System

Abstract

A system is provided.SOLUTION: A system comprising: means for receiving input data; means for pre-checking the received input data and detecting inappropriate content; means for sending input data determined to be appropriate to a generative artificial intelligence model and receiving a generated answer; means for pre-checking the generated answer and detecting inappropriate content; and means for sending the answer determined to be appropriate to a user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] In recent years, as generative artificial intelligence (AI) has become increasingly used in business, it is common for text data provided by users to be sent directly to the AI, and for the generated answers to be returned directly to the user. This situation poses a risk that confidential information that the user did not intend to disclose may be misused by the AI, increasing the risk of information leakage. Therefore, effective security measures to prevent the leakage of confidential information when using AI are required, but the current problem is that such measures have not been fully established. [Means for solving the problem]

[0005] In order to solve the above problems, the present invention provides a system that provides a security service between a user and a generative AI model, enabling the safe use of generative AI. The present invention is a system that includes the following means.

[0006] The system is configured to include a means for receiving input data, a means for checking the received input data in advance and detecting inappropriate content, a means for sending input data that is judged to be appropriate to a generative artificial intelligence model and receiving a generated answer, a means for checking the generated answer in advance and detecting inappropriate content, and a means for sending an answer that is judged to be appropriate to a user.

[0007] In particular, by having the checking means have the function of checking whether the input data contains specific keywords, and also including the function of checking whether the generated answer exceeds a predetermined number of characters, it is possible to significantly reduce the risk of leaking confidential information of users and provide a high level of security in the business use of generation AI.

[0008] "Input data" refers to text data that a user provides to a generative artificial intelligence model.

[0009] The "means for receiving" is a function for acquiring input data via a network.

[0010] "Pre-checking means" is a function for inspecting the content of input data and generated answers to detect inappropriate content.

[0011] "Inappropriate content" means confidential information not intended for public disclosure or data that violates established terms.

[0012] "Judgmented as appropriate" means that the item has been confirmed as not inappropriate through prior inspection using checking means.

[0013] A "generative artificial intelligence model" is an artificial intelligence algorithm that generates answers based on input data provided by the user.

[0014] The "generated answer" is response data generated by the generative artificial intelligence model in response to input data.

[0015] "Means for sending to user" is a function for returning the answer generated by the generative artificial intelligence model to the user.

[0016] A "security service" is an intermediate service placed between the user and the generative AI model to reduce risks such as information leaks. [Brief explanation of the drawings]

[0017] [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

[0018] 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.

[0019] First, the terms used in the following description will be explained.

[0020] 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).

[0021] 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.

[0022] 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.

[0023] 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.

[0024] 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."

[0025] [First embodiment]

[0026] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0027] 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.

[0028] 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).

[0029] 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.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0034] 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.

[0035] 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.

[0036] 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.

[0037] 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."

[0038] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[0039] overview

[0040] The present invention is a system that checks input data provided by users before sending it to a generative artificial intelligence model, and similarly checks the generated answers, providing security services to prevent the leakage of confidential information.

[0041] Basic system configuration

[0042] 1. Users

[0043] The user inputs text data through the terminal.

[0044] Example: User types, "This is a sensitive request."

[0045] 2. Terminal

[0046] It is responsible for sending text data entered by the user to the server.

[0047] Example: Submitting data by clicking a button on a web form or application.

[0048] 3. Server

[0049] The server creates an instance of the "SecurityService" class and receives input data.

[0050] The server first checks the received data using the "checking means." If the data contains inappropriate content, it returns an error message to the user saying "Input is inappropriate."

[0051] If the input data does not contain inappropriate content, the input data is passed to a generative artificial intelligence model and the generated answer is received.

[0052] The server also checks the generated answer using a "checking method." If the answer contains inappropriate content, it returns a message to the user saying, "The answer is inappropriate."

[0053] If it is determined to be an appropriate answer, it returns the answer to the user.

[0054] Specific examples

[0055] 1. Receiving and checking user input

[0056] The user types "This is a general request" into the terminal.

[0057] The terminal sends this input data to the server.

[0058] The server uses the "check_incoming_data" method to check the incoming data.

[0059] 2. Answer generation using generative AI models

[0060] The server passes input data that it deems appropriate to the generative artificial intelligence model.

[0061] The generative artificial intelligence model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[0062] 3. Check and submit the generated answers

[0063] The server checks the response using the "check_outgoing_data" method.

[0064] If the server determines that the answer is appropriate, it returns the answer to the terminal.

[0065] Example: "Generated AI response: This is a common request." is displayed to the user.

[0066] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[0067] The processing flow will be explained below.

[0068] Step 1:

[0069] A user inputs text data into a terminal.

[0070] Example: User types, "This is a general request."

[0071] Step 2:

[0072] The terminal transmits the user's input data to the server.

[0073] Example: The terminal clicks the send button and requests the input data from the server.

[0074] Step 3:

[0075] The server receives input data from the user.

[0076] Example: The server receives data through a REST API.

[0077] Step 4:

[0078] The input data received by the server is checked using the "check_incoming_data" method.

[0079] Example: This method checks for sensitive information or inappropriate keywords.

[0080] Step 5:

[0081] The server determines whether the input data is valid.

[0082] If deemed appropriate, the data is passed to a generative AI model.

[0083] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0084] Example: If the input data is "This is a request containing sensitive information," the server responds with "Invalid input."

[0085] Step 6:

[0086] The server passes input data that it deems appropriate to the generative AI model to generate an answer.

[0087] Example: Call the "generate" method of "AIModel" and pass in the user's input data.

[0088] Step 7:

[0089] A generative AI model generates answers based on user input data.

[0090] Example: Generate the answer "Generation AI answer: This is a common request."

[0091] Step 8:

[0092] The server receives the generated answer.

[0093] Example: Receiving a response from a generative AI model.

[0094] Step 9:

[0095] The server checks the generated answer with the "check_outgoing_data" method.

[0096] Example: This method checks if the answer is longer than a predefined number of characters.

[0097] Step 10:

[0098] The server determines whether the generated answer is appropriate.

[0099] If it is judged to be appropriate, it returns a response to the terminal.

[0100] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0101] Example: If the answer exceeds 500 characters, the server will respond with "The answer is inappropriate."

[0102] Step 11:

[0103] The terminal receives the appropriate response from the server and displays it to the user.

[0104] Example: The user will see the answer "Generated AI answer: This is a common request."

[0105] The above are the specific steps of the processing of the present invention, which can effectively reduce the risk of confidential information leakage when using generative AI models and provide users with safe and reliable services.

[0106] Example 1

[0107] 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."

[0108] While modern dialogue systems and data processing systems using generative artificial intelligence models are extremely useful, they pose a significant risk of confidential information leakage. Security measures are essential, particularly since both the input data provided by users and the generated answers may contain confidential information. However, these systems require effective means to pre-check the input data and generated answers and filter out inappropriate content.

[0109] 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.

[0110] In this invention, the server includes: means for receiving input data from a user; means for checking the received input data in advance to detect inappropriate content; means for sending input data determined to be appropriate to the generative AI model and receiving a generated answer; means for checking the generated answer in advance to detect inappropriate content; means for sending the answer determined to be appropriate to the user; means for converting the input data to be sent to the generative AI model into a prompt format in advance; means for converting the answer received from the generative AI model into a format displayable to the user; means for checking the input data and the generated answer by applying specific security rules to prevent leakage of confidential information; means for saving all data including the input data from the user and the answer from the generative AI model as a log; and means for periodically auditing the saved log data. This makes it possible to effectively prevent leakage of confidential information during the process of using the generative AI model.

[0111] "User" means a person who utilizes the System to input and receive data.

[0112] "Terminal" means a device operated by a user and used to input data and transmit data to a server.

[0113] "Server" refers to the central computer system that receives and validates input data from users and works with the generative artificial intelligence model to generate and validate answers.

[0114] "Input data" refers to data such as text sent by a user to a server via a terminal.

[0115] A "generative artificial intelligence model" refers to an AI system that generates appropriate answers based on input data from users.

[0116] "Prompt sentence" refers to an instruction sentence formatted to appropriately transmit input data to a generative artificial intelligence model.

[0117] "Checking means" refers to a function for analyzing input data and generated answers to check for inappropriate content or confidential information.

[0118] "Inappropriate content" refers to data that contains confidential or security-related information.

[0119] "Security rules" refer to analytical standards and procedures established to prevent the leakage of confidential information.

[0120] "Log" refers to a record of all data processed within the system, such as user input data and generated responses.

[0121] "Audit" refers to the process of periodically reviewing stored log data to verify the operation and security status of a system.

[0122] The present invention is a system for preventing the leakage of confidential information by checking input data provided by a user before sending it to a generative artificial intelligence model, and by similarly checking the generated answers.

[0123] Basic system configuration

[0124] 1. User:

[0125] The user is responsible for inputting and transmitting text data through the terminal.

[0126] Example: A user types, "This is a general request."

[0127] 2. Terminal:

[0128] The terminal is responsible for transmitting text data entered by the user to the server.

[0129] Example: Submitting data by clicking a button on a web form or application.

[0130] 3. Server:

[0131] The server performs the following process:

[0132] 1. Data reception:

[0133] The server creates an instance of the "SecurityService" class and receives input data from the user.

[0134] Example: Use the "check_incoming_data" method to check incoming data for inappropriate content.

[0135] 2. Input data validation:

[0136] If the data contains inappropriate content, the server sends an error message "Input is inappropriate" to the terminal.

[0137] Example: If the user types "This is confidential information", the server generates an error message.

[0138] 3. Data transfer to the generative AI model:

[0139] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[0140] Example: If the input is deemed appropriate, the data is passed to the generative AI model using the prompt "Please output the generative AI's answer."

[0141] 4. Receive a response from the generative AI model:

[0142] The server processes the answers received from the generative artificial intelligence model.

[0143] Example: A generative artificial intelligence model responds with, "Generative AI response: This is a common request."

[0144] 5. Check the generated data:

[0145] The server checks the generated data using the "check_outgoing_data" method.

[0146] Example: The server checks the generated data again using pre-filtering rules and databases.

[0147] 6. Error Handling:

[0148] If the answer contains inappropriate content, the server sends an error message "The answer is inappropriate" to the terminal.

[0149] Action: The server formats an appropriate error message and sends it to the terminal.

[0150] 7. Submitting the right response:

[0151] The server sends the answer it deems appropriate to the user.

[0152] Example: "Generated AI response: This is a common request." is sent to the user.

[0153] Examples and prompt formats

[0154] 1. Receive and check user input:

[0155] The user types "This is a general request" into the terminal.

[0156] The terminal sends this input data to the server.

[0157] The server checks the incoming data with the "check_incoming_data" method.

[0158] 2. Answer generation using generative AI models:

[0159] The server sends input data that it deems appropriate to the generative AI model.

[0160] The generative AI model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[0161] 3. Check and submit the generated answers:

[0162] The server checks the generated answer using the "check_outgoing_data" method.

[0163] The server returns an answer that it deems appropriate to the terminal.

[0164] Example: "Generated AI response: This is a common request." is displayed to the user.

[0165] Prompt Sentence Examples

[0166] Please output the generative AI's answer: This is a common request.

[0167] In this way, this system effectively reduces the risk of confidential information leaks when using generative AI. By inspecting the text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[0168] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0169] Step 1:

[0170] The user enters text data into the terminal.

[0171] Input: Text data entered by the user, e.g. "This is a general request."

[0172] Specific action: The user types characters on the keyboard and presses the Enter key.

[0173] Output: Input data is displayed on the terminal.

[0174] Step 2:

[0175] The terminal transmits the input text data to the server.

[0176] Input: Text data entered by the user, e.g., "This is a general request."

[0177] Specific operation: When you click the send button on the device, the browser generates an HTTP POST request and sends it to the server.

[0178] Output: Data is sent to the server.

[0179] Step 3:

[0180] The server receives the data from the user and checks the content using the "check_incoming_data" method.

[0181] Input: Text data sent from the terminal, e.g., "This is a general request."

[0182] Specific operation: The server creates an instance of the "SecurityService" class and analyzes the data using the database and filtering rules.

[0183] Data manipulation: Detecting inappropriate content using specific keywords and pattern matching.

[0184] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[0185] Step 4:

[0186] If the server determines that the data is invalid, it sends an error message to the terminal saying "Invalid input."

[0187] Input: Inspection results, e.g., inappropriate data "This is confidential information."

[0188] Specific behavior: The server generates an HTTP response and sends data including an error message.

[0189] Output: An error message is displayed on the terminal.

[0190] Step 5:

[0191] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[0192] Input: Inspection result, e.g. "This is a common request."

[0193] Specific operation: The server generates an appropriate prompt sentence and sends the data to the API of the generative artificial intelligence model.

[0194] Data processing: Convert the data into the format of the prompt statement.

[0195] Output: The prompt sent to the generative AI model.

[0196] Step 6:

[0197] The server receives the answer from the generative artificial intelligence model.

[0198] Input: An answer from a generative AI model, e.g., "Generative AI answer: This is a common request."

[0199] Specific operation: The server receives the HTTP response and parses the response body.

[0200] Output: The received response data.

[0201] Step 7:

[0202] The server checks the generated data using the "check_outgoing_data" method.

[0203] Input: Answer data from a generative AI model, e.g., "Generative AI answer: This is a common request."

[0204] Specific operation: The server examines the data using a filtering algorithm.

[0205] Data processing: Check that the response data does not contain inappropriate content.

[0206] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[0207] Step 8:

[0208] If the server generates an answer that contains inappropriate content, it sends an error message to the terminal saying "Answer is inappropriate."

[0209] Input: Test results, e.g., incorrect response data.

[0210] Specific behavior: The server generates an HTTP response and returns data including an error message.

[0211] Output: An error message is displayed on the terminal.

[0212] Step 9:

[0213] The server sends the answer that it deems appropriate to the terminal.

[0214] Input: Inspection result, e.g. "Generated AI response: This is a common request."

[0215] Specific behavior: The server generates an HTTP response and sends back data containing the appropriate answer.

[0216] Output: The answer of the generated AI is displayed on the terminal.

[0217] (Application example 1)

[0218] 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."

[0219] In systems using generative artificial intelligence models, if the input data or generated answers contain confidential information, it is important to prevent the leakage of that information. However, many current systems lack a means to automatically verify the appropriateness of the input data or generated answers, which poses a risk of confidential information leaking to the outside. The present invention aims to solve this problem and prevent the leakage of confidential information.

[0220] 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.

[0221] In this invention, the server includes means for receiving input data, means for checking the input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, and means for including a mobile application that runs on a smartphone and has a security function for preventing leakage of confidential information. This makes it possible to automatically perform security checks on the input data and the generated answers and prevent leakage of confidential information.

[0222] "Input Data" means text or other information provided to a system by a user.

[0223] "Means for receiving" refers to a device or method for receiving input data provided by a user.

[0224] "Pre-checking measures" refers to functions and methods for automatically inspecting input data and generated responses to check for inappropriate content or confidential information.

[0225] "Means for detecting inappropriate content" means a device or method for identifying inappropriate elements or information in input data or generated answers.

[0226] A "generative artificial intelligence model" refers to an artificial intelligence algorithm or system that generates answers based on input data provided by a user.

[0227] "Means for receiving a generated answer" refers to a device or method for receiving an answer generated from a generative artificial intelligence model.

[0228] "Means for sending answers that are judged to be appropriate to the user" refers to a device or method for returning a checked answer to the user if the checked answer is judged to be appropriate.

[0229] A "smartphone" is a type of mobile phone, a portable information terminal that can connect to the Internet and run various applications.

[0230] "Security features to prevent the leakage of confidential information" refers to features and methods to prevent confidential information from leaking to the outside if it is included in input data or generated answers.

[0231] "Mobile application" refers to a software application that runs on a mobile information device such as a smartphone or tablet.

[0232] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[0233] System Configuration

[0234] 1. Users

[0235] A user operates an application using a smartphone to input text data.

[0236] Example: User types "Summarize the meeting."

[0237] 2. Terminal

[0238] The terminal is responsible for receiving text data entered by the user and transmitting it to the server.

[0239] Specifically, data is submitted by clicking a button on a web form or application.

[0240] 3. Server

[0241] The server receives the input data and performs security checks by creating an instance of the "SecurityService" class and validating the input data.

[0242] The "check_incoming_data" method is used to check whether inappropriate content is included. If inappropriate content is included, the error message "Input is inappropriate" is returned to the terminal.

[0243] Input data that is deemed appropriate is sent to the generative artificial intelligence model, which then receives the generated answer.

[0244] The generated answer is inspected using the "check_outgoing_data" method. If it contains inappropriate content, the message "Answer is inappropriate" is returned to the terminal.

[0245] If the answer is determined to be appropriate, the answer is returned to the terminal and displayed to the user.

[0246] Data processing and hardware / software used

[0247] Hardware: Smartphone (iOS / Android)

[0248] Software: Python, Flask (web framework), TensorFlow / Transformers (generative AI models), SQL database

[0249] Data calculation: Input data checks and answer checks are performed by the "check_incoming_data()" and "check_outgoing_data()" methods in the "SecurityService" class.

[0250] Specific examples

[0251] When a user types "This is a general request" into a smartphone application, the content is sent to the server. The server checks this input data using the "check_incoming_data" method and sends it to the generative AI model only if it is deemed appropriate.

[0252] The generative AI model generates an answer, "Generative AI's answer: This is a typical request," and returns it to the server. The server then checks the answer again using the "check_outgoing_data" method, and if it is deemed appropriate, returns the answer to the user.

[0253] Example prompt sentence:

[0254] User input: Check if the content of 'Summarize the meeting' is typical, and if it is, run a function that provides appropriate feedback.

[0255] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[0256] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0257] Step 1:

[0258] The user opens the application on their smartphone and enters text. The input data is "Summarize the contents of the meeting." The device receives this user input and sends it to the server.

[0259] Input: User's text data "Summarize the contents of the meeting"

[0260] Output: Text data sent from the terminal to the server

[0261] Step 2:

[0262] The server checks the input data received from the terminal using the "check_incoming_data" method of the "SecurityService" class, which checks whether the input data contains confidential information or inappropriate content.

[0263] Input: Text data received by the server: "Summarize the contents of the meeting"

[0264] Data processing: Checks to see if there is any confidential information

[0265] Output: Check result (good or bad)

[0266] Specific behavior:

[0267] The text data received by the server is passed to the "check_incoming_data" method, which checks whether it contains specific keywords. For example, there is a rule that detects keywords such as "confidential" and "important."

[0268] Step 3:

[0269] If the server checks the results, it sends the text data to a generative AI model, which receives a specific prompt and generates an appropriate answer based on it.

[0270] Input: Text data that passed the check: "Summarize the contents of the meeting"

[0271] Data processing: Passing the data as prompts to the generative AI model

[0272] Output: The answer generated by the generative AI model

[0273] Specific behavior:

[0274] The server passes appropriate input data to the generative AI model as a prompt. For example, it requests the model to generate a summary of the user's input: 'Summarize the contents of the meeting.'

[0275] Step 4:

[0276] The generative AI model generates an answer based on the prompt and returns it to the server, for example, in the form of a "generated summary of the meeting content."

[0277] Input: Prompt sentence for generative AI model

[0278] Data computation: Generative AI models generate answers

[0279] Output: The generated answer

[0280] Specific behavior:

[0281] A generative AI model uses natural language processing based on the prompt to generate an appropriate answer.

[0282] Step 5:

[0283] The server checks the generated answer in advance using the "check_outgoing_data" method of the "SecurityService" class to ensure that the generated answer does not contain inappropriate content.

[0284] Input: Generated Answer

[0285] Data processing: Checking generated answers

[0286] Output: Check result (good or bad)

[0287] Specific behavior:

[0288] The server passes the generated answer to the "check_outgoing_data" method to check whether it contains specific keywords or phrases.

[0289] Step 6:

[0290] If the check result is appropriate, the server returns the generated answer to the terminal and displays it to the user. If the check result is inappropriate, the server returns an error message saying "The answer is inappropriate."

[0291] Input: Answer that passes the check

[0292] Data processing: Formatting answers (if necessary)

[0293] Output: The appropriate answer or error message that is displayed to the user

[0294] Specific behavior:

[0295] The server sends the answer that passes the check to the terminal, and the answer is displayed to the user by the smartphone application. For example, the terminal executes the operation "display the data received from the server."

[0296] 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.

[0297] This invention is a system that uses a generative artificial intelligence model to check the appropriateness of the user's input data and the generated answers, and also takes the user's emotions into consideration, thereby reducing the risk of confidential information leaks and providing more personalized responses. Below, we will explain in detail the main roles of the user, terminal, and server, and the specific processes performed by each.

[0298] overview

[0299] In addition to means for checking input data, means for using a generative artificial intelligence model, and means for checking generated answers, the present invention also includes an emotion engine that recognizes the user's emotions and reflects them in various checks and response generation.

[0300] Basic system configuration

[0301] 1. Users

[0302] The user inputs text data through the terminal.

[0303] Input can also include emotional states.

[0304] Example: A user types, "This is a general request," and their emotion is recognized as "calm."

[0305] 2. Terminal

[0306] It is responsible for sending the text data and emotional information entered by the user to the server.

[0307] Example: The device clicks the send button and requests emotion information along with the input data from the server.

[0308] 3. Server

[0309] The server has means for receiving the text data and the emotion information.

[0310] The received data is inspected using the "check_incoming_data" method, and the emotion engine analyzes the user's emotion information.

[0311] Example: The server further scrutinizes the input data based on a "calm" emotional state.

[0312] 4. Data appropriateness assessment

[0313] If the server determines that the input data is appropriate, it passes the data and emotional information to a generative AI model to generate an answer.

[0314] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0315] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[0316] 5. Use of generative AI models

[0317] The generative AI model generates answers based on the received text data and emotional information.

[0318] The emotion engine generates answers that reflect the user's current emotional state.

[0319] Example: Generated response: "Generated AI response: This is a common request. We're glad you're calm."

[0320] 6. Check the generated answers

[0321] The server checks the generated answer again with the "check_outgoing_data" method.

[0322] The emotion engine also evaluates the appropriateness of the generated answers based on emotional information.

[0323] For example: Make sure your answer is within the character limit and emotionally relevant.

[0324] 7. Sending the Response

[0325] If the answer is judged to be appropriate, it is returned to the device and displayed to the user in an appropriate state that reflects the emotional information.

[0326] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0327] Example: The user sees the message, "Generated AI response: This is a common request. We're glad you're calm."

[0328] Specific examples

[0329] 1. Receiving and checking user input

[0330] The user types "This is a general request" into the device, and the system recognizes the user's emotional state as "calm."

[0331] The device sends this input data and emotional information to the server.

[0332] 2. Answer generation using generative AI models

[0333] The server checks the data and emotion information using the "check_incoming_data" method.

[0334] Data and emotional information deemed appropriate are sent to the generative AI model.

[0335] The generative AI model generates an answer, "Generative AI's answer: This is a common request. We're glad you're calm." and returns it to the server.

[0336] 3. Check and submit the generated answers

[0337] The server checks the response with the "check_outgoing_data" method.

[0338] If it is determined to be an appropriate answer, it returns the answer to the terminal.

[0339] The generated answer is displayed to the user.

[0340] These are the specific steps of the process of the present invention. This system reduces the risk of leaking confidential user information when using generative AI models, and can provide personalized responses that take user emotions into account.

[0341] The processing flow will be explained below.

[0342] Step 1:

[0343] A user inputs text data into a terminal.

[0344] Example: User types, "This is a general request."

[0345] Step 2:

[0346] The terminal collects the user's input data as well as the emotional state.

[0347] Example: The device recognizes the user's emotion as "calm" through facial recognition and voice tone analysis.

[0348] Step 3:

[0349] The device transmits the collected text data and emotion information to a server.

[0350] Example: The device clicks the send button and sends the input data and emotion information as a request to the server.

[0351] Step 4:

[0352] The server receives input data and emotion information from the user.

[0353] Example: The server receives data and emotion information through a REST API.

[0354] Step 5:

[0355] The input data received by the server is checked using the "check_incoming_data" method.

[0356] Example: This method checks for sensitive information or inappropriate keywords.

[0357] Step 6:

[0358] The server analyzes the received emotion information using an emotion engine.

[0359] Example: The server analyzes the emotional information of "calm" and further scrutinizes the input data based on that information.

[0360] Step 7:

[0361] The server determines whether the input data is valid.

[0362] If deemed appropriate, the data and emotional information is passed to a generative AI model.

[0363] If it is determined to be inappropriate, an error message "Input is inappropriate" is returned and the process is terminated.

[0364] For example, "This is a request containing sensitive information." If the data is deemed inappropriate, the server will block it.

[0365] Step 8:

[0366] The server passes the input data and emotional information it deems appropriate to the generative AI model to generate an answer.

[0367] Example: Call the "generate" method of the generative AI model, passing in the user's input data and emotion information.

[0368] Step 9:

[0369] The generative AI model generates answers based on the user's input data and emotional information.

[0370] Example: The generative AI model generates the answer "Generative AI response: This is a common request. We're glad you're calm."

[0371] Step 10:

[0372] The server receives the generated answer.

[0373] Example: Receiving an answer as a response from a generative AI model.

[0374] Step 11:

[0375] The server checks the generated answer again with the "check_outgoing_data" method.

[0376] For example, this method checks whether the answer is within a predefined character limit and whether it is sentiment-based.

[0377] Step 12:

[0378] The server determines whether the generated answer is appropriate.

[0379] If it is determined to be appropriate, the answer is sent to the terminal.

[0380] If it is determined to be inappropriate, an error message "The answer is inappropriate" is returned and the process is terminated.

[0381] Example: If the answer exceeds 500 characters, the server will block it as an inappropriate answer.

[0382] Step 13:

[0383] The terminal receives the appropriate response from the server and displays it to the user.

[0384] Example: The user will see the answer, "Generator's response: This is a common request. We're glad you're calm."

[0385] These are the specific steps of the process of the present invention. This system reduces the risk of confidential information leakage when using generative AI models and can provide personalized responses that take user emotions into account.

[0386] Example 2

[0387] 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."

[0388] Conventional systems using generative artificial intelligence models have had difficulty accurately evaluating the appropriateness of user input data and generated answers. Furthermore, because they do not take the user's emotions into account, they are unable to provide personalized responses, resulting in reduced user satisfaction. Furthermore, there is a risk of data leakage due to inappropriate data content or emotional state. The present invention aims to solve these problems, improve the appropriateness of user input data and generated answers, and provide personalized responses that take the user's emotions into account.

[0389] 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.

[0390] In this invention, the server includes means for receiving text data and the user's emotional information from the user, means for analyzing the received text data and emotional information and detecting inappropriate content or an inappropriate emotional state, means for transmitting the text data and emotional information determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for analyzing the generated answer and the emotional information contained in the answer and detecting inappropriate content or an inappropriate emotional state, and means for transmitting the answer determined to be appropriate to the user. This makes it possible to improve the appropriateness of the user's input data and the generated answer and to provide a personalized response that takes the user's emotions into consideration.

[0391] A "user" is an entity that utilizes the system to provide input data and receive generated responses.

[0392] "Text data" refers to character string information that a user inputs through a terminal.

[0393] "Emotional information" refers to data that analyzes a user's emotional state and is information that the system uses to personalize the user's input data and responses.

[0394] "Means for receiving" refers to the modules and functions that allow a server or terminal to receive data from a user.

[0395] "Means for analyzing" refers to software or algorithms that examine the received data and check for appropriateness of content and emotional state.

[0396] "Inappropriate content" refers to data that the system does not tolerate, such as leaks of personal information, offensive language, or inappropriate language.

[0397] "Inappropriate emotional state" refers to a negative or unstable emotional state of the user detected by the system.

[0398] A "generative artificial intelligence model" refers to a machine learning model or algorithm that automatically generates appropriate responses based on received text data.

[0399] "Generated response" refers to a reply created by a generative artificial intelligence model.

[0400] "Transmission means" refers to the network infrastructure and communications protocols used by the server to transmit the generated response back to the user.

[0401] "Personalized responses" refer to replies that are individually tailored to take into account the user's emotional information and text data.

[0402] "System" refers to the overall technical environment and processing flow for receiving input data from a user, analyzing it, and generating and returning an appropriate response.

[0403] This system generates personalized responses based on user input data and associated emotional information, reducing the risk of confidential information leaks. The specific roles and operations of the user, device, and server are detailed below.

[0404] user

[0405] The user uses the device to input text data. At this time, the device is set up so that the user's emotional state is also input. For example, if the user inputs "This is a general request," the user's emotional state at that time is recognized as "calm."

[0406] Terminal

[0407] The terminal is responsible for transmitting the text data and emotion information input by the user to the server. Specifically, the terminal sends the input data and emotion information to the server through the network module, and then receives a response from the server and displays it to the user.

[0408] Example: The device clicks the send button and sends the input data "This is a general request" and emotional information "Calm" to the server.

[0409] server

[0410] The server provides the following specific functions:

[0411] 1. Data Reception

[0412] The server receives the text data and emotion information sent from the terminal. This reception is performed using a "receiving means."

[0413] 2. Data analysis and testing

[0414] The received data is analyzed using the "check_incoming_data" method to evaluate the appropriateness of the text data and the appropriateness of the emotional information. The server performs this analysis using an NLP (Natural Language Processing) module that includes an emotional engine.

[0415] Example: The server receives the text "This is a general request" and analyzes the emotional information "I feel calm."

[0416] 3. Data appropriateness assessment

[0417] The text data and emotion information that are deemed appropriate are sent to a generative AI model. If the content is inappropriate, the process ends and an error message is returned.

[0418] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[0419] 4. Response Generation

[0420] The generative AI model generates answers based on the received text data and emotional information, with the emotional engine creating personalized responses that reflect the user's current emotional state.

[0421] Example: A generative AI model generates the response, "Generative AI response: This is a common request. We're glad you're calm."

[0422] 5. Inspecting and Sending the Response

[0423] The generated answer is checked again by the "check_outgoing_data" method, and if it is deemed appropriate it is sent to the user. If it is an inappropriate answer, the process is terminated and an error message is returned.

[0424] Example: The server checks whether the generated answer is within the character limit and appropriate for the emotional state, and then sends the appropriate answer to the device.

[0425] The above is a specific example of the processing performed by the system of the present invention. This system ensures the appropriateness of the user's input data and the generated answers, and makes it possible to provide personalized responses that reflect the user's emotions.

[0426] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0427] Program processing flow

[0428] Step 1:

[0429] The user enters text data into the input field of the terminal and clicks the "Send" button. At this time, the emotion recognition software analyzes the user's emotional information and provides it to the terminal along with the input. For example, the input data is "This is a general request," and the emotional information is "calm."

[0430] input:

[0431] Text data entered by the user

[0432] Emotional information analyzed by emotion recognition software

[0433] output:

[0434] Text data and emotional information received by the device

[0435] Specific behavior:

[0436] User enters text and clicks submit button

[0437] Emotion recognition software analyzes user emotions

[0438] Step 2:

[0439] The terminal sends the text data and emotion information entered by the user to the server, and this data is transferred to the server using the network module.

[0440] input:

[0441] Text data and emotional information entered into the device

[0442] output:

[0443] Text data and emotional information sent to the server

[0444] Specific behavior:

[0445] The device's network module sends text data and emotional information to the server.

[0446] Step 3:

[0447] The server checks the received text data and emotion information using the "check_incoming_data" method. The server uses the NLP module and emotion engine to analyze the appropriateness of this data.

[0448] input:

[0449] Text data and emotion information received by the server

[0450] output:

[0451] Appropriateness evaluation results for text data and emotional information

[0452] Specific behavior:

[0453] The server analyzes the text using the NLP module

[0454] Emotion engine analyzes emotional information

[0455] Step 4:

[0456] The server determines the appropriateness of the data based on the analysis results. If the data is deemed appropriate, it sends it to the generative AI model. If the data is inappropriate, it generates an error message and returns it to the user.

[0457] input:

[0458] Text data and emotional information analyzed by the server

[0459] output:

[0460] When deemed appropriate: Data sent to generative AI models

[0461] If deemed inappropriate: Error message

[0462] Specific behavior:

[0463] Sending the right data to generative AI models

[0464] Invalid data generates an error message

[0465] Step 5:

[0466] The generative AI model generates answers based on the received text data and emotional information, and the emotion engine creates personalized responses that reflect the user's emotional state.

[0467] input:

[0468] Appropriate text data and emotional information sent from the server

[0469] output:

[0470] Generated Answer

[0471] Specific behavior:

[0472] Generative AI model generates responses based on text data and emotional information

[0473] Emotion engine creates responses that reflect the user's emotions

[0474] Step 6:

[0475] The server checks the generated answer using the "check_outgoing_data" method. If the check passes, the answer is sent to the user. If it is invalid, an error message is generated.

[0476] input:

[0477] Answers generated by generative AI models

[0478] output:

[0479] If the check passes: The answer sent to the user

[0480] If not passed: Error message

[0481] Specific behavior:

[0482] The server re-examines the generated answer

[0483] Sends appropriate response to user, may generate error message

[0484] Step 7:

[0485] The server sends an appropriate response to the user, and the device receives this response and displays it to the user. If the response is inappropriate, an error message is displayed.

[0486] input:

[0487] The response or error message sent by the server

[0488] output:

[0489] The answer or error message that the user sees

[0490] Specific behavior:

[0491] The device receives the response from the server and displays it to the user.

[0492] Display error messages if necessary

[0493] (Application example 2)

[0494] 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."

[0495] Conventional systems utilizing generative artificial intelligence models have the ability to determine the appropriateness of user input data and generated responses, but because they do not take user emotions into account, responses are not personalized and the user experience is insufficient. Furthermore, if emotional information is not taken into account, the appropriateness of the data cannot be adequately determined, which may increase the risk of confidential information being leaked. Therefore, an objective of the present invention is to provide a system that utilizes a user's emotional information to provide personalized responses based on emotions and reduce the risk of confidential information being leaked.

[0496] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving input data, means for checking the received input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, means for sending the answer determined to be appropriate to the user, emotion recognition means for receiving and analyzing the user's emotion information, and means for generating a personalized answer based on the emotion information. This enables a personalized response that is in line with the user's emotions, making it possible to reduce the risk of confidential information being leaked.

[0497] The "means for receiving input data" refers to a function or device for receiving text data or emotion information sent by a user.

[0498] "Means for checking received input data in advance and detecting inappropriate content" refers to functions or devices for analyzing input data and identifying content that does not conform to specified standards.

[0499] "Means for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer" refers to a function or device for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer.

[0500] "Means for checking generated answers in advance and detecting inappropriate content" refers to functions or devices for re-examining answers obtained from generative AI models to check whether they contain inappropriate content.

[0501] The "means for sending a response that is determined to be appropriate to the user" refers to a function or device for sending a response that is determined to not contain inappropriate content to the user.

[0502] "Emotion recognition means for receiving and analyzing user's emotional information" refers to a function or device for receiving information about the user's emotional state and analyzing that state.

[0503] The "means for generating a personalized answer based on emotional information" refers to a function or device for generating an answer optimized for a user, taking into account the emotional information of the user.

[0504] In order to put the present invention into practice, it is necessary to build a system that specifically separates the roles of users, terminals, and servers.

[0505] User Roles

[0506] The user's role is to access the online shopping site and input product inquiries and support requests in text format. As the user inputs information, the system automatically detects the user's emotional state. At this time, the user's emotional information is also transmitted along with the input data.

[0507] As a specific example, a user might enter, "I have a question about this product. I'm worried."

[0508] Device Role

[0509] The device receives input data and emotion information from the user and transmits it to the server. The device in this case can be a smartphone or desktop PC.

[0510] The request sent by the device will have the following format:

[0511] "I have a question about this product. I'm worried."

[0512] Server Roles

[0513] The server has multiple means for analyzing the received input data and emotional information and generating appropriate responses.

[0514] 1. Receipt and inspection of data

[0515] The server receives the input data sent by the user using the "means for receiving input data" and analyzes it using the "means for pre-checking the received input data and detecting inappropriate content." Specifically, it analyzes the text data using a natural language processing library and analyzes emotional information using an emotion recognition engine (e.g., VADER Sentiment Analysis).

[0516] 2. Submitting the right data and generating answers

[0517] The server uses the "means for sending input data judged appropriate to a generative AI model and receiving the generated answer" to send the input data judged appropriate to a generative AI model (e.g., GPT-3) and generate an answer. At this time, emotional information is taken into consideration to generate a personalized response.

[0518] 3. Check and submit the generated answers

[0519] The server re-examines the generated answer using "means for checking the generated answer in advance and detecting inappropriate content," evaluates it using "means for receiving and analyzing the user's emotional information," and then sends the answer that is judged to be appropriate to the user using "means for sending the answer to the user."

[0520] Hardware and software used

[0521] 1. Hardware

[0522] Server: A cloud server, such as an Amazon EC2 instance

[0523] Devices: Smartphones, desktop PCs

[0524] 2. Software

[0525] Natural language processing models: Generative artificial intelligence models such as GPT-3

[0526] Sentiment analysis library: VADER Sentiment Analysis

[0527] Specific examples

[0528] As a concrete example, let's say a user types, "I have a question about this product. I'm worried." In that case, the following process will be executed:

[0529] 1. The user types "I have a question about this product. I'm worried." into the device, and the emotion is recognized as "anxiety."

[0530] 2. The device sends the input data and emotion information to the server.

[0531] 3. The server receives and analyzes the text data and emotion information.

[0532] 4. If appropriate, the generative AI model generates an answer.

[0533] 5. The server re-examines the generated answers and rates them based on sentiment information.

[0534] 6. If the rating is deemed appropriate, the answer is sent to the user.

[0535] The above is a specific embodiment for carrying out the present invention. This system can provide personalized responses that take into account the user's emotional information, reducing the risk of confidential information being leaked.

[0536] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0537] Step 1:

[0538] The user uses a terminal to create input data and send it along with emotional information. At this time, the user enters the inquiry in text format, and the terminal automatically performs emotional analysis. For example, if the user enters "I have a question about this product. I'm worried," the terminal obtains this text data and the emotional information of "anxiety." The inputs are text data and emotional information. The output is sent to the server.

[0539] Step 2:

[0540] The device receives input data and sends emotional information to the server. The device's role is to relay data, and the input data and emotional information are sent to the server as is without conversion. The input includes the user's text data and emotional information. The output includes the data sent to the server.

[0541] Step 3:

[0542] The server checks the input data and emotional information it receives in advance to detect inappropriate content. Specifically, the server uses a natural language processing library to analyze the appropriateness of the text and an emotion recognition engine to analyze the emotional information. The input includes text data and emotional information sent from the device. Data processing involves natural language processing and emotional analysis. The output is either data deemed appropriate or an error message.

[0543] Step 4:

[0544] Input data deemed appropriate is sent to a generative AI model and the generated answer is received. To do this, the server sends text data and emotional information to the generative AI model and receives the generated answer. The inputs include text data and emotional information. Data calculation involves text generation by the generative AI model. The output is the generated answer.

[0545] Step 5:

[0546] The generated answer is checked in advance to detect inappropriate content. The server again analyzes the appropriateness of the answer using a natural language processing library and evaluates the emotional information using an emotion recognition engine. The input is the generated answer. Data processing involves natural language processing and emotional evaluation. The output is either an answer deemed appropriate or an error message.

[0547] Step 6:

[0548] The answer that is deemed appropriate is sent to the user. The server returns this answer to the terminal, which displays it to the user. The input is the answer that is deemed appropriate. The output is the answer that was sent to the user terminal.

[0549] Step 7:

[0550] The user checks the answer on the device and takes further action. The user checks the answer to the question through the device and asks again if necessary. The input is the answer from the server. The output is the user's satisfaction level and the next action.

[0551] The above is the flow of processing for the system program that realizes the application example.

[0552] 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.

[0553] 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.

[0554] 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.

[0555] [Second embodiment]

[0556] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0557] 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.

[0558] 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).

[0559] 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.

[0560] 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.

[0561] 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).

[0562] 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.

[0563] 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.

[0564] 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.

[0565] 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.

[0566] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0567] 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."

[0568] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[0569] overview

[0570] The present invention is a system that checks input data provided by users before sending it to a generative artificial intelligence model, and similarly checks the generated answers, providing security services to prevent the leakage of confidential information.

[0571] Basic system configuration

[0572] 1. Users

[0573] The user inputs text data through the terminal.

[0574] Example: User types, "This is a sensitive request."

[0575] 2. Terminal

[0576] It is responsible for sending text data entered by the user to the server.

[0577] Example: Submitting data by clicking a button on a web form or application.

[0578] 3. Server

[0579] The server creates an instance of the "SecurityService" class and receives input data.

[0580] The server first checks the received data using the "checking means." If the data contains inappropriate content, it returns an error message to the user saying "Input is inappropriate."

[0581] If the input data does not contain inappropriate content, the input data is passed to a generative artificial intelligence model and the generated answer is received.

[0582] The server also checks the generated answer using a "checking method." If the answer contains inappropriate content, it returns a message to the user saying, "The answer is inappropriate."

[0583] If it is determined to be an appropriate answer, it returns the answer to the user.

[0584] Specific examples

[0585] 1. Receiving and checking user input

[0586] The user types "This is a general request" into the terminal.

[0587] The terminal sends this input data to the server.

[0588] The server uses the "check_incoming_data" method to check the incoming data.

[0589] 2. Answer generation using generative AI models

[0590] The server passes input data that it deems appropriate to the generative artificial intelligence model.

[0591] The generative artificial intelligence model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[0592] 3. Check and submit the generated answers

[0593] The server checks the response using the "check_outgoing_data" method.

[0594] If the server determines that the answer is appropriate, it returns the answer to the terminal.

[0595] Example: "Generated AI response: This is a common request." is displayed to the user.

[0596] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[0597] The processing flow will be explained below.

[0598] Step 1:

[0599] A user inputs text data into a terminal.

[0600] Example: User types, "This is a general request."

[0601] Step 2:

[0602] The terminal transmits the user's input data to the server.

[0603] Example: The terminal clicks the send button and requests the input data from the server.

[0604] Step 3:

[0605] The server receives input data from the user.

[0606] Example: The server receives data through a REST API.

[0607] Step 4:

[0608] The input data received by the server is checked using the "check_incoming_data" method.

[0609] Example: This method checks for sensitive information or inappropriate keywords.

[0610] Step 5:

[0611] The server determines whether the input data is valid.

[0612] If deemed appropriate, the data is passed to a generative AI model.

[0613] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0614] Example: If the input data is "This is a request containing sensitive information," the server responds with "Invalid input."

[0615] Step 6:

[0616] The server passes input data that it deems appropriate to the generative AI model to generate an answer.

[0617] Example: Call the "generate" method of "AIModel" and pass in the user's input data.

[0618] Step 7:

[0619] A generative AI model generates answers based on user input data.

[0620] Example: Generate the answer "Generation AI answer: This is a common request."

[0621] Step 8:

[0622] The server receives the generated answer.

[0623] Example: Receiving a response from a generative AI model.

[0624] Step 9:

[0625] The server checks the generated answer with the "check_outgoing_data" method.

[0626] Example: This method checks if the answer is longer than a predefined number of characters.

[0627] Step 10:

[0628] The server determines whether the generated answer is appropriate.

[0629] If it is judged to be appropriate, it returns a response to the terminal.

[0630] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0631] Example: If the answer exceeds 500 characters, the server will respond with "The answer is inappropriate."

[0632] Step 11:

[0633] The terminal receives the appropriate response from the server and displays it to the user.

[0634] Example: The user will see the answer "Generated AI answer: This is a common request."

[0635] The above are the specific steps of the processing of the present invention, which can effectively reduce the risk of confidential information leakage when using generative AI models and provide users with safe and reliable services.

[0636] Example 1

[0637] 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."

[0638] While modern dialogue systems and data processing systems using generative artificial intelligence models are extremely useful, they pose a significant risk of confidential information leakage. Security measures are essential, particularly since both the input data provided by users and the generated answers may contain confidential information. However, these systems require effective means to pre-check the input data and generated answers and filter out inappropriate content.

[0639] 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.

[0640] In this invention, the server includes: means for receiving input data from a user; means for checking the received input data in advance to detect inappropriate content; means for sending input data determined to be appropriate to the generative AI model and receiving a generated answer; means for checking the generated answer in advance to detect inappropriate content; means for sending the answer determined to be appropriate to the user; means for converting the input data to be sent to the generative AI model into a prompt format in advance; means for converting the answer received from the generative AI model into a format displayable to the user; means for checking the input data and the generated answer by applying specific security rules to prevent leakage of confidential information; means for saving all data including the input data from the user and the answer from the generative AI model as a log; and means for periodically auditing the saved log data. This makes it possible to effectively prevent leakage of confidential information during the process of using the generative AI model.

[0641] "User" means a person who utilizes the System to input and receive data.

[0642] "Terminal" means a device operated by a user and used to input data and transmit data to a server.

[0643] "Server" refers to the central computer system that receives and validates input data from users and works with the generative artificial intelligence model to generate and validate answers.

[0644] "Input data" refers to data such as text sent by a user to a server via a terminal.

[0645] A "generative artificial intelligence model" refers to an AI system that generates appropriate answers based on input data from users.

[0646] "Prompt sentence" refers to an instruction sentence formatted to appropriately transmit input data to a generative artificial intelligence model.

[0647] "Checking means" refers to a function for analyzing input data and generated answers to check for inappropriate content or confidential information.

[0648] "Inappropriate content" refers to data that contains confidential or security-related information.

[0649] "Security rules" refer to analytical standards and procedures established to prevent the leakage of confidential information.

[0650] "Log" refers to a record of all data processed within the system, such as user input data and generated responses.

[0651] "Audit" refers to the process of periodically reviewing stored log data to verify the operation and security status of a system.

[0652] The present invention is a system for preventing the leakage of confidential information by checking input data provided by a user before sending it to a generative artificial intelligence model, and by similarly checking the generated answers.

[0653] Basic system configuration

[0654] 1. User:

[0655] The user is responsible for inputting and transmitting text data through the terminal.

[0656] Example: A user types, "This is a general request."

[0657] 2. Terminal:

[0658] The terminal is responsible for transmitting text data entered by the user to the server.

[0659] Example: Submitting data by clicking a button on a web form or application.

[0660] 3. Server:

[0661] The server performs the following process:

[0662] 1. Data reception:

[0663] The server creates an instance of the "SecurityService" class and receives input data from the user.

[0664] Example: Use the "check_incoming_data" method to check incoming data for inappropriate content.

[0665] 2. Input data validation:

[0666] If the data contains inappropriate content, the server sends an error message "Input is inappropriate" to the terminal.

[0667] Example: If the user types "This is confidential information", the server generates an error message.

[0668] 3. Data transfer to the generative AI model:

[0669] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[0670] Example: If the input is deemed appropriate, the data is passed to the generative AI model using the prompt "Please output the generative AI's answer."

[0671] 4. Receive a response from the generative AI model:

[0672] The server processes the answers received from the generative artificial intelligence model.

[0673] Example: A generative artificial intelligence model responds with, "Generative AI response: This is a common request."

[0674] 5. Check the generated data:

[0675] The server checks the generated data using the "check_outgoing_data" method.

[0676] Example: The server checks the generated data again using pre-filtering rules and databases.

[0677] 6. Error Handling:

[0678] If the answer contains inappropriate content, the server sends an error message "The answer is inappropriate" to the terminal.

[0679] Action: The server formats an appropriate error message and sends it to the terminal.

[0680] 7. Submitting the right response:

[0681] The server sends the answer it deems appropriate to the user.

[0682] Example: "Generated AI response: This is a common request." is sent to the user.

[0683] Examples and prompt formats

[0684] 1. Receive and check user input:

[0685] The user types "This is a general request" into the terminal.

[0686] The terminal sends this input data to the server.

[0687] The server checks the incoming data with the "check_incoming_data" method.

[0688] 2. Answer generation using generative AI models:

[0689] The server sends input data that it deems appropriate to the generative AI model.

[0690] The generative AI model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[0691] 3. Check and submit the generated answers:

[0692] The server checks the generated answer using the "check_outgoing_data" method.

[0693] The server returns an answer that it deems appropriate to the terminal.

[0694] Example: "Generated AI response: This is a common request." is displayed to the user.

[0695] Prompt Sentence Examples

[0696] Please output the generative AI's answer: This is a common request.

[0697] In this way, this system effectively reduces the risk of confidential information leaks when using generative AI. By inspecting the text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[0698] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0699] Step 1:

[0700] The user enters text data into the terminal.

[0701] Input: Text data entered by the user, e.g. "This is a general request."

[0702] Specific action: The user types characters on the keyboard and presses the Enter key.

[0703] Output: Input data is displayed on the terminal.

[0704] Step 2:

[0705] The terminal transmits the input text data to the server.

[0706] Input: Text data entered by the user, e.g., "This is a general request."

[0707] Specific operation: When you click the send button on the device, the browser generates an HTTP POST request and sends it to the server.

[0708] Output: Data is sent to the server.

[0709] Step 3:

[0710] The server receives the data from the user and checks the content using the "check_incoming_data" method.

[0711] Input: Text data sent from the terminal, e.g., "This is a general request."

[0712] Specific operation: The server creates an instance of the "SecurityService" class and analyzes the data using the database and filtering rules.

[0713] Data manipulation: Detecting inappropriate content using specific keywords and pattern matching.

[0714] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[0715] Step 4:

[0716] If the server determines that the data is invalid, it sends an error message to the terminal saying "Invalid input."

[0717] Input: Inspection results, e.g., inappropriate data "This is confidential information."

[0718] Specific behavior: The server generates an HTTP response and sends data including an error message.

[0719] Output: An error message is displayed on the terminal.

[0720] Step 5:

[0721] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[0722] Input: Inspection result, e.g. "This is a common request."

[0723] Specific operation: The server generates an appropriate prompt sentence and sends the data to the API of the generative artificial intelligence model.

[0724] Data processing: Convert the data into the format of the prompt statement.

[0725] Output: The prompt sent to the generative AI model.

[0726] Step 6:

[0727] The server receives the answer from the generative artificial intelligence model.

[0728] Input: An answer from a generative AI model, e.g., "Generative AI answer: This is a common request."

[0729] Specific operation: The server receives the HTTP response and parses the response body.

[0730] Output: The received response data.

[0731] Step 7:

[0732] The server checks the generated data using the "check_outgoing_data" method.

[0733] Input: Answer data from a generative AI model, e.g., "Generative AI answer: This is a common request."

[0734] Specific operation: The server examines the data using a filtering algorithm.

[0735] Data processing: Check that the response data does not contain inappropriate content.

[0736] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[0737] Step 8:

[0738] If the server generates an answer that contains inappropriate content, it sends an error message to the terminal saying "Answer is inappropriate."

[0739] Input: Test results, e.g., incorrect response data.

[0740] Specific behavior: The server generates an HTTP response and returns data including an error message.

[0741] Output: An error message is displayed on the terminal.

[0742] Step 9:

[0743] The server sends the answer that it deems appropriate to the terminal.

[0744] Input: Inspection result, e.g. "Generated AI response: This is a common request."

[0745] Specific behavior: The server generates an HTTP response and sends back data containing the appropriate answer.

[0746] Output: The answer of the generated AI is displayed on the terminal.

[0747] (Application example 1)

[0748] 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."

[0749] In systems using generative artificial intelligence models, if the input data or generated answers contain confidential information, it is important to prevent the leakage of that information. However, many current systems lack a means to automatically verify the appropriateness of the input data or generated answers, which poses a risk of confidential information leaking to the outside. The present invention aims to solve this problem and prevent the leakage of confidential information.

[0750] 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.

[0751] In this invention, the server includes means for receiving input data, means for checking the input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, and means for including a mobile application that runs on a smartphone and has a security function for preventing leakage of confidential information. This makes it possible to automatically perform security checks on the input data and the generated answers and prevent leakage of confidential information.

[0752] "Input Data" means text or other information provided to a system by a user.

[0753] "Means for receiving" refers to a device or method for receiving input data provided by a user.

[0754] "Pre-checking measures" refers to functions and methods for automatically inspecting input data and generated responses to check for inappropriate content or confidential information.

[0755] "Means for detecting inappropriate content" means a device or method for identifying inappropriate elements or information in input data or generated answers.

[0756] A "generative artificial intelligence model" refers to an artificial intelligence algorithm or system that generates answers based on input data provided by a user.

[0757] "Means for receiving a generated answer" refers to a device or method for receiving an answer generated from a generative artificial intelligence model.

[0758] "Means for sending answers that are judged to be appropriate to the user" refers to a device or method for returning a checked answer to the user if the checked answer is judged to be appropriate.

[0759] A "smartphone" is a type of mobile phone, a portable information terminal that can connect to the Internet and run various applications.

[0760] "Security features to prevent the leakage of confidential information" refers to features and methods to prevent confidential information from leaking to the outside if it is included in input data or generated answers.

[0761] "Mobile application" refers to a software application that runs on a mobile information device such as a smartphone or tablet.

[0762] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[0763] System Configuration

[0764] 1. Users

[0765] A user operates an application using a smartphone to input text data.

[0766] Example: User types "Summarize the meeting."

[0767] 2. Terminal

[0768] The terminal is responsible for receiving text data entered by the user and transmitting it to the server.

[0769] Specifically, data is submitted by clicking a button on a web form or application.

[0770] 3. Server

[0771] The server receives the input data and performs security checks by creating an instance of the "SecurityService" class and validating the input data.

[0772] The "check_incoming_data" method is used to check whether inappropriate content is included. If inappropriate content is included, the error message "Input is inappropriate" is returned to the terminal.

[0773] Input data that is deemed appropriate is sent to the generative artificial intelligence model, which then receives the generated answer.

[0774] The generated answer is inspected using the "check_outgoing_data" method. If it contains inappropriate content, the message "Answer is inappropriate" is returned to the terminal.

[0775] If the answer is determined to be appropriate, the answer is returned to the terminal and displayed to the user.

[0776] Data processing and hardware / software used

[0777] Hardware: Smartphone (iOS / Android)

[0778] Software: Python, Flask (web framework), TensorFlow / Transformers (generative AI models), SQL database

[0779] Data calculation: Input data checks and answer checks are performed by the "check_incoming_data()" and "check_outgoing_data()" methods in the "SecurityService" class.

[0780] Specific examples

[0781] When a user types "This is a general request" into a smartphone application, the content is sent to the server. The server checks this input data using the "check_incoming_data" method and sends it to the generative AI model only if it is deemed appropriate.

[0782] The generative AI model generates an answer, "Generative AI's answer: This is a typical request," and returns it to the server. The server then checks the answer again using the "check_outgoing_data" method, and if it is deemed appropriate, returns the answer to the user.

[0783] Example prompt sentence:

[0784] User input: Check if the content of 'Summarize the meeting' is typical, and if it is, run a function that provides appropriate feedback.

[0785] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[0786] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0787] Step 1:

[0788] The user opens the application on their smartphone and enters text. The input data is "Summarize the contents of the meeting." The device receives this user input and sends it to the server.

[0789] Input: User's text data "Summarize the contents of the meeting"

[0790] Output: Text data sent from the terminal to the server

[0791] Step 2:

[0792] The server checks the input data received from the terminal using the "check_incoming_data" method of the "SecurityService" class, which checks whether the input data contains confidential information or inappropriate content.

[0793] Input: Text data received by the server: "Summarize the contents of the meeting"

[0794] Data processing: Checks to see if there is any confidential information

[0795] Output: Check result (good or bad)

[0796] Specific behavior:

[0797] The text data received by the server is passed to the "check_incoming_data" method, which checks whether it contains specific keywords. For example, there is a rule that detects keywords such as "confidential" and "important."

[0798] Step 3:

[0799] If the server checks the results, it sends the text data to a generative AI model, which receives a specific prompt and generates an appropriate answer based on it.

[0800] Input: Text data that passed the check: "Summarize the contents of the meeting"

[0801] Data processing: Passing the data as prompts to the generative AI model

[0802] Output: The answer generated by the generative AI model

[0803] Specific behavior:

[0804] The server passes appropriate input data to the generative AI model as a prompt. For example, it requests the model to generate a summary of the user's input: 'Summarize the contents of the meeting.'

[0805] Step 4:

[0806] The generative AI model generates an answer based on the prompt and returns it to the server, for example, in the form of a "generated summary of the meeting content."

[0807] Input: Prompt sentence for generative AI model

[0808] Data computation: Generative AI models generate answers

[0809] Output: The generated answer

[0810] Specific behavior:

[0811] A generative AI model uses natural language processing based on the prompt to generate an appropriate answer.

[0812] Step 5:

[0813] The server checks the generated answer in advance using the "check_outgoing_data" method of the "SecurityService" class to ensure that the generated answer does not contain inappropriate content.

[0814] Input: Generated Answer

[0815] Data processing: Checking generated answers

[0816] Output: Check result (good or bad)

[0817] Specific behavior:

[0818] The server passes the generated answer to the "check_outgoing_data" method to check whether it contains specific keywords or phrases.

[0819] Step 6:

[0820] If the check result is appropriate, the server returns the generated answer to the terminal and displays it to the user. If the check result is inappropriate, the server returns an error message saying "The answer is inappropriate."

[0821] Input: Answer that passes the check

[0822] Data processing: Formatting answers (if necessary)

[0823] Output: The appropriate answer or error message that is displayed to the user

[0824] Specific behavior:

[0825] The server sends the answer that passes the check to the terminal, and the answer is displayed to the user by the smartphone application. For example, the terminal executes the operation "display the data received from the server."

[0826] 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.

[0827] This invention is a system that uses a generative artificial intelligence model to check the appropriateness of the user's input data and the generated answers, and also takes the user's emotions into consideration, thereby reducing the risk of confidential information leaks and providing more personalized responses. Below, we will explain in detail the main roles of the user, terminal, and server, and the specific processes performed by each.

[0828] overview

[0829] In addition to means for checking input data, means for using a generative artificial intelligence model, and means for checking generated answers, the present invention also includes an emotion engine that recognizes the user's emotions and reflects them in various checks and response generation.

[0830] Basic system configuration

[0831] 1. Users

[0832] The user inputs text data through the terminal.

[0833] Input can also include emotional states.

[0834] Example: A user types, "This is a general request," and their emotion is recognized as "calm."

[0835] 2. Terminal

[0836] It is responsible for sending the text data and emotional information entered by the user to the server.

[0837] Example: The device clicks the send button and requests emotion information along with the input data from the server.

[0838] 3. Server

[0839] The server has means for receiving the text data and the emotion information.

[0840] The received data is inspected using the "check_incoming_data" method, and the emotion engine analyzes the user's emotion information.

[0841] Example: The server further scrutinizes the input data based on a "calm" emotional state.

[0842] 4. Data appropriateness assessment

[0843] If the server determines that the input data is appropriate, it passes the data and emotional information to a generative AI model to generate an answer.

[0844] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0845] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[0846] 5. Use of generative AI models

[0847] The generative AI model generates answers based on the received text data and emotional information.

[0848] The emotion engine generates answers that reflect the user's current emotional state.

[0849] Example: Generated response: "Generated AI response: This is a common request. We're glad you're calm."

[0850] 6. Check the generated answers

[0851] The server checks the generated answer again with the "check_outgoing_data" method.

[0852] The emotion engine also evaluates the appropriateness of the generated answers based on emotional information.

[0853] For example: Make sure your answer is within the character limit and emotionally relevant.

[0854] 7. Sending the Response

[0855] If the answer is judged to be appropriate, it is returned to the device and displayed to the user in an appropriate state that reflects the emotional information.

[0856] If it is determined to be inappropriate, an error is returned and the process is terminated.

[0857] Example: The user sees the message, "Generated AI response: This is a common request. We're glad you're calm."

[0858] Specific examples

[0859] 1. Receiving and checking user input

[0860] The user types "This is a general request" into the device, and the system recognizes the user's emotional state as "calm."

[0861] The device sends this input data and emotional information to the server.

[0862] 2. Answer generation using generative AI models

[0863] The server checks the data and emotion information using the "check_incoming_data" method.

[0864] Data and emotional information deemed appropriate are sent to the generative AI model.

[0865] The generative AI model generates an answer, "Generative AI's answer: This is a common request. We're glad you're calm." and returns it to the server.

[0866] 3. Check and submit the generated answers

[0867] The server checks the response with the "check_outgoing_data" method.

[0868] If it is determined to be an appropriate answer, it returns the answer to the terminal.

[0869] The generated answer is displayed to the user.

[0870] These are the specific steps of the process of the present invention. This system reduces the risk of leaking confidential user information when using generative AI models, and can provide personalized responses that take user emotions into account.

[0871] The processing flow will be explained below.

[0872] Step 1:

[0873] A user inputs text data into a terminal.

[0874] Example: User types, "This is a general request."

[0875] Step 2:

[0876] The terminal collects the user's input data as well as the emotional state.

[0877] Example: The device recognizes the user's emotion as "calm" through facial recognition and voice tone analysis.

[0878] Step 3:

[0879] The device transmits the collected text data and emotion information to a server.

[0880] Example: The device clicks the send button and sends the input data and emotion information as a request to the server.

[0881] Step 4:

[0882] The server receives input data and emotion information from the user.

[0883] Example: The server receives data and emotion information through a REST API.

[0884] Step 5:

[0885] The input data received by the server is checked using the "check_incoming_data" method.

[0886] Example: This method checks for sensitive information or inappropriate keywords.

[0887] Step 6:

[0888] The server analyzes the received emotion information using an emotion engine.

[0889] Example: The server analyzes the emotional information of "calm" and further scrutinizes the input data based on that information.

[0890] Step 7:

[0891] The server determines whether the input data is valid.

[0892] If deemed appropriate, the data and emotional information is passed to a generative AI model.

[0893] If it is determined to be inappropriate, an error message "Input is inappropriate" is returned and the process is terminated.

[0894] For example, "This is a request containing sensitive information." If the data is deemed inappropriate, the server will block it.

[0895] Step 8:

[0896] The server passes the input data and emotional information it deems appropriate to the generative AI model to generate an answer.

[0897] Example: Call the "generate" method of the generative AI model, passing in the user's input data and emotion information.

[0898] Step 9:

[0899] The generative AI model generates answers based on the user's input data and emotional information.

[0900] Example: The generative AI model generates the answer "Generative AI response: This is a common request. We're glad you're calm."

[0901] Step 10:

[0902] The server receives the generated answer.

[0903] Example: Receiving an answer as a response from a generative AI model.

[0904] Step 11:

[0905] The server checks the generated answer again with the "check_outgoing_data" method.

[0906] For example, this method checks whether the answer is within a predefined character limit and whether it is sentiment-based.

[0907] Step 12:

[0908] The server determines whether the generated answer is appropriate.

[0909] If it is determined to be appropriate, the answer is sent to the terminal.

[0910] If it is determined to be inappropriate, an error message "The answer is inappropriate" is returned and the process is terminated.

[0911] Example: If the answer exceeds 500 characters, the server will block it as an inappropriate answer.

[0912] Step 13:

[0913] The terminal receives the appropriate response from the server and displays it to the user.

[0914] Example: The user will see the answer, "Generator's response: This is a common request. We're glad you're calm."

[0915] These are the specific steps of the process of the present invention. This system reduces the risk of confidential information leakage when using generative AI models and can provide personalized responses that take user emotions into account.

[0916] Example 2

[0917] 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."

[0918] Conventional systems using generative artificial intelligence models have had difficulty accurately evaluating the appropriateness of user input data and generated answers. Furthermore, because they do not take the user's emotions into account, they are unable to provide personalized responses, resulting in reduced user satisfaction. Furthermore, there is a risk of data leakage due to inappropriate data content or emotional state. The present invention aims to solve these problems, improve the appropriateness of user input data and generated answers, and provide personalized responses that take the user's emotions into account.

[0919] 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.

[0920] In this invention, the server includes means for receiving text data and the user's emotional information from the user, means for analyzing the received text data and emotional information and detecting inappropriate content or an inappropriate emotional state, means for transmitting the text data and emotional information determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for analyzing the generated answer and the emotional information contained in the answer and detecting inappropriate content or an inappropriate emotional state, and means for transmitting the answer determined to be appropriate to the user. This makes it possible to improve the appropriateness of the user's input data and the generated answer and to provide a personalized response that takes the user's emotions into consideration.

[0921] A "user" is an entity that utilizes the system to provide input data and receive generated responses.

[0922] "Text data" refers to character string information that a user inputs through a terminal.

[0923] "Emotional information" refers to data that analyzes a user's emotional state and is information that the system uses to personalize the user's input data and responses.

[0924] "Means for receiving" refers to the modules and functions that allow a server or terminal to receive data from a user.

[0925] "Means for analyzing" refers to software or algorithms that examine the received data and check for appropriateness of content and emotional state.

[0926] "Inappropriate content" refers to data that the system does not tolerate, such as leaks of personal information, offensive language, or inappropriate language.

[0927] "Inappropriate emotional state" refers to a negative or unstable emotional state of the user detected by the system.

[0928] A "generative artificial intelligence model" refers to a machine learning model or algorithm that automatically generates appropriate responses based on received text data.

[0929] "Generated response" refers to a reply created by a generative artificial intelligence model.

[0930] "Transmission means" refers to the network infrastructure and communications protocols used by the server to transmit the generated response back to the user.

[0931] "Personalized responses" refer to replies that are individually tailored to take into account the user's emotional information and text data.

[0932] "System" refers to the overall technical environment and processing flow for receiving input data from a user, analyzing it, and generating and returning an appropriate response.

[0933] This system generates personalized responses based on user input data and associated emotional information, reducing the risk of confidential information leaks. The specific roles and operations of the user, device, and server are detailed below.

[0934] user

[0935] The user uses the device to input text data. At this time, the device is set up so that the user's emotional state is also input. For example, if the user inputs "This is a general request," the user's emotional state at that time is recognized as "calm."

[0936] Terminal

[0937] The terminal is responsible for transmitting the text data and emotion information input by the user to the server. Specifically, the terminal sends the input data and emotion information to the server through the network module, and then receives a response from the server and displays it to the user.

[0938] Example: The device clicks the send button and sends the input data "This is a general request" and emotional information "Calm" to the server.

[0939] server

[0940] The server provides the following specific functions:

[0941] 1. Data Reception

[0942] The server receives the text data and emotion information sent from the terminal. This reception is performed using a "receiving means."

[0943] 2. Data analysis and testing

[0944] The received data is analyzed using the "check_incoming_data" method to evaluate the appropriateness of the text data and the appropriateness of the emotional information. The server performs this analysis using an NLP (Natural Language Processing) module that includes an emotional engine.

[0945] Example: The server receives the text "This is a general request" and analyzes the emotional information "I feel calm."

[0946] 3. Data appropriateness assessment

[0947] The text data and emotion information that are deemed appropriate are sent to a generative AI model. If the content is inappropriate, the process ends and an error message is returned.

[0948] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[0949] 4. Response Generation

[0950] The generative AI model generates answers based on the received text data and emotional information, with the emotional engine creating personalized responses that reflect the user's current emotional state.

[0951] Example: A generative AI model generates the response, "Generative AI response: This is a common request. We're glad you're calm."

[0952] 5. Inspecting and Sending the Response

[0953] The generated answer is checked again by the "check_outgoing_data" method, and if it is deemed appropriate it is sent to the user. If it is an inappropriate answer, the process is terminated and an error message is returned.

[0954] Example: The server checks whether the generated answer is within the character limit and appropriate for the emotional state, and then sends the appropriate answer to the device.

[0955] The above is a specific example of the processing performed by the system of the present invention. This system ensures the appropriateness of the user's input data and the generated answers, and makes it possible to provide personalized responses that reflect the user's emotions.

[0956] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0957] Program processing flow

[0958] Step 1:

[0959] The user enters text data into the input field of the terminal and clicks the "Send" button. At this time, the emotion recognition software analyzes the user's emotional information and provides it to the terminal along with the input. For example, the input data is "This is a general request," and the emotional information is "calm."

[0960] input:

[0961] Text data entered by the user

[0962] Emotional information analyzed by emotion recognition software

[0963] output:

[0964] Text data and emotional information received by the device

[0965] Specific behavior:

[0966] User enters text and clicks submit button

[0967] Emotion recognition software analyzes user emotions

[0968] Step 2:

[0969] The terminal sends the text data and emotion information entered by the user to the server, and this data is transferred to the server using the network module.

[0970] input:

[0971] Text data and emotional information entered into the device

[0972] output:

[0973] Text data and emotional information sent to the server

[0974] Specific behavior:

[0975] The device's network module sends text data and emotional information to the server.

[0976] Step 3:

[0977] The server checks the received text data and emotion information using the "check_incoming_data" method. The server uses the NLP module and emotion engine to analyze the appropriateness of this data.

[0978] input:

[0979] Text data and emotion information received by the server

[0980] output:

[0981] Appropriateness evaluation results for text data and emotional information

[0982] Specific behavior:

[0983] The server analyzes the text using the NLP module

[0984] Emotion engine analyzes emotional information

[0985] Step 4:

[0986] The server determines the appropriateness of the data based on the analysis results. If the data is deemed appropriate, it sends it to the generative AI model. If the data is inappropriate, it generates an error message and returns it to the user.

[0987] input:

[0988] Text data and emotional information analyzed by the server

[0989] output:

[0990] When deemed appropriate: Data sent to generative AI models

[0991] If deemed inappropriate: Error message

[0992] Specific behavior:

[0993] Sending the right data to generative AI models

[0994] Invalid data generates an error message

[0995] Step 5:

[0996] The generative AI model generates answers based on the received text data and emotional information, and the emotion engine creates personalized responses that reflect the user's emotional state.

[0997] input:

[0998] Appropriate text data and emotional information sent from the server

[0999] output:

[1000] Generated Answer

[1001] Specific behavior:

[1002] Generative AI model generates responses based on text data and emotional information

[1003] Emotion engine creates responses that reflect the user's emotions

[1004] Step 6:

[1005] The server checks the generated answer using the "check_outgoing_data" method. If the check passes, the answer is sent to the user. If it is invalid, an error message is generated.

[1006] input:

[1007] Answers generated by generative AI models

[1008] output:

[1009] If the check passes: The answer sent to the user

[1010] If not passed: Error message

[1011] Specific behavior:

[1012] The server re-examines the generated answer

[1013] Sends appropriate response to user, may generate error message

[1014] Step 7:

[1015] The server sends an appropriate response to the user, and the device receives this response and displays it to the user. If the response is inappropriate, an error message is displayed.

[1016] input:

[1017] The response or error message sent by the server

[1018] output:

[1019] The answer or error message that the user sees

[1020] Specific behavior:

[1021] The device receives the response from the server and displays it to the user.

[1022] Display error messages if necessary

[1023] (Application example 2)

[1024] 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."

[1025] Conventional systems utilizing generative artificial intelligence models have the ability to determine the appropriateness of user input data and generated responses, but because they do not take user emotions into account, responses are not personalized and the user experience is insufficient. Furthermore, if emotional information is not taken into account, the appropriateness of the data cannot be adequately determined, which may increase the risk of confidential information being leaked. Therefore, an objective of the present invention is to provide a system that utilizes a user's emotional information to provide personalized responses based on emotions and reduce the risk of confidential information being leaked.

[1026] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving input data, means for checking the received input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, means for sending the answer determined to be appropriate to the user, emotion recognition means for receiving and analyzing the user's emotion information, and means for generating a personalized answer based on the emotion information. This enables a personalized response that is in line with the user's emotions, making it possible to reduce the risk of confidential information being leaked.

[1027] The "means for receiving input data" refers to a function or device for receiving text data or emotion information sent by a user.

[1028] "Means for checking received input data in advance and detecting inappropriate content" refers to functions or devices for analyzing input data and identifying content that does not conform to specified standards.

[1029] "Means for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer" refers to a function or device for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer.

[1030] "Means for checking generated answers in advance and detecting inappropriate content" refers to functions or devices for re-examining answers obtained from generative AI models to check whether they contain inappropriate content.

[1031] The "means for sending a response that is determined to be appropriate to the user" refers to a function or device for sending a response that is determined to not contain inappropriate content to the user.

[1032] "Emotion recognition means for receiving and analyzing user's emotional information" refers to a function or device for receiving information about the user's emotional state and analyzing that state.

[1033] The "means for generating a personalized answer based on emotional information" refers to a function or device for generating an answer optimized for a user, taking into account the emotional information of the user.

[1034] In order to put the present invention into practice, it is necessary to build a system that specifically separates the roles of users, terminals, and servers.

[1035] User Roles

[1036] The user's role is to access the online shopping site and input product inquiries and support requests in text format. As the user inputs information, the system automatically detects the user's emotional state. At this time, the user's emotional information is also transmitted along with the input data.

[1037] As a specific example, a user might enter, "I have a question about this product. I'm worried."

[1038] Device Role

[1039] The device receives input data and emotion information from the user and transmits it to the server. The device in this case can be a smartphone or desktop PC.

[1040] The request sent by the device will have the following format:

[1041] "I have a question about this product. I'm worried."

[1042] Server Roles

[1043] The server has multiple means for analyzing the received input data and emotional information and generating appropriate responses.

[1044] 1. Receipt and inspection of data

[1045] The server receives the input data sent by the user using the "means for receiving input data" and analyzes it using the "means for pre-checking the received input data and detecting inappropriate content." Specifically, it analyzes the text data using a natural language processing library and analyzes emotional information using an emotion recognition engine (e.g., VADER Sentiment Analysis).

[1046] 2. Submitting the right data and generating answers

[1047] The server uses the "means for sending input data judged appropriate to a generative AI model and receiving the generated answer" to send the input data judged appropriate to a generative AI model (e.g., GPT-3) and generate an answer. At this time, emotional information is taken into consideration to generate a personalized response.

[1048] 3. Check and submit the generated answers

[1049] The server re-examines the generated answer using "means for checking the generated answer in advance and detecting inappropriate content," evaluates it using "means for receiving and analyzing the user's emotional information," and then sends the answer that is judged to be appropriate to the user using "means for sending the answer to the user."

[1050] Hardware and software used

[1051] 1. Hardware

[1052] Server: A cloud server, such as an Amazon EC2 instance

[1053] Devices: Smartphones, desktop PCs

[1054] 2. Software

[1055] Natural language processing models: Generative artificial intelligence models such as GPT-3

[1056] Sentiment analysis library: VADER Sentiment Analysis

[1057] Specific examples

[1058] As a concrete example, let's say a user types, "I have a question about this product. I'm worried." In that case, the following process will be executed:

[1059] 1. The user types "I have a question about this product. I'm worried." into the device, and the emotion is recognized as "anxiety."

[1060] 2. The device sends the input data and emotion information to the server.

[1061] 3. The server receives and analyzes the text data and emotion information.

[1062] 4. If appropriate, the generative AI model generates an answer.

[1063] 5. The server re-examines the generated answers and rates them based on sentiment information.

[1064] 6. If the rating is deemed appropriate, the answer is sent to the user.

[1065] The above is a specific embodiment for carrying out the present invention. This system can provide personalized responses that take into account the user's emotional information, reducing the risk of confidential information being leaked.

[1066] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1067] Step 1:

[1068] The user uses a terminal to create input data and send it along with emotional information. At this time, the user enters the inquiry in text format, and the terminal automatically performs emotional analysis. For example, if the user enters "I have a question about this product. I'm worried," the terminal obtains this text data and the emotional information of "anxiety." The inputs are text data and emotional information. The output is sent to the server.

[1069] Step 2:

[1070] The device receives input data and sends emotional information to the server. The device's role is to relay data, and the input data and emotional information are sent to the server as is without conversion. The input includes the user's text data and emotional information. The output includes the data sent to the server.

[1071] Step 3:

[1072] The server checks the input data and emotional information it receives in advance to detect inappropriate content. Specifically, the server uses a natural language processing library to analyze the appropriateness of the text and an emotion recognition engine to analyze the emotional information. The input includes text data and emotional information sent from the device. Data processing involves natural language processing and emotional analysis. The output is either data deemed appropriate or an error message.

[1073] Step 4:

[1074] Input data deemed appropriate is sent to a generative AI model and the generated answer is received. To do this, the server sends text data and emotional information to the generative AI model and receives the generated answer. The inputs include text data and emotional information. Data calculation involves text generation by the generative AI model. The output is the generated answer.

[1075] Step 5:

[1076] The generated answer is checked in advance to detect inappropriate content. The server again analyzes the appropriateness of the answer using a natural language processing library and evaluates the emotional information using an emotion recognition engine. The input is the generated answer. Data processing involves natural language processing and emotional evaluation. The output is either an answer deemed appropriate or an error message.

[1077] Step 6:

[1078] The answer that is deemed appropriate is sent to the user. The server returns this answer to the terminal, which displays it to the user. The input is the answer that is deemed appropriate. The output is the answer that was sent to the user terminal.

[1079] Step 7:

[1080] The user checks the answer on the device and takes further action. The user checks the answer to the question through the device and asks again if necessary. The input is the answer from the server. The output is the user's satisfaction level and the next action.

[1081] The above is the flow of processing for the system program that realizes the application example.

[1082] 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.

[1083] 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.

[1084] 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.

[1085] [Third embodiment]

[1086] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1087] 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.

[1088] 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).

[1089] 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.

[1090] 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.

[1091] 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).

[1092] 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.

[1093] 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.

[1094] 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.

[1095] 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.

[1096] 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.

[1097] 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."

[1098] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[1099] overview

[1100] The present invention is a system that checks input data provided by users before sending it to a generative artificial intelligence model, and similarly checks the generated answers, providing security services to prevent the leakage of confidential information.

[1101] Basic system configuration

[1102] 1. Users

[1103] The user inputs text data through the terminal.

[1104] Example: User types, "This is a sensitive request."

[1105] 2. Terminal

[1106] It is responsible for sending text data entered by the user to the server.

[1107] Example: Submitting data by clicking a button on a web form or application.

[1108] 3. Server

[1109] The server creates an instance of the "SecurityService" class and receives input data.

[1110] The server first checks the received data using the "checking means." If the data contains inappropriate content, it returns an error message to the user saying "Input is inappropriate."

[1111] If the input data does not contain inappropriate content, the input data is passed to a generative artificial intelligence model and the generated answer is received.

[1112] The server also checks the generated answer using a "checking method." If the answer contains inappropriate content, it returns a message to the user saying, "The answer is inappropriate."

[1113] If it is determined to be an appropriate answer, it returns the answer to the user.

[1114] Specific examples

[1115] 1. Receiving and checking user input

[1116] The user types "This is a general request" into the terminal.

[1117] The terminal sends this input data to the server.

[1118] The server uses the "check_incoming_data" method to check the incoming data.

[1119] 2. Answer generation using generative AI models

[1120] The server passes input data that it deems appropriate to the generative artificial intelligence model.

[1121] The generative artificial intelligence model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[1122] 3. Check and submit the generated answers

[1123] The server checks the response using the "check_outgoing_data" method.

[1124] If the server determines that the answer is appropriate, it returns the answer to the terminal.

[1125] Example: "Generated AI response: This is a common request." is displayed to the user.

[1126] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[1127] The processing flow will be explained below.

[1128] Step 1:

[1129] A user inputs text data into a terminal.

[1130] Example: User types, "This is a general request."

[1131] Step 2:

[1132] The terminal transmits the user's input data to the server.

[1133] Example: The terminal clicks the send button and requests the input data from the server.

[1134] Step 3:

[1135] The server receives input data from the user.

[1136] Example: The server receives data through a REST API.

[1137] Step 4:

[1138] The input data received by the server is checked using the "check_incoming_data" method.

[1139] Example: This method checks for sensitive information or inappropriate keywords.

[1140] Step 5:

[1141] The server determines whether the input data is valid.

[1142] If deemed appropriate, the data is passed to a generative AI model.

[1143] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1144] Example: If the input data is "This is a request containing sensitive information," the server responds with "Invalid input."

[1145] Step 6:

[1146] The server passes input data that it deems appropriate to the generative AI model to generate an answer.

[1147] Example: Call the "generate" method of "AIModel" and pass in the user's input data.

[1148] Step 7:

[1149] A generative AI model generates answers based on user input data.

[1150] Example: Generate the answer "Generation AI answer: This is a common request."

[1151] Step 8:

[1152] The server receives the generated answer.

[1153] Example: Receiving a response from a generative AI model.

[1154] Step 9:

[1155] The server checks the generated answer with the "check_outgoing_data" method.

[1156] Example: This method checks if the answer is longer than a predefined number of characters.

[1157] Step 10:

[1158] The server determines whether the generated answer is appropriate.

[1159] If it is judged to be appropriate, it returns a response to the terminal.

[1160] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1161] Example: If the answer exceeds 500 characters, the server will respond with "The answer is inappropriate."

[1162] Step 11:

[1163] The terminal receives the appropriate response from the server and displays it to the user.

[1164] Example: The user will see the answer "Generated AI answer: This is a common request."

[1165] The above are the specific steps of the processing of the present invention, which can effectively reduce the risk of confidential information leakage when using generative AI models and provide users with safe and reliable services.

[1166] Example 1

[1167] 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."

[1168] While modern dialogue systems and data processing systems using generative artificial intelligence models are extremely useful, they pose a significant risk of confidential information leakage. Security measures are essential, particularly since both the input data provided by users and the generated answers may contain confidential information. However, these systems require effective means to pre-check the input data and generated answers and filter out inappropriate content.

[1169] 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.

[1170] In this invention, the server includes: means for receiving input data from a user; means for checking the received input data in advance to detect inappropriate content; means for sending input data determined to be appropriate to the generative AI model and receiving a generated answer; means for checking the generated answer in advance to detect inappropriate content; means for sending the answer determined to be appropriate to the user; means for converting the input data to be sent to the generative AI model into a prompt format in advance; means for converting the answer received from the generative AI model into a format displayable to the user; means for checking the input data and the generated answer by applying specific security rules to prevent leakage of confidential information; means for saving all data including the input data from the user and the answer from the generative AI model as a log; and means for periodically auditing the saved log data. This makes it possible to effectively prevent leakage of confidential information during the process of using the generative AI model.

[1171] "User" means a person who utilizes the System to input and receive data.

[1172] "Terminal" means a device operated by a user and used to input data and transmit data to a server.

[1173] "Server" refers to the central computer system that receives and validates input data from users and works with the generative artificial intelligence model to generate and validate answers.

[1174] "Input data" refers to data such as text sent by a user to a server via a terminal.

[1175] A "generative artificial intelligence model" refers to an AI system that generates appropriate answers based on input data from users.

[1176] "Prompt sentence" refers to an instruction sentence formatted to appropriately transmit input data to a generative artificial intelligence model.

[1177] "Checking means" refers to a function for analyzing input data and generated answers to check for inappropriate content or confidential information.

[1178] "Inappropriate content" refers to data that contains confidential or security-related information.

[1179] "Security rules" refer to analytical standards and procedures established to prevent the leakage of confidential information.

[1180] "Log" refers to a record of all data processed within the system, such as user input data and generated responses.

[1181] "Audit" refers to the process of periodically reviewing stored log data to verify the operation and security status of a system.

[1182] The present invention is a system for preventing the leakage of confidential information by checking input data provided by a user before sending it to a generative artificial intelligence model, and by similarly checking the generated answers.

[1183] Basic system configuration

[1184] 1. User:

[1185] The user is responsible for inputting and transmitting text data through the terminal.

[1186] Example: A user types, "This is a general request."

[1187] 2. Terminal:

[1188] The terminal is responsible for transmitting text data entered by the user to the server.

[1189] Example: Submitting data by clicking a button on a web form or application.

[1190] 3. Server:

[1191] The server performs the following process:

[1192] 1. Data reception:

[1193] The server creates an instance of the "SecurityService" class and receives input data from the user.

[1194] Example: Use the "check_incoming_data" method to check incoming data for inappropriate content.

[1195] 2. Input data validation:

[1196] If the data contains inappropriate content, the server sends an error message "Input is inappropriate" to the terminal.

[1197] Example: If the user types "This is confidential information", the server generates an error message.

[1198] 3. Data transfer to the generative AI model:

[1199] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[1200] Example: If the input is deemed appropriate, the data is passed to the generative AI model using the prompt "Please output the generative AI's answer."

[1201] 4. Receive a response from the generative AI model:

[1202] The server processes the answers received from the generative artificial intelligence model.

[1203] Example: A generative artificial intelligence model responds with, "Generative AI response: This is a common request."

[1204] 5. Check the generated data:

[1205] The server checks the generated data using the "check_outgoing_data" method.

[1206] Example: The server checks the generated data again using pre-filtering rules and databases.

[1207] 6. Error Handling:

[1208] If the answer contains inappropriate content, the server sends an error message "The answer is inappropriate" to the terminal.

[1209] Action: The server formats an appropriate error message and sends it to the terminal.

[1210] 7. Submitting the right response:

[1211] The server sends the answer it deems appropriate to the user.

[1212] Example: "Generated AI response: This is a common request." is sent to the user.

[1213] Examples and prompt formats

[1214] 1. Receive and check user input:

[1215] The user types "This is a general request" into the terminal.

[1216] The terminal sends this input data to the server.

[1217] The server checks the incoming data with the "check_incoming_data" method.

[1218] 2. Answer generation using generative AI models:

[1219] The server sends input data that it deems appropriate to the generative AI model.

[1220] The generative AI model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[1221] 3. Check and submit the generated answers:

[1222] The server checks the generated answer using the "check_outgoing_data" method.

[1223] The server returns an answer that it deems appropriate to the terminal.

[1224] Example: "Generated AI response: This is a common request." is displayed to the user.

[1225] Prompt Sentence Examples

[1226] Please output the generative AI's answer: This is a common request.

[1227] In this way, this system effectively reduces the risk of confidential information leaks when using generative AI. By inspecting the text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[1228] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1229] Step 1:

[1230] The user enters text data into the terminal.

[1231] Input: Text data entered by the user, e.g. "This is a general request."

[1232] Specific action: The user types characters on the keyboard and presses the Enter key.

[1233] Output: Input data is displayed on the terminal.

[1234] Step 2:

[1235] The terminal transmits the input text data to the server.

[1236] Input: Text data entered by the user, e.g., "This is a general request."

[1237] Specific operation: When you click the send button on the device, the browser generates an HTTP POST request and sends it to the server.

[1238] Output: Data is sent to the server.

[1239] Step 3:

[1240] The server receives the data from the user and checks the content using the "check_incoming_data" method.

[1241] Input: Text data sent from the terminal, e.g., "This is a general request."

[1242] Specific operation: The server creates an instance of the "SecurityService" class and analyzes the data using the database and filtering rules.

[1243] Data manipulation: Detecting inappropriate content using specific keywords and pattern matching.

[1244] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[1245] Step 4:

[1246] If the server determines that the data is invalid, it sends an error message to the terminal saying "Invalid input."

[1247] Input: Inspection results, e.g., inappropriate data "This is confidential information."

[1248] Specific behavior: The server generates an HTTP response and sends data including an error message.

[1249] Output: An error message is displayed on the terminal.

[1250] Step 5:

[1251] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[1252] Input: Inspection result, e.g. "This is a common request."

[1253] Specific operation: The server generates an appropriate prompt sentence and sends the data to the API of the generative artificial intelligence model.

[1254] Data processing: Convert the data into the format of the prompt statement.

[1255] Output: The prompt sent to the generative AI model.

[1256] Step 6:

[1257] The server receives the answer from the generative artificial intelligence model.

[1258] Input: An answer from a generative AI model, e.g., "Generative AI answer: This is a common request."

[1259] Specific operation: The server receives the HTTP response and parses the response body.

[1260] Output: The received response data.

[1261] Step 7:

[1262] The server checks the generated data using the "check_outgoing_data" method.

[1263] Input: Answer data from a generative AI model, e.g., "Generative AI answer: This is a common request."

[1264] Specific operation: The server examines the data using a filtering algorithm.

[1265] Data processing: Check that the response data does not contain inappropriate content.

[1266] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[1267] Step 8:

[1268] If the server generates an answer that contains inappropriate content, it sends an error message to the terminal saying "Answer is inappropriate."

[1269] Input: Test results, e.g., incorrect response data.

[1270] Specific behavior: The server generates an HTTP response and returns data including an error message.

[1271] Output: An error message is displayed on the terminal.

[1272] Step 9:

[1273] The server sends the answer that it deems appropriate to the terminal.

[1274] Input: Inspection result, e.g. "Generated AI response: This is a common request."

[1275] Specific behavior: The server generates an HTTP response and sends back data containing the appropriate answer.

[1276] Output: The answer of the generated AI is displayed on the terminal.

[1277] (Application example 1)

[1278] 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."

[1279] In systems using generative artificial intelligence models, if the input data or generated answers contain confidential information, it is important to prevent the leakage of that information. However, many current systems lack a means to automatically verify the appropriateness of the input data or generated answers, which poses a risk of confidential information leaking to the outside. The present invention aims to solve this problem and prevent the leakage of confidential information.

[1280] 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.

[1281] In this invention, the server includes means for receiving input data, means for checking the input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, and means for including a mobile application that runs on a smartphone and has a security function for preventing leakage of confidential information. This makes it possible to automatically perform security checks on the input data and the generated answers and prevent leakage of confidential information.

[1282] "Input Data" means text or other information provided to a system by a user.

[1283] "Means for receiving" refers to a device or method for receiving input data provided by a user.

[1284] "Pre-checking measures" refers to functions and methods for automatically inspecting input data and generated responses to check for inappropriate content or confidential information.

[1285] "Means for detecting inappropriate content" means a device or method for identifying inappropriate elements or information in input data or generated answers.

[1286] A "generative artificial intelligence model" refers to an artificial intelligence algorithm or system that generates answers based on input data provided by a user.

[1287] "Means for receiving a generated answer" refers to a device or method for receiving an answer generated from a generative artificial intelligence model.

[1288] "Means for sending answers that are judged to be appropriate to the user" refers to a device or method for returning a checked answer to the user if the checked answer is judged to be appropriate.

[1289] A "smartphone" is a type of mobile phone, a portable information terminal that can connect to the Internet and run various applications.

[1290] "Security features to prevent the leakage of confidential information" refers to features and methods to prevent confidential information from leaking to the outside if it is included in input data or generated answers.

[1291] "Mobile application" refers to a software application that runs on a mobile information device such as a smartphone or tablet.

[1292] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[1293] System Configuration

[1294] 1. Users

[1295] A user operates an application using a smartphone to input text data.

[1296] Example: User types "Summarize the meeting."

[1297] 2. Terminal

[1298] The terminal is responsible for receiving text data entered by the user and transmitting it to the server.

[1299] Specifically, data is submitted by clicking a button on a web form or application.

[1300] 3. Server

[1301] The server receives the input data and performs security checks by creating an instance of the "SecurityService" class and validating the input data.

[1302] The "check_incoming_data" method is used to check whether inappropriate content is included. If inappropriate content is included, the error message "Input is inappropriate" is returned to the terminal.

[1303] Input data that is deemed appropriate is sent to the generative artificial intelligence model, which then receives the generated answer.

[1304] The generated answer is inspected using the "check_outgoing_data" method. If it contains inappropriate content, the message "Answer is inappropriate" is returned to the terminal.

[1305] If the answer is determined to be appropriate, the answer is returned to the terminal and displayed to the user.

[1306] Data processing and hardware / software used

[1307] Hardware: Smartphone (iOS / Android)

[1308] Software: Python, Flask (web framework), TensorFlow / Transformers (generative AI models), SQL database

[1309] Data calculation: Input data checks and answer checks are performed by the "check_incoming_data()" and "check_outgoing_data()" methods in the "SecurityService" class.

[1310] Specific examples

[1311] When a user types "This is a general request" into a smartphone application, the content is sent to the server. The server checks this input data using the "check_incoming_data" method and sends it to the generative AI model only if it is deemed appropriate.

[1312] The generative AI model generates an answer, "Generative AI's answer: This is a typical request," and returns it to the server. The server then checks the answer again using the "check_outgoing_data" method, and if it is deemed appropriate, returns the answer to the user.

[1313] Example prompt sentence:

[1314] User input: Check if the content of 'Summarize the meeting' is typical, and if it is, run a function that provides appropriate feedback.

[1315] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[1316] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1317] Step 1:

[1318] The user opens the application on their smartphone and enters text. The input data is "Summarize the contents of the meeting." The device receives this user input and sends it to the server.

[1319] Input: User's text data "Summarize the contents of the meeting"

[1320] Output: Text data sent from the terminal to the server

[1321] Step 2:

[1322] The server checks the input data received from the terminal using the "check_incoming_data" method of the "SecurityService" class, which checks whether the input data contains confidential information or inappropriate content.

[1323] Input: Text data received by the server: "Summarize the contents of the meeting"

[1324] Data processing: Checks to see if there is any confidential information

[1325] Output: Check result (good or bad)

[1326] Specific behavior:

[1327] The text data received by the server is passed to the "check_incoming_data" method, which checks whether it contains specific keywords. For example, there is a rule that detects keywords such as "confidential" and "important."

[1328] Step 3:

[1329] If the server checks the results, it sends the text data to a generative AI model, which receives a specific prompt and generates an appropriate answer based on it.

[1330] Input: Text data that passed the check: "Summarize the contents of the meeting"

[1331] Data processing: Passing the data as prompts to the generative AI model

[1332] Output: The answer generated by the generative AI model

[1333] Specific behavior:

[1334] The server passes appropriate input data to the generative AI model as a prompt. For example, it requests the model to generate a summary of the user's input: 'Summarize the contents of the meeting.'

[1335] Step 4:

[1336] The generative AI model generates an answer based on the prompt and returns it to the server, for example, in the form of a "generated summary of the meeting content."

[1337] Input: Prompt sentence for generative AI model

[1338] Data computation: Generative AI models generate answers

[1339] Output: The generated answer

[1340] Specific behavior:

[1341] A generative AI model uses natural language processing based on the prompt to generate an appropriate answer.

[1342] Step 5:

[1343] The server checks the generated answer in advance using the "check_outgoing_data" method of the "SecurityService" class to ensure that the generated answer does not contain inappropriate content.

[1344] Input: Generated Answer

[1345] Data processing: Checking generated answers

[1346] Output: Check result (good or bad)

[1347] Specific behavior:

[1348] The server passes the generated answer to the "check_outgoing_data" method to check whether it contains specific keywords or phrases.

[1349] Step 6:

[1350] If the check result is appropriate, the server returns the generated answer to the terminal and displays it to the user. If the check result is inappropriate, the server returns an error message saying "The answer is inappropriate."

[1351] Input: Answer that passes the check

[1352] Data processing: Formatting answers (if necessary)

[1353] Output: The appropriate answer or error message that is displayed to the user

[1354] Specific behavior:

[1355] The server sends the answer that passes the check to the terminal, and the answer is displayed to the user by the smartphone application. For example, the terminal executes the operation "display the data received from the server."

[1356] 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.

[1357] This invention is a system that uses a generative artificial intelligence model to check the appropriateness of the user's input data and the generated answers, and also takes the user's emotions into consideration, thereby reducing the risk of confidential information leaks and providing more personalized responses. Below, we will explain in detail the main roles of the user, terminal, and server, and the specific processes performed by each.

[1358] overview

[1359] In addition to means for checking input data, means for using a generative artificial intelligence model, and means for checking generated answers, the present invention also includes an emotion engine that recognizes the user's emotions and reflects them in various checks and response generation.

[1360] Basic system configuration

[1361] 1. Users

[1362] The user inputs text data through the terminal.

[1363] Input can also include emotional states.

[1364] Example: A user types, "This is a general request," and their emotion is recognized as "calm."

[1365] 2. Terminal

[1366] It is responsible for sending the text data and emotional information entered by the user to the server.

[1367] Example: The device clicks the send button and requests emotion information along with the input data from the server.

[1368] 3. Server

[1369] The server has means for receiving the text data and the emotion information.

[1370] The received data is inspected using the "check_incoming_data" method, and the emotion engine analyzes the user's emotion information.

[1371] Example: The server further scrutinizes the input data based on a "calm" emotional state.

[1372] 4. Data appropriateness assessment

[1373] If the server determines that the input data is appropriate, it passes the data and emotional information to a generative AI model to generate an answer.

[1374] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1375] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[1376] 5. Use of generative AI models

[1377] The generative AI model generates answers based on the received text data and emotional information.

[1378] The emotion engine generates answers that reflect the user's current emotional state.

[1379] Example: Generated response: "Generated AI response: This is a common request. We're glad you're calm."

[1380] 6. Check the generated answers

[1381] The server checks the generated answer again with the "check_outgoing_data" method.

[1382] The emotion engine also evaluates the appropriateness of the generated answers based on emotional information.

[1383] For example: Make sure your answer is within the character limit and emotionally relevant.

[1384] 7. Sending the Response

[1385] If the answer is judged to be appropriate, it is returned to the device and displayed to the user in an appropriate state that reflects the emotional information.

[1386] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1387] Example: The user sees the message, "Generated AI response: This is a common request. We're glad you're calm."

[1388] Specific examples

[1389] 1. Receiving and checking user input

[1390] The user types "This is a general request" into the device, and the system recognizes the user's emotional state as "calm."

[1391] The device sends this input data and emotional information to the server.

[1392] 2. Answer generation using generative AI models

[1393] The server checks the data and emotion information using the "check_incoming_data" method.

[1394] Data and emotional information deemed appropriate are sent to the generative AI model.

[1395] The generative AI model generates an answer, "Generative AI's answer: This is a common request. We're glad you're calm." and returns it to the server.

[1396] 3. Check and submit the generated answers

[1397] The server checks the response with the "check_outgoing_data" method.

[1398] If it is determined to be an appropriate answer, it returns the answer to the terminal.

[1399] The generated answer is displayed to the user.

[1400] These are the specific steps of the process of the present invention. This system reduces the risk of leaking confidential user information when using generative AI models, and can provide personalized responses that take user emotions into account.

[1401] The processing flow will be explained below.

[1402] Step 1:

[1403] A user inputs text data into a terminal.

[1404] Example: User types, "This is a general request."

[1405] Step 2:

[1406] The terminal collects the user's input data as well as the emotional state.

[1407] Example: The device recognizes the user's emotion as "calm" through facial recognition and voice tone analysis.

[1408] Step 3:

[1409] The device transmits the collected text data and emotion information to a server.

[1410] Example: The device clicks the send button and sends the input data and emotion information as a request to the server.

[1411] Step 4:

[1412] The server receives input data and emotion information from the user.

[1413] Example: The server receives data and emotion information through a REST API.

[1414] Step 5:

[1415] The input data received by the server is checked using the "check_incoming_data" method.

[1416] Example: This method checks for sensitive information or inappropriate keywords.

[1417] Step 6:

[1418] The server analyzes the received emotion information using an emotion engine.

[1419] Example: The server analyzes the emotional information of "calm" and further scrutinizes the input data based on that information.

[1420] Step 7:

[1421] The server determines whether the input data is valid.

[1422] If deemed appropriate, the data and emotional information is passed to a generative AI model.

[1423] If it is determined to be inappropriate, an error message "Input is inappropriate" is returned and the process is terminated.

[1424] For example, "This is a request containing sensitive information." If the data is deemed inappropriate, the server will block it.

[1425] Step 8:

[1426] The server passes the input data and emotional information it deems appropriate to the generative AI model to generate an answer.

[1427] Example: Call the "generate" method of the generative AI model, passing in the user's input data and emotion information.

[1428] Step 9:

[1429] The generative AI model generates answers based on the user's input data and emotional information.

[1430] Example: The generative AI model generates the answer "Generative AI response: This is a common request. We're glad you're calm."

[1431] Step 10:

[1432] The server receives the generated answer.

[1433] Example: Receiving an answer as a response from a generative AI model.

[1434] Step 11:

[1435] The server checks the generated answer again with the "check_outgoing_data" method.

[1436] For example, this method checks whether the answer is within a predefined character limit and whether it is sentiment-based.

[1437] Step 12:

[1438] The server determines whether the generated answer is appropriate.

[1439] If it is determined to be appropriate, the answer is sent to the terminal.

[1440] If it is determined to be inappropriate, an error message "The answer is inappropriate" is returned and the process is terminated.

[1441] Example: If the answer exceeds 500 characters, the server will block it as an inappropriate answer.

[1442] Step 13:

[1443] The terminal receives the appropriate response from the server and displays it to the user.

[1444] Example: The user will see the answer, "Generator's response: This is a common request. We're glad you're calm."

[1445] These are the specific steps of the process of the present invention. This system reduces the risk of confidential information leakage when using generative AI models and can provide personalized responses that take user emotions into account.

[1446] Example 2

[1447] 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."

[1448] Conventional systems using generative artificial intelligence models have had difficulty accurately evaluating the appropriateness of user input data and generated answers. Furthermore, because they do not take the user's emotions into account, they are unable to provide personalized responses, resulting in reduced user satisfaction. Furthermore, there is a risk of data leakage due to inappropriate data content or emotional state. The present invention aims to solve these problems, improve the appropriateness of user input data and generated answers, and provide personalized responses that take the user's emotions into account.

[1449] 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.

[1450] In this invention, the server includes means for receiving text data and the user's emotional information from the user, means for analyzing the received text data and emotional information and detecting inappropriate content or an inappropriate emotional state, means for transmitting the text data and emotional information determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for analyzing the generated answer and the emotional information contained in the answer and detecting inappropriate content or an inappropriate emotional state, and means for transmitting the answer determined to be appropriate to the user. This makes it possible to improve the appropriateness of the user's input data and the generated answer and to provide a personalized response that takes the user's emotions into consideration.

[1451] A "user" is an entity that utilizes the system to provide input data and receive generated responses.

[1452] "Text data" refers to character string information that a user inputs through a terminal.

[1453] "Emotional information" refers to data that analyzes a user's emotional state and is information that the system uses to personalize the user's input data and responses.

[1454] "Means for receiving" refers to the modules and functions that allow a server or terminal to receive data from a user.

[1455] "Means for analyzing" refers to software or algorithms that examine the received data and check for appropriateness of content and emotional state.

[1456] "Inappropriate content" refers to data that the system does not tolerate, such as leaks of personal information, offensive language, or inappropriate language.

[1457] "Inappropriate emotional state" refers to a negative or unstable emotional state of the user detected by the system.

[1458] A "generative artificial intelligence model" refers to a machine learning model or algorithm that automatically generates appropriate responses based on received text data.

[1459] "Generated response" refers to a reply created by a generative artificial intelligence model.

[1460] "Transmission means" refers to the network infrastructure and communications protocols used by the server to transmit the generated response back to the user.

[1461] "Personalized responses" refer to replies that are individually tailored to take into account the user's emotional information and text data.

[1462] "System" refers to the overall technical environment and processing flow for receiving input data from a user, analyzing it, and generating and returning an appropriate response.

[1463] This system generates personalized responses based on user input data and associated emotional information, reducing the risk of confidential information leaks. The specific roles and operations of the user, device, and server are detailed below.

[1464] user

[1465] The user uses the device to input text data. At this time, the device is set up so that the user's emotional state is also input. For example, if the user inputs "This is a general request," the user's emotional state at that time is recognized as "calm."

[1466] Terminal

[1467] The terminal is responsible for transmitting the text data and emotion information input by the user to the server. Specifically, the terminal sends the input data and emotion information to the server through the network module, and then receives a response from the server and displays it to the user.

[1468] Example: The device clicks the send button and sends the input data "This is a general request" and emotional information "Calm" to the server.

[1469] server

[1470] The server provides the following specific functions:

[1471] 1. Data Reception

[1472] The server receives the text data and emotion information sent from the terminal. This reception is performed using a "receiving means."

[1473] 2. Data analysis and testing

[1474] The received data is analyzed using the "check_incoming_data" method to evaluate the appropriateness of the text data and the appropriateness of the emotional information. The server performs this analysis using an NLP (Natural Language Processing) module that includes an emotional engine.

[1475] Example: The server receives the text "This is a general request" and analyzes the emotional information "I feel calm."

[1476] 3. Data appropriateness assessment

[1477] The text data and emotion information that are deemed appropriate are sent to a generative AI model. If the content is inappropriate, the process ends and an error message is returned.

[1478] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[1479] 4. Response Generation

[1480] The generative AI model generates answers based on the received text data and emotional information, with the emotional engine creating personalized responses that reflect the user's current emotional state.

[1481] Example: A generative AI model generates the response, "Generative AI response: This is a common request. We're glad you're calm."

[1482] 5. Inspecting and Sending the Response

[1483] The generated answer is checked again by the "check_outgoing_data" method, and if it is deemed appropriate it is sent to the user. If it is an inappropriate answer, the process is terminated and an error message is returned.

[1484] Example: The server checks whether the generated answer is within the character limit and appropriate for the emotional state, and then sends the appropriate answer to the device.

[1485] The above is a specific example of the processing performed by the system of the present invention. This system ensures the appropriateness of the user's input data and the generated answers, and makes it possible to provide personalized responses that reflect the user's emotions.

[1486] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1487] Program processing flow

[1488] Step 1:

[1489] The user enters text data into the input field of the terminal and clicks the "Send" button. At this time, the emotion recognition software analyzes the user's emotional information and provides it to the terminal along with the input. For example, the input data is "This is a general request," and the emotional information is "calm."

[1490] input:

[1491] Text data entered by the user

[1492] Emotional information analyzed by emotion recognition software

[1493] output:

[1494] Text data and emotional information received by the device

[1495] Specific behavior:

[1496] User enters text and clicks submit button

[1497] Emotion recognition software analyzes user emotions

[1498] Step 2:

[1499] The terminal sends the text data and emotion information entered by the user to the server, and this data is transferred to the server using the network module.

[1500] input:

[1501] Text data and emotional information entered into the device

[1502] output:

[1503] Text data and emotional information sent to the server

[1504] Specific behavior:

[1505] The device's network module sends text data and emotional information to the server.

[1506] Step 3:

[1507] The server checks the received text data and emotion information using the "check_incoming_data" method. The server uses the NLP module and emotion engine to analyze the appropriateness of this data.

[1508] input:

[1509] Text data and emotion information received by the server

[1510] output:

[1511] Appropriateness evaluation results for text data and emotional information

[1512] Specific behavior:

[1513] The server analyzes the text using the NLP module

[1514] Emotion engine analyzes emotional information

[1515] Step 4:

[1516] The server determines the appropriateness of the data based on the analysis results. If the data is deemed appropriate, it sends it to the generative AI model. If the data is inappropriate, it generates an error message and returns it to the user.

[1517] input:

[1518] Text data and emotional information analyzed by the server

[1519] output:

[1520] When deemed appropriate: Data sent to generative AI models

[1521] If deemed inappropriate: Error message

[1522] Specific behavior:

[1523] Sending the right data to generative AI models

[1524] Invalid data generates an error message

[1525] Step 5:

[1526] The generative AI model generates answers based on the received text data and emotional information, and the emotion engine creates personalized responses that reflect the user's emotional state.

[1527] input:

[1528] Appropriate text data and emotional information sent from the server

[1529] output:

[1530] Generated Answer

[1531] Specific behavior:

[1532] Generative AI model generates responses based on text data and emotional information

[1533] Emotion engine creates responses that reflect the user's emotions

[1534] Step 6:

[1535] The server checks the generated answer using the "check_outgoing_data" method. If the check passes, the answer is sent to the user. If it is invalid, an error message is generated.

[1536] input:

[1537] Answers generated by generative AI models

[1538] output:

[1539] If the check passes: The answer sent to the user

[1540] If not passed: Error message

[1541] Specific behavior:

[1542] The server re-examines the generated answer

[1543] Sends appropriate response to user, may generate error message

[1544] Step 7:

[1545] The server sends an appropriate response to the user, and the device receives this response and displays it to the user. If the response is inappropriate, an error message is displayed.

[1546] input:

[1547] The response or error message sent by the server

[1548] output:

[1549] The answer or error message that the user sees

[1550] Specific behavior:

[1551] The device receives the response from the server and displays it to the user.

[1552] Display error messages if necessary

[1553] (Application example 2)

[1554] 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."

[1555] Conventional systems utilizing generative artificial intelligence models have the ability to determine the appropriateness of user input data and generated responses, but because they do not take user emotions into account, responses are not personalized and the user experience is insufficient. Furthermore, if emotional information is not taken into account, the appropriateness of the data cannot be adequately determined, which may increase the risk of confidential information being leaked. Therefore, an objective of the present invention is to provide a system that utilizes a user's emotional information to provide personalized responses based on emotions and reduce the risk of confidential information being leaked.

[1556] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving input data, means for checking the received input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, means for sending the answer determined to be appropriate to the user, emotion recognition means for receiving and analyzing the user's emotion information, and means for generating a personalized answer based on the emotion information. This enables a personalized response that is in line with the user's emotions, making it possible to reduce the risk of confidential information being leaked.

[1557] The "means for receiving input data" refers to a function or device for receiving text data or emotion information sent by a user.

[1558] "Means for checking received input data in advance and detecting inappropriate content" refers to functions or devices for analyzing input data and identifying content that does not conform to specified standards.

[1559] "Means for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer" refers to a function or device for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer.

[1560] "Means for checking generated answers in advance and detecting inappropriate content" refers to functions or devices for re-examining answers obtained from generative AI models to check whether they contain inappropriate content.

[1561] The "means for sending a response that is determined to be appropriate to the user" refers to a function or device for sending a response that is determined to not contain inappropriate content to the user.

[1562] "Emotion recognition means for receiving and analyzing user's emotional information" refers to a function or device for receiving information about the user's emotional state and analyzing that state.

[1563] The "means for generating a personalized answer based on emotional information" refers to a function or device for generating an answer optimized for a user, taking into account the emotional information of the user.

[1564] In order to put the present invention into practice, it is necessary to build a system that specifically separates the roles of users, terminals, and servers.

[1565] User Roles

[1566] The user's role is to access the online shopping site and input product inquiries and support requests in text format. As the user inputs information, the system automatically detects the user's emotional state. At this time, the user's emotional information is also transmitted along with the input data.

[1567] As a specific example, a user might enter, "I have a question about this product. I'm worried."

[1568] Device Role

[1569] The device receives input data and emotion information from the user and transmits it to the server. The device in this case can be a smartphone or desktop PC.

[1570] The request sent by the device will have the following format:

[1571] "I have a question about this product. I'm worried."

[1572] Server Roles

[1573] The server has multiple means for analyzing the received input data and emotional information and generating appropriate responses.

[1574] 1. Receipt and inspection of data

[1575] The server receives the input data sent by the user using the "means for receiving input data" and analyzes it using the "means for pre-checking the received input data and detecting inappropriate content." Specifically, it analyzes the text data using a natural language processing library and analyzes emotional information using an emotion recognition engine (e.g., VADER Sentiment Analysis).

[1576] 2. Submitting the right data and generating answers

[1577] The server uses the "means for sending input data judged appropriate to a generative AI model and receiving the generated answer" to send the input data judged appropriate to a generative AI model (e.g., GPT-3) and generate an answer. At this time, emotional information is taken into consideration to generate a personalized response.

[1578] 3. Check and submit the generated answers

[1579] The server re-examines the generated answer using "means for checking the generated answer in advance and detecting inappropriate content," evaluates it using "means for receiving and analyzing the user's emotional information," and then sends the answer that is judged to be appropriate to the user using "means for sending the answer to the user."

[1580] Hardware and software used

[1581] 1. Hardware

[1582] Server: A cloud server, such as an Amazon EC2 instance

[1583] Devices: Smartphones, desktop PCs

[1584] 2. Software

[1585] Natural language processing models: Generative artificial intelligence models such as GPT-3

[1586] Sentiment analysis library: VADER Sentiment Analysis

[1587] Specific examples

[1588] As a concrete example, let's say a user types, "I have a question about this product. I'm worried." In that case, the following process will be executed:

[1589] 1. The user types "I have a question about this product. I'm worried." into the device, and the emotion is recognized as "anxiety."

[1590] 2. The device sends the input data and emotion information to the server.

[1591] 3. The server receives and analyzes the text data and emotion information.

[1592] 4. If appropriate, the generative AI model generates an answer.

[1593] 5. The server re-examines the generated answers and rates them based on sentiment information.

[1594] 6. If the rating is deemed appropriate, the answer is sent to the user.

[1595] The above is a specific embodiment for carrying out the present invention. This system can provide personalized responses that take into account the user's emotional information, reducing the risk of confidential information being leaked.

[1596] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1597] Step 1:

[1598] The user uses a terminal to create input data and send it along with emotional information. At this time, the user enters the inquiry in text format, and the terminal automatically performs emotional analysis. For example, if the user enters "I have a question about this product. I'm worried," the terminal obtains this text data and the emotional information of "anxiety." The inputs are text data and emotional information. The output is sent to the server.

[1599] Step 2:

[1600] The device receives input data and sends emotional information to the server. The device's role is to relay data, and the input data and emotional information are sent to the server as is without conversion. The input includes the user's text data and emotional information. The output includes the data sent to the server.

[1601] Step 3:

[1602] The server checks the input data and emotional information it receives in advance to detect inappropriate content. Specifically, the server uses a natural language processing library to analyze the appropriateness of the text and an emotion recognition engine to analyze the emotional information. The input includes text data and emotional information sent from the device. Data processing involves natural language processing and emotional analysis. The output is either data deemed appropriate or an error message.

[1603] Step 4:

[1604] Input data deemed appropriate is sent to a generative AI model and the generated answer is received. To do this, the server sends text data and emotional information to the generative AI model and receives the generated answer. The inputs include text data and emotional information. Data calculation involves text generation by the generative AI model. The output is the generated answer.

[1605] Step 5:

[1606] The generated answer is checked in advance to detect inappropriate content. The server again analyzes the appropriateness of the answer using a natural language processing library and evaluates the emotional information using an emotion recognition engine. The input is the generated answer. Data processing involves natural language processing and emotional evaluation. The output is either an answer deemed appropriate or an error message.

[1607] Step 6:

[1608] The answer that is deemed appropriate is sent to the user. The server returns this answer to the terminal, which displays it to the user. The input is the answer that is deemed appropriate. The output is the answer that was sent to the user terminal.

[1609] Step 7:

[1610] The user checks the answer on the device and takes further action. The user checks the answer to the question through the device and asks again if necessary. The input is the answer from the server. The output is the user's satisfaction level and the next action.

[1611] The above is the flow of processing for the system program that realizes the application example.

[1612] 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.

[1613] 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.

[1614] 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.

[1615] [Fourth embodiment]

[1616] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1617] 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.

[1618] 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).

[1619] 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.

[1620] 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.

[1621] 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).

[1622] 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.

[1623] 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.

[1624] 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.

[1625] 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.

[1626] 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.

[1627] 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.

[1628] 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."

[1629] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[1630] overview

[1631] The present invention is a system that checks input data provided by users before sending it to a generative artificial intelligence model, and similarly checks the generated answers, providing security services to prevent the leakage of confidential information.

[1632] Basic system configuration

[1633] 1. Users

[1634] The user inputs text data through the terminal.

[1635] Example: User types, "This is a sensitive request."

[1636] 2. Terminal

[1637] It is responsible for sending text data entered by the user to the server.

[1638] Example: Submitting data by clicking a button on a web form or application.

[1639] 3. Server

[1640] The server creates an instance of the "SecurityService" class and receives input data.

[1641] The server first checks the received data using the "checking means." If the data contains inappropriate content, it returns an error message to the user saying "Input is inappropriate."

[1642] If the input data does not contain inappropriate content, the input data is passed to a generative artificial intelligence model and the generated answer is received.

[1643] The server also checks the generated answer using a "checking method." If the answer contains inappropriate content, it returns a message to the user saying, "The answer is inappropriate."

[1644] If it is determined to be an appropriate answer, it returns the answer to the user.

[1645] Specific examples

[1646] 1. Receiving and checking user input

[1647] The user types "This is a general request" into the terminal.

[1648] The terminal sends this input data to the server.

[1649] The server uses the "check_incoming_data" method to check the incoming data.

[1650] 2. Answer generation using generative AI models

[1651] The server passes input data that it deems appropriate to the generative artificial intelligence model.

[1652] The generative artificial intelligence model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[1653] 3. Check and submit the generated answers

[1654] The server checks the response using the "check_outgoing_data" method.

[1655] If the server determines that the answer is appropriate, it returns the answer to the terminal.

[1656] Example: "Generated AI response: This is a common request." is displayed to the user.

[1657] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[1658] The processing flow will be explained below.

[1659] Step 1:

[1660] A user inputs text data into a terminal.

[1661] Example: User types, "This is a general request."

[1662] Step 2:

[1663] The terminal transmits the user's input data to the server.

[1664] Example: The terminal clicks the send button and requests the input data from the server.

[1665] Step 3:

[1666] The server receives input data from the user.

[1667] Example: The server receives data through a REST API.

[1668] Step 4:

[1669] The input data received by the server is checked using the "check_incoming_data" method.

[1670] Example: This method checks for sensitive information or inappropriate keywords.

[1671] Step 5:

[1672] The server determines whether the input data is valid.

[1673] If deemed appropriate, the data is passed to a generative AI model.

[1674] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1675] Example: If the input data is "This is a request containing sensitive information," the server responds with "Invalid input."

[1676] Step 6:

[1677] The server passes input data that it deems appropriate to the generative AI model to generate an answer.

[1678] Example: Call the "generate" method of "AIModel" and pass in the user's input data.

[1679] Step 7:

[1680] A generative AI model generates answers based on user input data.

[1681] Example: Generate the answer "Generation AI answer: This is a common request."

[1682] Step 8:

[1683] The server receives the generated answer.

[1684] Example: Receiving a response from a generative AI model.

[1685] Step 9:

[1686] The server checks the generated answer with the "check_outgoing_data" method.

[1687] Example: This method checks if the answer is longer than a predefined number of characters.

[1688] Step 10:

[1689] The server determines whether the generated answer is appropriate.

[1690] If it is judged to be appropriate, it returns a response to the terminal.

[1691] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1692] Example: If the answer exceeds 500 characters, the server will respond with "The answer is inappropriate."

[1693] Step 11:

[1694] The terminal receives the appropriate response from the server and displays it to the user.

[1695] Example: The user will see the answer "Generated AI answer: This is a common request."

[1696] The above are the specific steps of the processing of the present invention, which can effectively reduce the risk of confidential information leakage when using generative AI models and provide users with safe and reliable services.

[1697] Example 1

[1698] 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."

[1699] While modern dialogue systems and data processing systems using generative artificial intelligence models are extremely useful, they pose a significant risk of confidential information leakage. Security measures are essential, particularly since both the input data provided by users and the generated answers may contain confidential information. However, these systems require effective means to pre-check the input data and generated answers and filter out inappropriate content.

[1700] 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.

[1701] In this invention, the server includes: means for receiving input data from a user; means for checking the received input data in advance to detect inappropriate content; means for sending input data determined to be appropriate to the generative AI model and receiving a generated answer; means for checking the generated answer in advance to detect inappropriate content; means for sending the answer determined to be appropriate to the user; means for converting the input data to be sent to the generative AI model into a prompt format in advance; means for converting the answer received from the generative AI model into a format displayable to the user; means for checking the input data and the generated answer by applying specific security rules to prevent leakage of confidential information; means for saving all data including the input data from the user and the answer from the generative AI model as a log; and means for periodically auditing the saved log data. This makes it possible to effectively prevent leakage of confidential information during the process of using the generative AI model.

[1702] "User" means a person who utilizes the System to input and receive data.

[1703] "Terminal" means a device operated by a user and used to input data and transmit data to a server.

[1704] "Server" refers to the central computer system that receives and validates input data from users and works with the generative artificial intelligence model to generate and validate answers.

[1705] "Input data" refers to data such as text sent by a user to a server via a terminal.

[1706] A "generative artificial intelligence model" refers to an AI system that generates appropriate answers based on input data from users.

[1707] "Prompt sentence" refers to an instruction sentence formatted to appropriately transmit input data to a generative artificial intelligence model.

[1708] "Checking means" refers to a function for analyzing input data and generated answers to check for inappropriate content or confidential information.

[1709] "Inappropriate content" refers to data that contains confidential or security-related information.

[1710] "Security rules" refer to analytical standards and procedures established to prevent the leakage of confidential information.

[1711] "Log" refers to a record of all data processed within the system, such as user input data and generated responses.

[1712] "Audit" refers to the process of periodically reviewing stored log data to verify the operation and security status of a system.

[1713] The present invention is a system for preventing the leakage of confidential information by checking input data provided by a user before sending it to a generative artificial intelligence model, and by similarly checking the generated answers.

[1714] Basic system configuration

[1715] 1. User:

[1716] The user is responsible for inputting and transmitting text data through the terminal.

[1717] Example: A user types, "This is a general request."

[1718] 2. Terminal:

[1719] The terminal is responsible for transmitting text data entered by the user to the server.

[1720] Example: Submitting data by clicking a button on a web form or application.

[1721] 3. Server:

[1722] The server performs the following process:

[1723] 1. Data reception:

[1724] The server creates an instance of the "SecurityService" class and receives input data from the user.

[1725] Example: Use the "check_incoming_data" method to check incoming data for inappropriate content.

[1726] 2. Input data validation:

[1727] If the data contains inappropriate content, the server sends an error message "Input is inappropriate" to the terminal.

[1728] Example: If the user types "This is confidential information", the server generates an error message.

[1729] 3. Data transfer to the generative AI model:

[1730] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[1731] Example: If the input is deemed appropriate, the data is passed to the generative AI model using the prompt "Please output the generative AI's answer."

[1732] 4. Receive a response from the generative AI model:

[1733] The server processes the answers received from the generative artificial intelligence model.

[1734] Example: A generative artificial intelligence model responds with, "Generative AI response: This is a common request."

[1735] 5. Check the generated data:

[1736] The server checks the generated data using the "check_outgoing_data" method.

[1737] Example: The server checks the generated data again using pre-filtering rules and databases.

[1738] 6. Error Handling:

[1739] If the answer contains inappropriate content, the server sends an error message "The answer is inappropriate" to the terminal.

[1740] Action: The server formats an appropriate error message and sends it to the terminal.

[1741] 7. Submitting the right response:

[1742] The server sends the answer it deems appropriate to the user.

[1743] Example: "Generated AI response: This is a common request." is sent to the user.

[1744] Examples and prompt formats

[1745] 1. Receive and check user input:

[1746] The user types "This is a general request" into the terminal.

[1747] The terminal sends this input data to the server.

[1748] The server checks the incoming data with the "check_incoming_data" method.

[1749] 2. Answer generation using generative AI models:

[1750] The server sends input data that it deems appropriate to the generative AI model.

[1751] The generative AI model generates an answer such as "Generative AI's answer: This is a typical request," and returns it to the server.

[1752] 3. Check and submit the generated answers:

[1753] The server checks the generated answer using the "check_outgoing_data" method.

[1754] The server returns an answer that it deems appropriate to the terminal.

[1755] Example: "Generated AI response: This is a common request." is displayed to the user.

[1756] Prompt Sentence Examples

[1757] Please output the generative AI's answer: This is a common request.

[1758] In this way, this system effectively reduces the risk of confidential information leaks when using generative AI. By inspecting the text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[1759] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1760] Step 1:

[1761] The user enters text data into the terminal.

[1762] Input: Text data entered by the user, e.g. "This is a general request."

[1763] Specific action: The user types characters on the keyboard and presses the Enter key.

[1764] Output: Input data is displayed on the terminal.

[1765] Step 2:

[1766] The terminal transmits the input text data to the server.

[1767] Input: Text data entered by the user, e.g., "This is a general request."

[1768] Specific operation: When you click the send button on the device, the browser generates an HTTP POST request and sends it to the server.

[1769] Output: Data is sent to the server.

[1770] Step 3:

[1771] The server receives the data from the user and checks the content using the "check_incoming_data" method.

[1772] Input: Text data sent from the terminal, e.g., "This is a general request."

[1773] Specific operation: The server creates an instance of the "SecurityService" class and analyzes the data using the database and filtering rules.

[1774] Data manipulation: Detecting inappropriate content using specific keywords and pattern matching.

[1775] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[1776] Step 4:

[1777] If the server determines that the data is invalid, it sends an error message to the terminal saying "Invalid input."

[1778] Input: Inspection results, e.g., inappropriate data "This is confidential information."

[1779] Specific behavior: The server generates an HTTP response and sends data including an error message.

[1780] Output: An error message is displayed on the terminal.

[1781] Step 5:

[1782] The server sends input data that it deems appropriate to the generative artificial intelligence model.

[1783] Input: Inspection result, e.g. "This is a common request."

[1784] Specific operation: The server generates an appropriate prompt sentence and sends the data to the API of the generative artificial intelligence model.

[1785] Data processing: Convert the data into the format of the prompt statement.

[1786] Output: The prompt sent to the generative AI model.

[1787] Step 6:

[1788] The server receives the answer from the generative artificial intelligence model.

[1789] Input: An answer from a generative AI model, e.g., "Generative AI answer: This is a common request."

[1790] Specific operation: The server receives the HTTP response and parses the response body.

[1791] Output: The received response data.

[1792] Step 7:

[1793] The server checks the generated data using the "check_outgoing_data" method.

[1794] Input: Answer data from a generative AI model, e.g., "Generative AI answer: This is a common request."

[1795] Specific operation: The server examines the data using a filtering algorithm.

[1796] Data processing: Check that the response data does not contain inappropriate content.

[1797] Output: Inspection results, error message if inappropriate, proceed to next step if appropriate.

[1798] Step 8:

[1799] If the server generates an answer that contains inappropriate content, it sends an error message to the terminal saying "Answer is inappropriate."

[1800] Input: Test results, e.g., incorrect response data.

[1801] Specific behavior: The server generates an HTTP response and returns data including an error message.

[1802] Output: An error message is displayed on the terminal.

[1803] Step 9:

[1804] The server sends the answer that it deems appropriate to the terminal.

[1805] Input: Inspection result, e.g. "Generated AI response: This is a common request."

[1806] Specific behavior: The server generates an HTTP response and sends back data containing the appropriate answer.

[1807] Output: The answer of the generated AI is displayed on the terminal.

[1808] (Application example 1)

[1809] 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."

[1810] In systems using generative artificial intelligence models, if the input data or generated answers contain confidential information, it is important to prevent the leakage of that information. However, many current systems lack a means to automatically verify the appropriateness of the input data or generated answers, which poses a risk of confidential information leaking to the outside. The present invention aims to solve this problem and prevent the leakage of confidential information.

[1811] 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.

[1812] In this invention, the server includes means for receiving input data, means for checking the input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, and means for including a mobile application that runs on a smartphone and has a security function for preventing leakage of confidential information. This makes it possible to automatically perform security checks on the input data and the generated answers and prevent leakage of confidential information.

[1813] "Input Data" means text or other information provided to a system by a user.

[1814] "Means for receiving" refers to a device or method for receiving input data provided by a user.

[1815] "Pre-checking measures" refers to functions and methods for automatically inspecting input data and generated responses to check for inappropriate content or confidential information.

[1816] "Means for detecting inappropriate content" means a device or method for identifying inappropriate elements or information in input data or generated answers.

[1817] A "generative artificial intelligence model" refers to an artificial intelligence algorithm or system that generates answers based on input data provided by a user.

[1818] "Means for receiving a generated answer" refers to a device or method for receiving an answer generated from a generative artificial intelligence model.

[1819] "Means for sending answers that are judged to be appropriate to the user" refers to a device or method for returning a checked answer to the user if the checked answer is judged to be appropriate.

[1820] A "smartphone" is a type of mobile phone, a portable information terminal that can connect to the Internet and run various applications.

[1821] "Security features to prevent the leakage of confidential information" refers to features and methods to prevent confidential information from leaking to the outside if it is included in input data or generated answers.

[1822] "Mobile application" refers to a software application that runs on a mobile information device such as a smartphone or tablet.

[1823] The following describes an embodiment of the present invention, specifically showing the main roles of the user, terminal, and server, and the processes performed by each of them.

[1824] System Configuration

[1825] 1. Users

[1826] A user operates an application using a smartphone to input text data.

[1827] Example: User types "Summarize the meeting."

[1828] 2. Terminal

[1829] The terminal is responsible for receiving text data entered by the user and transmitting it to the server.

[1830] Specifically, data is submitted by clicking a button on a web form or application.

[1831] 3. Server

[1832] The server receives the input data and performs security checks by creating an instance of the "SecurityService" class and validating the input data.

[1833] The "check_incoming_data" method is used to check whether inappropriate content is included. If inappropriate content is included, the error message "Input is inappropriate" is returned to the terminal.

[1834] Input data that is deemed appropriate is sent to the generative artificial intelligence model, which then receives the generated answer.

[1835] The generated answer is inspected using the "check_outgoing_data" method. If it contains inappropriate content, the message "Answer is inappropriate" is returned to the terminal.

[1836] If the answer is determined to be appropriate, the answer is returned to the terminal and displayed to the user.

[1837] Data processing and hardware / software used

[1838] Hardware: Smartphone (iOS / Android)

[1839] Software: Python, Flask (web framework), TensorFlow / Transformers (generative AI models), SQL database

[1840] Data calculation: Input data checks and answer checks are performed by the "check_incoming_data()" and "check_outgoing_data()" methods in the "SecurityService" class.

[1841] Specific examples

[1842] When a user types "This is a general request" into a smartphone application, the content is sent to the server. The server checks this input data using the "check_incoming_data" method and sends it to the generative AI model only if it is deemed appropriate.

[1843] The generative AI model generates an answer, "Generative AI's answer: This is a typical request," and returns it to the server. The server then checks the answer again using the "check_outgoing_data" method, and if it is deemed appropriate, returns the answer to the user.

[1844] Example prompt sentence:

[1845] User input: Check if the content of 'Summarize the meeting' is typical, and if it is, run a function that provides appropriate feedback.

[1846] In this way, the present invention reduces the risk of confidential information leaks when using generative AI and provides a high level of security. By inspecting text data provided by users using appropriate means, it is possible to prevent the generation or transmission of inappropriate data.

[1847] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1848] Step 1:

[1849] The user opens the application on their smartphone and enters text. The input data is "Summarize the contents of the meeting." The device receives this user input and sends it to the server.

[1850] Input: User's text data "Summarize the contents of the meeting"

[1851] Output: Text data sent from the terminal to the server

[1852] Step 2:

[1853] The server checks the input data received from the terminal using the "check_incoming_data" method of the "SecurityService" class, which checks whether the input data contains confidential information or inappropriate content.

[1854] Input: Text data received by the server: "Summarize the contents of the meeting"

[1855] Data processing: Checks to see if there is any confidential information

[1856] Output: Check result (good or bad)

[1857] Specific behavior:

[1858] The text data received by the server is passed to the "check_incoming_data" method, which checks whether it contains specific keywords. For example, there is a rule that detects keywords such as "confidential" and "important."

[1859] Step 3:

[1860] If the server checks the results, it sends the text data to a generative AI model, which receives a specific prompt and generates an appropriate answer based on it.

[1861] Input: Text data that passed the check: "Summarize the contents of the meeting"

[1862] Data processing: Passing the data as prompts to the generative AI model

[1863] Output: The answer generated by the generative AI model

[1864] Specific behavior:

[1865] The server passes appropriate input data to the generative AI model as a prompt. For example, it requests the model to generate a summary of the user's input: 'Summarize the contents of the meeting.'

[1866] Step 4:

[1867] The generative AI model generates an answer based on the prompt and returns it to the server, for example, in the form of a "generated summary of the meeting content."

[1868] Input: Prompt sentence for generative AI model

[1869] Data computation: Generative AI models generate answers

[1870] Output: The generated answer

[1871] Specific behavior:

[1872] A generative AI model uses natural language processing based on the prompt to generate an appropriate answer.

[1873] Step 5:

[1874] The server checks the generated answer in advance using the "check_outgoing_data" method of the "SecurityService" class to ensure that the generated answer does not contain inappropriate content.

[1875] Input: Generated Answer

[1876] Data processing: Checking generated answers

[1877] Output: Check result (good or bad)

[1878] Specific behavior:

[1879] The server passes the generated answer to the "check_outgoing_data" method to check whether it contains specific keywords or phrases.

[1880] Step 6:

[1881] If the check result is appropriate, the server returns the generated answer to the terminal and displays it to the user. If the check result is inappropriate, the server returns an error message saying "The answer is inappropriate."

[1882] Input: Answer that passes the check

[1883] Data processing: Formatting answers (if necessary)

[1884] Output: The appropriate answer or error message that is displayed to the user

[1885] Specific behavior:

[1886] The server sends the answer that passes the check to the terminal, and the answer is displayed to the user by the smartphone application. For example, the terminal executes the operation "display the data received from the server."

[1887] 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.

[1888] This invention is a system that uses a generative artificial intelligence model to check the appropriateness of the user's input data and the generated answers, and also takes the user's emotions into consideration, thereby reducing the risk of confidential information leaks and providing more personalized responses. Below, we will explain in detail the main roles of the user, terminal, and server, and the specific processes performed by each.

[1889] overview

[1890] In addition to means for checking input data, means for using a generative artificial intelligence model, and means for checking generated answers, the present invention also includes an emotion engine that recognizes the user's emotions and reflects them in various checks and response generation.

[1891] Basic system configuration

[1892] 1. Users

[1893] The user inputs text data through the terminal.

[1894] Input can also include emotional states.

[1895] Example: A user types, "This is a general request," and their emotion is recognized as "calm."

[1896] 2. Terminal

[1897] It is responsible for sending the text data and emotional information entered by the user to the server.

[1898] Example: The device clicks the send button and requests emotion information along with the input data from the server.

[1899] 3. Server

[1900] The server has means for receiving the text data and the emotion information.

[1901] The received data is inspected using the "check_incoming_data" method, and the emotion engine analyzes the user's emotion information.

[1902] Example: The server further scrutinizes the input data based on a "calm" emotional state.

[1903] 4. Data appropriateness assessment

[1904] If the server determines that the input data is appropriate, it passes the data and emotional information to a generative AI model to generate an answer.

[1905] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1906] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[1907] 5. Use of generative AI models

[1908] The generative AI model generates answers based on the received text data and emotional information.

[1909] The emotion engine generates answers that reflect the user's current emotional state.

[1910] Example: Generated response: "Generated AI response: This is a common request. We're glad you're calm."

[1911] 6. Check the generated answers

[1912] The server checks the generated answer again with the "check_outgoing_data" method.

[1913] The emotion engine also evaluates the appropriateness of the generated answers based on emotional information.

[1914] For example: Make sure your answer is within the character limit and emotionally relevant.

[1915] 7. Sending the Response

[1916] If the answer is judged to be appropriate, it is returned to the device and displayed to the user in an appropriate state that reflects the emotional information.

[1917] If it is determined to be inappropriate, an error is returned and the process is terminated.

[1918] Example: The user sees the message, "Generated AI response: This is a common request. We're glad you're calm."

[1919] Specific examples

[1920] 1. Receiving and checking user input

[1921] The user types "This is a general request" into the device, and the system recognizes the user's emotional state as "calm."

[1922] The device sends this input data and emotional information to the server.

[1923] 2. Answer generation using generative AI models

[1924] The server checks the data and emotion information using the "check_incoming_data" method.

[1925] Data and emotional information deemed appropriate are sent to the generative AI model.

[1926] The generative AI model generates an answer, "Generative AI's answer: This is a common request. We're glad you're calm." and returns it to the server.

[1927] 3. Check and submit the generated answers

[1928] The server checks the response with the "check_outgoing_data" method.

[1929] If it is determined to be an appropriate answer, it returns the answer to the terminal.

[1930] The generated answer is displayed to the user.

[1931] These are the specific steps of the process of the present invention. This system reduces the risk of leaking confidential user information when using generative AI models, and can provide personalized responses that take user emotions into account.

[1932] The processing flow will be explained below.

[1933] Step 1:

[1934] A user inputs text data into a terminal.

[1935] Example: User types, "This is a general request."

[1936] Step 2:

[1937] The terminal collects the user's input data as well as the emotional state.

[1938] Example: The device recognizes the user's emotion as "calm" through facial recognition and voice tone analysis.

[1939] Step 3:

[1940] The device transmits the collected text data and emotion information to a server.

[1941] Example: The device clicks the send button and sends the input data and emotion information as a request to the server.

[1942] Step 4:

[1943] The server receives input data and emotion information from the user.

[1944] Example: The server receives data and emotion information through a REST API.

[1945] Step 5:

[1946] The input data received by the server is checked using the "check_incoming_data" method.

[1947] Example: This method checks for sensitive information or inappropriate keywords.

[1948] Step 6:

[1949] The server analyzes the received emotion information using an emotion engine.

[1950] Example: The server analyzes the emotional information of "calm" and further scrutinizes the input data based on that information.

[1951] Step 7:

[1952] The server determines whether the input data is valid.

[1953] If deemed appropriate, the data and emotional information is passed to a generative AI model.

[1954] If it is determined to be inappropriate, an error message "Input is inappropriate" is returned and the process is terminated.

[1955] For example, "This is a request containing sensitive information." If the data is deemed inappropriate, the server will block it.

[1956] Step 8:

[1957] The server passes the input data and emotional information it deems appropriate to the generative AI model to generate an answer.

[1958] Example: Call the "generate" method of the generative AI model, passing in the user's input data and emotion information.

[1959] Step 9:

[1960] The generative AI model generates answers based on the user's input data and emotional information.

[1961] Example: The generative AI model generates the answer "Generative AI response: This is a common request. We're glad you're calm."

[1962] Step 10:

[1963] The server receives the generated answer.

[1964] Example: Receiving an answer as a response from a generative AI model.

[1965] Step 11:

[1966] The server checks the generated answer again with the "check_outgoing_data" method.

[1967] For example, this method checks whether the answer is within a predefined character limit and whether it is sentiment-based.

[1968] Step 12:

[1969] The server determines whether the generated answer is appropriate.

[1970] If it is determined to be appropriate, the answer is sent to the terminal.

[1971] If it is determined to be inappropriate, an error message "The answer is inappropriate" is returned and the process is terminated.

[1972] Example: If the answer exceeds 500 characters, the server will block it as an inappropriate answer.

[1973] Step 13:

[1974] The terminal receives the appropriate response from the server and displays it to the user.

[1975] Example: The user will see the answer, "Generator's response: This is a common request. We're glad you're calm."

[1976] These are the specific steps of the process of the present invention. This system reduces the risk of confidential information leakage when using generative AI models and can provide personalized responses that take user emotions into account.

[1977] Example 2

[1978] 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."

[1979] Conventional systems using generative artificial intelligence models have had difficulty accurately evaluating the appropriateness of user input data and generated answers. Furthermore, because they do not take the user's emotions into account, they are unable to provide personalized responses, resulting in reduced user satisfaction. Furthermore, there is a risk of data leakage due to inappropriate data content or emotional state. The present invention aims to solve these problems, improve the appropriateness of user input data and generated answers, and provide personalized responses that take the user's emotions into account.

[1980] 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.

[1981] In this invention, the server includes means for receiving text data and the user's emotional information from the user, means for analyzing the received text data and emotional information and detecting inappropriate content or an inappropriate emotional state, means for transmitting the text data and emotional information determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for analyzing the generated answer and the emotional information contained in the answer and detecting inappropriate content or an inappropriate emotional state, and means for transmitting the answer determined to be appropriate to the user. This makes it possible to improve the appropriateness of the user's input data and the generated answer and to provide a personalized response that takes the user's emotions into consideration.

[1982] A "user" is an entity that utilizes the system to provide input data and receive generated responses.

[1983] "Text data" refers to character string information that a user inputs through a terminal.

[1984] "Emotional information" refers to data that analyzes a user's emotional state and is information that the system uses to personalize the user's input data and responses.

[1985] "Means for receiving" refers to the modules and functions that allow a server or terminal to receive data from a user.

[1986] "Means for analyzing" refers to software or algorithms that examine the received data and check for appropriateness of content and emotional state.

[1987] "Inappropriate content" refers to data that the system does not tolerate, such as leaks of personal information, offensive language, or inappropriate language.

[1988] "Inappropriate emotional state" refers to a negative or unstable emotional state of the user detected by the system.

[1989] A "generative artificial intelligence model" refers to a machine learning model or algorithm that automatically generates appropriate responses based on received text data.

[1990] "Generated response" refers to a reply created by a generative artificial intelligence model.

[1991] "Transmission means" refers to the network infrastructure and communications protocols used by the server to transmit the generated response back to the user.

[1992] "Personalized responses" refer to replies that are individually tailored to take into account the user's emotional information and text data.

[1993] "System" refers to the overall technical environment and processing flow for receiving input data from a user, analyzing it, and generating and returning an appropriate response.

[1994] This system generates personalized responses based on user input data and associated emotional information, reducing the risk of confidential information leaks. The specific roles and operations of the user, device, and server are detailed below.

[1995] user

[1996] The user uses the device to input text data. At this time, the device is set up so that the user's emotional state is also input. For example, if the user inputs "This is a general request," the user's emotional state at that time is recognized as "calm."

[1997] Terminal

[1998] The terminal is responsible for transmitting the text data and emotion information input by the user to the server. Specifically, the terminal sends the input data and emotion information to the server through the network module, and then receives a response from the server and displays it to the user.

[1999] Example: The device clicks the send button and sends the input data "This is a general request" and emotional information "Calm" to the server.

[2000] server

[2001] The server provides the following specific functions:

[2002] 1. Data Reception

[2003] The server receives the text data and emotion information sent from the terminal. This reception is performed using a "receiving means."

[2004] 2. Data analysis and testing

[2005] The received data is analyzed using the "check_incoming_data" method to evaluate the appropriateness of the text data and the appropriateness of the emotional information. The server performs this analysis using an NLP (Natural Language Processing) module that includes an emotional engine.

[2006] Example: The server receives the text "This is a general request" and analyzes the emotional information "I feel calm."

[2007] 3. Data appropriateness assessment

[2008] The text data and emotion information that are deemed appropriate are sent to a generative AI model. If the content is inappropriate, the process ends and an error message is returned.

[2009] Example: The input "This is a common request" is deemed acceptable and sent to the generative AI model.

[2010] 4. Response Generation

[2011] The generative AI model generates answers based on the received text data and emotional information, with the emotional engine creating personalized responses that reflect the user's current emotional state.

[2012] Example: A generative AI model generates the response, "Generative AI response: This is a common request. We're glad you're calm."

[2013] 5. Inspecting and Sending the Response

[2014] The generated answer is checked again by the "check_outgoing_data" method, and if it is deemed appropriate it is sent to the user. If it is an inappropriate answer, the process is terminated and an error message is returned.

[2015] Example: The server checks whether the generated answer is within the character limit and appropriate for the emotional state, and then sends the appropriate answer to the device.

[2016] The above is a specific example of the processing performed by the system of the present invention. This system ensures the appropriateness of the user's input data and the generated answers, and makes it possible to provide personalized responses that reflect the user's emotions.

[2017] The flow of the identification process in the second embodiment will be described with reference to FIG.

[2018] Program processing flow

[2019] Step 1:

[2020] The user enters text data into the input field of the terminal and clicks the "Send" button. At this time, the emotion recognition software analyzes the user's emotional information and provides it to the terminal along with the input. For example, the input data is "This is a general request," and the emotional information is "calm."

[2021] input:

[2022] Text data entered by the user

[2023] Emotional information analyzed by emotion recognition software

[2024] output:

[2025] Text data and emotional information received by the device

[2026] Specific behavior:

[2027] User enters text and clicks submit button

[2028] Emotion recognition software analyzes user emotions

[2029] Step 2:

[2030] The terminal sends the text data and emotion information entered by the user to the server, and this data is transferred to the server using the network module.

[2031] input:

[2032] Text data and emotional information entered into the device

[2033] output:

[2034] Text data and emotional information sent to the server

[2035] Specific behavior:

[2036] The device's network module sends text data and emotional information to the server.

[2037] Step 3:

[2038] The server checks the received text data and emotion information using the "check_incoming_data" method. The server uses the NLP module and emotion engine to analyze the appropriateness of this data.

[2039] input:

[2040] Text data and emotion information received by the server

[2041] output:

[2042] Appropriateness evaluation results for text data and emotional information

[2043] Specific behavior:

[2044] The server analyzes the text using the NLP module

[2045] Emotion engine analyzes emotional information

[2046] Step 4:

[2047] The server determines the appropriateness of the data based on the analysis results. If the data is deemed appropriate, it sends it to the generative AI model. If the data is inappropriate, it generates an error message and returns it to the user.

[2048] input:

[2049] Text data and emotional information analyzed by the server

[2050] output:

[2051] When deemed appropriate: Data sent to generative AI models

[2052] If deemed inappropriate: Error message

[2053] Specific behavior:

[2054] Sending the right data to generative AI models

[2055] Invalid data generates an error message

[2056] Step 5:

[2057] The generative AI model generates answers based on the received text data and emotional information, and the emotion engine creates personalized responses that reflect the user's emotional state.

[2058] input:

[2059] Appropriate text data and emotional information sent from the server

[2060] output:

[2061] Generated Answer

[2062] Specific behavior:

[2063] Generative AI model generates responses based on text data and emotional information

[2064] Emotion engine creates responses that reflect the user's emotions

[2065] Step 6:

[2066] The server checks the generated answer using the "check_outgoing_data" method. If the check passes, the answer is sent to the user. If it is invalid, an error message is generated.

[2067] input:

[2068] Answers generated by generative AI models

[2069] output:

[2070] If the check passes: The answer sent to the user

[2071] If not passed: Error message

[2072] Specific behavior:

[2073] The server re-examines the generated answer

[2074] Sends appropriate response to user, may generate error message

[2075] Step 7:

[2076] The server sends an appropriate response to the user, and the device receives this response and displays it to the user. If the response is inappropriate, an error message is displayed.

[2077] input:

[2078] The response or error message sent by the server

[2079] output:

[2080] The answer or error message that the user sees

[2081] Specific behavior:

[2082] The device receives the response from the server and displays it to the user.

[2083] Display error messages if necessary

[2084] (Application example 2)

[2085] 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."

[2086] Conventional systems utilizing generative artificial intelligence models have the ability to determine the appropriateness of user input data and generated responses, but because they do not take user emotions into account, responses are not personalized and the user experience is insufficient. Furthermore, if emotional information is not taken into account, the appropriateness of the data cannot be adequately determined, which may increase the risk of confidential information being leaked. Therefore, an objective of the present invention is to provide a system that utilizes a user's emotional information to provide personalized responses based on emotions and reduce the risk of confidential information being leaked.

[2087] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving input data, means for checking the received input data in advance and detecting inappropriate content, means for sending input data determined to be appropriate to the generative artificial intelligence model and receiving a generated answer, means for checking the generated answer in advance and detecting inappropriate content, means for sending the answer determined to be appropriate to the user, emotion recognition means for receiving and analyzing the user's emotion information, and means for generating a personalized answer based on the emotion information. This enables a personalized response that is in line with the user's emotions, making it possible to reduce the risk of confidential information being leaked.

[2088] The "means for receiving input data" refers to a function or device for receiving text data or emotion information sent by a user.

[2089] "Means for checking received input data in advance and detecting inappropriate content" refers to functions or devices for analyzing input data and identifying content that does not conform to specified standards.

[2090] "Means for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer" refers to a function or device for sending input data judged appropriate to a generative artificial intelligence model and receiving the generated answer.

[2091] "Means for checking generated answers in advance and detecting inappropriate content" refers to functions or devices for re-examining answers obtained from generative AI models to check whether they contain inappropriate content.

[2092] The "means for sending a response that is determined to be appropriate to the user" refers to a function or device for sending a response that is determined to not contain inappropriate content to the user.

[2093] "Emotion recognition means for receiving and analyzing user's emotional information" refers to a function or device for receiving information about the user's emotional state and analyzing that state.

[2094] The "means for generating a personalized answer based on emotional information" refers to a function or device for generating an answer optimized for a user, taking into account the emotional information of the user.

[2095] In order to put the present invention into practice, it is necessary to build a system that specifically separates the roles of users, terminals, and servers.

[2096] User Roles

[2097] The user's role is to access the online shopping site and input product inquiries and support requests in text format. As the user inputs information, the system automatically detects the user's emotional state. At this time, the user's emotional information is also transmitted along with the input data.

[2098] As a specific example, a user might enter, "I have a question about this product. I'm worried."

[2099] Device Role

[2100] The device receives input data and emotion information from the user and transmits it to the server. The device in this case can be a smartphone or desktop PC.

[2101] The request sent by the device will have the following format:

[2102] "I have a question about this product. I'm worried."

[2103] Server Roles

[2104] The server has multiple means for analyzing the received input data and emotional information and generating appropriate responses.

[2105] 1. Receipt and inspection of data

[2106] The server receives the input data sent by the user using the "means for receiving input data" and analyzes it using the "means for pre-checking the received input data and detecting inappropriate content." Specifically, it analyzes the text data using a natural language processing library and analyzes emotional information using an emotion recognition engine (e.g., VADER Sentiment Analysis).

[2107] 2. Submitting the right data and generating answers

[2108] The server uses the "means for sending input data judged appropriate to a generative AI model and receiving the generated answer" to send the input data judged appropriate to a generative AI model (e.g., GPT-3) and generate an answer. At this time, emotional information is taken into consideration to generate a personalized response.

[2109] 3. Check and submit the generated answers

[2110] The server re-examines the generated answer using "means for checking the generated answer in advance and detecting inappropriate content," evaluates it using "means for receiving and analyzing the user's emotional information," and then sends the answer that is judged to be appropriate to the user using "means for sending the answer to the user."

[2111] Hardware and software used

[2112] 1. Hardware

[2113] Server: A cloud server, such as an Amazon EC2 instance

[2114] Devices: Smartphones, desktop PCs

[2115] 2. Software

[2116] Natural language processing models: Generative artificial intelligence models such as GPT-3

[2117] Sentiment analysis library: VADER Sentiment Analysis

[2118] Specific examples

[2119] As a concrete example, let's say a user types, "I have a question about this product. I'm worried." In that case, the following process will be executed:

[2120] 1. The user types "I have a question about this product. I'm worried." into the device, and the emotion is recognized as "anxiety."

[2121] 2. The device sends the input data and emotion information to the server.

[2122] 3. The server receives and analyzes the text data and emotion information.

[2123] 4. If appropriate, the generative AI model generates an answer.

[2124] 5. The server re-examines the generated answers and rates them based on sentiment information.

[2125] 6. If the rating is deemed appropriate, the answer is sent to the user.

[2126] The above is a specific embodiment for carrying out the present invention. This system can provide personalized responses that take into account the user's emotional information, reducing the risk of confidential information being leaked.

[2127] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[2128] Step 1:

[2129] The user uses a terminal to create input data and send it along with emotional information. At this time, the user enters the inquiry in text format, and the terminal automatically performs emotional analysis. For example, if the user enters "I have a question about this product. I'm worried," the terminal obtains this text data and the emotional information of "anxiety." The inputs are text data and emotional information. The output is sent to the server.

[2130] Step 2:

[2131] The device receives input data and sends emotional information to the server. The device's role is to relay data, and the input data and emotional information are sent to the server as is without conversion. The input includes the user's text data and emotional information. The output includes the data sent to the server.

[2132] Step 3:

[2133] The server checks the input data and emotional information it receives in advance to detect inappropriate content. Specifically, the server uses a natural language processing library to analyze the appropriateness of the text and an emotion recognition engine to analyze the emotional information. The input includes text data and emotional information sent from the device. Data processing involves natural language processing and emotional analysis. The output is either data deemed appropriate or an error message.

[2134] Step 4:

[2135] Input data deemed appropriate is sent to a generative AI model and the generated answer is received. To do this, the server sends text data and emotional information to the generative AI model and receives the generated answer. The inputs include text data and emotional information. Data calculation involves text generation by the generative AI model. The output is the generated answer.

[2136] Step 5:

[2137] The generated answer is checked in advance to detect inappropriate content. The server again analyzes the appropriateness of the answer using a natural language processing library and evaluates the emotional information using an emotion recognition engine. The input is the generated answer. Data processing involves natural language processing and emotional evaluation. The output is either an answer deemed appropriate or an error message.

[2138] Step 6:

[2139] The answer that is deemed appropriate is sent to the user. The server returns this answer to the terminal, which displays it to the user. The input is the answer that is deemed appropriate. The output is the answer that was sent to the user terminal.

[2140] Step 7:

[2141] The user checks the answer on the device and takes further action. The user checks the answer to the question through the device and asks again if necessary. The input is the answer from the server. The output is the user's satisfaction level and the next action.

[2142] The above is the flow of processing for the system program that realizes the application example.

[2143] 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.

[2144] 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.

[2145] 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.

[2146] 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.

[2147] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[2148] 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.

[2149] 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).

[2150] 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.

[2151] 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."

[2152] 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.

[2153] 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).

[2154] 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.

[2155] 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.

[2156] 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.

[2157] 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.

[2158] 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.

[2159] 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.

[2160] 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.

[2161] 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.

[2162] 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.

[2163] 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.

[2164] The following is further disclosed regarding the above embodiment.

[2165] (Claim 1)

[2166] means for receiving input data;

[2167] a means for proactively checking received input data to detect inappropriate content;

[2168] means for transmitting input data determined to be appropriate to a generative artificial intelligence model and receiving a generated answer;

[2169] A means of pre-checking generated answers to detect inappropriate content;

[2170] a means for transmitting the response determined to be appropriate to the user;

[2171] A system including:

[2172] (Claim 2)

[2173] 2. The system according to claim 1, wherein the checking means checks whether the input data contains a specific keyword.

[2174] (Claim 3)

[2175] 2. The system according to claim 1, wherein the checking means checks whether the generated answer does not exceed a predetermined number of characters.

[2176] "Example 1"

[2177] (Claim 1)

[2178] means for receiving input data from a user;

[2179] a means for proactively checking received input data to detect inappropriate content;

[2180] means for transmitting input data determined to be appropriate to a generative artificial intelligence model and receiving a generated answer;

[2181] A means of pre-checking generated answers to detect inappropriate content;

[2182] a means for transmitting the response determined to be appropriate to the user;

[2183] A system including:

[2184] (Claim 2)

[2185] 2. The system according to claim 1, wherein the checking means checks whether the input data contains a specific keyword.

[2186] (Claim 3)

[2187] 2. The system according to claim 1, wherein the checking means checks whether the generated answer does not exceed a predetermined number of characters.

[2188] (Claim 4)

[2189] A means for converting input data to be sent to the generative artificial intelligence model into a prompt sentence format in advance;

[2190] means for converting the answers received from the generative artificial intelligence model into a format that is displayable to a user;

[2191] 10. The system of claim 1, comprising:

[2192] (Claim 5)

[2193] 2. The system according to claim 1, wherein the checking means checks the input data and the generated answer by applying specific security rules to prevent leakage of confidential information.

[2194] (Claim 6)

[2195] A means of storing all data, including input data from users and answers from generative AI models, as logs; and

[2196] a means of periodically auditing the stored log data;

[2197] 10. The system of claim 1, comprising:

[2198] "Application Example 1"

[2199] (Claim 1)

[2200] means for receiving input data;

[2201] a means for proactively checking received input data to detect inappropriate content;

[2202] means for transmitting input data determined to be appropriate to a generative artificial intelligence model and receiving a generated answer;

[2203] A means of pre-checking generated answers to detect inappropriate content;

[2204] a means for transmitting the response determined to be appropriate to the user;

[2205] a means including a mobile application that operates on a smartphone and has a security function for preventing leakage of confidential information;

[2206] A system including:

[2207] (Claim 2)

[2208] 2. The system according to claim 1, wherein the checking means checks whether the input data contains a specific keyword.

[2209] (Claim 3)

[2210] 2. The system according to claim 1, wherein the checking means checks whether the generated answer does not exceed a predetermined number of characters.

[2211] "Example 2: Combining Emotion Engines"

[2212] (Claim 1)

[2213] means for receiving text data and user emotion information from a user;

[2214] means for analyzing the received text data and emotional information to detect inappropriate content or an inappropriate emotional state;

[2215] means for transmitting the text data and emotion information determined to be appropriate to a generative artificial intelligence model and receiving a generated answer;

[2216] means for analyzing the generated answers and the emotional information contained in the answers to detect inappropriate content or inappropriate emotional states;

[2217] a means for transmitting the response determined to be appropriate to the user;

[2218] A system including:

[2219] (Claim 2)

[2220] 2. The system according to claim 1, wherein the analyzing means checks whether the input text data contains a specific keyword and whether the emotional information matches a predetermined state.

[2221] (Claim 3)

[2222] 2. The system of claim 1, wherein the analysis means checks whether the generated answer does not exceed a predetermined number of characters and whether the emotional information matches a predetermined state.

[2223] "Application example 2 when combining emotion engines"

[2224] (Claim 1)

[2225] means for receiving input data;

[2226] a means for proactively checking received input data to detect inappropriate content;

[2227] means for transmitting input data determined to be appropriate to a generative artificial intelligence model and receiving a generated answer;

[2228] A means of pre-checking generated answers to detect inappropriate content;

[2229] a means for transmitting the response determined to be appropriate to the user;

[2230] emotion recognition means for receiving and analyzing emotion information of a user;

[2231] a means for generating personalized responses based on emotional information;

[2232] A system including:

[2233] (Claim 2)

[2234] 2. The system according to claim 1, wherein the checking means checks whether the input data contains specific keywords and adjusts the response based on the user's emotional information.

[2235] (Claim 3)

[2236] 2. The system according to claim 1, wherein the checking means checks whether the generated answer does not exceed a predetermined number of characters and evaluates the appropriateness of the answer based on emotional information. [Explanation of symbols]

[2237] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. means for receiving input data; a means for proactively checking received input data to detect inappropriate content; means for transmitting input data determined to be appropriate to a generative artificial intelligence model and receiving a generated answer; A means of pre-checking generated answers to detect inappropriate content; a means for transmitting the response determined to be appropriate to the user; A system including:

2. 2. The system according to claim 1, wherein the checking means checks whether the input data contains a specific keyword.

3. 2. The system according to claim 1, wherein the checking means checks whether the generated answer does not exceed a predetermined number of characters.

Images