.

[1662] This invention is a system that identifies errors and suspicious points in text created by users of a text generation service, and further recognizes the user's emotions to generate more accurate and reliable text. Specific components and operations of this system will be described below.

[1663] System Configuration

[1664] The system consists of the following main components:

[1665] 1. User Device

[1666] A device that allows a user to input text and receive feedback, such as a PC, smartphone, or tablet.

[1667] The input interface is built with front-end frameworks such as React or Vue.js.

[1668] 2. Server

[1669] This is a central processing unit that receives and analyzes text data entered by the user and executes the generative AI model. The server uses a cloud server such as AWS (Amazon Web Services) or Google Cloud Platform.

[1670] 3. Database

[1671] This is data storage for saving authentication information, text data, analysis results, etc. SQL databases and NoSQL databases are used.

[1672] 4. Generative AI Models

[1673] It is an artificial intelligence model used to analyze text for grammatical errors, typos, and suspicious points. Specifically, it uses large-scale language models such as GPT-3.

[1674] 5. Emotion Recognition Engine

[1675] It is an engine for recognizing user emotions in real time and adjusting the tone of the text. TensorFlow and OpenCV are often used.

[1676] 6. RAG technology

[1677] This technology evaluates the reliability of the findings extracted by the generative AI model by comparing them with public databases and closed data.

[1678] Operation explanation

[1679] Below, we will explain the specific steps involved in the operation of the system and how to implement them.

[1680] 1. The user logs in

[1681] Users use the login interface on their devices to enter their ID and password, which is then encrypted by SSL / TLS and sent to the server.

[1682] 2. The server authenticates the user

[1683] The server authenticates the user by checking the received ID and password against the authentication information in the database. If authentication is successful, a session ID is generated and returned to the user.

[1684] 3. The user enters text

[1685] The user enters the content of a business email or document into the device interface, and when the input is complete, clicks the send button.

[1686] 4. The device sends the entered text to the server

[1687] The entered text data is sent to the server in JSON format, along with the session ID.

[1688] 5. The server passes the text to the generative AI model

[1689] The received text data is passed to the generative AI model, and a request is sent to analyze it for grammatical errors, typos, and suspicious points.

[1690] 6. Generative AI models analyze text

[1691] A generative AI model analyzes text data and extracts insights using natural language processing (NLP) techniques.

[1692] 7. Emotion recognition engine recognizes user emotions

[1693] Recognize and analyze the user's emotions in real time while they are entering text.

[1694] 8. Adjust the analysis results of generative AI models based on emotional information

[1695] Based on the results of the emotion recognition engine, the analysis results are optimized according to the user's emotions.

[1696] 9. The server collates the data using RAG technology

[1697] The findings extracted by the generative AI model are verified using RAG technology to assess their reliability, for example by comparing them with public databases or closed data.

[1698] 10. The server generates a list of issues and sends it to the user's device.

[1699] A list of the verified issues is generated and transmitted to the user terminal based on this information.

[1700] 11. The user checks the issues and makes corrections

[1701] The user can review the list of issues and manually correct each one, or click on the suggested fixes to have them applied automatically.

[1702] 12. Check and save the final text

[1703] The user confirms the final text after correction and clicks the save button, and the server saves this final text in its database and associates it with the user's account.

[1704] Specific examples

[1705] For example, consider a case where a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, optimal feedback based on the user's emotions is provided. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[1706] Prompt Sentence Examples

[1707] Please review this email and let us know about any grammatical or typo errors or anything suspicious. Also, please adjust the tone of the email to a more calm one.

[1708] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1709] Step 1:

[1710] The user performs a login operation.

[1711] Input: User ID, Password

[1712] Operation: The user accesses the login screen of the text validation UI from the terminal and enters their ID and password.

[1713] Output: Login information is sent to the server.

[1714] Step 2:

[1715] The server authenticates the user

[1716] Input: Login information (user ID, password)

[1717] How it works: The server checks the received ID and password against the information in its database.

[1718] Output: If authentication is successful, a session ID is generated and returned to the user. At the same time, the user information is loaded into memory or cache.

[1719] Step 3:

[1720] The user enters text

[1721] Input: Text of business emails and documents

[1722] Action: A user enters text into the interface on their device and clicks the send button.

[1723] Output: The input text data is stored in the terminal.

[1724] Step 4:

[1725] The device sends the entered text to the server

[1726] Input: Text data, session ID

[1727] What happens: The device sends text data in JSON format to the server, which is sent as an HTTP POST request.

[1728] Output: The server receives the text data and the session ID.

[1729] Step 5:

[1730] The server passes the text to the generative AI model

[1731] Input: Text data

[1732] How it works: The server formats the text data and sends a request to the generative AI model endpoint.

[1733] Output: The generative AI model receives the text data.

[1734] Step 6:

[1735] Generative AI models parse text

[1736] Input: Text data

[1737] How it works: Generative AI models use natural language processing (NLP) techniques to analyze text data for grammatical errors, typos, and other suspicious content.

[1738] Output: A list of issues is generated and returned to the server.

[1739] Step 7:

[1740] Emotion recognition engine recognizes user emotions

[1741] Input: User input status, facial expression data, etc.

[1742] How it works: The emotion recognition engine recognizes the user's emotions in real time, analyzing images and input speed.

[1743] Output: User emotion data is generated and sent to the server.

[1744] Step 8:

[1745] Adjusting the analysis results of generative AI models based on emotional information

[1746] Input: List of points to be pointed out, user's emotional data

[1747] How it works: The server optimizes the analysis results of the generative AI model based on the results of the emotion recognition engine, adjusting the tone of the feedback, etc.

[1748] Output: A reconciled list of issues is generated.

[1749] Step 9:

[1750] The server collates data using RAG technology

[1751] Input: Reconciled issue list

[1752] How it works: The server uses RAG technology to match items in the findings list against public and closed databases.

[1753] Output: The reliability evaluation result is generated.

[1754] Step 10:

[1755] The server generates a list of issues and sends it to the user's device.

[1756] Input: Reliability evaluation results, adjustment results

[1757] Operation: The server generates a final list of issues and sends it to the user's terminal.

[1758] Output: The final issue list is displayed on the user's terminal.

[1759] Step 11:

[1760] The user checks the issues and makes corrections

[1761] Input: Final list of issues

[1762] How it works: The user reviews the list of issues on their device and either manually fixes each one or clicks on the suggested fixes to automatically apply them.

[1763] Output: The modified text data is saved to the terminal.

[1764] Step 12:

[1765] User confirms and saves the final text

[1766] Input: Modified text data

[1767] Action: The user reviews the final text after corrections and clicks the save button.

[1768] Output: The final text is sent to the server.

[1769] Step 13:

[1770] The server saves the final text

[1771] Input: Final text

[1772] How it works: The server stores the final text in a database and associates it with the user's account.

[1773] Output: A confirmation message is displayed to the user indicating that the save was successful.

[1774] (Application example 2)

[1775] 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."

[1776] Conventional text generation services have limited functionality for pointing out errors or suspicious points in text entered by users, making it difficult to provide feedback that takes emotions into account. Furthermore, even in multilingual translation, they have been unable to detect grammatical errors, typos, and suspicious points, or provide translations with a natural tone that reflects the user's emotions. Therefore, the objective of this invention is to provide an advanced text validation system that not only points out errors and suspicious points in text, but also includes multilingual translation and user sentiment analysis.

[1777] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1778] In this invention, the server includes a means for translating text entered by a user into a different language, a means for detecting grammatical errors, typos, and suspicious points in the translated text, and a means for analyzing the user's emotions and adjusting the tone of the text, thereby enabling the generation of grammatically correct, natural-looking text that reflects the user's emotions even in multiple languages.

[1779] "User" is a person who uses the system to input text and perform translation and validation.

[1780] "Login operation" refers to the authentication procedure required when a user accesses a system.

[1781] A "generative AI model" refers to an artificial intelligence model that analyzes received text data and extracts grammatical errors, typos, and suspicious points.

[1782] A "database" is a data storage device that stores information used to verify the reliability of candidates extracted by a generative AI model.

[1783] "Translation facility" refers to a facility that converts user-entered text into a different language.

[1784] "Grammar error detector" refers to a function that detects grammatical errors present in the translated text.

[1785] "Sentiment analysis means" refers to the function that analyzes the user's emotions when typing and adjusts the tone of the text.

[1786] The "list of points of criticism" is a list of information summarizing errors and corrections detected by the generative AI model, grammar error detection means, and sentiment analysis means.

[1787] The "means for selecting correction content" refers to a function that allows the user to select or newly input the content to be corrected based on the list of pointed out points.

[1788] "Final text" refers to text data that has been corrected by the user, finally confirmed, and saved.

[1789] "RAG technology" refers to a technology that compares candidates extracted by a generative AI model with existing data and evaluates their reliability.

[1790] MODE FOR CARRYING OUT THE INVENTION

[1791] System Program

[1792] The system that realizes this application example includes the following programs that perform a series of functions: user authentication, text input, text processing, sentiment analysis, grammar error detection, translation, presenting a list of corrections to the user, and saving the corrections.

[1793] Program processing

[1794] The server provides a means for users to log in when they access the system. The login operation uses a user authentication system using Django to verify the ID and password.

[1795] When the user enters text, the terminal sends the entered text data to the backend, which uses React for the frontend, making the user interface intuitive and responsive.

[1796] The server passes the received text data to a generative AI model, which can use BERT, for example, to detect grammatical errors, typos, and other suspicious aspects of the text. The output from the generative AI model is then sent back to the server.

[1797] The server then compares the candidates extracted by the generative AI model against a database of existing data and uses RAG technology to assess the reliability of the candidates.

[1798] Additionally, the server provides a means to analyze the user's emotions, using the Google Cloud Natural Language API to calculate an emotion score in real time as the user types.

[1799] Based on the results of the sentiment analysis, the server adjusts the tone of the text and performs the translation function, which uses the Google Cloud Translation API to translate the text entered by the user into different languages.

[1800] Based on the results of these processes, the server generates a list of issues and presents it to the user. The user can check the list on their device and carefully examine the details of each correction. If there are grammatical errors or misleading expressions, the user can correct them or select the correction suggested by the system.

[1801] Finally, the final text that the user has confirmed is stored on the server, and this final text is highly reliable and in line with the user's sentiment and intent.

[1802] Hardware and software used

[1803] Hardware: Smartphones, computers, and cloud-based server environments.

[1804] software:

[1805] Frontend: React

[1806] Backend: Django

[1807] Translation API: Google Cloud Translation API

[1808] Sentiment analysis: Google Cloud Natural Language API

[1809] Grammar Error Detection: BERT (TensorFlow / PyTorch)

[1810] Specific examples

[1811] For example, a user might enter the text "Good morning, how can I help you today?" in Japanese. The application translates this text into English and displays "Good morning, how can I help you today?". At the same time, it performs sentiment analysis to ensure the user maintains a friendly tone. The generative AI model detects grammatical errors, typos, and other suspicious points and presents a list of issues to the user.

[1812] Prompt example

[1813] User Input: "Good morning, how are you doing today?"

[1814] -> Translate to English

[1815] -> Detect grammar issues

[1816] -> Analyze sentiment score

[1817] In this way, the system provides grammatically accurate text generation across multiple languages ​​and feedback that reflects the user's sentiment, making user communication more efficient and reliable.

[1818] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1819] Step 1:

[1820] The server provides a login screen for the user to log in. The user enters their ID and password and clicks the login button. The server receives this input data and checks it against the authentication information in the database. If authentication is successful, a session ID is generated and the user information is loaded into the system.

[1821] Input: User ID and password

[1822] Output: Session ID and user information

[1823] Step 2:

[1824] The terminal displays a text input interface to the authenticated user. The user enters text and clicks the send button to send the input text to the server. The terminal then sends the entered text data to the server in JSON format, including the session ID.

[1825] Input: Text entered by the user

[1826] Output: JSON format text data and session ID

[1827] Step 3:

[1828] The server passes the received text data to a generative AI model. The generative AI model (e.g., BERT) analyzes the text data and extracts grammatical errors, typos, and suspicious points. The server then sends a request to the generative AI model's endpoint to obtain the analysis results.

[1829] Input: Received JSON format text data

[1830] Output: Analysis results from the generative AI model

[1831] Step 4:

[1832] The server compares the candidates extracted by the generative AI model with a database, using RAG technology to compare the candidates with public and closed databases and evaluate their reliability.

[1833] Input: Analysis results from generative AI model

[1834] Output: Matching results and confidence scores for each candidate

[1835] Step 5:

[1836] The server passes the received text data to the emotion engine to analyze the user's emotions. The emotion engine calculates the user's emotion score in real time using Google Cloud Natural Language API and other tools.

[1837] Input: Received text data

[1838] Output: Sentiment score

[1839] Step 6:

[1840] The server adjusts the tone of the text based on the results of sentiment analysis, and uses the Google Cloud Translation API to translate the text into different languages. Finally, it generates the evaluated translation results and adjusted text.

[1841] Input: Sentiment scores and received text data

[1842] Output: Adjusted translated text

[1843] Step 7:

[1844] The server generates and presents a list of issues to the user, including grammar, typos, suspicious points, and feedback based on sentiment scores, and sends this information to the front-end in an appropriate format.

[1845] Input: Evaluated translated text, sentiment score, and analysis results

[1846] Output: List of issues

[1847] Step 8:

[1848] The user can view a list of issues on their device, review the details of each correction, and make any necessary changes. They can either manually correct the text or click on the system's suggested corrections to automatically apply them.

[1849] Input: List of issues

[1850] Output: User's corrected text

[1851] Step 9:

[1852] The user finally checks the corrected text and clicks the save button. The server saves the final text in its database and associates it with the user's account.

[1853] Input: Final text as modified by the user

[1854] Output: Final text stored in the database

[1855] 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.

[1856] 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.

[1857] 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.

[1858] [Fourth embodiment]

[1859] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1860] 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.

[1861] 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).

[1862] 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.

[1863] 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.

[1864] 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).

[1865] 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.

[1866] 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.

[1867] 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.

[1868] 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.

[1869] 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.

[1870] 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.

[1871] 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."

[1872] This invention is a text validation UI system that points out errors and suspicious points in text created by users of a text generation service and generates accurate and reliable text. Specific embodiments of this system are described below.

[1873] Initial Setup

[1874] 1. The user logs in

[1875] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[1876] 2. The server authenticates the user

[1877] The server authenticates the user by checking the ID and password against a database. If authentication is successful, the user session is started and the user information is loaded into the system.

[1878] Text Input

[1879] 1. The user enters text

[1880] Users enter the text of business email drafts and documents into the device's interface.

[1881] 2. The device sends the entered text to the server

[1882] When the user clicks the input button, the terminal sends the entered text data to the server in an appropriate format, such as JSON format.

[1883] Text analytics

[1884] 1. The server passes the text to the generative AI model

[1885] The server passes the received text data to an existing generative AI model, which then receives the input text and begins analyzing it.

[1886] 2. A generative AI model analyzes the text

[1887] The generative AI model analyzes the content of the text and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[1888] Validation

[1889] 1. The server collates data using RAG technology

[1890] The server uses RAG technology to compare candidates extracted by the generative AI model with public and closed databases, and evaluates the reliability of the candidates.

[1891] 2. The server generates a list of issues

[1892] Based on the matching results, the server generates a list of issues, such as "misspelled company name" or "incorrect date format."

[1893] feedback

[1894] 1. The server sends the list of issues to the user's device.

[1895] The list of issues is sent to the user's device in an appropriate format, where the user can view it.

[1896] 2. The user confirms the issue

[1897] The user reviews the presented list of issues and carefully examines the points that need to be corrected.

[1898] 3. User makes corrections

[1899] Users can make corrections based on the suggestions, either manually or by simply clicking on the suggested corrections.

[1900] Final confirmation and saving

[1901] 1. The user sees the final text

[1902] The user checks the final text after correction and confirms the content.

[1903] 2. The user clicks the save button

[1904] When the user clicks the save button, the final edited text is sent to the server.

[1905] 3. The server saves the final text

[1906] The server stores the final text in a database and associates it with the user's account.

[1907] Specific examples

[1908] For example, suppose a user is composing a business email containing information about a new product. First, the user logs in to the system and enters the email content. The entered text is sent to the server and analyzed by the generative AI model. The user then receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the user saves the email with the corrections completed, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the diverse needs of enterprise users.

[1909] The processing flow will be explained below.

[1910] Step 1:

[1911] The user performs a login operation.

[1912] The user accesses the login screen of the text validation UI from the terminal, enters their ID and password, and clicks the login button to send the entered information to the server.

[1913] Step 2:

[1914] The server authenticates the user

[1915] The server checks the received ID and password against the authentication information in the database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[1916] Step 3:

[1917] The user enters text

[1918] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[1919] Step 4:

[1920] The device sends the entered text to the server

[1921] The device sends the text data entered by the user to the server in JSON format or an appropriate data format, including the session ID.

[1922] Step 5:

[1923] The server passes the text to the generative AI model

[1924] The server processes the received text data and sends a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[1925] Step 6:

[1926] Generative AI models parse text

[1927] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, and suspicious points, using natural language processing (NLP) techniques.

[1928] Step 7:

[1929] The server collates data using RAG technology

[1930] The server receives the candidates extracted by the generative AI model and compares them with public and closed databases using RAG technology, thereby evaluating the reliability of each candidate.

[1931] Step 8:

[1932] The server generates a list of issues

[1933] The server generates a list of issues based on the results of the match, including specific corrections such as grammar errors and incorrect data.

[1934] Step 9:

[1935] The server sends the list of issues to the user's device.

[1936] The server sends the list of issues to the user's device in JSON format, etc., which the user can view on the interface.

[1937] Step 10:

[1938] The user confirms the points

[1939] The user can check the list of issues on their device and carefully examine the details of each correction.

[1940] Step 11:

[1941] User makes corrections

[1942] The user can then make the necessary corrections based on the list, either by manually entering the corrections or by clicking on the suggested corrections provided to automatically incorporate them.

[1943] Step 12:

[1944] User confirms final text

[1945] After completing the corrections, the user checks the final text to ensure there are no problems with the content.

[1946] Step 13:

[1947] The user clicks the save button

[1948] When the user clicks the save button, the final text is sent to the server.

[1949] Step 14:

[1950] The server saves the final text

[1951] The server stores the final text in a database and associates it with the user's account, completing the process.

[1952] Example 1

[1953] 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."

[1954] Conventional text generation systems have difficulty efficiently detecting errors and inconsistencies and providing appropriate feedback to users. Furthermore, they lack an interface that allows users to review corrections and quickly and accurately correct the text. This can lead to problems such as a decrease in the reliability and quality of the generated text.

[1955] 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.

[1956] In this invention, the server includes a means for user authentication, a means for passing received text data to the generative AI model, and a means for comparing candidates extracted by the generative AI model with a database using RAG technology, which enables reliable evaluation of errors and suspicious points in the text detected by the generative AI model and provides accurate feedback and suggested corrections to the user.

[1957] "User authentication" is the process in which the server checks the authentication information (such as ID and password) entered by the user against a database to verify whether the user is legitimate.

[1958] A "terminal" is a device, such as a computer or smartphone, that a user uses to enter text or perform login operations.

[1959] "Text data" refers to character string information such as sentences or words entered by the user.

[1960] A "generative AI model" is an algorithm that uses natural language processing technology to analyze input text and extract grammatical errors, typos, and suspicious points.

[1961] "RAG technology" stands for Retrieval-Augmented Generation, and is a technology that compares generated candidates with existing databases to evaluate their reliability.

[1962] A "database" is a structured collection of data that is accessed by a server and stores user authentication information and data for verification.

[1963] The "list of points of concern" is a list of grammatical errors, typos, suspicious points, and other areas that need to be corrected, generated as a result of analysis by the generative AI model.

[1964] "Saving" is the process by which the server stores the text, once the user has completed the modifications, in a database for later access.

[1965] "Corrections" refer to corrections suggested by the generative AI model or corrections manually entered by the user.

[1966] This invention provides a system for pointing out errors and suspicious points in text created by a user using a text generation service, and generating accurate and reliable text. Specific embodiments of this system are described below.

[1967] Initial Setup

[1968] A user opens a browser on their device (e.g., a PC or smartphone) and accesses the text validation UI. They enter their ID and password on the login screen and perform the login operation. The server checks the ID and password entered by the user against the database to confirm whether they are a legitimate user. If authentication is successful, the server starts a user session, loads related user information, and provides the user with the dashboard or main operation screen.

[1969] Text Input

[1970] For example, when composing a business email containing new product information, a user can enter text into the device interface or copy and paste it from another source. When the user clicks the "Done" button, the device converts the entered text data into JSON format and sends it to the server via the API.

[1971] Text analytics

[1972] The server then passes the received text data to a generative AI model (e.g., GPT-3 or BERT). This generative AI model receives the input text and analyzes it to identify grammatical errors, typos, and other suspicious features. Specifically, it uses natural language processing techniques to tokenize each sentence and extract anomalous patterns.

[1973] Validation

[1974] The server uses Retrieval-Augmented Generation (RAG) technology to compare the suspicious points and candidates extracted by the generative AI model with public databases and internal closed data. This comparison evaluates the reliability and accuracy of the candidates. Based on the comparison results, the server creates a list of issues to be addressed. Examples of issues on this list include "misspellings in company names" and "date format errors."

[1975] feedback

[1976] The server converts the list of issues into an appropriate format (e.g., JSON) and sends it to the user's device. The user's device displays the received data on an interface. The user reviews the list of issues and carefully examines which parts are incorrect and how they should be corrected. The user corrects the text based on the issues, either manually or by applying the system's suggested corrections with one click.

[1977] Final confirmation and saving

[1978] The user then finalizes the edited text to ensure it is correct. Once the user clicks the save button, the final edited text is sent to the server, which stores the edited text in a database and associates it with the user's account.

[1979] Specific examples

[1980] For example, a user creates a business email containing new product information and enters the content into the system. The entered text is sent to the server and analyzed by the generative AI model. The user receives a list of issues and corrects the text based on the automatically generated correction suggestions. Finally, the corrected email is saved, and the system stores it in the user's account. Through this process, users can easily create reliable text and meet the needs of various enterprise users.

[1981] Prompt Sentence Examples

[1982] The following prompts can be fed to the generative AI model:

[1983] Enter the paragraph "Please check this document." Point out the errors in the text output by the generative AI model and provide the best possible corrections.

[1984] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1985] Step 1:

[1986] The user performs a login operation.

[1987] Input: The user enters their ID and password from the device.

[1988] Data processing: The terminal converts the entered authentication information into a standardized format (e.g., JSON).

[1989] Output: The device sends the authentication information to the server.

[1990] What happens: The user opens a browser, accesses the login screen, and enters the required credentials.

[1991] Step 2:

[1992] The server authenticates the user

[1993] Input: The authentication information sent from the device.

[1994] Data processing: The server queries the database for authentication information and verifies whether the user's ID and password are correct.

[1995] Output: Prepares the authentication result (success or failure).

[1996] Specific operation: The server references the database and starts a user session based on the authentication result.

[1997] Step 3:

[1998] The user enters text

[1999] Input: A user types text for a business email or document.

[2000] Data processing: The entered text is temporarily stored on the terminal as string data.

[2001] Output: The input text data.

[2002] What happens: The user types text directly into the device interface or copies and pastes it from another source.

[2003] Step 4:

[2004] The device sends the entered text to the server

[2005] Input: The user completes text input.

[2006] Data processing: The terminal converts the input text into JSON format.

[2007] Output: The converted text data is sent to the server.

[2008] Specific operation: When the user clicks the "Complete input" button, the terminal sends the text data to the server.

[2009] Step 5:

[2010] The server passes the text to the generative AI model

[2011] Input: Text data sent from the terminal.

[2012] Data processing: The server formats the received text data into a format suitable for the generative AI model.

[2013] Output: The formatted text data is fed into a generative AI model.

[2014] How it works: The server passes the text data to a generative AI model (e.g., GPT-3 or BERT).

[2015] Step 6:

[2016] Generative AI models parse text

[2017] Input: Text data passed from the server.

[2018] Data processing: The generative AI model analyzes the text and extracts grammatical errors, typos, and suspicious points.

[2019] Output: A list of candidates as the analysis result.

[2020] How it works: The generative AI model uses natural language processing techniques to analyze text for tokens and extract anomalous patterns.

[2021] Step 7:

[2022] The server collates data using RAG technology

[2023] Input: A list of candidates extracted by a generative AI model.

[2024] Data processing: The server uses RAG technology to check the candidate list against the database.

[2025] Output: A list of candidates with a confidence rating.

[2026] Specific operation: The server checks candidates against public databases and closed data.

[2027] Step 8:

[2028] The server generates a list of issues

[2029] Input: A list of candidates with confidence ratings.

[2030] Data processing: The server generates a list of findings.

[2031] Output: A list of issues.

[2032] Specific operation: Based on the matching results, the server creates a list of issues such as "misspelled company name" or "date format error."

[2033] Step 9:

[2034] The server sends the list of issues to the user's device.

[2035] Input: The generated issue list.

[2036] Data processing: The server converts the list of findings into an appropriate format (e.g., JSON).

[2037] Output: Send the converted issue list to the user's terminal.

[2038] Specific operation: The server sends a list of issues to the user's device and displays it on the user's interface.

[2039] Step 10:

[2040] The user confirms the points

[2041] Input: A list of issues displayed on the terminal.

[2042] Data processing: The user confirms the indicated content.

[2043] Output: The confirmed issue.

[2044] Specific action: The user looks at the list of issues on the interface and identifies the problem.

[2045] Step 11:

[2046] User makes corrections

[2047] Input: The confirmed issue.

[2048] Data processing: The user corrects the text based on the suggestions.

[2049] Output: The corrected text.

[2050] What happens: The user manually corrects the text or clicks to apply the suggested correction.

[2051] Step 12:

[2052] User confirms final text

[2053] Input: The corrected text.

[2054] Data processing: The user makes a final check of the revised text.

[2055] Output: The final text that was checked.

[2056] Specific operation: The user reviews the revised text in detail and performs a final check.

[2057] Step 13:

[2058] The user clicks the save button

[2059] Input: Final text confirmed.

[2060] Data processing: The user clicks the save button.

[2061] Output: The final text is sent to the server.

[2062] Specific behavior: When the user clicks the save button, the device sends the final text to the server.

[2063] Step 14:

[2064] The server saves the final text

[2065] Input: The final text sent from the terminal.

[2066] Data processing: The server stores the final text in a database.

[2067] Output: The final text saved.

[2068] What it does: The server stores the final text it receives in a database and associates it with the user's account.

[2069] (Application example 1)

[2070] 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."

[2071] When creating business texts or security documents, it is difficult for users to quickly and accurately identify grammatical errors, typographical errors, and even security risk-related errors to create reliable final text. In addition, real-time analysis and correction suggestions are required, while ensuring compliance with industry regulations and internal rules. An efficient system to solve these challenges is needed.

[2072] 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.

[2073] In this invention, the server includes means for a user to perform authentication operations, means for passing received text data to a generative AI model, means for comparing candidates extracted by the generative AI model with a data storage location, means for generating a list of points to be pointed out and presenting it to the user, means for the user to select or input corrections, means for saving the final text, means for analyzing voice-input text in real time and pointing out errors and security risks, means for displaying correction suggestions in real time, and means for checking compliance with industry regulations and internal rules.This enables users to check and correct text errors and security risks in real time and efficiently create reliable final text.

[2074] "Means for users to perform authentication operations" refers to a function for verifying the identity of a user based on the authentication information entered by the user.

[2075] "Means for passing received text data to a generative AI model" refers to a function that transfers text data obtained from a user to a generative AI model that analyzes the data and suggests corrections.

[2076] "Means for comparing candidates extracted by the generative AI model with data storage locations" refers to a function that compares and verifies candidates for errors and corrections in text extracted by the generative AI model with an existing database.

[2077] "Means for generating a list of points of concern and presenting it to the user" refers to a function that creates a list of errors and suspicious points in the text based on the generative AI model and the matching results, and displays it to the user.

[2078] The "means for the user to select or input corrections" refers to an interface or function that allows the user to select or newly input corrections to the pointed out points in the text.

[2079] The "means for saving the final text" is a function for saving the final text after correction in a database or a specific storage location.

[2080] "Means for analyzing voice-input text in real time and pointing out errors and security risks" is a function that instantly analyzes voice-input text, detects grammatical errors and security risks, and points them out in real time.

[2081] "Means for displaying suggested revisions in real time" refers to a function that instantly presents users with suggested revisions based on the analysis results.

[2082] The "means for checking compliance with industry regulations and internal rules" is a function for checking whether the generated text complies with the relevant industry regulations and internal rules.

[2083] The "Secure Text Guard" system allows users to create text using voice input or text input, and detects grammatical errors, typos, suspicious points, and security risks in the text in real time and makes correction suggestions. Specific embodiments of the system are described below.

[2084] Hardware Configuration

[2085] 1. Smart Glasses

[2086] A device for voice input and text display.

[2087] It has a voice recognition function that converts voice input data into text.

[2088] 2. Server

[2089] A central server responsible for user authentication, text analysis, generating correction suggestions, and data storage.

[2090] Integrate with databases of data locations and industry regulations.

[2091] Software Configuration

[2092] 1. Generative AI Models

[2093] An AI model that analyzes user-entered text to detect grammatical errors, typos, and security risks.

[2094] 2. NLP technology

[2095] As a natural language processing technology, it is used to analyze input text and extract suspicious points.

[2096] 3. RAG technology

[2097] A technology that compares the results of text analysis with a database and evaluates their reliability.

[2098] System operation flow

[2099] 1. Authentication Operation

[2100] The user puts on the smart glasses and enters authentication information (user ID and password) using voice or touch.

[2101] The server checks this authentication information against a database to authenticate the user.

[2102] 2. Text Input

[2103] After successful authentication, the user can then use voice input or direct text input.

[2104] The entered text data is sent to the server in real time.

[2105] 3. Text analysis and suggested revisions

[2106] The server passes the received text to the generative AI model and begins analysis.

[2107] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[2108] Meanwhile, RAG technology is used to compare the analysis results with the data storage location and evaluate the reliability of the analysis.

[2109] 4. Corrections and Final Text Generation

[2110] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[2111] The user selects or inputs the corrections based on this list.

[2112] The server applies the modifications and generates the final text.

[2113] 5. Save

[2114] The final text, after corrections are complete, is stored in a data repository by the server.

[2115] Specific examples

[2116] For example, a user can use smart glasses to voice-input "important meeting materials." The system converts this into text in real time and begins analysis. If grammatical errors or risks of leaking confidential information are detected, they are immediately presented to the user and suggested corrections are displayed. The user can select or enter the suggested corrections, and the final text is generated.

[2117] Example prompt sentence:

[2118] "Analyzes the entered text and points out errors and suspicious points such as grammar mistakes, typos, and risk of leaking confidential information."

[2119] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[2120] Step 1:

[2121] The user puts on the smart glasses and enters authentication information (user ID and password) by voice or touch.

[2122] Input: User ID, Password

[2123] Output: Authentication request

[2124] Step 2:

[2125] The server receives the authentication information and authenticates the user against a database.

[2126] Input:Authentication request

[2127] Data processing and calculation: Verification with authentication information in the database

[2128] Output: Authentication result (success / failure)

[2129] Step 3:

[2130] After successful authentication, the user can enter voice or text.

[2131] Input: Audio or text data

[2132] Output: Input text data

[2133] Step 4:

[2134] The entered text data is sent to the server in real time.

[2135] Input: Input text data

[2136] Output: Text data sent to the server

[2137] Step 5:

[2138] The server passes the received text data to the generative AI model and begins analysis.

[2139] Input: Text data sent to the server

[2140] Data processing and computation: Text analysis with generative AI models

[2141] Output: Analysis results (potential errors or suspicious points)

[2142] Step 6:

[2143] It uses NLP technology to detect grammatical errors, typos, and suspicious points and generate correction suggestions.

[2144] Input: Analysis results (candidates for errors or suspicious points)

[2145] Data processing and calculation: Additional detection and correction suggestion generation using NLP technology

[2146] Output: List of suggested fixes

[2147] Step 7:

[2148] Using RAG technology, the analysis results are compared with the data storage location to evaluate the reliability of the analysis.

[2149] Input: list of suggested revisions

[2150] Data processing and calculation: Verification with data storage location

[2151] Output: Reliability evaluation results

[2152] Step 8:

[2153] The server generates a list of issues and suggests corrections, which are then displayed on the smart glasses.

[2154] Input: Reliability evaluation results

[2155] Output: A list of issues and suggested fixes displayed on the user's screen

[2156] Step 9:

[2157] The user selects or enters corrections based on this list.

[2158] Input: List of issues and suggested fixes

[2159] Output: Modifications selected or entered

[2160] Step 10:

[2161] The server applies the modifications and generates the final text.

[2162] Input: The selected or entered correction

[2163] Data processing and calculation: Reflection of corrections

[2164] Output: Final text

[2165] Step 11:

[2166] The final text, after corrections are complete, is stored in a data repository by the server.

[2167] Input: Final text

[2168] Output: Saved text data

[2169] 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.

[2170] This invention is a text validation UI system that identifies errors and suspicious points in text created by users of a text generation service, recognizes the user's emotions, and generates more accurate and reliable text. A specific embodiment of this system is described below.

[2171] Initial Setup

[2172] 1. The user logs in

[2173] The user accesses the login screen of the text validation UI from the terminal and enters their ID and password, which starts user authentication.

[2174] 2. The server authenticates the user

[2175] The server authenticates the user by checking the ID and password entered by the user against the authentication information in the database. If authentication is successful, a user session is started, a session ID is generated, and returned to the user. At the same time, the user information is loaded into the system.

[2176] Text Input

[2177] 1. The user enters text

[2178] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[2179] 2. The device sends the entered text to the server

[2180] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[2181] Text analytics

[2182] 1. The server passes the text to the generative AI model

[2183] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[2184] 2. A generative AI model analyzes the text

[2185] The generative AI model analyzes text data and extracts candidates for grammatical errors, typos, suspicious points, etc. This analysis uses natural language processing (NLP) techniques.

[2186] emotion recognition

[2187] 1. Emotion engine recognizes user emotions

[2188] The emotion engine recognizes users' emotions in real time as they type, and this emotional information is reflected in the tone of the text and suggested corrections.

[2189] 2. The emotion engine uses emotional information for analysis

[2190] The emotion engine analyzes the user's emotions as they type and combines them with the results of a generative AI model, which adjusts the tone of the text.

[2191] Validation

[2192] 1. The server collates data using RAG technology

[2193] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[2194] 2. The server generates a list of issues

[2195] Based on the matching results, the server generates a list of issues to be fixed, including specific corrections such as grammar errors and incorrect data.

[2196] feedback

[2197] 1. The server sends the list of issues to the user's device.

[2198] The list of issues is sent to the user's terminal in an appropriate format, and the user can view it on the interface.

[2199] 2. The user confirms the issue

[2200] Users can review the list of issues and review the details of each correction on their device, and the emotion engine will provide feedback tailored to the user's emotions.

[2201] 3. User makes corrections

[2202] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[2203] Final confirmation and saving

[2204] 1. The user checks the final text

[2205] The user checks the final text after corrections and confirms that there are no problems with the content.

[2206] 2. The user clicks the save button

[2207] When the user clicks the save button, the final text is sent to the server.

[2208] 3. The server saves the final text

[2209] The server stores the final text in a database and associates it with the user's account.

[2210] Specific examples

[2211] For example, imagine a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, it provides optimal feedback based on the user's emotions. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[2212] The processing flow will be explained below.

[2213] Step 1:

[2214] The user performs a login operation.

[2215] The user accesses the login screen of the text validation UI from the terminal, enters their ID and password, and clicks the login button to send the entered information to the server.

[2216] Step 2:

[2217] The server authenticates the user

[2218] The server checks the received ID and password against the authentication information in its database. If authentication is successful, it starts a user session, generates a session ID, and returns it to the user. At the same time, the user information is loaded into the system.

[2219] Step 3:

[2220] The user enters text

[2221] Users enter the text of a business email or document into the device's interface, and when they're done, click the send button.

[2222] Step 4:

[2223] The device sends the entered text to the server

[2224] The device sends the text data entered by the user to the server in JSON format or other appropriate data format, including the session ID.

[2225] Step 5:

[2226] The server passes the text to the generative AI model

[2227] The server processes the received text data and sends it as a request to the generative AI model's endpoint, where the text data is passed to the generative AI model.

[2228] Step 6:

[2229] Generative AI models parse text

[2230] The generative AI model analyzes the text data, extracts candidates for grammatical errors, typos, and suspicious points, and returns the results to the server. This analysis uses natural language processing (NLP) technology.

[2231] Step 7:

[2232] Emotion engine recognizes user emotions

[2233] The emotion engine collects and analyzes real-time emotion data while the user is entering text, thereby identifying the user's emotional state.

[2234] Step 8:

[2235] The emotion engine uses emotional information for analysis

[2236] The emotion engine combines the results returned by the generative AI model with the user's emotional information, allowing the tone of the text and suggested revisions to be adjusted based on the user's emotions.

[2237] Step 9:

[2238] The server collates data using RAG technology

[2239] The server receives the candidates extracted by the generative AI model and uses RAG technology to compare them with public and closed databases, thereby evaluating the reliability of each candidate.

[2240] Step 10:

[2241] The server generates a list of issues

[2242] Based on the results of the match, the server generates a list of issues, including specific corrections such as "grammatical errors" or "incorrect data."

[2243] Step 11:

[2244] The server sends the list of issues to the user's device.

[2245] The list of issues is sent to the user's device in an appropriate format, where the user can view it in the interface.

[2246] Step 12:

[2247] The user confirms the points

[2248] Users can view a list of issues and review the details of each correction on their device, and the emotion engine provides feedback tailored to the user's emotions.

[2249] Step 13:

[2250] User makes corrections

[2251] Users can make corrections based on the suggestions, either by manually entering them or by clicking on the suggested corrections to automatically apply them.

[2252] Step 14:

[2253] User confirms final text

[2254] The user checks the final text after corrections and confirms that there are no problems with the content.

[2255] Step 15:

[2256] The user clicks the save button

[2257] When the user clicks the save button, the final text is sent to the server.

[2258] Step 16:

[2259] The server saves the final text

[2260] The server stores the final text in a database and associates it with the user's account, completing the process.

[2261] Example 2

[2262] 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."

[2263] Modern text generation services require a method to effectively identify errors and suspicious points in user-generated text. However, current systems are unable to provide feedback that takes into account the user's emotions, and the reliability of the generated feedback is not sufficiently guaranteed. As a result, users spend a lot of time correcting the text, and the quality of the final text is difficult to guarantee.

[2264] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[2265] In this invention, the server includes an emotion recognition means for recognizing a user's emotions in real time, a means for adjusting the analysis results of the generative AI model based on the emotion information, and a means for reflecting emotion-based feedback in the list of points to be pointed out. This makes it possible to provide feedback that takes the user's emotions into consideration and improve the reliability and quality of the text.

[2266] "Means for users to log in" refers to the means by which users access the system and perform authentication using their own account information.

[2267] "Means for passing received text data to the generative AI model" refers to the means by which the server receives text data entered by the user and sends it to the generative AI model for analysis.

[2268] "Means for comparing candidates extracted by the generative AI model with a database" refers to a means for comparing grammatical errors, typos, and suspicious points pointed out by the generative AI model with information in a database and evaluating their reliability.

[2269] "Means for generating a list of issues and presenting it to the user" refers to a means for compiling correction suggestions obtained from a generative AI model or database matching into a list and displaying it to the user.

[2270] The "means for the user to select or input correction content" refers to the means by which the user selects the necessary correction from the presented list of issues or inputs the correction themselves.

[2271] The "means for saving the final text" is a means for saving the text after the user has completed the corrections in a database and associating it with the user's account information.

[2272] The "emotion recognition means for recognizing a user's emotions in real time" is a means for recognizing and analyzing the emotions of a user while the user is inputting text.

[2273] "Means for adjusting the analysis results of a generative AI model based on emotional information" refers to a means for optimizing the analysis results of a generative AI model by taking into account the user's emotions and providing more appropriate feedback.

[2274] The "means for reflecting emotion-based feedback in the points to be pointed out list" is a means for adjusting the feedback content displayed in the points to be pointed out list based on the user's emotion information.

[2275] This invention is a system that identifies errors and suspicious points in text created by users of a text generation service, and further recognizes the user's emotions to generate more accurate and reliable text. Specific components and operations of this system will be described below.

[2276] System Configuration

[2277] The system consists of the following main components:

[2278] 1. User Device

[2279] A device that allows a user to input text and receive feedback, such as a PC, smartphone, or tablet.

[2280] The input interface is built with front-end frameworks such as React or Vue.js.

[2281] 2. Server

[2282] This is a central processing unit that receives and analyzes text data entered by the user and executes the generative AI model. The server uses a cloud server such as AWS (Amazon Web Services) or Google Cloud Platform.

[2283] 3. Database

[2284] This is data storage for saving authentication information, text data, analysis results, etc. SQL databases and NoSQL databases are used.

[2285] 4. Generative AI Models

[2286] It is an artificial intelligence model used to analyze text for grammatical errors, typos, and suspicious points. Specifically, it uses large-scale language models such as GPT-3.

[2287] 5. Emotion Recognition Engine

[2288] It is an engine for recognizing user emotions in real time and adjusting the tone of the text. TensorFlow and OpenCV are often used.

[2289] 6. RAG technology

[2290] This technology evaluates the reliability of the findings extracted by the generative AI model by comparing them with public databases and closed data.

[2291] Operation explanation

[2292] Below, we will explain the specific steps involved in the operation of the system and how to implement them.

[2293] 1. The user logs in

[2294] Users use the login interface on their devices to enter their ID and password, which is then encrypted by SSL / TLS and sent to the server.

[2295] 2. The server authenticates the user

[2296] The server authenticates the user by checking the received ID and password against the authentication information in the database. If authentication is successful, a session ID is generated and returned to the user.

[2297] 3. The user enters text

[2298] The user enters the content of a business email or document into the device interface, and when the input is complete, clicks the send button.

[2299] 4. The device sends the entered text to the server

[2300] The entered text data is sent to the server in JSON format, along with the session ID.

[2301] 5. The server passes the text to the generative AI model

[2302] The received text data is passed to the generative AI model, and a request is sent to analyze it for grammatical errors, typos, and suspicious points.

[2303] 6. Generative AI models analyze text

[2304] A generative AI model analyzes text data and extracts insights using natural language processing (NLP) techniques.

[2305] 7. Emotion recognition engine recognizes user emotions

[2306] Recognize and analyze the user's emotions in real time while they are entering text.

[2307] 8. Adjust the analysis results of generative AI models based on emotional information

[2308] Based on the results of the emotion recognition engine, the analysis results are optimized according to the user's emotions.

[2309] 9. The server collates the data using RAG technology

[2310] The findings extracted by the generative AI model are verified using RAG technology to assess their reliability, for example by comparing them with public databases or closed data.

[2311] 10. The server generates a list of issues and sends it to the user's device.

[2312] A list of the verified issues is generated and transmitted to the user terminal based on this information.

[2313] 11. The user checks the issues and makes corrections

[2314] The user can review the list of issues and manually correct each one, or click on the suggested fixes to have them applied automatically.

[2315] 12. Check and save the final text

[2316] The user confirms the final text after correction and clicks the save button, and the server saves this final text in its database and associates it with the user's account.

[2317] Specific examples

[2318] For example, consider a case where a user is composing a business email containing information about a new product. The user first logs in to the system and enters the content of the email. At this time, the emotion engine recognizes the user's emotions in real time and analyzes the tone of the text. Based on the analysis results and emotion information, the generative AI model extracts grammatical errors, typos, and suspicious points. Meanwhile, the server uses RAG technology to compare data and evaluate the reliability of candidates. Along with the generated list of issues, optimal feedback based on the user's emotions is provided. The user can make corrections based on this feedback and save the final text, easily creating reliable business emails.

[2319] Prompt Sentence Examples

[2320] Please review this email and let us know about any grammatical or typo errors or anything suspicious. Also, please adjust the tone of the email to a more calm one.

[2321] The flow of the identification process in the second embodiment will be described with reference to FIG.

[2322] Step 1:

[2323] The user performs a login operation.

[2324] Input: User ID, Password

[2325] Operation: The user accesses the login screen of the text validation UI from the terminal and enters their ID and password.

[2326] Output: Login information is sent to the server.

[2327] Step 2:

[2328] The server authenticates the user

[2329] Input: Login information (user ID, password)

[2330] How it works: The server checks the received ID and password against the information in its database.

[2331] Output: If authentication is successful, a session ID is generated and returned to the user. At the same time, the user information is loaded into memory or cache.

[2332] Step 3:

[2333] The user enters text

[2334] Input: Text of business emails and documents

[2335] Action: A user enters text into the interface on their device and clicks the send button.

[2336] Output: The input text data is stored in the terminal.

[2337] Step 4:

[2338] The device sends the entered text to the server

[2339] Input: Text data, session ID

[2340] What happens: The device sends text data in JSON format to the server, which is sent as an HTTP POST request.

[2341] Output: The server receives the text data and the session ID.

[2342] Step 5:

[2343] The server passes the text to the generative AI model

[2344] Input: Text data

[2345] How it works: The server formats the text data and sends a request to the generative AI model endpoint.

[2346] Output: The generative AI model receives the text data.

[2347] Step 6:

[2348] Generative AI models parse text

[2349] Input: Text data

[2350] How it works: Generative AI models use natural language processing (NLP) techniques to analyze text data for grammatical errors, typos, and other suspicious content.

[2351] Output: A list of issues is generated and returned to the server.

[2352] Step 7:

[2353] Emotion recognition engine recognizes user emotions

[2354] Input: User input status, facial expression data, etc.

[2355] How it works: The emotion recognition engine recognizes the user's emotions in real time, analyzing images and input speed.

[2356] Output: User emotion data is generated and sent to the server.

[2357] Step 8:

[2358] Adjusting the analysis results of generative AI models based on emotional information

[2359] Input: List of points to be pointed out, user's emotional data

[2360] How it works: The server optimizes the analysis results of the generative AI model based on the results of the emotion recognition engine, adjusting the tone of the feedback, etc.

[2361] Output: A reconciled list of issues is generated.

[2362] Step 9:

[2363] The server collates data using RAG technology

[2364] Input: Reconciled issue list

[2365] How it works: The server uses RAG technology to match items in the findings list against public and closed databases.

[2366] Output: The reliability evaluation result is generated.

[2367] Step 10:

[2368] The server generates a list of issues and sends it to the user's device.

[2369] Input: Reliability evaluation results, adjustment results

[2370] Operation: The server generates a final list of issues and sends it to the user's terminal.

[2371] Output: The final issue list is displayed on the user's terminal.

[2372] Step 11:

[2373] The user checks the issues and makes corrections

[2374] Input: Final list of issues

[2375] How it works: The user reviews the list of issues on their device and either manually fixes each one or clicks on the suggested fixes to automatically apply them.

[2376] Output: The modified text data is saved to the terminal.

[2377] Step 12:

[2378] User confirms and saves the final text

[2379] Input: Modified text data

[2380] Action: The user reviews the final text after corrections and clicks the save button.

[2381] Output: The final text is sent to the server.

[2382] Step 13:

[2383] The server saves the final text

[2384] Input: Final text

[2385] How it works: The server stores the final text in a database and associates it with the user's account.

[2386] Output: A confirmation message is displayed to the user indicating that the save was successful.

[2387] (Application example 2)

[2388] 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."

[2389] Conventional text generation services have limited functionality for pointing out errors or suspicious points in text entered by users, making it difficult to provide feedback that takes emotions into account. Furthermore, even in multilingual translation, they have been unable to detect grammatical errors, typos, and suspicious points, or provide translations with a natural tone that reflects the user's emotions. Therefore, the objective of this invention is to provide an advanced text validation system that not only points out errors and suspicious points in text, but also includes multilingual translation and user sentiment analysis.

[2390] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[2391] In this invention, the server includes a means for translating text entered by a user into a different language, a means for detecting grammatical errors, typos, and suspicious points in the translated text, and a means for analyzing the user's emotions and adjusting the tone of the text, thereby enabling the generation of grammatically correct, natural-looking text that reflects the user's emotions even in multiple languages.

[2392] "User" is a person who uses the system to input text and perform translation and validation.

[2393] "Login operation" refers to the authentication procedure required when a user accesses a system.

[2394] A "generative AI model" refers to an artificial intelligence model that analyzes received text data and extracts grammatical errors, typos, and suspicious points.

[2395] A "database" is a data storage device that stores information used to verify the reliability of candidates extracted by a generative AI model.

[2396] "Translation facility" refers to a facility that converts user-entered text into a different language.

[2397] "Grammar error detector" refers to a function that detects grammatical errors present in the translated text.

[2398] "Sentiment analysis means" refers to the function that analyzes the user's emotions when typing and adjusts the tone of the text.

[2399] The "list of points of criticism" is a list of information summarizing errors and corrections detected by the generative AI model, grammar error detection means, and sentiment analysis means.

[2400] The "means for selecting correction content" refers to a function that allows the user to select or newly input the content to be corrected based on the list of pointed out points.

[2401] "Final text" refers to text data that has been corrected by the user, finally confirmed, and saved.

[2402] "RAG technology" refers to a technology that compares candidates extracted by a generative AI model with existing data and evaluates their reliability.

[2403] MODE FOR CARRYING OUT THE INVENTION

[2404] System Program

[2405] The system that realizes this application example includes the following programs that perform a series of functions: user authentication, text input, text processing, sentiment analysis, grammar error detection, translation, presenting a list of corrections to the user, and saving the corrections.

[2406] Program processing

[2407] The server provides a means for users to log in when they access the system. The login operation uses a user authentication system using Django to verify the ID and password.

[2408] When the user enters text, the terminal sends the entered text data to the backend, which uses React for the frontend, making the user interface intuitive and responsive.

[2409] The server passes the received text data to a generative AI model, which can use BERT, for example, to detect grammatical errors, typos, and other suspicious aspects of the text. The output from the generative AI model is then sent back to the server.

[2410] The server then compares the candidates extracted by the generative AI model against a database of existing data and uses RAG technology to assess the reliability of the candidates.

[2411] Additionally, the server provides a means to analyze the user's emotions, using the Google Cloud Natural Language API to calculate an emotion score in real time as the user types.

[2412] Based on the results of the sentiment analysis, the server adjusts the tone of the text and performs the translation function, which uses the Google Cloud Translation API to translate the text entered by the user into different languages.

[2413] Based on the results of these processes, the server generates a list of issues and presents it to the user. The user can check the list on their device and carefully examine the details of each correction. If there are grammatical errors or misleading expressions, the user can correct them or select the correction suggested by the system.

[2414] Finally, the final text that the user has confirmed is stored on the server, and this final text is highly reliable and in line with the user's sentiment and intent.

[2415] Hardware and software used

[2416] Hardware: Smartphones, computers, and cloud-based server environments.

[2417] software:

[2418] Frontend: React

[2419] Backend: Django

[2420] Translation API: Google Cloud Translation API

[2421] Sentiment analysis: Google Cloud Natural Language API

[2422] Grammar Error Detection: BERT (TensorFlow / PyTorch)

[2423] Specific examples

[2424] For example, a user might enter the text "Good morning, how can I help you today?" in Japanese. The application translates this text into English and displays "Good morning, how can I help you today?". At the same time, it performs sentiment analysis to ensure the user maintains a friendly tone. The generative AI model detects grammatical errors, typos, and other suspicious points and presents a list of issues to the user.

[2425] Prompt example

[2426] User Input: "Good morning, how are you doing today?"

[2427] -> Translate to English

[2428] -> Detect grammar issues

[2429] -> Analyze sentiment score

[2430] In this way, the system provides grammatically accurate text generation across multiple languages ​​and feedback that reflects the user's sentiment, making user communication more efficient and reliable.

[2431] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[2432] Step 1:

[2433] The server provides a login screen for the user to log in. The user enters their ID and password and clicks the login button. The server receives this input data and checks it against the authentication information in the database. If authentication is successful, a session ID is generated and the user information is loaded into the system.

[2434] Input: User ID and password

[2435] Output: Session ID and user information

[2436] Step 2:

[2437] The terminal displays a text input interface to the authenticated user. The user enters text and clicks the send button to send the input text to the server. The terminal then sends the entered text data to the server in JSON format, including the session ID.

[2438] Input: Text entered by the user

[2439] Output: JSON format text data and session ID

[2440] Step 3:

[2441] The server passes the received text data to a generative AI model. The generative AI model (e.g., BERT) analyzes the text data and extracts grammatical errors, typos, and suspicious points. The server then sends a request to the generative AI model's endpoint to obtain the analysis results.

[2442] Input: Received JSON format text data

[2443] Output: Analysis results from the generative AI model

[2444] Step 4:

[2445] The server compares the candidates extracted by the generative AI model with a database, using RAG technology to compare the candidates with public and closed databases and evaluate their reliability.

[2446] Input: Analysis results from generative AI model

[2447] Output: Matching results and confidence scores for each candidate

[2448] Step 5:

[2449] The server passes the received text data to the emotion engine to analyze the user's emotions. The emotion engine calculates the user's emotion score in real time using Google Cloud Natural Language API and other tools.

[2450] Input: Received text data

[2451] Output: Sentiment score

[2452] Step 6:

[2453] The server adjusts the tone of the text based on the results of sentiment analysis, and uses the Google Cloud Translation API to translate the text into different languages. Finally, it generates the evaluated translation results and adjusted text.

[2454] Input: Sentiment scores and received text data

[2455] Output: Adjusted translated text

[2456] Step 7:

[2457] The server generates and presents a list of issues to the user, including grammar, typos, suspicious points, and feedback based on sentiment scores, and sends this information to the front-end in an appropriate format.

[2458] Input: Evaluated translated text, sentiment score, and analysis results

[2459] Output: List of issues

[2460] Step 8:

[2461] The user can view a list of issues on their device, review the details of each correction, and make any necessary changes. They can either manually correct the text or click on the system's suggested corrections to automatically apply them.

[2462] Input: List of issues

[2463] Output: User's corrected text

[2464] Step 9:

[2465] The user finally checks the corrected text and clicks the save button. The server saves the final text in its database and associates it with the user's account.

[2466] Input: Final text as modified by the user

[2467] Output: Final text stored in the database

[2468] 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.

[2469] 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.

[2470] 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.

[2471] 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.

[2472] FIG. 9 illustrates 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 behaviors 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.

[2473] 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.

[2474] 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).

[2475] 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.

[2476] 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."

[2477] 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.

[2478] 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).

[2479] 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.

[2480] 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.

[2481] 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.

[2482] 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.

[2483] 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.

[2484] 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.

[2485] 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.

[2486] 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.

[2487] 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.

[2488] 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.

[2489] The following is further disclosed regarding the above embodiment.

[2490] (Claim 1)

[2491] A means by which a user can log in;

[2492] A means for passing the received text data to the generative AI model;

[2493] A means for matching candidates extracted by the generative AI model with a database; and

[2494] means for generating a list of issues and presenting it to a user;

[2495] a means for the user to select or input modifications;

[2496] a means for saving the final text;

[2497] A system including:

[2498] (Claim 2)

[2499] 10. The system of claim 1, further comprising means for matching candidates extracted by the generative AI model using RAG techniques.

[2500] (Claim 3)

[2501] 10. The system of claim 1, further comprising means for generating final text that reflects user selected or entered modifications.

[2502] "Example 1"

[2503] (Claim 1)

[2504] A means for users to log in from their terminals;

[2505] a means by which the server authenticates the user;

[2506] a means for a user to input text into the terminal's interface;

[2507] means for the terminal to transmit the entered text to a server;

[2508] A means for the server to pass the received text data to the generative AI model;

[2509] A generative AI model analyzes text and extracts grammatical errors, typos, and suspicious points.

[2510] A means for the server to compare the candidates extracted by the generated AI model with a database using RAG technology;

[2511] A means for the server to generate a list of points to be pointed out based on the collation result;

[2512] A means for the server to transmit the list of points to the user's terminal;

[2513] A means for users to check and correct the issues pointed out on their device, and

[2514] A means for the user to review and save the final text after correction;

[2515] a means for the server to store the final text in a database;

[2516] A system including:

[2517] (Claim 2)

[2518] 2. The system of claim 1, wherein the server includes means for using RAG technology to match candidates extracted by the generative AI model with a database.

[2519] (Claim 3)

[2520] 10. The system of claim 1, further comprising means for generating final text that reflects user selected or entered modifications.

[2521] "Application Example 1"

[2522] (Claim 1)

[2523] A means for a user to perform authentication operations;

[2524] A means for passing the received text data to the generative AI model;

[2525] A means for matching the candidates extracted by the generative AI model with data storage locations; and

[2526] means for generating a list of issues and presenting it to a user;

[2527] a means for the user to select or input modifications;

[2528] a means for saving the final text;

[2529] A means to analyze voice-entered text in real time and point out errors and security risks,

[2530] a means of displaying suggested revisions in real time;

[2531] A means of checking compliance with industry regulations and internal rules,

[2532] A system including:

[2533] (Claim 2)

[2534] 10. The system of claim 1, further comprising means for matching candidates extracted by the generative AI model using RAG techniques.

[2535] (Claim 3)

[2536] 10. The system of claim 1, further comprising means for generating final text that reflects user selected or entered modifications.

[2537] "Example 2: Combining Emotion Engines"

[2538] (Claim 1)

[2539] A means by which a user can log in;

[2540] A means for passing the received text data to the generative AI model;

[2541] A means for matching candidates extracted by the generative AI model with a database; and

[2542] means for generating a list of issues and presenting it to a user;

[2543] a means for the user to select or input modifications;

[2544] a means for saving the final text;

[2545] An emotion recognition means for recognizing a user's emotion in real time;

[2546] A means for adjusting the analysis results of the generative AI model based on emotional information; and

[2547] A means of incorporating emotional feedback into the list of issues;

[2548] A system including:

[2549] (Claim 2)

[2550] 10. The system of claim 1, further comprising means for matching candidates extracted by the generative AI model using RAG techniques.

[2551] (Claim 3)

[2552] 10. The system of claim 1, further comprising means for generating final text that reflects user selected or entered modifications.

[2553] "Application example 2 when combining emotion engines"

[2554] (Claim 1)

[2555] A means by which a user can log in;

[2556] A means for passing the received text data to the generative AI model;

[2557] A means for matching candidates extracted by the generative AI model with a database; and

[2558] A means for translating user-entered text into different languages;

[2559] A means of detecting grammatical errors, typos, and suspicious points in the translated text,

[2560] A means of analyzing user sentiment and adjusting the tone of the text;

[2561] means for generating a list of issues and presenting it to a user;

[2562] a means for the user to select or input modifications;

[2563] a means for saving the final text;

[2564] A system including:

[2565] (Claim 2)

[2566] 10. The system of claim 1, further comprising means for matching candidates extracted by the generative AI model using RAG techniques.

[2567] (Claim 3)

[2568] 10. The system of claim 1, further comprising means for generating final text that reflects user selected or entered modifications. [Explanation of symbols]

[2569] 10, 210, 310, 410 Data Processing Systems 12 Data Process...

Claims

1. A means by which a user can log in; A means for passing the received text data to the generative AI model; A means for matching candidates extracted by the generative AI model with a database; and means for generating a list of issues and presenting it to a user; a means for the user to select or input modifications; a means for saving the final text; A system including:

2. The system of claim 1 , further comprising means for matching candidates extracted by the generative AI model using RAG techniques.

3. 2. The system of claim 1, further comprising means for generating final text that reflects modifications selected or entered by the user.

Images