System

Abstract

A system is provided.SOLUTION: A system comprising: means for receiving request data from a terminal by a receiving means; means for detecting and correcting or replacing inappropriate words or phrases in the request data by a filtering means; means for generating text to be generated based on the filtered request by a generating means; means for re-detecting and correcting or replacing inappropriate words or phrases in the generated text by a re-filtering means; and means for transmitting the filtered generated text to the terminal by a transmitting means.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, natural language text generation technology has advanced and is now being used in a variety of situations. However, the generated text may not be appropriate for all users, and may contain inappropriate content, especially when used by children. This has led to a growing need for generative AI systems that can be safely used by children. Furthermore, existing filtering systems often have difficulty completely removing inappropriate content in a single check, requiring repeated confirmation. An objective of the present invention is to provide a system that efficiently and reliably removes inappropriate content from generative AI systems, ensuring safe use by children. [Means for solving the problem]

[0005] The present invention solves the above problems by the following means. First, request data is received from a terminal by a receiving means, and inappropriate words and phrases in the request data are detected by a filtering means, and corrected or replaced as necessary. Next, text is generated based on the filtered request data by a generating means, and inappropriate words and phrases are again detected in the generated text by a re-filtering means, and corrected or replaced. Finally, the filtered generated text is sent to the terminal by a transmitting means, allowing the user to use the generative AI system with peace of mind. This realizes a generative AI system that provides only content appropriate for children and can be used with peace of mind.

[0006] The "receiving means" is a function or device for receiving request data sent from a terminal.

[0007] "Filtering means" means a function or device for detecting inappropriate words or phrases in request data or generated text and correcting or replacing them as necessary.

[0008] "Generation means" refers to a function or device for generating new text based on filtered request data.

[0009] The "re-filtering means" is a function or device for detecting inappropriate words or phrases again in the content of the generated text and correcting or replacing them.

[0010] The "transmission means" is a function or device for transmitting the filtered generated text to the terminal.

[0011] A "terminal" is an electronic device that allows a user to input requests and view generated text.

[0012] "User" means an individual who utilizes the system to enter requests and receive and use the generated text.

[0013] "Request data" refers to the content and information of requests made to a generative AI system by a user that they input into their device.

[0014] "Inappropriate words and phrases" are those that are objectionable or may cause offense or danger to children. [Brief explanation of the drawings]

[0015] [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

[0016] 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.

[0017] First, the terms used in the following description will be explained.

[0018] 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).

[0019] 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.

[0020] 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.

[0021] 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.

[0022] 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."

[0023] [First embodiment]

[0024] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0025] 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.

[0026] 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).

[0027] 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.

[0028] 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.

[0029] 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.

[0030] 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.

[0031] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0032] 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.

[0033] 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.

[0034] 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.

[0035] 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."

[0036] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[0037] A natural language explanation of the program's processing

[0038] Server-side processing

[0039] The server first receives request data sent from the device. This request data includes the text entered by the user and the content they want to generate. For example, if the user enters "Tell me an interesting story," the device sends the request to the server.

[0040] Next, the server uses a filtering mechanism to detect inappropriate words or phrases in the request data. The filtering mechanism scans the request data using regular expressions or a list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate, and the request is modified to "Tell me a funny story."

[0041] Based on the filtered request data, the server generates text using a generation means. For example, a generative AI model (such as GPT-3 (registered trademark)) is used as this generation means. The generation means takes the filtered request data as input and generates appropriate text.

[0042] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0043] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0044] Processing on the device (kids' phone)

[0045] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[0046] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0047] User (child) operations

[0048] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[0049] Specific examples

[0050] 1. First request:

[0051] The user (child) types "tell me a scary story" into the terminal.

[0052] The terminal sends the request to the server.

[0053] The server receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[0054] The server generates interesting stories using generative AI.

[0055] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0056] The server sends the text of an interesting story to the device.

[0057] The terminal displays the received text on its screen so that the user can read it.

[0058] 2. Harmless Requests:

[0059] The user (child) types "Tell me a story about a cute animal" into the terminal.

[0060] The terminal sends the request to the server.

[0061] The server receives the request, filters it, and because it contains no profanity, the request is kept as is.

[0062] The server generates cute animal stories using generative AI.

[0063] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0064] The server sends text of cute animal stories to the device.

[0065] The terminal displays the received text on its screen so that the user can read it.

[0066] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[0067] The processing flow will be explained below.

[0068] Server-side processing

[0069] Step 1:

[0070] The server receives the request data sent from the terminal.

[0071] Specific behavior: Listens for and receives HTTP requests, extracting user-entered text and generated request content from the request body.

[0072] Step 2:

[0073] The server uses filtering means to detect inappropriate words or phrases in the received request data.

[0074] What it does: Scans request data using regular expressions and profanity lists, and fixes or replaces any profanity it finds.

[0075] Step 3:

[0076] The server generates text using the filtered request data with the generation means.

[0077] What it does: Passes the filtered request to a generative AI model (e.g., GPT-3) and makes an API request to generate text that meets the request.

[0078] Step 4:

[0079] The server again applies filtering measures to the generated text to detect inappropriate words and phrases.

[0080] What it does: Scan the generated text again against the regular expressions and profanity list, making corrections or substitutions if necessary.

[0081] Step 5:

[0082] The server sends the filtered generated text to the terminal as an HTTP response.

[0083] Specific behavior: Constructs an HTTP response and sends the generated text to the terminal in the response body.

[0084] Processing on the device (kids' phone)

[0085] Step 1:

[0086] The terminal receives input from the user and sends the content to the server as a request.

[0087] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[0088] Step 2:

[0089] The terminal waits for and receives a response from the server.

[0090] Specific operation: Asynchronously waits for an HTTP response from the server, and retrieves the content of the response when it arrives.

[0091] Step 3:

[0092] The terminal displays the received generated text on the screen.

[0093] Specific operation: The generated text received is reflected in the text display area, making it viewable by the user.

[0094] User (child) operations

[0095] Step 1:

[0096] The user inputs the text content that he or she wants to generate into the terminal.

[0097] Example: Using a text field or voice input, type "Draw me a funny picture."

[0098] Step 2:

[0099] The user checks the generated text and takes the following action:

[0100] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[0101] Example 1

[0102] 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."

[0103] With conventional generative AI systems, the generated text may contain inappropriate language, making it difficult to provide an environment that children can use with confidence. Specifically, if inappropriate words or phrases are included in the request data or generated text, they are output as is, posing a risk that children may come into contact with harmful information. There is a need to solve these issues and provide a generative AI system that children can use with confidence.

[0104] 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.

[0105] In this invention, the server includes: [means for receiving request data from the information processing device by a receiving means; [means for detecting inappropriate words and standard phrases in the request data by a filtering means and correcting or replacing them;] [means for creating text to be generated based on the filtered request data by a generating means; [means for re-detecting inappropriate words and standard phrases in the generated text by a re-filtering means and correcting or replacing them; and [means for transmitting the filtered generated text to the information processing device by a transmitting means.] This enables a generative AI system that can be used safely by children by doubly filtering inappropriate expressions in both the request data and the generated text.

[0106] The "receiving means" is a means for receiving request data from an information processing device.

[0107] "Filtering means" refers to means for detecting inappropriate words or boilerplate phrases in request data or generated text and correcting or replacing them.

[0108] The "generator" is a means for generating text based on the filtered request data.

[0109] The "re-filtering means" is a means for detecting inappropriate words or boilerplate phrases again in the generated text and correcting or replacing them.

[0110] The "transmitting means" is a means for transmitting the filtered generated text to the information processing device.

[0111] An "information processing device" is a device for inputting request data from a user and transmitting it to a server.

[0112] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[0113] Server-side processing

[0114] The server receives request data sent from the device. This request data includes the text entered by the user and the content to be generated. For example, if the user enters "tell me an interesting story," the device sends the request to the server.

[0115] Next, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. The filtering mechanism scans the request data using regular expressions or a predefined list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," if "scary" is determined to be inappropriate, the request is modified to "Tell me a funny story."

[0116] Based on the filtered request data, the server generates text using a generation means. For example, a generative AI model (such as GPT-3) is used as the generation means. The generation means takes the filtered request data as input and generates appropriate text.

[0117] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0118] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0119] Processing on the terminal (information processing device) side

[0120] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[0121] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0122] User (child) operations

[0123] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[0124] Specific examples

[0125] 1. First time request

[0126] User: Type "tell me a scary story" into the terminal.

[0127] Terminal: Sends an HTTP request to the server.

[0128] Server: Receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[0129] Server: Generate text using a generative AI model using the modified request data.

[0130] Server: Re-filter the generated text to ensure it is free of profanity.

[0131] Server: Sends the filtered generated text to the terminal as an HTTP response.

[0132] Terminal: Displays the received text on the screen, displaying the story "Once upon a time..." for the user to read.

[0133] 2. Harmless Requests

[0134] User: Type "Tell me a cute animal story" into the terminal.

[0135] Terminal: Sends an HTTP request to the server.

[0136] Server: Receives the request and filters it. Since it contains no profanity, the request is kept as is.

[0137] Server: Generate text using a generative AI model using the filtered request data.

[0138] Server: Re-filter the generated text to ensure it is free of profanity.

[0139] Server: Sends the filtered generated text to the terminal as an HTTP response.

[0140] Terminal: Displays the received text on the screen. It displays "cute animal stories" for the user to read.

[0141] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[0142] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0143] Step 1: Receiving request data

[0144] The server receives request data sent from the device. Input is text entered by the user into the device or content they want to generate. For example, the user might enter "Tell me an interesting story." The device sends this input content to the server as an HTTP request. The server receives the HTTP request and obtains the request data.

[0145] Step 2: Filtering the request data

[0146] The server scans the received request data using a filtering method. The input is the received request data. The filtering method uses regular expressions and a predefined list of inappropriate words. Specifically, the server detects inappropriate words and fixed phrases in the request data. For example, in the case of a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate and the request is corrected to "Tell me a funny story." The output is the filtered request data.

[0147] Step 3: Generate text

[0148] The server generates text using a generative method based on the filtered request data. The input is the filtered request data ("Tell me a funny story"). The server passes this input to a generative AI model (e.g., GPT-3) as a prompt. The generative AI model generates appropriate text corresponding to the input prompt, e.g., "Once upon a time...". The output is the generated text.

[0149] Step 4: Re-filtering the generated text

[0150] The server scans the generated text again using the filtering method. The input is the generated text. Specifically, the server detects inappropriate words or boilerplate phrases in the generated text and corrects or replaces them. For example, if the generated text contains the word "scary," it replaces it with "funny." The output is the re-filtered generated text.

[0151] Step 5: Sending generated text

[0152] The server sends the re-filtered generated text to the terminal using a transmission means. The input is the re-filtered generated text. In specific operations, the server constructs an HTTP response and includes the generated text in the response body. Then, the server sends the HTTP response to the terminal. The terminal displays the received generated text on the screen so that the user can read it. The output is the generated text displayed on the terminal.

[0153] (Application example 1)

[0154] 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."

[0155] Current generative AI systems can sometimes be difficult to guarantee safety so that children can use them with confidence. Measures are needed to properly filter requests and generated text that contain inappropriate words or phrases, but specific methods for achieving this have not yet been established. Furthermore, there is a lack of interfaces that allow children to enjoy the generated text with peace of mind. A solution to these issues is needed.

[0156] 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.

[0157] In this invention, the server includes means for receiving request data from the terminal using a receiving means, means for detecting and correcting or replacing inappropriate words and phrases in the request data using a filtering means, means for creating text to be generated based on the request filtered by the generative AI model, means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text using a re-filtering means, means for transmitting the filtered generated text to the terminal using a transmitting means, and means for providing an interface that allows a user to check and enjoy the re-filtered generated text. This realizes a generative AI system that children can use safely, allowing children to safely enjoy generated text that does not contain inappropriate words or phrases.

[0158] The "receiving means" is a means for receiving request data from a terminal.

[0159] "Filtering means" means means for detecting and modifying or replacing inappropriate words or phrases in request data.

[0160] A "generative AI model" is an artificial intelligence model for generating appropriate text based on filtered request data.

[0161] A "re-filtering means" is a means for re-detecting and correcting or replacing inappropriate words or phrases in the generated text.

[0162] The "transmitting means" is a means for transmitting the filtered generated text to the terminal.

[0163] An "interface" is a display means and operating means that allows a user to view and enjoy the re-filtered generated text.

[0164] The embodiment of this invention provides a generative AI system that children can use safely. The system mainly consists of a server, a terminal, and a user.

[0165] System configuration

[0166] Server-side processing

[0167] The server comprises a "receiving means," a "filtering means," a "generative AI model," a "re-filtering means," and a "transmitting means."

[0168] 1. Receiving means: Receives request data sent from the terminal. The request data includes text entered by the user and the content to be generated.

[0169] 2. Filtering methods: Detect inappropriate words or phrases in the request data and modify or replace them. Use regular expressions or profanity lists.

[0170] 3. Generative AI model: Generate appropriate text based on the filtered request data. Here, we use a generative AI model such as OpenAI's GPT-3.

[0171] 4. Re-filtering: The generated text is checked again for inappropriate language and corrected or replaced as necessary.

[0172] 5. Transmission method: Send the filtered generated text to the terminal. Construct an HTTP response and send the generated text in the response body.

[0173] Terminal side processing

[0174] The terminal receives input from the user, sends the content to the server as a request, and displays the generated text on the screen.

[0175] Example

[0176] User (child) operations

[0177] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "tell me a funny story," the request is sent to the server, and the generated funny story is displayed on the terminal.

[0178] Specific examples

[0179] First request:

[0180] The user types "tell me a scary story."

[0181] The terminal sends the request to the server.

[0182] The server uses filtering means to modify "scary" to "interesting."

[0183] A generative AI model generates "interesting stories" and a re-filtering method checks their appropriateness.

[0184] An interesting story text will be sent to your device and displayed.

[0185] Harmless requests:

[0186] A user types, "Tell me about cute animals."

[0187] The terminal sends the request to the server.

[0188] The server filters the request and generates a story using a generative AI model.

[0189] The generated stories are filtered and displayed on the device.

[0190] Prompt Sentence Examples

[0191] Prompt: Tell me a funny story

[0192] This system provides an environment where children can use generative AI safely, and ensures that the generated text is appropriate for the user, allowing them to enjoy it with peace of mind.

[0193] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0194] Step 1:

[0195] The user enters text into the terminal.

[0196] Input: Text entered by the user using a keyboard or touchscreen

[0197] What it does: A user types the text "tell me a funny story" into their device.

[0198] Output: The entered text is saved on the terminal as request data.

[0199] Step 2:

[0200] The terminal transmits the request data to the server.

[0201] Input: Request data stored on the device

[0202] How it works: The device constructs an HTTP request and sends the request data to the server.

[0203] Output: The server receives the request data from the HTTP request.

[0204] Step 3:

[0205] Filters the request data received by the server.

[0206] Input: Received request data (e.g., "Tell me a scary story")

[0207] Data Modification: Filtering methods are used to detect and modify or replace inappropriate words or phrases in request data, such as using regular expressions or profanity lists.

[0208] Action: The request data "Tell me a scary story" is modified to "Tell me a funny story."

[0209] Output: Filtered request data (e.g., "Tell me a funny story").

[0210] Step 4:

[0211] The server generates text based on the filtered request data.

[0212] Input: Filtered request data (e.g., "Tell me a funny story")

[0213] Data calculation: Text generation using a generative AI model. The prompt sentence "Tell me an interesting story" is input into a generative AI (e.g., GPT-3) to generate appropriate text.

[0214] How it works: A generative AI model generates text that begins with "Once upon a time..."

[0215] Output: The generated text (e.g., "Once upon a time...").

[0216] Step 5:

[0217] The server re-filters the generated text.

[0218] Input: Generated text (e.g. "Once upon a time...")

[0219] Data processing: Re-filtering measures are used to detect and correct or replace inappropriate words or phrases in the generated text.

[0220] Action: Checks the generated text for profanity and corrects it if necessary.

[0221] Output: The refiltered generated text (e.g., "Once upon a time...").

[0222] Step 6:

[0223] The server sends the re-filtered generated text to the terminal.

[0224] Input: Re-filtered generated text (e.g., "Once upon a time...")

[0225] Behavior: The server sends the re-filtered generated text to the terminal as an HTTP response and includes it in the response body.

[0226] Output: The device receives the re-filtered generated text from the server.

[0227] Step 7:

[0228] The terminal displays the refiltered generated text to the user.

[0229] Input: Re-filtered generated text received from the server (e.g., "Once upon a time...")

[0230] Action: The terminal displays the resulting refiltered text on the screen, presenting the text in a user-friendly format.

[0231] Output: The generated text displayed on the screen (e.g., "Once upon a time...").

[0232] Step 8:

[0233] The user reviews and enjoys the re-filtered generated text.

[0234] Input: Refiltered generated text displayed in terminal

[0235] Actions: The user reads and enjoys the generated text, and optionally enters new requests.

[0236] Output: User satisfaction and new request data.

[0237] 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.

[0238] This invention combines an emotion engine with a generative AI system that children can use safely to generate appropriate text based on the user's emotions. The system includes a series of processes: receiving request data, recognizing emotions, filtering, generating text, re-filtering, and sending the generated text.

[0239] A natural language explanation of the program's processing

[0240] Server-side processing

[0241] The server first receives request data sent from the terminal. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[0242] Next, the server uses an emotion engine to recognize the user's emotion in the received request data. The emotion engine uses natural language processing technology to extract the user's emotion from the request data. For example, a request such as "Tell me a fun story" would recognize a positive emotion.

[0243] Based on the emotions recognized by the emotion engine, the server uses a filtering means to detect inappropriate words and phrases in the request data. The filtering means adaptively adjusts rules according to the emotions and scans the request data using regular expressions and inappropriate word lists. If an inappropriate expression is detected, it is corrected or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," "scary" is determined to be inappropriate, and the request is corrected to "Tell me a funny story."

[0244] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator generates appropriate text while adjusting the content and tone based on the user's emotions recognized by the emotion engine.

[0245] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0246] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0247] Processing on the device (kids' phone)

[0248] The terminal receives input from the user and sends the content to the server as a request. For example, if the user inputs "tell me an interesting story," the terminal sends an HTTP request containing the request content to the server.

[0249] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a funny story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0250] User (child) operations

[0251] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[0252] Specific examples

[0253] 1. Requests that made me happy:

[0254] The user (child) types "Tell me a happy story" into the terminal.

[0255] The terminal sends the request to the server.

[0256] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[0257] The server adjusts rules and filters according to emotions.

[0258] The server generates happy stories using generative AI.

[0259] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0260] The server sends the text of the happy story to the device.

[0261] The terminal displays the received text on its screen for the user to read.

[0262] 2. A request that made me feel uneasy:

[0263] The user (child) types into the terminal, "Tell me a story that will ease my anxiety."

[0264] The terminal sends the request to the server.

[0265] The server receives the request and uses an emotion engine to recognize the emotion "anxiety."

[0266] The server adjusts filtering rules according to the emotion and scans the request.

[0267] The server uses generative AI to generate stories that ease anxiety.

[0268] The generated stories are then refiltered to ensure they are free of inappropriate language.

[0269] The server sends text messages to the device to ease anxiety.

[0270] The terminal displays the received text on its screen for the user to read.

[0271] In this way, by combining emotion engines, it is possible to realize a generative AI system that can adaptively adjust the request content and generated text according to the user's emotions, and that can be used safely by children.

[0272] The processing flow will be explained below.

[0273] Server-side processing

[0274] Step 1:

[0275] The server receives the request data sent from the terminal.

[0276] Specific behavior: Waits for and receives HTTP requests, and extracts the generated request from the request body, such as the text entered by the user, "Tell me an interesting story."

[0277] Step 2:

[0278] The server uses an emotion engine to recognize the user's emotion in the received request data.

[0279] What it does: Uses natural language processing techniques to extract emotion tags (e.g., happy, sad, joyful, anxious) from request data.

[0280] Step 3:

[0281] The server adjusts the rules of the filtering means based on the user's emotions recognized by the emotion engine.

[0282] Specific behavior: If the emotion tag is recognized as "fun," apply the filtering rules according to that emotion.

[0283] Step 4:

[0284] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data.

[0285] Specific behavior: Scans request data using regular expressions and inappropriate word lists, and modifies inappropriate expressions to appropriate ones. For example, "Tell me a scary story" is modified to "Tell me a funny story."

[0286] Step 5:

[0287] The server generates text using the filtered request data with the generation means.

[0288] What it does: A generative AI model (e.g., GPT-3) is fed with the filtered request data and recognized emotion tags, and generates appropriate text based on the request and emotion.

[0289] Step 6:

[0290] The server then applies filtering measures to the generated text to detect and correct or replace inappropriate words or phrases.

[0291] Specific behavior: The generated text is scanned again against regular expressions and profanity lists, and profanity is corrected to proper expressions.

[0292] Step 7:

[0293] The server sends the filtered generated text to the terminal as an HTTP response.

[0294] Specific behavior: Constructs an HTTP response and sends the generated appropriate text to the device in the response body.

[0295] Processing on the device (kids' phone)

[0296] Step 1:

[0297] The terminal receives input from the user and sends the content to the server as a request.

[0298] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[0299] Step 2:

[0300] The terminal waits for and receives a response from the server.

[0301] Specific operation: Asynchronously waits for the HTTP response sent from the server, receives it, and obtains the generated text from the response body.

[0302] Step 3:

[0303] The terminal displays the received generated text on the screen.

[0304] Specific behavior: The generated text received is reflected in the text display area so that the user can read it.

[0305] User (child) operations

[0306] Step 1:

[0307] The user inputs the text content that he or she wants to generate into the terminal.

[0308] Example: Using a text field or voice input, type "Tell me a fun story."

[0309] Step 2:

[0310] The user checks the generated text and takes the following action:

[0311] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[0312] Example 2

[0313] 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."

[0314] Conventional generative AI systems have safety issues due to the risk of inappropriate content being included when used by children. Furthermore, they lack a mechanism for generating appropriate content that adapts to the user's emotions, creating a need for a system that children can use safely. This invention aims to provide a generative AI system that generates text based on the user's emotions and filters out inappropriate content, making it safe for children to use.

[0315] 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.

[0316] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means;] [means for recognizing the user's emotions from the received request data using an emotion engine;] [means for detecting and correcting or replacing inappropriate words and phrases in the request data based on the emotions recognized by a filtering means;] [means for creating generated text in accordance with the user's emotions based on the request data filtered by a generating means;] [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; and [means for transmitting the filtered generated text to the terminal by a transmitting means]. This makes it possible [to realize a generative AI system that generates text that adapts to the user's emotions while eliminating inappropriate content, and is safe for children to use].

[0317] The "receiving means" is a device or function that receives request data sent from a terminal.

[0318] An "emotion engine" is a technology that includes software and algorithms for recognizing and analyzing user emotions from received data.

[0319] A "filtering means" is a device or function that detects and modifies or replaces inappropriate words or phrases in data.

[0320] "Generator" means a device or function for generating appropriate text based on filtered request data, including, for example, a generative AI model.

[0321] A "re-filtering means" is a device or function that re-detects and corrects or replaces inappropriate words or phrases in the generated text.

[0322] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[0323] A "generative AI model" refers to an artificial intelligence model that generates natural language text based on specific input data.

[0324] This invention provides a generative AI system that generates text that is adaptive to the user's emotions while eliminating inappropriate content, making it safe for children to use. This system includes three main components: a server, a terminal, and a user.

[0325] Server-side processing

[0326] Receiving means

[0327] The server first receives request data sent from the terminal using the receiving means. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[0328] Emotion Engine

[0329] The server uses an emotion engine to recognize the user's emotion from the received request data. The emotion engine uses natural language processing technologies such as VADER and TextBlob. These technologies extract emotions such as positive, negative, and neutral from the request data. For example, a request such as "Tell me a fun story" will recognize a positive emotion.

[0330] Filtering Methods

[0331] Based on the emotions recognized by the emotion engine, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. This filtering mechanism scans the request data using regular expressions and inappropriate word lists, and if an inappropriate expression is found, it modifies or replaces it with an appropriate expression. For example, if a request such as "Tell me a scary story" is determined to be "scary," it will be modified to "Tell me a funny story."

[0332] generation means

[0333] After filtering, the server generates text using a generative method based on the request data. This method uses a generative AI model (such as GPT-3). The generative AI model generates appropriate text by adjusting the content and tone based on the user's emotions recognized by the emotion engine. For example, the generated text for "Tell me a fun story" is a fun story that begins with "Once upon a time..."

[0334] Re-filtering means

[0335] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0336] Transmission method

[0337] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated text in the response body, and sends it to the terminal. In this way, the generated text is delivered to the user.

[0338] Terminal side processing

[0339] Receiving User Input

[0340] The terminal receives input from the user and prepares it to be sent as a request to the server, for example, "Tell me a funny story."

[0341] Sending request data

[0342] The terminal includes the request content entered by the user in an HTTP request and sends it to the server. After the server returns a response, the terminal receives the response.

[0343] Viewing generated text

[0344] The generated text is then displayed on the screen, allowing the user to read it. For example, a funny story like "Once upon a time..." is displayed on the screen.

[0345] User operation

[0346] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[0347] Specific examples

[0348] 1. User Request: "Tell me a happy story"

[0349] The terminal sends the request to the server.

[0350] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[0351] The server uses filtering measures to detect and correct inappropriate content in requests.

[0352] The server uses a generative AI model to generate happy stories.

[0353] The generated stories are then filtered again to ensure they are free of inappropriate content.

[0354] The server sends the text of the happy story to the device.

[0355] The terminal displays the received text on its screen for the user to read.

[0356] In this way, the system achieves safe text generation while adapting to the user's emotions.

[0357] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0358] Step 1:

[0359] Receiving User Input

[0360] The user inputs the text they want to generate into the device. For example, they input "Tell me a fun story." The input request is saved on the device, and an HTTP request is prepared. The input in this case is "Tell me a fun story." The output is the request data saved on the device.

[0361] Step 2:

[0362] Sending request data

[0363] The device constructs an HTTP request that includes the user's input and sends it to the server. For example, a request containing the content "Tell me a fun story" is sent to the server. The input at this time is the request data saved on the device. The output is the request data sent to the server.

[0364] Step 3:

[0365] Conducting sentiment analysis

[0366] The server passes the received request data to the emotion engine, which recognizes the user's emotion. The emotion engine uses natural language processing technology to extract emotion from the request data. For example, it can recognize positive emotion from a request such as "Tell me a fun story." The input in this case is the request data "Tell me a fun story," and the output is emotion data of "positive."

[0367] Step 4:

[0368] Filtering Request Data

[0369] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data based on the emotions recognized by the emotion engine. For example, a request such as "Tell me a scary story" is determined to be "scary" and modified to "Tell me a funny story." The input in this case is emotion data of "positive" and request data of "Tell me a fun story," and the output is the filtered request data.

[0370] Step 5:

[0371] Text Generation

[0372] The server generates appropriate text using a generative AI model (e.g., GPT-3) based on the filtered request data. For example, in response to the request "Tell me a fun story," a fun story such as "Once upon a time..." is generated. The input in this case is the filtered request data, and the output is the generated text.

[0373] Step 6:

[0374] Re-filtering the generated text

[0375] The server then applies the filtering measures to the generated text again. For example, the generated text "Once upon a time..." is filtered again to check for inappropriate language, correcting or replacing it as necessary. The input is the generated text, and the output is the re-filtered generated text.

[0376] Step 7:

[0377] Sending generated text

[0378] Finally, the server constructs the filtered generated text as an HTTP response and sends it to the terminal. The server sends the generated text in the response body to the terminal. The input at this time is the re-filtered generated text, and the output is the response data sent to the terminal.

[0379] Step 8:

[0380] Viewing generated text

[0381] The terminal receives the response from the server and displays the generated text on the screen. For example, a fun story like "Once upon a time..." is displayed on the screen. The input is the response data received by the terminal, and the output is the generated text displayed on the screen.

[0382] The above is the specific processing flow of this system.

[0383] (Application example 2)

[0384] 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."

[0385] Current generative AI systems lack appropriate filtering and emotion-based adjustment to ensure safe use by children. This means that generated content may contain inappropriate content, and it is difficult to provide content that quickly and accurately reflects the user's emotions. There is a need for a system that can solve this issue and realize a content distribution service that children can enjoy safely.

[0386] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0387] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means; [means for recognizing the user's emotions in the request data by an emotion recognition means; [means for detecting and correcting or replacing inappropriate words and phrases in the request data by a filtering means; [means for creating generated text according to the emotions based on the request filtered by the generating means; [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; [means for transmitting the filtered generated text to the terminal by a transmitting means; and [means for displaying the generated text on the terminal.] This makes it possible to provide an emotion-based generative AI system that children can use safely.

[0388] The "receiving means" is a device or function that receives request data from a terminal.

[0389] An "emotion recognition means" is a device or algorithm for recognizing a user's emotion in request data.

[0390] A "filtering means" is a device or function that detects inappropriate words or phrases in request data and modifies or replaces them.

[0391] A "generator" is a device or algorithm that creates emotion-sensitive generated text based on filtered request data.

[0392] A "re-filtering means" is a device or function that re-detects inappropriate words or phrases in the generated text and corrects or replaces them.

[0393] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[0394] A "terminal" is an electronic device that allows a user to input request data and display the generated text.

[0395] This invention relates to the implementation of the smartphone application "Emotory" as an example of an application of a content distribution service that can be used safely by children. "Emotory" is an application that generates stories based on emotions, and provides content appropriate to the user through a series of processes.

[0396] First, the user (child) inputs their feelings and interests into their smartphone. The device then sends this input to the server as request data. For example, the user might input "Tell me a fun story." The server then processes the data as follows:

[0397] Server-side processing

[0398] 1. Receiving request data: The server receives the request data sent from the terminal. This is realized by the "receiving means."

[0399] 2. Emotion Recognition: To recognize the user's emotions from the received request data, we use an "emotion recognition method." This method uses natural language processing technology, such as NLP libraries like "spaCy" and "NLTK."

[0400] 3. Request Data Filtering: Using "filtering methods" to detect and modify or replace inappropriate words and phrases based on recognized sentiment. Filtering methods may use regular expressions or predefined lists of inappropriate words.

[0401] 4. Generating generated text: A "generator" is used to generate sentiment-based generated text based on the filtered request data. This generator uses a large-scale generative AI model, such as OpenAI's "GPT-3."

[0402] 5. Re-filtering the generated text: A "re-filtering method" is used to again detect and correct or replace inappropriate words or phrases in the generated text.

[0403] 6. Sending the generated text: Finally, the filtered generated text is sent to the terminal by the "sending means."

[0404] Terminal side processing

[0405] The terminal receives the generated text sent from the server and has the function of "displaying the generated text on the terminal" in a format that is easy for the user to read. The user (child) can enjoy reading this generated text.

[0406] Specific hardware / software usage

[0407] NLP library: "spaCy" and "NLTK" are used for emotion recognition.

[0408] Generative AI model: OpenAI's "GPT-3" is used for text generation.

[0409] Filtering libraries: Use your own implementation or third-party libraries that use regular expressions and profanity lists.

[0410] Examples of concrete examples and prompts

[0411] Examples:

[0412] 1. If you feel happy:

[0413] User input: Child types "Tell me a happy story."

[0414] Server processing: After emotion recognition, filtering, generation, and re-filtering, a positive story is generated: "Once upon a time, there was a very happy village..."

[0415] Terminal display: "Once upon a time, there was a very happy village..."

[0416] 2. To relieve anxiety:

[0417] User input: Child types, "Tell me a story to calm my anxiety."

[0418] Server processing: After emotion recognition, filtering, generation, and re-filtering, a soothing story is generated: "Listen to this story. In a very peaceful forest..."

[0419] Terminal display: "Listen to this story. In a very peaceful forest..."

[0420] Example prompt sentence:

[0421] A child types "Tell me a fun story" into their smartphone.

[0422] As a result, "Emotory" is an application that allows children to spend time safely and enjoyably, and can provide appropriate content according to the user's emotions.

[0423] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0424] Step 1:

[0425] User Input

[0426] The user (child) inputs the text they want to generate into their smartphone. For example, they might input "Tell me a fun story." The device then sends this input to the server as request data.

[0427] Input: Text entered by the user (e.g., "Tell me a nice story")

[0428] Output: Request data sent to the server

[0429] Step 2:

[0430] Receiving request data

[0431] The server receives the request data sent from the terminal, and provides the server with information about the content the user is looking for.

[0432] Input: Request data

[0433] Output: Request data stored on the server

[0434] Step 3:

[0435] Emotion recognition

[0436] The server recognizes the user's emotion from the received request data. For emotion recognition, it uses an NLP library (e.g., "spaCy" or "NLTK"). For example, if the request is "Tell me a fun story," the user's emotion is recognized as "positive."

[0437] Input: Request data

[0438] Output: Perceived emotion (e.g., "positive")

[0439] Step 4:

[0440] Filtering Request Data

[0441] The server detects and modifies or replaces inappropriate words and phrases in the request data based on sentiment. Regular expressions and inappropriate word lists are used as filtering methods. For example, if a request is made to "tell me a scary story," the word "scary" is deemed "inappropriate" and the request is modified to "tell me a funny story."

[0442] Input: Request data, recognized emotion

[0443] Output: Filtered request data

[0444] Step 5:

[0445] Creating generated text

[0446] The server generates text based on the filtered request data, according to the sentiment. A large-scale generative AI model (e.g., GPT-3) is used as the generation method. For example, if the filtered data is "Tell me a happy story" and the sentiment is "positive," the generated text would be something like "Once upon a time, there was a very happy village..."

[0447] Input: filtered request data, recognized sentiment

[0448] Output: The generated text

[0449] Step 6:

[0450] Re-filtering the generated text

[0451] The server again detects and corrects or replaces inappropriate words or phrases in the generated text. For example, if the generated text contains an "inappropriate word," it replaces it with an appropriate expression. This process is also performed using filtering methods.

[0452] Input: Generated text

[0453] Output: Refiltered generated text

[0454] Step 7:

[0455] Sending generated text

[0456] The server sends the filtered generated text to the terminal using an HTTP response as a transmission means, and includes the text in the response body.

[0457] Input: Refiltered generated text

[0458] Output: The generated text sent to the terminal

[0459] Step 8:

[0460] Viewing generated text

[0461] The terminal receives the generated text sent from the server and displays it on the screen, for example, the generated text "Once upon a time, there was a very pleasant village..." is displayed on the screen for the user to read.

[0462] Input: Generated text sent from the server

[0463] Output: The generated text displayed on the terminal.

[0464] 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.

[0465] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0466] 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.

[0467] [Second embodiment]

[0468] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0469] 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.

[0470] 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).

[0471] 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.

[0472] 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.

[0473] 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).

[0474] 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.

[0475] 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.

[0476] 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.

[0477] 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.

[0478] 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.

[0479] 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."

[0480] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[0481] A natural language explanation of the program's processing

[0482] Server-side processing

[0483] The server first receives request data sent from the device. This request data includes the text entered by the user and the content they want to generate. For example, if the user enters "Tell me an interesting story," the device sends the request to the server.

[0484] Next, the server uses a filtering mechanism to detect inappropriate words or phrases in the request data. The filtering mechanism scans the request data using regular expressions or a list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate, and the request is modified to "Tell me a funny story."

[0485] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator takes the filtered request data as input and generates appropriate text.

[0486] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0487] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0488] Processing on the device (kids' phone)

[0489] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[0490] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0491] User (child) operations

[0492] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[0493] Specific examples

[0494] 1. First request:

[0495] The user (child) types "tell me a scary story" into the terminal.

[0496] The terminal sends the request to the server.

[0497] The server receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[0498] The server generates interesting stories using generative AI.

[0499] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0500] The server sends the text of an interesting story to the device.

[0501] The terminal displays the received text on its screen so that the user can read it.

[0502] 2. Harmless Requests:

[0503] The user (child) types "Tell me a story about a cute animal" into the terminal.

[0504] The terminal sends the request to the server.

[0505] The server receives the request, filters it, and because it contains no profanity, the request is kept as is.

[0506] The server generates cute animal stories using generative AI.

[0507] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0508] The server sends text of cute animal stories to the device.

[0509] The terminal displays the received text on its screen so that the user can read it.

[0510] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[0511] The processing flow will be explained below.

[0512] Server-side processing

[0513] Step 1:

[0514] The server receives the request data sent from the terminal.

[0515] Specific behavior: Listens for and receives HTTP requests, extracting user-entered text and generated request content from the request body.

[0516] Step 2:

[0517] The server uses filtering means to detect inappropriate words or phrases in the received request data.

[0518] What it does: Scans request data using regular expressions and profanity lists, and fixes or replaces any profanity it finds.

[0519] Step 3:

[0520] The server generates text using the filtered request data with the generation means.

[0521] What it does: Passes the filtered request to a generative AI model (e.g., GPT-3) and makes an API request to generate text that meets the request.

[0522] Step 4:

[0523] The server again applies filtering measures to the generated text to detect inappropriate words and phrases.

[0524] What it does: Scan the generated text again against the regular expressions and profanity list, making corrections or substitutions if necessary.

[0525] Step 5:

[0526] The server sends the filtered generated text to the terminal as an HTTP response.

[0527] Specific behavior: Constructs an HTTP response and sends the generated text to the terminal in the response body.

[0528] Processing on the device (kids' phone)

[0529] Step 1:

[0530] The terminal receives input from the user and sends the content to the server as a request.

[0531] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[0532] Step 2:

[0533] The terminal waits for and receives a response from the server.

[0534] Specific operation: Asynchronously waits for an HTTP response from the server, and retrieves the content of the response when it arrives.

[0535] Step 3:

[0536] The terminal displays the received generated text on the screen.

[0537] Specific operation: The generated text received is reflected in the text display area, making it viewable by the user.

[0538] User (child) operations

[0539] Step 1:

[0540] The user inputs the text content that he or she wants to generate into the terminal.

[0541] Example: Using a text field or voice input, type "Draw me a funny picture."

[0542] Step 2:

[0543] The user checks the generated text and takes the following action:

[0544] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[0545] Example 1

[0546] 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."

[0547] With conventional generative AI systems, the generated text may contain inappropriate language, making it difficult to provide an environment that children can use with confidence. Specifically, if inappropriate words or phrases are included in the request data or generated text, they are output as is, posing a risk that children may come into contact with harmful information. There is a need to solve these issues and provide a generative AI system that children can use with confidence.

[0548] 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.

[0549] In this invention, the server includes: [means for receiving request data from the information processing device by a receiving means; [means for detecting inappropriate words and standard phrases in the request data by a filtering means and correcting or replacing them;] [means for creating text to be generated based on the filtered request data by a generating means; [means for re-detecting inappropriate words and standard phrases in the generated text by a re-filtering means and correcting or replacing them; and [means for transmitting the filtered generated text to the information processing device by a transmitting means.] This enables a generative AI system that can be used safely by children by doubly filtering inappropriate expressions in both the request data and the generated text.

[0550] The "receiving means" is a means for receiving request data from an information processing device.

[0551] "Filtering means" refers to means for detecting inappropriate words or boilerplate phrases in request data or generated text and correcting or replacing them.

[0552] The "generator" is a means for generating text based on the filtered request data.

[0553] The "re-filtering means" is a means for detecting inappropriate words or boilerplate phrases again in the generated text and correcting or replacing them.

[0554] The "transmitting means" is a means for transmitting the filtered generated text to the information processing device.

[0555] An "information processing device" is a device for inputting request data from a user and transmitting it to a server.

[0556] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[0557] Server-side processing

[0558] The server receives request data sent from the device. This request data includes the text entered by the user and the content to be generated. For example, if the user enters "tell me an interesting story," the device sends the request to the server.

[0559] Next, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. The filtering mechanism scans the request data using regular expressions or a predefined list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," if "scary" is determined to be inappropriate, the request is modified to "Tell me a funny story."

[0560] Based on the filtered request data, the server generates text using a generation means. For example, a generative AI model (such as GPT-3) is used as the generation means. The generation means takes the filtered request data as input and generates appropriate text.

[0561] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0562] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0563] Processing on the terminal (information processing device) side

[0564] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[0565] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0566] User (child) operations

[0567] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[0568] Specific examples

[0569] 1. First time request

[0570] User: Type "tell me a scary story" into the terminal.

[0571] Terminal: Sends an HTTP request to the server.

[0572] Server: Receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[0573] Server: Generate text using a generative AI model using the modified request data.

[0574] Server: Re-filter the generated text to ensure it is free of profanity.

[0575] Server: Sends the filtered generated text to the terminal as an HTTP response.

[0576] Terminal: Displays the received text on the screen, displaying the story "Once upon a time..." for the user to read.

[0577] 2. Harmless Requests

[0578] User: Type "Tell me a cute animal story" into the terminal.

[0579] Terminal: Sends an HTTP request to the server.

[0580] Server: Receives the request and filters it. Since it contains no profanity, the request is kept as is.

[0581] Server: Generate text using a generative AI model using the filtered request data.

[0582] Server: Re-filter the generated text to ensure it is free of profanity.

[0583] Server: Sends the filtered generated text to the terminal as an HTTP response.

[0584] Terminal: Displays the received text on the screen. It displays "cute animal stories" for the user to read.

[0585] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[0586] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0587] Step 1: Receiving request data

[0588] The server receives request data sent from the device. Input is text entered by the user into the device or content they want to generate. For example, the user might enter "Tell me an interesting story." The device sends this input content to the server as an HTTP request. The server receives the HTTP request and obtains the request data.

[0589] Step 2: Filtering the request data

[0590] The server scans the received request data using a filtering method. The input is the received request data. The filtering method uses regular expressions and a predefined list of inappropriate words. Specifically, the server detects inappropriate words and fixed phrases in the request data. For example, in the case of a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate and the request is corrected to "Tell me a funny story." The output is the filtered request data.

[0591] Step 3: Generate text

[0592] The server generates text using a generative method based on the filtered request data. The input is the filtered request data ("Tell me a funny story"). The server passes this input to a generative AI model (e.g., GPT-3) as a prompt. The generative AI model generates appropriate text corresponding to the input prompt, e.g., "Once upon a time...". The output is the generated text.

[0593] Step 4: Re-filtering the generated text

[0594] The server scans the generated text again using the filtering method. The input is the generated text. Specifically, the server detects inappropriate words or boilerplate phrases in the generated text and corrects or replaces them. For example, if the generated text contains the word "scary," it replaces it with "funny." The output is the re-filtered generated text.

[0595] Step 5: Sending generated text

[0596] The server sends the re-filtered generated text to the terminal using a transmission means. The input is the re-filtered generated text. In specific operations, the server constructs an HTTP response and includes the generated text in the response body. Then, the server sends the HTTP response to the terminal. The terminal displays the received generated text on the screen so that the user can read it. The output is the generated text displayed on the terminal.

[0597] (Application example 1)

[0598] 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."

[0599] Current generative AI systems can sometimes be difficult to guarantee safety so that children can use them with confidence. Measures are needed to properly filter requests and generated text that contain inappropriate words or phrases, but specific methods for achieving this have not yet been established. Furthermore, there is a lack of interfaces that allow children to enjoy the generated text with peace of mind. A solution to these issues is needed.

[0600] 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.

[0601] In this invention, the server includes means for receiving request data from the terminal using a receiving means, means for detecting and correcting or replacing inappropriate words and phrases in the request data using a filtering means, means for creating text to be generated based on the request filtered by the generative AI model, means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text using a re-filtering means, means for transmitting the filtered generated text to the terminal using a transmitting means, and means for providing an interface that allows a user to check and enjoy the re-filtered generated text. This realizes a generative AI system that children can use safely, allowing children to safely enjoy generated text that does not contain inappropriate words or phrases.

[0602] The "receiving means" is a means for receiving request data from a terminal.

[0603] "Filtering means" means means for detecting and modifying or replacing inappropriate words or phrases in request data.

[0604] A "generative AI model" is an artificial intelligence model for generating appropriate text based on filtered request data.

[0605] A "re-filtering means" is a means for re-detecting and correcting or replacing inappropriate words or phrases in the generated text.

[0606] The "transmitting means" is a means for transmitting the filtered generated text to the terminal.

[0607] An "interface" is a display means and operating means that allows a user to view and enjoy the re-filtered generated text.

[0608] The embodiment of this invention provides a generative AI system that children can use safely. The system mainly consists of a server, a terminal, and a user.

[0609] System configuration

[0610] Server-side processing

[0611] The server comprises a "receiving means," a "filtering means," a "generative AI model," a "re-filtering means," and a "transmitting means."

[0612] 1. Receiving means: Receives request data sent from the terminal. The request data includes text entered by the user and the content to be generated.

[0613] 2. Filtering methods: Detect inappropriate words or phrases in the request data and modify or replace them. Use regular expressions or profanity lists.

[0614] 3. Generative AI model: Generate appropriate text based on the filtered request data. Here, we use a generative AI model such as OpenAI's GPT-3.

[0615] 4. Re-filtering: The generated text is checked again for inappropriate language and corrected or replaced as necessary.

[0616] 5. Transmission method: Send the filtered generated text to the terminal. Construct an HTTP response and send the generated text in the response body.

[0617] Terminal side processing

[0618] The terminal receives input from the user, sends the content to the server as a request, and displays the generated text on the screen.

[0619] Example

[0620] User (child) operations

[0621] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "tell me a funny story," the request is sent to the server, and the generated funny story is displayed on the terminal.

[0622] Specific examples

[0623] First request:

[0624] The user types "tell me a scary story."

[0625] The terminal sends the request to the server.

[0626] The server uses filtering means to modify "scary" to "interesting."

[0627] A generative AI model generates "interesting stories" and a re-filtering method checks their appropriateness.

[0628] An interesting story text will be sent to your device and displayed.

[0629] Harmless requests:

[0630] A user types, "Tell me about cute animals."

[0631] The terminal sends the request to the server.

[0632] The server filters the request and generates a story using a generative AI model.

[0633] The generated stories are filtered and displayed on the device.

[0634] Prompt Sentence Examples

[0635] Prompt: Tell me a funny story

[0636] This system provides an environment where children can use generative AI safely, and ensures that the generated text is appropriate for the user, allowing them to enjoy it with peace of mind.

[0637] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0638] Step 1:

[0639] The user enters text into the terminal.

[0640] Input: Text entered by the user using a keyboard or touchscreen

[0641] What it does: A user types the text "tell me a funny story" into their device.

[0642] Output: The entered text is saved on the terminal as request data.

[0643] Step 2:

[0644] The terminal transmits the request data to the server.

[0645] Input: Request data stored on the device

[0646] How it works: The device constructs an HTTP request and sends the request data to the server.

[0647] Output: The server receives the request data from the HTTP request.

[0648] Step 3:

[0649] Filters the request data received by the server.

[0650] Input: Received request data (e.g., "Tell me a scary story")

[0651] Data Modification: Filtering methods are used to detect and modify or replace inappropriate words or phrases in request data, such as using regular expressions or profanity lists.

[0652] Action: The request data "Tell me a scary story" is modified to "Tell me a funny story."

[0653] Output: Filtered request data (e.g., "Tell me a funny story").

[0654] Step 4:

[0655] The server generates text based on the filtered request data.

[0656] Input: Filtered request data (e.g., "Tell me a funny story")

[0657] Data calculation: Text generation using a generative AI model. The prompt sentence "Tell me an interesting story" is input into a generative AI (e.g., GPT-3) to generate appropriate text.

[0658] How it works: A generative AI model generates text that begins with "Once upon a time..."

[0659] Output: The generated text (e.g., "Once upon a time...").

[0660] Step 5:

[0661] The server re-filters the generated text.

[0662] Input: Generated text (e.g. "Once upon a time...")

[0663] Data processing: Re-filtering measures are used to detect and correct or replace inappropriate words or phrases in the generated text.

[0664] Action: Checks the generated text for profanity and corrects it if necessary.

[0665] Output: The refiltered generated text (e.g., "Once upon a time...").

[0666] Step 6:

[0667] The server sends the re-filtered generated text to the terminal.

[0668] Input: Re-filtered generated text (e.g., "Once upon a time...")

[0669] Behavior: The server sends the re-filtered generated text to the terminal as an HTTP response and includes it in the response body.

[0670] Output: The device receives the re-filtered generated text from the server.

[0671] Step 7:

[0672] The terminal displays the refiltered generated text to the user.

[0673] Input: Re-filtered generated text received from the server (e.g., "Once upon a time...")

[0674] Action: The terminal displays the resulting refiltered text on the screen, presenting the text in a user-friendly format.

[0675] Output: The generated text displayed on the screen (e.g., "Once upon a time...").

[0676] Step 8:

[0677] The user reviews and enjoys the re-filtered generated text.

[0678] Input: Refiltered generated text displayed in terminal

[0679] Actions: The user reads and enjoys the generated text, and optionally enters new requests.

[0680] Output: User satisfaction and new request data.

[0681] 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.

[0682] This invention combines an emotion engine with a generative AI system that children can use safely to generate appropriate text based on the user's emotions. The system includes a series of processes: receiving request data, recognizing emotions, filtering, generating text, re-filtering, and sending the generated text.

[0683] A natural language explanation of the program's processing

[0684] Server-side processing

[0685] The server first receives request data sent from the terminal. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[0686] Next, the server uses an emotion engine to recognize the user's emotion in the received request data. The emotion engine uses natural language processing technology to extract the user's emotion from the request data. For example, a request such as "Tell me a fun story" would recognize a positive emotion.

[0687] Based on the emotions recognized by the emotion engine, the server uses a filtering means to detect inappropriate words and phrases in the request data. The filtering means adaptively adjusts rules according to the emotions and scans the request data using regular expressions and inappropriate word lists. If an inappropriate expression is detected, it is corrected or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," "scary" is determined to be inappropriate, and the request is corrected to "Tell me a funny story."

[0688] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator generates appropriate text while adjusting the content and tone based on the user's emotions recognized by the emotion engine.

[0689] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0690] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0691] Processing on the device (kids' phone)

[0692] The terminal receives input from the user and sends the content to the server as a request. For example, if the user inputs "tell me an interesting story," the terminal sends an HTTP request containing the request content to the server.

[0693] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a funny story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0694] User (child) operations

[0695] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[0696] Specific examples

[0697] 1. Requests that made me happy:

[0698] The user (child) types "Tell me a happy story" into the terminal.

[0699] The terminal sends the request to the server.

[0700] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[0701] The server adjusts rules and filters according to emotions.

[0702] The server generates happy stories using generative AI.

[0703] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0704] The server sends the text of the happy story to the device.

[0705] The terminal displays the received text on its screen for the user to read.

[0706] 2. A request that made me feel uneasy:

[0707] The user (child) types into the terminal, "Tell me a story that will ease my anxiety."

[0708] The terminal sends the request to the server.

[0709] The server receives the request and uses an emotion engine to recognize the emotion "anxiety."

[0710] The server adjusts filtering rules according to the emotion and scans the request.

[0711] The server uses generative AI to generate stories that ease anxiety.

[0712] The generated stories are then refiltered to ensure they are free of inappropriate language.

[0713] The server sends text messages to the device to ease anxiety.

[0714] The terminal displays the received text on its screen for the user to read.

[0715] In this way, by combining emotion engines, it is possible to realize a generative AI system that can adaptively adjust the request content and generated text according to the user's emotions, and that can be used safely by children.

[0716] The processing flow will be explained below.

[0717] Server-side processing

[0718] Step 1:

[0719] The server receives the request data sent from the terminal.

[0720] Specific behavior: Waits for and receives HTTP requests, and extracts the generated request from the request body, such as the text entered by the user, "Tell me an interesting story."

[0721] Step 2:

[0722] The server uses an emotion engine to recognize the user's emotion in the received request data.

[0723] What it does: Uses natural language processing techniques to extract emotion tags (e.g., happy, sad, joyful, anxious) from request data.

[0724] Step 3:

[0725] The server adjusts the rules of the filtering means based on the user's emotions recognized by the emotion engine.

[0726] Specific behavior: If the emotion tag is recognized as "fun," apply the filtering rules according to that emotion.

[0727] Step 4:

[0728] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data.

[0729] Specific behavior: Scans request data using regular expressions and inappropriate word lists, and modifies inappropriate expressions to appropriate ones. For example, "Tell me a scary story" is modified to "Tell me a funny story."

[0730] Step 5:

[0731] The server generates text using the filtered request data with the generation means.

[0732] What it does: A generative AI model (e.g., GPT-3) is fed with the filtered request data and recognized emotion tags, and generates appropriate text based on the request and emotion.

[0733] Step 6:

[0734] The server then applies filtering measures to the generated text to detect and correct or replace inappropriate words or phrases.

[0735] Specific behavior: The generated text is scanned again against regular expressions and profanity lists, and profanity is corrected to proper expressions.

[0736] Step 7:

[0737] The server sends the filtered generated text to the terminal as an HTTP response.

[0738] Specific behavior: Constructs an HTTP response and sends the generated appropriate text to the device in the response body.

[0739] Processing on the device (kids' phone)

[0740] Step 1:

[0741] The terminal receives input from the user and sends the content to the server as a request.

[0742] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[0743] Step 2:

[0744] The terminal waits for and receives a response from the server.

[0745] Specific operation: Asynchronously waits for the HTTP response sent from the server, receives it, and obtains the generated text from the response body.

[0746] Step 3:

[0747] The terminal displays the received generated text on the screen.

[0748] Specific behavior: The generated text received is reflected in the text display area so that the user can read it.

[0749] User (child) operations

[0750] Step 1:

[0751] The user inputs the text content that he or she wants to generate into the terminal.

[0752] Example: Using a text field or voice input, type "Tell me a fun story."

[0753] Step 2:

[0754] The user checks the generated text and takes the following action:

[0755] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[0756] Example 2

[0757] 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."

[0758] Conventional generative AI systems have safety issues due to the risk of inappropriate content being included when used by children. Furthermore, they lack a mechanism for generating appropriate content that adapts to the user's emotions, creating a need for a system that children can use safely. This invention aims to provide a generative AI system that generates text based on the user's emotions and filters out inappropriate content, making it safe for children to use.

[0759] 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.

[0760] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means;] [means for recognizing the user's emotions from the received request data using an emotion engine;] [means for detecting and correcting or replacing inappropriate words and phrases in the request data based on the emotions recognized by a filtering means;] [means for creating generated text in accordance with the user's emotions based on the request data filtered by a generating means;] [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; and [means for transmitting the filtered generated text to the terminal by a transmitting means]. This makes it possible [to realize a generative AI system that generates text that adapts to the user's emotions while eliminating inappropriate content, and is safe for children to use].

[0761] The "receiving means" is a device or function that receives request data sent from a terminal.

[0762] An "emotion engine" is a technology that includes software and algorithms for recognizing and analyzing user emotions from received data.

[0763] A "filtering means" is a device or function that detects and modifies or replaces inappropriate words or phrases in data.

[0764] "Generator" means a device or function for generating appropriate text based on filtered request data, including, for example, a generative AI model.

[0765] A "re-filtering means" is a device or function that re-detects and corrects or replaces inappropriate words or phrases in the generated text.

[0766] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[0767] A "generative AI model" refers to an artificial intelligence model that generates natural language text based on specific input data.

[0768] This invention provides a generative AI system that generates text that is adaptive to the user's emotions while eliminating inappropriate content, making it safe for children to use. This system includes three main components: a server, a terminal, and a user.

[0769] Server-side processing

[0770] Receiving means

[0771] The server first receives request data sent from the terminal using the receiving means. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[0772] Emotion Engine

[0773] The server uses an emotion engine to recognize the user's emotion from the received request data. The emotion engine uses natural language processing technologies such as VADER and TextBlob. These technologies extract emotions such as positive, negative, and neutral from the request data. For example, a request such as "Tell me a fun story" will recognize a positive emotion.

[0774] Filtering Methods

[0775] Based on the emotions recognized by the emotion engine, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. This filtering mechanism scans the request data using regular expressions and inappropriate word lists, and if an inappropriate expression is found, it modifies or replaces it with an appropriate expression. For example, if a request such as "Tell me a scary story" is determined to be "scary," it will be modified to "Tell me a funny story."

[0776] generation means

[0777] After filtering, the server generates text using a generative method based on the request data. This method uses a generative AI model (such as GPT-3). The generative AI model generates appropriate text by adjusting the content and tone based on the user's emotions recognized by the emotion engine. For example, the generated text for "Tell me a fun story" is a fun story that begins with "Once upon a time..."

[0778] Re-filtering means

[0779] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0780] Transmission method

[0781] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated text in the response body, and sends it to the terminal. In this way, the generated text is delivered to the user.

[0782] Terminal side processing

[0783] Receiving User Input

[0784] The terminal receives input from the user and prepares it to be sent as a request to the server, for example, "Tell me a funny story."

[0785] Sending request data

[0786] The terminal includes the request content entered by the user in an HTTP request and sends it to the server. After the server returns a response, the terminal receives the response.

[0787] Viewing generated text

[0788] The generated text is then displayed on the screen, allowing the user to read it. For example, a funny story like "Once upon a time..." is displayed on the screen.

[0789] User operation

[0790] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[0791] Specific examples

[0792] 1. User Request: "Tell me a happy story"

[0793] The terminal sends the request to the server.

[0794] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[0795] The server uses filtering measures to detect and correct inappropriate content in requests.

[0796] The server uses a generative AI model to generate happy stories.

[0797] The generated stories are then filtered again to ensure they are free of inappropriate content.

[0798] The server sends the text of the happy story to the device.

[0799] The terminal displays the received text on its screen for the user to read.

[0800] In this way, the system achieves safe text generation while adapting to the user's emotions.

[0801] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0802] Step 1:

[0803] Receiving User Input

[0804] The user inputs the text they want to generate into the device. For example, they input "Tell me a fun story." The input request is saved on the device, and an HTTP request is prepared. The input in this case is "Tell me a fun story." The output is the request data saved on the device.

[0805] Step 2:

[0806] Sending request data

[0807] The device constructs an HTTP request that includes the user's input and sends it to the server. For example, a request containing the content "Tell me a fun story" is sent to the server. The input at this time is the request data saved on the device. The output is the request data sent to the server.

[0808] Step 3:

[0809] Conducting sentiment analysis

[0810] The server passes the received request data to the emotion engine, which recognizes the user's emotion. The emotion engine uses natural language processing technology to extract emotion from the request data. For example, it can recognize positive emotion from a request such as "Tell me a fun story." The input in this case is the request data "Tell me a fun story," and the output is emotion data of "positive."

[0811] Step 4:

[0812] Filtering Request Data

[0813] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data based on the emotions recognized by the emotion engine. For example, a request such as "Tell me a scary story" is determined to be "scary" and modified to "Tell me a funny story." The input in this case is emotion data of "positive" and request data of "Tell me a fun story," and the output is the filtered request data.

[0814] Step 5:

[0815] Text Generation

[0816] The server generates appropriate text using a generative AI model (e.g., GPT-3) based on the filtered request data. For example, in response to the request "Tell me a fun story," a fun story such as "Once upon a time..." is generated. The input in this case is the filtered request data, and the output is the generated text.

[0817] Step 6:

[0818] Re-filtering the generated text

[0819] The server then applies the filtering measures to the generated text again. For example, the generated text "Once upon a time..." is filtered again to check for inappropriate language, correcting or replacing it as necessary. The input is the generated text, and the output is the re-filtered generated text.

[0820] Step 7:

[0821] Sending generated text

[0822] Finally, the server constructs the filtered generated text as an HTTP response and sends it to the terminal. The server sends the generated text in the response body to the terminal. The input at this time is the re-filtered generated text, and the output is the response data sent to the terminal.

[0823] Step 8:

[0824] Viewing generated text

[0825] The terminal receives the response from the server and displays the generated text on the screen. For example, a fun story like "Once upon a time..." is displayed on the screen. The input is the response data received by the terminal, and the output is the generated text displayed on the screen.

[0826] The above is the specific processing flow of this system.

[0827] (Application example 2)

[0828] 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."

[0829] Current generative AI systems lack appropriate filtering and emotion-based adjustment to ensure safe use by children. This means that generated content may contain inappropriate content, and it is difficult to provide content that quickly and accurately reflects the user's emotions. There is a need for a system that can solve this issue and realize a content distribution service that children can enjoy safely.

[0830] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0831] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means; [means for recognizing the user's emotions in the request data by an emotion recognition means; [means for detecting and correcting or replacing inappropriate words and phrases in the request data by a filtering means; [means for creating generated text according to the emotions based on the request filtered by the generating means; [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; [means for transmitting the filtered generated text to the terminal by a transmitting means; and [means for displaying the generated text on the terminal.] This makes it possible to provide an emotion-based generative AI system that children can use safely.

[0832] The "receiving means" is a device or function that receives request data from a terminal.

[0833] An "emotion recognition means" is a device or algorithm for recognizing a user's emotion in request data.

[0834] A "filtering means" is a device or function that detects inappropriate words or phrases in request data and modifies or replaces them.

[0835] A "generator" is a device or algorithm that creates emotion-sensitive generated text based on filtered request data.

[0836] A "re-filtering means" is a device or function that re-detects inappropriate words or phrases in the generated text and corrects or replaces them.

[0837] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[0838] A "terminal" is an electronic device that allows a user to input request data and display the generated text.

[0839] This invention relates to the implementation of the smartphone application "Emotory" as an example of an application of a content distribution service that can be used safely by children. "Emotory" is an application that generates stories based on emotions, and provides content appropriate to the user through a series of processes.

[0840] First, the user (child) inputs their feelings and interests into their smartphone. The device then sends this input to the server as request data. For example, the user might input "Tell me a fun story." The server then processes the data as follows:

[0841] Server-side processing

[0842] 1. Receiving request data: The server receives the request data sent from the terminal. This is realized by the "receiving means."

[0843] 2. Emotion Recognition: To recognize the user's emotions from the received request data, we use an "emotion recognition method." This method uses natural language processing technology, such as NLP libraries like "spaCy" and "NLTK."

[0844] 3. Request Data Filtering: Using "filtering methods" to detect and modify or replace inappropriate words and phrases based on recognized sentiment. Filtering methods may use regular expressions or predefined lists of inappropriate words.

[0845] 4. Generating generated text: A "generator" is used to generate sentiment-based generated text based on the filtered request data. This generator uses a large-scale generative AI model, such as OpenAI's "GPT-3."

[0846] 5. Re-filtering the generated text: A "re-filtering method" is used to again detect and correct or replace inappropriate words or phrases in the generated text.

[0847] 6. Sending the generated text: Finally, the filtered generated text is sent to the terminal by the "sending means."

[0848] Terminal side processing

[0849] The terminal receives the generated text sent from the server and has the function of "displaying the generated text on the terminal" in a format that is easy for the user to read. The user (child) can enjoy reading this generated text.

[0850] Specific hardware / software usage

[0851] NLP library: "spaCy" and "NLTK" are used for emotion recognition.

[0852] Generative AI model: OpenAI's "GPT-3" is used for text generation.

[0853] Filtering libraries: Use your own implementation or third-party libraries that use regular expressions and profanity lists.

[0854] Examples of concrete examples and prompts

[0855] Examples:

[0856] 1. If you feel happy:

[0857] User input: Child types "Tell me a happy story."

[0858] Server processing: After emotion recognition, filtering, generation, and re-filtering, a positive story is generated: "Once upon a time, there was a very happy village..."

[0859] Terminal display: "Once upon a time, there was a very happy village..."

[0860] 2. To relieve anxiety:

[0861] User input: Child types, "Tell me a story to calm my anxiety."

[0862] Server processing: After emotion recognition, filtering, generation, and re-filtering, a soothing story is generated: "Listen to this story. In a very peaceful forest..."

[0863] Terminal display: "Listen to this story. In a very peaceful forest..."

[0864] Example prompt sentence:

[0865] A child types "Tell me a fun story" into their smartphone.

[0866] As a result, "Emotory" is an application that allows children to spend time safely and enjoyably, and can provide appropriate content according to the user's emotions.

[0867] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0868] Step 1:

[0869] User Input

[0870] The user (child) inputs the text they want to generate into their smartphone. For example, they might input "Tell me a fun story." The device then sends this input to the server as request data.

[0871] Input: Text entered by the user (e.g., "Tell me a nice story")

[0872] Output: Request data sent to the server

[0873] Step 2:

[0874] Receiving request data

[0875] The server receives the request data sent from the terminal, and provides the server with information about the content the user is looking for.

[0876] Input: Request data

[0877] Output: Request data stored on the server

[0878] Step 3:

[0879] Emotion recognition

[0880] The server recognizes the user's emotion from the received request data. For emotion recognition, it uses an NLP library (e.g., "spaCy" or "NLTK"). For example, if the request is "Tell me a fun story," the user's emotion is recognized as "positive."

[0881] Input: Request data

[0882] Output: Perceived emotion (e.g., "positive")

[0883] Step 4:

[0884] Filtering Request Data

[0885] The server detects and modifies or replaces inappropriate words and phrases in the request data based on sentiment. Regular expressions and inappropriate word lists are used as filtering methods. For example, if a request is made to "tell me a scary story," the word "scary" is deemed "inappropriate" and the request is modified to "tell me a funny story."

[0886] Input: Request data, recognized emotion

[0887] Output: Filtered request data

[0888] Step 5:

[0889] Creating generated text

[0890] The server generates text based on the filtered request data, according to the sentiment. A large-scale generative AI model (e.g., GPT-3) is used as the generation method. For example, if the filtered data is "Tell me a happy story" and the sentiment is "positive," the generated text would be something like "Once upon a time, there was a very happy village..."

[0891] Input: filtered request data, recognized sentiment

[0892] Output: The generated text

[0893] Step 6:

[0894] Re-filtering the generated text

[0895] The server again detects and corrects or replaces inappropriate words or phrases in the generated text. For example, if the generated text contains an "inappropriate word," it replaces it with an appropriate expression. This process is also performed using filtering methods.

[0896] Input: Generated text

[0897] Output: Refiltered generated text

[0898] Step 7:

[0899] Sending generated text

[0900] The server sends the filtered generated text to the terminal using an HTTP response as a transmission means, and includes the text in the response body.

[0901] Input: Refiltered generated text

[0902] Output: The generated text sent to the terminal

[0903] Step 8:

[0904] Viewing generated text

[0905] The terminal receives the generated text sent from the server and displays it on the screen, for example, the generated text "Once upon a time, there was a very pleasant village..." is displayed on the screen for the user to read.

[0906] Input: Generated text sent from the server

[0907] Output: The generated text displayed on the terminal.

[0908] 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.

[0909] 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.

[0910] 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.

[0911] [Third embodiment]

[0912] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0913] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0914] 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).

[0915] 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.

[0916] 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.

[0917] 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).

[0918] 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.

[0919] 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.

[0920] 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.

[0921] 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.

[0922] 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.

[0923] 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."

[0924] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[0925] A natural language explanation of the program's processing

[0926] Server-side processing

[0927] The server first receives request data sent from the device. This request data includes the text entered by the user and the content they want to generate. For example, if the user enters "Tell me an interesting story," the device sends the request to the server.

[0928] Next, the server uses a filtering mechanism to detect inappropriate words or phrases in the request data. The filtering mechanism scans the request data using regular expressions or a list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate, and the request is modified to "Tell me a funny story."

[0929] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator takes the filtered request data as input and generates appropriate text.

[0930] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[0931] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[0932] Processing on the device (kids' phone)

[0933] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[0934] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[0935] User (child) operations

[0936] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[0937] Specific examples

[0938] 1. First request:

[0939] The user (child) types "tell me a scary story" into the terminal.

[0940] The terminal sends the request to the server.

[0941] The server receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[0942] The server generates interesting stories using generative AI.

[0943] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0944] The server sends the text of an interesting story to the device.

[0945] The terminal displays the received text on its screen so that the user can read it.

[0946] 2. Harmless Requests:

[0947] The user (child) types "Tell me a story about a cute animal" into the terminal.

[0948] The terminal sends the request to the server.

[0949] The server receives the request, filters it, and because it contains no profanity, the request is kept as is.

[0950] The server generates cute animal stories using generative AI.

[0951] The generated stories are then filtered again to ensure they are free of inappropriate language.

[0952] The server sends text of cute animal stories to the device.

[0953] The terminal displays the received text on its screen so that the user can read it.

[0954] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[0955] The processing flow will be explained below.

[0956] Server-side processing

[0957] Step 1:

[0958] The server receives the request data sent from the terminal.

[0959] Specific behavior: Listens for and receives HTTP requests, extracting user-entered text and generated request content from the request body.

[0960] Step 2:

[0961] The server uses filtering means to detect inappropriate words or phrases in the received request data.

[0962] What it does: Scans request data using regular expressions and profanity lists, and fixes or replaces any profanity it finds.

[0963] Step 3:

[0964] The server generates text using the filtered request data with the generation means.

[0965] What it does: Passes the filtered request to a generative AI model (e.g., GPT-3) and makes an API request to generate text that meets the request.

[0966] Step 4:

[0967] The server again applies filtering measures to the generated text to detect inappropriate words and phrases.

[0968] What it does: Scan the generated text again against the regular expressions and profanity list, making corrections or substitutions if necessary.

[0969] Step 5:

[0970] The server sends the filtered generated text to the terminal as an HTTP response.

[0971] Specific behavior: Constructs an HTTP response and sends the generated text to the terminal in the response body.

[0972] Processing on the device (kids' phone)

[0973] Step 1:

[0974] The terminal receives input from the user and sends the content to the server as a request.

[0975] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[0976] Step 2:

[0977] The terminal waits for and receives a response from the server.

[0978] Specific operation: Asynchronously waits for an HTTP response from the server, and retrieves the content of the response when it arrives.

[0979] Step 3:

[0980] The terminal displays the received generated text on the screen.

[0981] Specific operation: The generated text received is reflected in the text display area, making it viewable by the user.

[0982] User (child) operations

[0983] Step 1:

[0984] The user inputs the text content that he or she wants to generate into the terminal.

[0985] Example: Using a text field or voice input, type "Draw me a funny picture."

[0986] Step 2:

[0987] The user checks the generated text and takes the following action:

[0988] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[0989] Example 1

[0990] 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."

[0991] With conventional generative AI systems, the generated text may contain inappropriate language, making it difficult to provide an environment that children can use with confidence. Specifically, if inappropriate words or phrases are included in the request data or generated text, they are output as is, posing a risk that children may come into contact with harmful information. There is a need to solve these issues and provide a generative AI system that children can use with confidence.

[0992] 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.

[0993] In this invention, the server includes: [means for receiving request data from the information processing device by a receiving means; [means for detecting inappropriate words and standard phrases in the request data by a filtering means and correcting or replacing them;] [means for creating text to be generated based on the filtered request data by a generating means; [means for re-detecting inappropriate words and standard phrases in the generated text by a re-filtering means and correcting or replacing them; and [means for transmitting the filtered generated text to the information processing device by a transmitting means.] This enables a generative AI system that can be used safely by children by doubly filtering inappropriate expressions in both the request data and the generated text.

[0994] The "receiving means" is a means for receiving request data from an information processing device.

[0995] "Filtering means" refers to means for detecting inappropriate words or boilerplate phrases in request data or generated text and correcting or replacing them.

[0996] The "generator" is a means for generating text based on the filtered request data.

[0997] The "re-filtering means" is a means for detecting inappropriate words or boilerplate phrases again in the generated text and correcting or replacing them.

[0998] The "transmitting means" is a means for transmitting the filtered generated text to the information processing device.

[0999] An "information processing device" is a device for inputting request data from a user and transmitting it to a server.

[1000] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[1001] Server-side processing

[1002] The server receives request data sent from the device. This request data includes the text entered by the user and the content to be generated. For example, if the user enters "tell me an interesting story," the device sends the request to the server.

[1003] Next, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. The filtering mechanism scans the request data using regular expressions or a predefined list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," if "scary" is determined to be inappropriate, the request is modified to "Tell me a funny story."

[1004] Based on the filtered request data, the server generates text using a generation means. For example, a generative AI model (such as GPT-3) is used as the generation means. The generation means takes the filtered request data as input and generates appropriate text.

[1005] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1006] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[1007] Processing on the terminal (information processing device) side

[1008] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[1009] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[1010] User (child) operations

[1011] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[1012] Specific examples

[1013] 1. First time request

[1014] User: Type "tell me a scary story" into the terminal.

[1015] Terminal: Sends an HTTP request to the server.

[1016] Server: Receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[1017] Server: Generate text using a generative AI model using the modified request data.

[1018] Server: Re-filter the generated text to ensure it is free of profanity.

[1019] Server: Sends the filtered generated text to the terminal as an HTTP response.

[1020] Terminal: Displays the received text on the screen, displaying the story "Once upon a time..." for the user to read.

[1021] 2. Harmless Requests

[1022] User: Type "Tell me a cute animal story" into the terminal.

[1023] Terminal: Sends an HTTP request to the server.

[1024] Server: Receives the request and filters it. Since it contains no profanity, the request is kept as is.

[1025] Server: Generate text using a generative AI model using the filtered request data.

[1026] Server: Re-filter the generated text to ensure it is free of profanity.

[1027] Server: Sends the filtered generated text to the terminal as an HTTP response.

[1028] Terminal: Displays the received text on the screen. It displays "cute animal stories" for the user to read.

[1029] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[1030] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1031] Step 1: Receiving request data

[1032] The server receives request data sent from the device. Input is text entered by the user into the device or content they want to generate. For example, the user might enter "Tell me an interesting story." The device sends this input content to the server as an HTTP request. The server receives the HTTP request and obtains the request data.

[1033] Step 2: Filtering the request data

[1034] The server scans the received request data using a filtering method. The input is the received request data. The filtering method uses regular expressions and a predefined list of inappropriate words. Specifically, the server detects inappropriate words and fixed phrases in the request data. For example, in the case of a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate and the request is corrected to "Tell me a funny story." The output is the filtered request data.

[1035] Step 3: Generate text

[1036] The server generates text using a generative method based on the filtered request data. The input is the filtered request data ("Tell me a funny story"). The server passes this input to a generative AI model (e.g., GPT-3) as a prompt. The generative AI model generates appropriate text corresponding to the input prompt, e.g., "Once upon a time...". The output is the generated text.

[1037] Step 4: Re-filtering the generated text

[1038] The server scans the generated text again using the filtering method. The input is the generated text. Specifically, the server detects inappropriate words or boilerplate phrases in the generated text and corrects or replaces them. For example, if the generated text contains the word "scary," it replaces it with "funny." The output is the re-filtered generated text.

[1039] Step 5: Sending generated text

[1040] The server sends the re-filtered generated text to the terminal using a transmission means. The input is the re-filtered generated text. In specific operations, the server constructs an HTTP response and includes the generated text in the response body. Then, the server sends the HTTP response to the terminal. The terminal displays the received generated text on the screen so that the user can read it. The output is the generated text displayed on the terminal.

[1041] (Application example 1)

[1042] 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."

[1043] Current generative AI systems can sometimes be difficult to guarantee safety so that children can use them with confidence. Measures are needed to properly filter requests and generated text that contain inappropriate words or phrases, but specific methods for achieving this have not yet been established. Furthermore, there is a lack of interfaces that allow children to enjoy the generated text with peace of mind. A solution to these issues is needed.

[1044] 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.

[1045] In this invention, the server includes means for receiving request data from the terminal using a receiving means, means for detecting and correcting or replacing inappropriate words and phrases in the request data using a filtering means, means for creating text to be generated based on the request filtered by the generative AI model, means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text using a re-filtering means, means for transmitting the filtered generated text to the terminal using a transmitting means, and means for providing an interface that allows a user to check and enjoy the re-filtered generated text. This realizes a generative AI system that children can use safely, allowing children to safely enjoy generated text that does not contain inappropriate words or phrases.

[1046] The "receiving means" is a means for receiving request data from a terminal.

[1047] "Filtering means" means means for detecting and modifying or replacing inappropriate words or phrases in request data.

[1048] A "generative AI model" is an artificial intelligence model for generating appropriate text based on filtered request data.

[1049] A "re-filtering means" is a means for re-detecting and correcting or replacing inappropriate words or phrases in the generated text.

[1050] The "transmitting means" is a means for transmitting the filtered generated text to the terminal.

[1051] An "interface" is a display means and operating means that allows a user to view and enjoy the re-filtered generated text.

[1052] The embodiment of this invention provides a generative AI system that children can use safely. The system mainly consists of a server, a terminal, and a user.

[1053] System configuration

[1054] Server-side processing

[1055] The server comprises a "receiving means," a "filtering means," a "generative AI model," a "re-filtering means," and a "transmitting means."

[1056] 1. Receiving means: Receives request data sent from the terminal. The request data includes text entered by the user and the content to be generated.

[1057] 2. Filtering methods: Detect inappropriate words or phrases in the request data and modify or replace them. Use regular expressions or profanity lists.

[1058] 3. Generative AI model: Generate appropriate text based on the filtered request data. Here, we use a generative AI model such as OpenAI's GPT-3.

[1059] 4. Re-filtering: The generated text is checked again for inappropriate language and corrected or replaced as necessary.

[1060] 5. Transmission method: Send the filtered generated text to the terminal. Construct an HTTP response and send the generated text in the response body.

[1061] Terminal side processing

[1062] The terminal receives input from the user, sends the content to the server as a request, and displays the generated text on the screen.

[1063] Example

[1064] User (child) operations

[1065] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "tell me a funny story," the request is sent to the server, and the generated funny story is displayed on the terminal.

[1066] Specific examples

[1067] First request:

[1068] The user types "tell me a scary story."

[1069] The terminal sends the request to the server.

[1070] The server uses filtering means to modify "scary" to "interesting."

[1071] A generative AI model generates "interesting stories" and a re-filtering method checks their appropriateness.

[1072] An interesting story text will be sent to your device and displayed.

[1073] Harmless requests:

[1074] A user types, "Tell me about cute animals."

[1075] The terminal sends the request to the server.

[1076] The server filters the request and generates a story using a generative AI model.

[1077] The generated stories are filtered and displayed on the device.

[1078] Prompt Sentence Examples

[1079] Prompt: Tell me a funny story

[1080] This system provides an environment where children can use generative AI safely, and ensures that the generated text is appropriate for the user, allowing them to enjoy it with peace of mind.

[1081] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1082] Step 1:

[1083] The user enters text into the terminal.

[1084] Input: Text entered by the user using a keyboard or touchscreen

[1085] What it does: A user types the text "tell me a funny story" into their device.

[1086] Output: The entered text is saved on the terminal as request data.

[1087] Step 2:

[1088] The terminal transmits the request data to the server.

[1089] Input: Request data stored on the device

[1090] How it works: The device constructs an HTTP request and sends the request data to the server.

[1091] Output: The server receives the request data from the HTTP request.

[1092] Step 3:

[1093] Filters the request data received by the server.

[1094] Input: Received request data (e.g., "Tell me a scary story")

[1095] Data Modification: Filtering methods are used to detect and modify or replace inappropriate words or phrases in request data, such as using regular expressions or profanity lists.

[1096] Action: The request data "Tell me a scary story" is modified to "Tell me a funny story."

[1097] Output: Filtered request data (e.g., "Tell me a funny story").

[1098] Step 4:

[1099] The server generates text based on the filtered request data.

[1100] Input: Filtered request data (e.g., "Tell me a funny story")

[1101] Data calculation: Text generation using a generative AI model. The prompt sentence "Tell me an interesting story" is input into a generative AI (e.g., GPT-3) to generate appropriate text.

[1102] How it works: A generative AI model generates text that begins with "Once upon a time..."

[1103] Output: The generated text (e.g., "Once upon a time...").

[1104] Step 5:

[1105] The server re-filters the generated text.

[1106] Input: Generated text (e.g. "Once upon a time...")

[1107] Data processing: Re-filtering measures are used to detect and correct or replace inappropriate words or phrases in the generated text.

[1108] Action: Checks the generated text for profanity and corrects it if necessary.

[1109] Output: The refiltered generated text (e.g., "Once upon a time...").

[1110] Step 6:

[1111] The server sends the re-filtered generated text to the terminal.

[1112] Input: Re-filtered generated text (e.g., "Once upon a time...")

[1113] Behavior: The server sends the re-filtered generated text to the terminal as an HTTP response and includes it in the response body.

[1114] Output: The device receives the re-filtered generated text from the server.

[1115] Step 7:

[1116] The terminal displays the refiltered generated text to the user.

[1117] Input: Re-filtered generated text received from the server (e.g., "Once upon a time...")

[1118] Action: The terminal displays the resulting refiltered text on the screen, presenting the text in a user-friendly format.

[1119] Output: The generated text displayed on the screen (e.g., "Once upon a time...").

[1120] Step 8:

[1121] The user reviews and enjoys the re-filtered generated text.

[1122] Input: Refiltered generated text displayed in terminal

[1123] Actions: The user reads and enjoys the generated text, and optionally enters new requests.

[1124] Output: User satisfaction and new request data.

[1125] 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.

[1126] This invention combines an emotion engine with a generative AI system that children can use safely to generate appropriate text based on the user's emotions. The system includes a series of processes: receiving request data, recognizing emotions, filtering, generating text, re-filtering, and sending the generated text.

[1127] A natural language explanation of the program's processing

[1128] Server-side processing

[1129] The server first receives request data sent from the terminal. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[1130] Next, the server uses an emotion engine to recognize the user's emotion in the received request data. The emotion engine uses natural language processing technology to extract the user's emotion from the request data. For example, a request such as "Tell me a fun story" would recognize a positive emotion.

[1131] Based on the emotions recognized by the emotion engine, the server uses a filtering means to detect inappropriate words and phrases in the request data. The filtering means adaptively adjusts rules according to the emotions and scans the request data using regular expressions and inappropriate word lists. If an inappropriate expression is detected, it is corrected or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," "scary" is determined to be inappropriate, and the request is corrected to "Tell me a funny story."

[1132] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator generates appropriate text while adjusting the content and tone based on the user's emotions recognized by the emotion engine.

[1133] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1134] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[1135] Processing on the device (kids' phone)

[1136] The terminal receives input from the user and sends the content to the server as a request. For example, if the user inputs "tell me an interesting story," the terminal sends an HTTP request containing the request content to the server.

[1137] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a funny story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[1138] User (child) operations

[1139] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[1140] Specific examples

[1141] 1. Requests that made me happy:

[1142] The user (child) types "Tell me a happy story" into the terminal.

[1143] The terminal sends the request to the server.

[1144] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[1145] The server adjusts rules and filters according to emotions.

[1146] The server generates happy stories using generative AI.

[1147] The generated stories are then filtered again to ensure they are free of inappropriate language.

[1148] The server sends the text of the happy story to the device.

[1149] The terminal displays the received text on its screen for the user to read.

[1150] 2. A request that made me feel uneasy:

[1151] The user (child) types into the terminal, "Tell me a story that will ease my anxiety."

[1152] The terminal sends the request to the server.

[1153] The server receives the request and uses an emotion engine to recognize the emotion "anxiety."

[1154] The server adjusts filtering rules according to the emotion and scans the request.

[1155] The server uses generative AI to generate stories that ease anxiety.

[1156] The generated stories are then refiltered to ensure they are free of inappropriate language.

[1157] The server sends text messages to the device to ease anxiety.

[1158] The terminal displays the received text on its screen for the user to read.

[1159] In this way, by combining emotion engines, it is possible to realize a generative AI system that can adaptively adjust the request content and generated text according to the user's emotions, and that can be used safely by children.

[1160] The processing flow will be explained below.

[1161] Server-side processing

[1162] Step 1:

[1163] The server receives the request data sent from the terminal.

[1164] Specific behavior: Waits for and receives HTTP requests, and extracts the generated request from the request body, such as the text entered by the user, "Tell me an interesting story."

[1165] Step 2:

[1166] The server uses an emotion engine to recognize the user's emotion in the received request data.

[1167] What it does: Uses natural language processing techniques to extract emotion tags (e.g., happy, sad, joyful, anxious) from request data.

[1168] Step 3:

[1169] The server adjusts the rules of the filtering means based on the user's emotions recognized by the emotion engine.

[1170] Specific behavior: If the emotion tag is recognized as "fun," apply the filtering rules according to that emotion.

[1171] Step 4:

[1172] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data.

[1173] Specific behavior: Scans request data using regular expressions and inappropriate word lists, and modifies inappropriate expressions to appropriate ones. For example, "Tell me a scary story" is modified to "Tell me a funny story."

[1174] Step 5:

[1175] The server generates text using the filtered request data with the generation means.

[1176] What it does: A generative AI model (e.g., GPT-3) is fed with the filtered request data and recognized emotion tags, and generates appropriate text based on the request and emotion.

[1177] Step 6:

[1178] The server then applies filtering measures to the generated text to detect and correct or replace inappropriate words or phrases.

[1179] Specific behavior: The generated text is scanned again against regular expressions and profanity lists, and profanity is corrected to proper expressions.

[1180] Step 7:

[1181] The server sends the filtered generated text to the terminal as an HTTP response.

[1182] Specific behavior: Constructs an HTTP response and sends the generated appropriate text to the device in the response body.

[1183] Processing on the device (kids' phone)

[1184] Step 1:

[1185] The terminal receives input from the user and sends the content to the server as a request.

[1186] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[1187] Step 2:

[1188] The terminal waits for and receives a response from the server.

[1189] Specific operation: Asynchronously waits for the HTTP response sent from the server, receives it, and obtains the generated text from the response body.

[1190] Step 3:

[1191] The terminal displays the received generated text on the screen.

[1192] Specific behavior: The generated text received is reflected in the text display area so that the user can read it.

[1193] User (child) operations

[1194] Step 1:

[1195] The user inputs the text content that he or she wants to generate into the terminal.

[1196] Example: Using a text field or voice input, type "Tell me a fun story."

[1197] Step 2:

[1198] The user checks the generated text and takes the following action:

[1199] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[1200] Example 2

[1201] 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."

[1202] Conventional generative AI systems have safety issues due to the risk of inappropriate content being included when used by children. Furthermore, they lack a mechanism for generating appropriate content that adapts to the user's emotions, creating a need for a system that children can use safely. This invention aims to provide a generative AI system that generates text based on the user's emotions and filters out inappropriate content, making it safe for children to use.

[1203] 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.

[1204] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means;] [means for recognizing the user's emotions from the received request data using an emotion engine;] [means for detecting and correcting or replacing inappropriate words and phrases in the request data based on the emotions recognized by a filtering means;] [means for creating generated text in accordance with the user's emotions based on the request data filtered by a generating means;] [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; and [means for transmitting the filtered generated text to the terminal by a transmitting means]. This makes it possible [to realize a generative AI system that generates text that adapts to the user's emotions while eliminating inappropriate content, and is safe for children to use].

[1205] The "receiving means" is a device or function that receives request data sent from a terminal.

[1206] An "emotion engine" is a technology that includes software and algorithms for recognizing and analyzing user emotions from received data.

[1207] A "filtering means" is a device or function that detects and modifies or replaces inappropriate words or phrases in data.

[1208] "Generator" means a device or function for generating appropriate text based on filtered request data, including, for example, a generative AI model.

[1209] A "re-filtering means" is a device or function that re-detects and corrects or replaces inappropriate words or phrases in the generated text.

[1210] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[1211] A "generative AI model" refers to an artificial intelligence model that generates natural language text based on specific input data.

[1212] This invention provides a generative AI system that generates text that is adaptive to the user's emotions while eliminating inappropriate content, making it safe for children to use. This system includes three main components: a server, a terminal, and a user.

[1213] Server-side processing

[1214] Receiving means

[1215] The server first receives request data sent from the terminal using the receiving means. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[1216] Emotion Engine

[1217] The server uses an emotion engine to recognize the user's emotion from the received request data. The emotion engine uses natural language processing technologies such as VADER and TextBlob. These technologies extract emotions such as positive, negative, and neutral from the request data. For example, a request such as "Tell me a fun story" will recognize a positive emotion.

[1218] Filtering Methods

[1219] Based on the emotions recognized by the emotion engine, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. This filtering mechanism scans the request data using regular expressions and inappropriate word lists, and if an inappropriate expression is found, it modifies or replaces it with an appropriate expression. For example, if a request such as "Tell me a scary story" is determined to be "scary," it will be modified to "Tell me a funny story."

[1220] generation means

[1221] After filtering, the server generates text using a generative method based on the request data. This method uses a generative AI model (such as GPT-3). The generative AI model generates appropriate text by adjusting the content and tone based on the user's emotions recognized by the emotion engine. For example, the generated text for "Tell me a fun story" is a fun story that begins with "Once upon a time..."

[1222] Re-filtering means

[1223] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1224] Transmission method

[1225] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated text in the response body, and sends it to the terminal. In this way, the generated text is delivered to the user.

[1226] Terminal side processing

[1227] Receiving User Input

[1228] The terminal receives input from the user and prepares it to be sent as a request to the server, for example, "Tell me a funny story."

[1229] Sending request data

[1230] The terminal includes the request content entered by the user in an HTTP request and sends it to the server. After the server returns a response, the terminal receives the response.

[1231] Viewing generated text

[1232] The generated text is then displayed on the screen, allowing the user to read it. For example, a funny story like "Once upon a time..." is displayed on the screen.

[1233] User operation

[1234] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[1235] Specific examples

[1236] 1. User Request: "Tell me a happy story"

[1237] The terminal sends the request to the server.

[1238] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[1239] The server uses filtering measures to detect and correct inappropriate content in requests.

[1240] The server uses a generative AI model to generate happy stories.

[1241] The generated stories are then filtered again to ensure they are free of inappropriate content.

[1242] The server sends the text of the happy story to the device.

[1243] The terminal displays the received text on its screen for the user to read.

[1244] In this way, the system achieves safe text generation while adapting to the user's emotions.

[1245] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1246] Step 1:

[1247] Receiving User Input

[1248] The user inputs the text they want to generate into the device. For example, they input "Tell me a fun story." The input request is saved on the device, and an HTTP request is prepared. The input in this case is "Tell me a fun story." The output is the request data saved on the device.

[1249] Step 2:

[1250] Sending request data

[1251] The device constructs an HTTP request that includes the user's input and sends it to the server. For example, a request containing the content "Tell me a fun story" is sent to the server. The input at this time is the request data saved on the device. The output is the request data sent to the server.

[1252] Step 3:

[1253] Conducting sentiment analysis

[1254] The server passes the received request data to the emotion engine, which recognizes the user's emotion. The emotion engine uses natural language processing technology to extract emotion from the request data. For example, it can recognize positive emotion from a request such as "Tell me a fun story." The input in this case is the request data "Tell me a fun story," and the output is emotion data of "positive."

[1255] Step 4:

[1256] Filtering Request Data

[1257] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data based on the emotions recognized by the emotion engine. For example, a request such as "Tell me a scary story" is determined to be "scary" and modified to "Tell me a funny story." The input in this case is emotion data of "positive" and request data of "Tell me a fun story," and the output is the filtered request data.

[1258] Step 5:

[1259] Text Generation

[1260] The server generates appropriate text using a generative AI model (e.g., GPT-3) based on the filtered request data. For example, in response to the request "Tell me a fun story," a fun story such as "Once upon a time..." is generated. The input in this case is the filtered request data, and the output is the generated text.

[1261] Step 6:

[1262] Re-filtering the generated text

[1263] The server then applies the filtering measures to the generated text again. For example, the generated text "Once upon a time..." is filtered again to check for inappropriate language, correcting or replacing it as necessary. The input is the generated text, and the output is the re-filtered generated text.

[1264] Step 7:

[1265] Sending generated text

[1266] Finally, the server constructs the filtered generated text as an HTTP response and sends it to the terminal. The server sends the generated text in the response body to the terminal. The input at this time is the re-filtered generated text, and the output is the response data sent to the terminal.

[1267] Step 8:

[1268] Viewing generated text

[1269] The terminal receives the response from the server and displays the generated text on the screen. For example, a fun story like "Once upon a time..." is displayed on the screen. The input is the response data received by the terminal, and the output is the generated text displayed on the screen.

[1270] The above is the specific processing flow of this system.

[1271] (Application example 2)

[1272] 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."

[1273] Current generative AI systems lack appropriate filtering and emotion-based adjustment to ensure safe use by children. This means that generated content may contain inappropriate content, and it is difficult to provide content that quickly and accurately reflects the user's emotions. There is a need for a system that can solve this issue and realize a content distribution service that children can enjoy safely.

[1274] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1275] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means; [means for recognizing the user's emotions in the request data by an emotion recognition means; [means for detecting and correcting or replacing inappropriate words and phrases in the request data by a filtering means; [means for creating generated text according to the emotions based on the request filtered by the generating means; [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; [means for transmitting the filtered generated text to the terminal by a transmitting means; and [means for displaying the generated text on the terminal.] This makes it possible to provide an emotion-based generative AI system that children can use safely.

[1276] The "receiving means" is a device or function that receives request data from a terminal.

[1277] An "emotion recognition means" is a device or algorithm for recognizing a user's emotion in request data.

[1278] A "filtering means" is a device or function that detects inappropriate words or phrases in request data and modifies or replaces them.

[1279] A "generator" is a device or algorithm that creates emotion-sensitive generated text based on filtered request data.

[1280] A "re-filtering means" is a device or function that re-detects inappropriate words or phrases in the generated text and corrects or replaces them.

[1281] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[1282] A "terminal" is an electronic device that allows a user to input request data and display the generated text.

[1283] This invention relates to the implementation of the smartphone application "Emotory" as an example of an application of a content distribution service that can be used safely by children. "Emotory" is an application that generates stories based on emotions, and provides content appropriate to the user through a series of processes.

[1284] First, the user (child) inputs their feelings and interests into their smartphone. The device then sends this input to the server as request data. For example, the user might input "Tell me a fun story." The server then processes the data as follows:

[1285] Server-side processing

[1286] 1. Receiving request data: The server receives the request data sent from the terminal. This is realized by the "receiving means."

[1287] 2. Emotion Recognition: To recognize the user's emotions from the received request data, we use an "emotion recognition method." This method uses natural language processing technology, such as NLP libraries like "spaCy" and "NLTK."

[1288] 3. Request Data Filtering: Using "filtering methods" to detect and modify or replace inappropriate words and phrases based on recognized sentiment. Filtering methods may use regular expressions or predefined lists of inappropriate words.

[1289] 4. Generating generated text: A "generator" is used to generate sentiment-based generated text based on the filtered request data. This generator uses a large-scale generative AI model, such as OpenAI's "GPT-3."

[1290] 5. Re-filtering the generated text: A "re-filtering method" is used to again detect and correct or replace inappropriate words or phrases in the generated text.

[1291] 6. Sending the generated text: Finally, the filtered generated text is sent to the terminal by the "sending means."

[1292] Terminal side processing

[1293] The terminal receives the generated text sent from the server and has the function of "displaying the generated text on the terminal" in a format that is easy for the user to read. The user (child) can enjoy reading this generated text.

[1294] Specific hardware / software usage

[1295] NLP library: "spaCy" and "NLTK" are used for emotion recognition.

[1296] Generative AI model: OpenAI's "GPT-3" is used for text generation.

[1297] Filtering libraries: Use your own implementation or third-party libraries that use regular expressions and profanity lists.

[1298] Examples of concrete examples and prompts

[1299] Examples:

[1300] 1. If you feel happy:

[1301] User input: Child types "Tell me a happy story."

[1302] Server processing: After emotion recognition, filtering, generation, and re-filtering, a positive story is generated: "Once upon a time, there was a very happy village..."

[1303] Terminal display: "Once upon a time, there was a very happy village..."

[1304] 2. To relieve anxiety:

[1305] User input: Child types, "Tell me a story to calm my anxiety."

[1306] Server processing: After emotion recognition, filtering, generation, and re-filtering, a soothing story is generated: "Listen to this story. In a very peaceful forest..."

[1307] Terminal display: "Listen to this story. In a very peaceful forest..."

[1308] Example prompt sentence:

[1309] A child types "Tell me a fun story" into their smartphone.

[1310] As a result, "Emotory" is an application that allows children to spend time safely and enjoyably, and can provide appropriate content according to the user's emotions.

[1311] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1312] Step 1:

[1313] User Input

[1314] The user (child) inputs the text they want to generate into their smartphone. For example, they might input "Tell me a fun story." The device then sends this input to the server as request data.

[1315] Input: Text entered by the user (e.g., "Tell me a nice story")

[1316] Output: Request data sent to the server

[1317] Step 2:

[1318] Receiving request data

[1319] The server receives the request data sent from the terminal, and provides the server with information about the content the user is looking for.

[1320] Input: Request data

[1321] Output: Request data stored on the server

[1322] Step 3:

[1323] Emotion recognition

[1324] The server recognizes the user's emotion from the received request data. For emotion recognition, it uses an NLP library (e.g., "spaCy" or "NLTK"). For example, if the request is "Tell me a fun story," the user's emotion is recognized as "positive."

[1325] Input: Request data

[1326] Output: Perceived emotion (e.g., "positive")

[1327] Step 4:

[1328] Filtering Request Data

[1329] The server detects and modifies or replaces inappropriate words and phrases in the request data based on sentiment. Regular expressions and inappropriate word lists are used as filtering methods. For example, if a request is made to "tell me a scary story," the word "scary" is deemed "inappropriate" and the request is modified to "tell me a funny story."

[1330] Input: Request data, recognized emotion

[1331] Output: Filtered request data

[1332] Step 5:

[1333] Creating generated text

[1334] The server generates text based on the filtered request data, according to the sentiment. A large-scale generative AI model (e.g., GPT-3) is used as the generation method. For example, if the filtered data is "Tell me a happy story" and the sentiment is "positive," the generated text would be something like "Once upon a time, there was a very happy village..."

[1335] Input: filtered request data, recognized sentiment

[1336] Output: The generated text

[1337] Step 6:

[1338] Re-filtering the generated text

[1339] The server again detects and corrects or replaces inappropriate words or phrases in the generated text. For example, if the generated text contains an "inappropriate word," it replaces it with an appropriate expression. This process is also performed using filtering methods.

[1340] Input: Generated text

[1341] Output: Refiltered generated text

[1342] Step 7:

[1343] Sending generated text

[1344] The server sends the filtered generated text to the terminal using an HTTP response as a transmission means, and includes the text in the response body.

[1345] Input: Refiltered generated text

[1346] Output: The generated text sent to the terminal

[1347] Step 8:

[1348] Viewing generated text

[1349] The terminal receives the generated text sent from the server and displays it on the screen, for example, the generated text "Once upon a time, there was a very pleasant village..." is displayed on the screen for the user to read.

[1350] Input: Generated text sent from the server

[1351] Output: The generated text displayed on the terminal.

[1352] 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.

[1353] 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.

[1354] 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.

[1355] [Fourth embodiment]

[1356] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1357] 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.

[1358] 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).

[1359] 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.

[1360] 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.

[1361] 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).

[1362] 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.

[1363] 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.

[1364] 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.

[1365] 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.

[1366] 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.

[1367] 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.

[1368] 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."

[1369] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[1370] A natural language explanation of the program's processing

[1371] Server-side processing

[1372] The server first receives request data sent from the device. This request data includes the text entered by the user and the content they want to generate. For example, if the user enters "Tell me an interesting story," the device sends the request to the server.

[1373] Next, the server uses a filtering mechanism to detect inappropriate words or phrases in the request data. The filtering mechanism scans the request data using regular expressions or a list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate, and the request is modified to "Tell me a funny story."

[1374] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator takes the filtered request data as input and generates appropriate text.

[1375] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1376] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[1377] Processing on the device (kids' phone)

[1378] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[1379] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[1380] User (child) operations

[1381] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[1382] Specific examples

[1383] 1. First request:

[1384] The user (child) types "tell me a scary story" into the terminal.

[1385] The terminal sends the request to the server.

[1386] The server receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[1387] The server generates interesting stories using generative AI.

[1388] The generated stories are then filtered again to ensure they are free of inappropriate language.

[1389] The server sends the text of an interesting story to the device.

[1390] The terminal displays the received text on its screen so that the user can read it.

[1391] 2. Harmless Requests:

[1392] The user (child) types "Tell me a story about a cute animal" into the terminal.

[1393] The terminal sends the request to the server.

[1394] The server receives the request, filters it, and because it contains no profanity, the request is kept as is.

[1395] The server generates cute animal stories using generative AI.

[1396] The generated stories are then filtered again to ensure they are free of inappropriate language.

[1397] The server sends text of cute animal stories to the device.

[1398] The terminal displays the received text on its screen so that the user can read it.

[1399] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[1400] The processing flow will be explained below.

[1401] Server-side processing

[1402] Step 1:

[1403] The server receives the request data sent from the terminal.

[1404] Specific behavior: Listens for and receives HTTP requests, extracting user-entered text and generated request content from the request body.

[1405] Step 2:

[1406] The server uses filtering means to detect inappropriate words or phrases in the received request data.

[1407] What it does: Scans request data using regular expressions and profanity lists, and fixes or replaces any profanity it finds.

[1408] Step 3:

[1409] The server generates text using the filtered request data with the generation means.

[1410] What it does: Passes the filtered request to a generative AI model (e.g., GPT-3) and makes an API request to generate text that meets the request.

[1411] Step 4:

[1412] The server again applies filtering measures to the generated text to detect inappropriate words and phrases.

[1413] What it does: Scan the generated text again against the regular expressions and profanity list, making corrections or substitutions if necessary.

[1414] Step 5:

[1415] The server sends the filtered generated text to the terminal as an HTTP response.

[1416] Specific behavior: Constructs an HTTP response and sends the generated text to the terminal in the response body.

[1417] Processing on the device (kids' phone)

[1418] Step 1:

[1419] The terminal receives input from the user and sends the content to the server as a request.

[1420] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[1421] Step 2:

[1422] The terminal waits for and receives a response from the server.

[1423] Specific operation: Asynchronously waits for an HTTP response from the server, and retrieves the content of the response when it arrives.

[1424] Step 3:

[1425] The terminal displays the received generated text on the screen.

[1426] Specific operation: The generated text received is reflected in the text display area, making it viewable by the user.

[1427] User (child) operations

[1428] Step 1:

[1429] The user inputs the text content that he or she wants to generate into the terminal.

[1430] Example: Using a text field or voice input, type "Draw me a funny picture."

[1431] Step 2:

[1432] The user checks the generated text and takes the following action:

[1433] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[1434] Example 1

[1435] 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."

[1436] With conventional generative AI systems, the generated text may contain inappropriate language, making it difficult to provide an environment that children can use with confidence. Specifically, if inappropriate words or phrases are included in the request data or generated text, they are output as is, posing a risk that children may come into contact with harmful information. There is a need to solve these issues and provide a generative AI system that children can use with confidence.

[1437] 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.

[1438] In this invention, the server includes: [means for receiving request data from the information processing device by a receiving means; [means for detecting inappropriate words and standard phrases in the request data by a filtering means and correcting or replacing them;] [means for creating text to be generated based on the filtered request data by a generating means; [means for re-detecting inappropriate words and standard phrases in the generated text by a re-filtering means and correcting or replacing them; and [means for transmitting the filtered generated text to the information processing device by a transmitting means.] This enables a generative AI system that can be used safely by children by doubly filtering inappropriate expressions in both the request data and the generated text.

[1439] The "receiving means" is a means for receiving request data from an information processing device.

[1440] "Filtering means" refers to means for detecting inappropriate words or boilerplate phrases in request data or generated text and correcting or replacing them.

[1441] The "generator" is a means for generating text based on the filtered request data.

[1442] The "re-filtering means" is a means for detecting inappropriate words or boilerplate phrases again in the generated text and correcting or replacing them.

[1443] The "transmitting means" is a means for transmitting the filtered generated text to the information processing device.

[1444] An "information processing device" is a device for inputting request data from a user and transmitting it to a server.

[1445] This invention aims to provide a generative AI system that can be used safely by children. The system includes a series of processes: receiving request data, filtering, generating text, re-filtering, and sending the generated text.

[1446] Server-side processing

[1447] The server receives request data sent from the device. This request data includes the text entered by the user and the content to be generated. For example, if the user enters "tell me an interesting story," the device sends the request to the server.

[1448] Next, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. The filtering mechanism scans the request data using regular expressions or a predefined list of inappropriate words. If an inappropriate expression is detected, it is modified or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," if "scary" is determined to be inappropriate, the request is modified to "Tell me a funny story."

[1449] Based on the filtered request data, the server generates text using a generation means. For example, a generative AI model (such as GPT-3) is used as the generation means. The generation means takes the filtered request data as input and generates appropriate text.

[1450] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1451] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[1452] Processing on the terminal (information processing device) side

[1453] The terminal receives input from the user and sends the content of the input to the server as a request. For example, if the user inputs "Tell me a fun story," the terminal sends an HTTP request including the request content to the server.

[1454] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a fun story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[1455] User (child) operations

[1456] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "draw a funny picture," the request is sent to the server, and the generated picture's explanation and instructions are displayed on the terminal. The user can check and enjoy it.

[1457] Specific examples

[1458] 1. First time request

[1459] User: Type "tell me a scary story" into the terminal.

[1460] Terminal: Sends an HTTP request to the server.

[1461] Server: Receives the request and filters it. "Scary" is determined to be inappropriate and is corrected to "Tell me a funny story."

[1462] Server: Generate text using a generative AI model using the modified request data.

[1463] Server: Re-filter the generated text to ensure it is free of profanity.

[1464] Server: Sends the filtered generated text to the terminal as an HTTP response.

[1465] Terminal: Displays the received text on the screen, displaying the story "Once upon a time..." for the user to read.

[1466] 2. Harmless Requests

[1467] User: Type "Tell me a cute animal story" into the terminal.

[1468] Terminal: Sends an HTTP request to the server.

[1469] Server: Receives the request and filters it. Since it contains no profanity, the request is kept as is.

[1470] Server: Generate text using a generative AI model using the filtered request data.

[1471] Server: Re-filter the generated text to ensure it is free of profanity.

[1472] Server: Sends the filtered generated text to the terminal as an HTTP response.

[1473] Terminal: Displays the received text on the screen. It displays "cute animal stories" for the user to read.

[1474] The above is a concrete example of how to implement the invention. This system provides an environment where children can use generative AI with peace of mind.

[1475] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1476] Step 1: Receiving request data

[1477] The server receives request data sent from the device. Input is text entered by the user into the device or content they want to generate. For example, the user might enter "Tell me an interesting story." The device sends this input content to the server as an HTTP request. The server receives the HTTP request and obtains the request data.

[1478] Step 2: Filtering the request data

[1479] The server scans the received request data using a filtering method. The input is the received request data. The filtering method uses regular expressions and a predefined list of inappropriate words. Specifically, the server detects inappropriate words and fixed phrases in the request data. For example, in the case of a request such as "Tell me a scary story," the word "scary" is determined to be inappropriate and the request is corrected to "Tell me a funny story." The output is the filtered request data.

[1480] Step 3: Generate text

[1481] The server generates text using a generative method based on the filtered request data. The input is the filtered request data ("Tell me a funny story"). The server passes this input to a generative AI model (e.g., GPT-3) as a prompt. The generative AI model generates appropriate text corresponding to the input prompt, e.g., "Once upon a time...". The output is the generated text.

[1482] Step 4: Re-filtering the generated text

[1483] The server scans the generated text again using the filtering method. The input is the generated text. Specifically, the server detects inappropriate words or boilerplate phrases in the generated text and corrects or replaces them. For example, if the generated text contains the word "scary," it replaces it with "funny." The output is the re-filtered generated text.

[1484] Step 5: Sending generated text

[1485] The server sends the re-filtered generated text to the terminal using a transmission means. The input is the re-filtered generated text. In specific operations, the server constructs an HTTP response and includes the generated text in the response body. Then, the server sends the HTTP response to the terminal. The terminal displays the received generated text on the screen so that the user can read it. The output is the generated text displayed on the terminal.

[1486] (Application example 1)

[1487] 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."

[1488] Current generative AI systems can sometimes be difficult to guarantee safety so that children can use them with confidence. Measures are needed to properly filter requests and generated text that contain inappropriate words or phrases, but specific methods for achieving this have not yet been established. Furthermore, there is a lack of interfaces that allow children to enjoy the generated text with peace of mind. A solution to these issues is needed.

[1489] 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.

[1490] In this invention, the server includes means for receiving request data from the terminal using a receiving means, means for detecting and correcting or replacing inappropriate words and phrases in the request data using a filtering means, means for creating text to be generated based on the request filtered by the generative AI model, means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text using a re-filtering means, means for transmitting the filtered generated text to the terminal using a transmitting means, and means for providing an interface that allows a user to check and enjoy the re-filtered generated text. This realizes a generative AI system that children can use safely, allowing children to safely enjoy generated text that does not contain inappropriate words or phrases.

[1491] The "receiving means" is a means for receiving request data from a terminal.

[1492] "Filtering means" means means for detecting and modifying or replacing inappropriate words or phrases in request data.

[1493] A "generative AI model" is an artificial intelligence model for generating appropriate text based on filtered request data.

[1494] A "re-filtering means" is a means for re-detecting and correcting or replacing inappropriate words or phrases in the generated text.

[1495] The "transmitting means" is a means for transmitting the filtered generated text to the terminal.

[1496] An "interface" is a display means and operating means that allows a user to view and enjoy the re-filtered generated text.

[1497] The embodiment of this invention provides a generative AI system that children can use safely. The system mainly consists of a server, a terminal, and a user.

[1498] System configuration

[1499] Server-side processing

[1500] The server comprises a "receiving means," a "filtering means," a "generative AI model," a "re-filtering means," and a "transmitting means."

[1501] 1. Receiving means: Receives request data sent from the terminal. The request data includes text entered by the user and the content to be generated.

[1502] 2. Filtering methods: Detect inappropriate words or phrases in the request data and modify or replace them. Use regular expressions or profanity lists.

[1503] 3. Generative AI model: Generate appropriate text based on the filtered request data. Here, we use a generative AI model such as OpenAI's GPT-3.

[1504] 4. Re-filtering: The generated text is checked again for inappropriate language and corrected or replaced as necessary.

[1505] 5. Transmission method: Send the filtered generated text to the terminal. Construct an HTTP response and send the generated text in the response body.

[1506] Terminal side processing

[1507] The terminal receives input from the user, sends the content to the server as a request, and displays the generated text on the screen.

[1508] Example

[1509] User (child) operations

[1510] The user inputs text into the terminal and checks the generated text. For example, if the user inputs "tell me a funny story," the request is sent to the server, and the generated funny story is displayed on the terminal.

[1511] Specific examples

[1512] First request:

[1513] The user types "tell me a scary story."

[1514] The terminal sends the request to the server.

[1515] The server uses filtering means to modify "scary" to "interesting."

[1516] A generative AI model generates "interesting stories" and a re-filtering method checks their appropriateness.

[1517] An interesting story text will be sent to your device and displayed.

[1518] Harmless requests:

[1519] A user types, "Tell me about cute animals."

[1520] The terminal sends the request to the server.

[1521] The server filters the request and generates a story using a generative AI model.

[1522] The generated stories are filtered and displayed on the device.

[1523] Prompt Sentence Examples

[1524] Prompt: Tell me a funny story

[1525] This system provides an environment where children can use generative AI safely, and ensures that the generated text is appropriate for the user, allowing them to enjoy it with peace of mind.

[1526] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1527] Step 1:

[1528] The user enters text into the terminal.

[1529] Input: Text entered by the user using a keyboard or touchscreen

[1530] What it does: A user types the text "tell me a funny story" into their device.

[1531] Output: The entered text is saved on the terminal as request data.

[1532] Step 2:

[1533] The terminal transmits the request data to the server.

[1534] Input: Request data stored on the device

[1535] How it works: The device constructs an HTTP request and sends the request data to the server.

[1536] Output: The server receives the request data from the HTTP request.

[1537] Step 3:

[1538] Filters the request data received by the server.

[1539] Input: Received request data (e.g., "Tell me a scary story")

[1540] Data Modification: Filtering methods are used to detect and modify or replace inappropriate words or phrases in request data, such as using regular expressions or profanity lists.

[1541] Action: The request data "Tell me a scary story" is modified to "Tell me a funny story."

[1542] Output: Filtered request data (e.g., "Tell me a funny story").

[1543] Step 4:

[1544] The server generates text based on the filtered request data.

[1545] Input: Filtered request data (e.g., "Tell me a funny story")

[1546] Data calculation: Text generation using a generative AI model. The prompt sentence "Tell me an interesting story" is input into a generative AI (e.g., GPT-3) to generate appropriate text.

[1547] How it works: A generative AI model generates text that begins with "Once upon a time..."

[1548] Output: The generated text (e.g., "Once upon a time...").

[1549] Step 5:

[1550] The server re-filters the generated text.

[1551] Input: Generated text (e.g. "Once upon a time...")

[1552] Data processing: Re-filtering measures are used to detect and correct or replace inappropriate words or phrases in the generated text.

[1553] Action: Checks the generated text for profanity and corrects it if necessary.

[1554] Output: The refiltered generated text (e.g., "Once upon a time...").

[1555] Step 6:

[1556] The server sends the re-filtered generated text to the terminal.

[1557] Input: Re-filtered generated text (e.g., "Once upon a time...")

[1558] Behavior: The server sends the re-filtered generated text to the terminal as an HTTP response and includes it in the response body.

[1559] Output: The device receives the re-filtered generated text from the server.

[1560] Step 7:

[1561] The terminal displays the refiltered generated text to the user.

[1562] Input: Re-filtered generated text received from the server (e.g., "Once upon a time...")

[1563] Action: The terminal displays the resulting refiltered text on the screen, presenting the text in a user-friendly format.

[1564] Output: The generated text displayed on the screen (e.g., "Once upon a time...").

[1565] Step 8:

[1566] The user reviews and enjoys the re-filtered generated text.

[1567] Input: Refiltered generated text displayed in terminal

[1568] Actions: The user reads and enjoys the generated text, and optionally enters new requests.

[1569] Output: User satisfaction and new request data.

[1570] 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.

[1571] This invention combines an emotion engine with a generative AI system that children can use safely to generate appropriate text based on the user's emotions. The system includes a series of processes: receiving request data, recognizing emotions, filtering, generating text, re-filtering, and sending the generated text.

[1572] A natural language explanation of the program's processing

[1573] Server-side processing

[1574] The server first receives request data sent from the terminal. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[1575] Next, the server uses an emotion engine to recognize the user's emotion in the received request data. The emotion engine uses natural language processing technology to extract the user's emotion from the request data. For example, a request such as "Tell me a fun story" would recognize a positive emotion.

[1576] Based on the emotions recognized by the emotion engine, the server uses a filtering means to detect inappropriate words and phrases in the request data. The filtering means adaptively adjusts rules according to the emotions and scans the request data using regular expressions and inappropriate word lists. If an inappropriate expression is detected, it is corrected or replaced with an appropriate expression. For example, in response to a request such as "Tell me a scary story," "scary" is determined to be inappropriate, and the request is corrected to "Tell me a funny story."

[1577] Based on the filtered request data, the server generates text using a generator. For example, a generative AI model (such as GPT-3) is used as the generator. The generator generates appropriate text while adjusting the content and tone based on the user's emotions recognized by the emotion engine.

[1578] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1579] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated appropriate text in the response body, and sends it to the terminal.

[1580] Processing on the device (kids' phone)

[1581] The terminal receives input from the user and sends the content to the server as a request. For example, if the user inputs "tell me an interesting story," the terminal sends an HTTP request containing the request content to the server.

[1582] The terminal that receives the response from the server displays the generated text on the screen. For example, if the text is a funny story such as "Once upon a time...", it will be displayed on the screen so that the user can read it.

[1583] User (child) operations

[1584] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[1585] Specific examples

[1586] 1. Requests that made me happy:

[1587] The user (child) types "Tell me a happy story" into the terminal.

[1588] The terminal sends the request to the server.

[1589] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[1590] The server adjusts rules and filters according to emotions.

[1591] The server generates happy stories using generative AI.

[1592] The generated stories are then filtered again to ensure they are free of inappropriate language.

[1593] The server sends the text of the happy story to the device.

[1594] The terminal displays the received text on its screen for the user to read.

[1595] 2. A request that made me feel uneasy:

[1596] The user (child) types into the terminal, "Tell me a story that will ease my anxiety."

[1597] The terminal sends the request to the server.

[1598] The server receives the request and uses an emotion engine to recognize the emotion "anxiety."

[1599] The server adjusts filtering rules according to the emotion and scans the request.

[1600] The server uses generative AI to generate stories that ease anxiety.

[1601] The generated stories are then refiltered to ensure they are free of inappropriate language.

[1602] The server sends text messages to the device to ease anxiety.

[1603] The terminal displays the received text on its screen for the user to read.

[1604] In this way, by combining emotion engines, it is possible to realize a generative AI system that can adaptively adjust the request content and generated text according to the user's emotions, and that can be used safely by children.

[1605] The processing flow will be explained below.

[1606] Server-side processing

[1607] Step 1:

[1608] The server receives the request data sent from the terminal.

[1609] Specific behavior: Waits for and receives HTTP requests, and extracts the generated request from the request body, such as the text entered by the user, "Tell me an interesting story."

[1610] Step 2:

[1611] The server uses an emotion engine to recognize the user's emotion in the received request data.

[1612] What it does: Uses natural language processing techniques to extract emotion tags (e.g., happy, sad, joyful, anxious) from request data.

[1613] Step 3:

[1614] The server adjusts the rules of the filtering means based on the user's emotions recognized by the emotion engine.

[1615] Specific behavior: If the emotion tag is recognized as "fun," apply the filtering rules according to that emotion.

[1616] Step 4:

[1617] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data.

[1618] Specific behavior: Scans request data using regular expressions and inappropriate word lists, and modifies inappropriate expressions to appropriate ones. For example, "Tell me a scary story" is modified to "Tell me a funny story."

[1619] Step 5:

[1620] The server generates text using the filtered request data with the generation means.

[1621] What it does: A generative AI model (e.g., GPT-3) is fed with the filtered request data and recognized emotion tags, and generates appropriate text based on the request and emotion.

[1622] Step 6:

[1623] The server then applies filtering measures to the generated text to detect and correct or replace inappropriate words or phrases.

[1624] Specific behavior: The generated text is scanned again against regular expressions and profanity lists, and profanity is corrected to proper expressions.

[1625] Step 7:

[1626] The server sends the filtered generated text to the terminal as an HTTP response.

[1627] Specific behavior: Constructs an HTTP response and sends the generated appropriate text to the device in the response body.

[1628] Processing on the device (kids' phone)

[1629] Step 1:

[1630] The terminal receives input from the user and sends the content to the server as a request.

[1631] What it does: It takes what the user types into the text input field, constructs an HTTP request containing that content, and sends it to the server.

[1632] Step 2:

[1633] The terminal waits for and receives a response from the server.

[1634] Specific operation: Asynchronously waits for the HTTP response sent from the server, receives it, and obtains the generated text from the response body.

[1635] Step 3:

[1636] The terminal displays the received generated text on the screen.

[1637] Specific behavior: The generated text received is reflected in the text display area so that the user can read it.

[1638] User (child) operations

[1639] Step 1:

[1640] The user inputs the text content that he or she wants to generate into the terminal.

[1641] Example: Using a text field or voice input, type "Tell me a fun story."

[1642] Step 2:

[1643] The user checks the generated text and takes the following action:

[1644] Example: Read the generated text displayed on the screen and enter your next request or act on what is generated.

[1645] Example 2

[1646] 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."

[1647] Conventional generative AI systems have safety issues due to the risk of inappropriate content being included when used by children. Furthermore, they lack a mechanism for generating appropriate content that adapts to the user's emotions, creating a need for a system that children can use safely. This invention aims to provide a generative AI system that generates text based on the user's emotions and filters out inappropriate content, making it safe for children to use.

[1648] 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.

[1649] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means;] [means for recognizing the user's emotions from the received request data using an emotion engine;] [means for detecting and correcting or replacing inappropriate words and phrases in the request data based on the emotions recognized by a filtering means;] [means for creating generated text in accordance with the user's emotions based on the request data filtered by a generating means;] [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; and [means for transmitting the filtered generated text to the terminal by a transmitting means]. This makes it possible [to realize a generative AI system that generates text that adapts to the user's emotions while eliminating inappropriate content, and is safe for children to use].

[1650] The "receiving means" is a device or function that receives request data sent from a terminal.

[1651] An "emotion engine" is a technology that includes software and algorithms for recognizing and analyzing user emotions from received data.

[1652] A "filtering means" is a device or function that detects and modifies or replaces inappropriate words or phrases in data.

[1653] "Generator" means a device or function for generating appropriate text based on filtered request data, including, for example, a generative AI model.

[1654] A "re-filtering means" is a device or function that re-detects and corrects or replaces inappropriate words or phrases in the generated text.

[1655] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[1656] A "generative AI model" refers to an artificial intelligence model that generates natural language text based on specific input data.

[1657] This invention provides a generative AI system that generates text that is adaptive to the user's emotions while eliminating inappropriate content, making it safe for children to use. This system includes three main components: a server, a terminal, and a user.

[1658] Server-side processing

[1659] Receiving means

[1660] The server first receives request data sent from the terminal using the receiving means. This request data includes the text entered by the user and the generated request content. For example, if the user enters "Tell me a fun story," the terminal sends the request to the server.

[1661] Emotion Engine

[1662] The server uses an emotion engine to recognize the user's emotion from the received request data. The emotion engine uses natural language processing technologies such as VADER and TextBlob. These technologies extract emotions such as positive, negative, and neutral from the request data. For example, a request such as "Tell me a fun story" will recognize a positive emotion.

[1663] Filtering Methods

[1664] Based on the emotions recognized by the emotion engine, the server uses a filtering mechanism to detect inappropriate words and phrases in the request data. This filtering mechanism scans the request data using regular expressions and inappropriate word lists, and if an inappropriate expression is found, it modifies or replaces it with an appropriate expression. For example, if a request such as "Tell me a scary story" is determined to be "scary," it will be modified to "Tell me a funny story."

[1665] generation means

[1666] After filtering, the server generates text using a generative method based on the request data. This method uses a generative AI model (such as GPT-3). The generative AI model generates appropriate text by adjusting the content and tone based on the user's emotions recognized by the emotion engine. For example, the generated text for "Tell me a fun story" is a fun story that begins with "Once upon a time..."

[1667] Re-filtering means

[1668] The server then applies the filtering measures to the generated text again. In this step, the generated text is again checked for inappropriate content and corrected or replaced if necessary. For example, if the generated text contains inappropriate language, it is replaced with appropriate language.

[1669] Transmission method

[1670] Finally, the filtered generated text is sent to the terminal via a transmission means. The server constructs an HTTP response, includes the generated text in the response body, and sends it to the terminal. In this way, the generated text is delivered to the user.

[1671] Terminal side processing

[1672] Receiving User Input

[1673] The terminal receives input from the user and prepares it to be sent as a request to the server, for example, "Tell me a funny story."

[1674] Sending request data

[1675] The terminal includes the request content entered by the user in an HTTP request and sends it to the server. After the server returns a response, the terminal receives the response.

[1676] Viewing generated text

[1677] The generated text is then displayed on the screen, allowing the user to read it. For example, a funny story like "Once upon a time..." is displayed on the screen.

[1678] User operation

[1679] The user (child) inputs the text they want to generate into the device. For example, if they input "Tell me a fun story," the request is sent to the server, and the generated fun story text is displayed on the device. The user can check the generated text and enjoy it.

[1680] Specific examples

[1681] 1. User Request: "Tell me a happy story"

[1682] The terminal sends the request to the server.

[1683] The server receives the request and uses an emotion engine to recognize the emotion "happy."

[1684] The server uses filtering measures to detect and correct inappropriate content in requests.

[1685] The server uses a generative AI model to generate happy stories.

[1686] The generated stories are then filtered again to ensure they are free of inappropriate content.

[1687] The server sends the text of the happy story to the device.

[1688] The terminal displays the received text on its screen for the user to read.

[1689] In this way, the system achieves safe text generation while adapting to the user's emotions.

[1690] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1691] Step 1:

[1692] Receiving User Input

[1693] The user inputs the text they want to generate into the device. For example, they input "Tell me a fun story." The input request is saved on the device, and an HTTP request is prepared. The input in this case is "Tell me a fun story." The output is the request data saved on the device.

[1694] Step 2:

[1695] Sending request data

[1696] The device constructs an HTTP request that includes the user's input and sends it to the server. For example, a request containing the content "Tell me a fun story" is sent to the server. The input at this time is the request data saved on the device. The output is the request data sent to the server.

[1697] Step 3:

[1698] Conducting sentiment analysis

[1699] The server passes the received request data to the emotion engine, which recognizes the user's emotion. The emotion engine uses natural language processing technology to extract emotion from the request data. For example, it can recognize positive emotion from a request such as "Tell me a fun story." The input in this case is the request data "Tell me a fun story," and the output is emotion data of "positive."

[1700] Step 4:

[1701] Filtering Request Data

[1702] The server uses filtering means to detect and modify or replace inappropriate words or phrases in the request data based on the emotions recognized by the emotion engine. For example, a request such as "Tell me a scary story" is determined to be "scary" and modified to "Tell me a funny story." The input in this case is emotion data of "positive" and request data of "Tell me a fun story," and the output is the filtered request data.

[1703] Step 5:

[1704] Text Generation

[1705] The server generates appropriate text using a generative AI model (e.g., GPT-3) based on the filtered request data. For example, in response to the request "Tell me a fun story," a fun story such as "Once upon a time..." is generated. The input in this case is the filtered request data, and the output is the generated text.

[1706] Step 6:

[1707] Re-filtering the generated text

[1708] The server then applies the filtering measures to the generated text again. For example, the generated text "Once upon a time..." is filtered again to check for inappropriate language, correcting or replacing it as necessary. The input is the generated text, and the output is the re-filtered generated text.

[1709] Step 7:

[1710] Sending generated text

[1711] Finally, the server constructs the filtered generated text as an HTTP response and sends it to the terminal. The server sends the generated text in the response body to the terminal. The input at this time is the re-filtered generated text, and the output is the response data sent to the terminal.

[1712] Step 8:

[1713] Viewing generated text

[1714] The terminal receives the response from the server and displays the generated text on the screen. For example, a fun story like "Once upon a time..." is displayed on the screen. The input is the response data received by the terminal, and the output is the generated text displayed on the screen.

[1715] The above is the specific processing flow of this system.

[1716] (Application example 2)

[1717] 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."

[1718] Current generative AI systems lack appropriate filtering and emotion-based adjustment to ensure safe use by children. This means that generated content may contain inappropriate content, and it is difficult to provide content that quickly and accurately reflects the user's emotions. There is a need for a system that can solve this issue and realize a content distribution service that children can enjoy safely.

[1719] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1720] In this invention, the server includes: [means for receiving request data from the terminal by a receiving means; [means for recognizing the user's emotions in the request data by an emotion recognition means; [means for detecting and correcting or replacing inappropriate words and phrases in the request data by a filtering means; [means for creating generated text according to the emotions based on the request filtered by the generating means; [means for re-detecting and correcting or replacing inappropriate words and phrases in the generated text by a re-filtering means; [means for transmitting the filtered generated text to the terminal by a transmitting means; and [means for displaying the generated text on the terminal.] This makes it possible to provide an emotion-based generative AI system that children can use safely.

[1721] The "receiving means" is a device or function that receives request data from a terminal.

[1722] An "emotion recognition means" is a device or algorithm for recognizing a user's emotion in request data.

[1723] A "filtering means" is a device or function that detects inappropriate words or phrases in request data and modifies or replaces them.

[1724] A "generator" is a device or algorithm that creates emotion-sensitive generated text based on filtered request data.

[1725] A "re-filtering means" is a device or function that re-detects inappropriate words or phrases in the generated text and corrects or replaces them.

[1726] The "transmitting means" is a device or function that transmits the filtered generated text to the terminal.

[1727] A "terminal" is an electronic device that allows a user to input request data and display the generated text.

[1728] This invention relates to the implementation of the smartphone application "Emotory" as an example of an application of a content distribution service that can be used safely by children. "Emotory" is an application that generates stories based on emotions, and provides content appropriate to the user through a series of processes.

[1729] First, the user (child) inputs their feelings and interests into their smartphone. The device then sends this input to the server as request data. For example, the user might input "Tell me a fun story." The server then processes the data as follows:

[1730] Server-side processing

[1731] 1. Receiving request data: The server receives the request data sent from the terminal. This is realized by the "receiving means."

[1732] 2. Emotion Recognition: To recognize the user's emotions from the received request data, we use an "emotion recognition method." This method uses natural language processing technology, such as NLP libraries like "spaCy" and "NLTK."

[1733] 3. Request Data Filtering: Using "filtering methods" to detect and modify or replace inappropriate words and phrases based on recognized sentiment. Filtering methods may use regular expressions or predefined lists of inappropriate words.

[1734] 4. Generating generated text: A "generator" is used to generate sentiment-based generated text based on the filtered request data. This generator uses a large-scale generative AI model, such as OpenAI's "GPT-3."

[1735] 5. Re-filtering the generated text: A "re-filtering method" is used to again detect and correct or replace inappropriate words or phrases in the generated text.

[1736] 6. Sending the generated text: Finally, the filtered generated text is sent to the terminal by the "sending means."

[1737] Terminal side processing

[1738] The terminal receives the generated text sent from the server and has the function of "displaying the generated text on the terminal" in a format that is easy for the user to read. The user (child) can enjoy reading this generated text.

[1739] Specific hardware / software usage

[1740] NLP library: "spaCy" and "NLTK" are used for emotion recognition.

[1741] Generative AI model: OpenAI's "GPT-3" is used for text generation.

[1742] Filtering libraries: Use your own implementation or third-party libraries that use regular expressions and profanity lists.

[1743] Examples of concrete examples and prompts

[1744] Examples:

[1745] 1. If you feel happy:

[1746] User input: Child types "Tell me a happy story."

[1747] Server processing: After emotion recognition, filtering, generation, and re-filtering, a positive story is generated: "Once upon a time, there was a very happy village..."

[1748] Terminal display: "Once upon a time, there was a very happy village..."

[1749] 2. To relieve anxiety:

[1750] User input: Child types, "Tell me a story to calm my anxiety."

[1751] Server processing: After emotion recognition, filtering, generation, and re-filtering, a soothing story is generated: "Listen to this story. In a very peaceful forest..."

[1752] Terminal display: "Listen to this story. In a very peaceful forest..."

[1753] Example prompt sentence:

[1754] A child types "Tell me a fun story" into their smartphone.

[1755] As a result, "Emotory" is an application that allows children to spend time safely and enjoyably, and can provide appropriate content according to the user's emotions.

[1756] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1757] Step 1:

[1758] User Input

[1759] The user (child) inputs the text they want to generate into their smartphone. For example, they might input "Tell me a fun story." The device then sends this input to the server as request data.

[1760] Input: Text entered by the user (e.g., "Tell me a nice story")

[1761] Output: Request data sent to the server

[1762] Step 2:

[1763] Receiving request data

[1764] The server receives the request data sent from the terminal, and provides the server with information about the content the user is looking for.

[1765] Input: Request data

[1766] Output: Request data stored on the server

[1767] Step 3:

[1768] Emotion recognition

[1769] The server recognizes the user's emotion from the received request data. For emotion recognition, it uses an NLP library (e.g., "spaCy" or "NLTK"). For example, if the request is "Tell me a fun story," the user's emotion is recognized as "positive."

[1770] Input: Request data

[1771] Output: Perceived emotion (e.g., "positive")

[1772] Step 4:

[1773] Filtering Request Data

[1774] The server detects and modifies or replaces inappropriate words and phrases in the request data based on sentiment. Regular expressions and inappropriate word lists are used as filtering methods. For example, if a request is made to "tell me a scary story," the word "scary" is deemed "inappropriate" and the request is modified to "tell me a funny story."

[1775] Input: Request data, recognized emotion

[1776] Output: Filtered request data

[1777] Step 5:

[1778] Creating generated text

[1779] The server generates text based on the filtered request data, according to the sentiment. A large-scale generative AI model (e.g., GPT-3) is used as the generation method. For example, if the filtered data is "Tell me a happy story" and the sentiment is "positive," the generated text would be something like "Once upon a time, there was a very happy village..."

[1780] Input: filtered request data, recognized sentiment

[1781] Output: The generated text

[1782] Step 6:

[1783] Re-filtering the generated text

[1784] The server again detects and corrects or replaces inappropriate words or phrases in the generated text. For example, if the generated text contains an "inappropriate word," it replaces it with an appropriate expression. This process is also performed using filtering methods.

[1785] Input: Generated text

[1786] Output: Refiltered generated text

[1787] Step 7:

[1788] Sending generated text

[1789] The server sends the filtered generated text to the terminal using an HTTP response as a transmission means, and includes the text in the response body.

[1790] Input: Refiltered generated text

[1791] Output: The generated text sent to the terminal

[1792] Step 8:

[1793] Viewing generated text

[1794] The terminal receives the generated text sent from the server and displays it on the screen, for example, the generated text "Once upon a time, there was a very pleasant village..." is displayed on the screen for the user to read.

[1795] Input: Generated text sent from the server

[1796] Output: The generated text displayed on the terminal.

[1797] 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.

[1798] 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.

[1799] 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.

[1800] 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.

[1801] 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.

[1802] 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.

[1803] 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).

[1804] 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.

[1805] 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."

[1806] 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.

[1807] 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).

[1808] 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.

[1809] 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.

[1810] 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.

[1811] 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.

[1812] 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.

[1813] 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.

[1814] 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.

[1815] 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.

[1816] 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.

[1817] 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.

[1818] The following is further disclosed regarding the above embodiment.

[1819] (Claim 1)

[1820] [Means for receiving request data from a terminal by a receiving means;

[1821] [Means for detecting and correcting or replacing inappropriate words or phrases in the request data through filtering means; and

[1822] [means for generating text based on the requests filtered by the generating means;

[1823] [means for re-detecting and correcting or replacing inappropriate words or phrases in the text produced by the re-filtering means; and

[1824] [Means for transmitting the generated text filtered by the transmitting means to the terminal;

[1825] A system including:

[1826] (Claim 2)

[1827] [Means for receiving request data from a terminal by a receiving means, and filtering the request data;

[1828] [means for generating generated text based on the request data filtered by the generating means;

[1829] [Means for transmitting the text generated by the transmitting means to a terminal;

[1830] 10. The system of claim 1, comprising:

[1831] (Claim 3)

[1832] The system of claim 1, wherein the filtering means filters the request data and the generated text using regular expressions or a predefined list of inappropriate words.

[1833] "Example 1"

[1834] (Claim 1)

[1835] [Means for receiving request data from an information processing device by a receiving means;

[1836] [Means for detecting and correcting or replacing inappropriate words or phrases in the request data by filtering means;

[1837] [means for generating text based on the request data filtered by the generating means;

[1838] [means for re-detecting and correcting or replacing inappropriate words or boilerplate phrases in the text generated by the re-filtering means; and

[1839] [Means for transmitting the generated text filtered by the transmitting means to the information processing device;

[1840] A system including:

[1841] (Claim 2)

[1842] [Means for receiving request data from an information processing device by a receiving means, and filtering the request data;

[1843] [means for generating generated text based on the request data filtered by the generating means;

[1844] [Means for transmitting the text generated by the transmitting means to the information processing device;

[1845] 10. The system of claim 1, comprising:

[1846] (Claim 3)

[1847] The system of claim 1, wherein the filtering means filters the request data and the generated text using regular expressions or a predefined list of inappropriate words.

[1848] "Application Example 1"

[1849] (Claim 1)

[1850] [Means for receiving request data from a terminal by a receiving means;

[1851] [Means for detecting and correcting or replacing inappropriate words or phrases in the request data through filtering means; and

[1852] [means for generating text based on the requests filtered by the generative AI model; and

[1853] [means for re-detecting and correcting or replacing inappropriate words or phrases in the text produced by the re-filtering means; and

[1854] [Means for transmitting the generated text filtered by the transmitting means to the terminal;

[1855] [means for providing an interface that allows the user to review and enjoy the re-filtered generated text; and

[1856] A system including:

[1857] (Claim 2)

[1858] [Means for receiving request data from a terminal by a receiving means, and filtering the request data;

[1859] [means for generating generated text based on the request data filtered by the generative AI model;

[1860] [Means for transmitting the text generated by the transmitting means to a terminal;

[1861] [Providing an interface on the terminal that allows the user to view and enjoy the generated text; and

[1862] 10. The system of claim 1, comprising:

[1863] (Claim 3)

[1864] The system of claim 1, wherein the filtering means filters the request data and the generated text using regular expressions or a predefined list of inappropriate words.

[1865] "Example 2: Combining Emotion Engines"

[1866] (Claim 1)

[1867] [Means for receiving request data from a terminal by a receiving means;

[1868] [Means for recognizing a user's emotion from the received request data using an emotion engine;

[1869] [means for detecting and modifying or replacing inappropriate words or phrases in the request data based on the sentiment recognized by the filtering means;

[1870] [Means for generating text according to the user's emotions based on the request data filtered by the generating means;

[1871] [means for re-detecting and correcting or replacing inappropriate words or phrases in the text produced by the re-filtering means; and

[1872] [Means for transmitting the generated text filtered by the transmitting means to the terminal;

[1873] A system including:

[1874] (Claim 2)

[1875] [Means for receiving request data from a terminal by a receiving means, and recognizing and filtering emotions from the request data by an emotion engine;

[1876] [Means for generating generated text according to the user's emotions based on the request data filtered by the generating means;

[1877] [Means for transmitting the text generated by the transmitting means to a terminal;

[1878] 10. The system of claim 1, comprising:

[1879] (Claim 3)

[1880] The system of claim 1, wherein the filtering means filters the request data and the generated text using regular expressions or a predefined list of inappropriate words.

[1881] "Application example 2 when combining emotion engines"

[1882] (Claim 1)

[1883] [Means for receiving request data from a terminal by a receiving means;

[1884] [Means for recognizing a user's emotion in request data by an emotion recognition means;

[1885] [Means for detecting and correcting or replacing inappropriate words or phrases in the request data through filtering means; and

[1886] [Means for generating sentiment-aware generated text based on the requests filtered by the generating means;

[1887] [means for re-detecting and correcting or replacing inappropriate words or phrases in the text produced by the re-filtering means; and

[1888] [Means for transmitting the generated text filtered by the transmitting means to the terminal;

[1889] [means for displaying the generated text at the terminal;

[1890] A system including:

[1891] (Claim 2)

[1892] [Means for receiving request data from a terminal by a receiving means, and filtering the request data;

[1893] [Means for generating generated text according to emotions based on the request data filtered by the generating means;

[1894] [Means for transmitting the text generated by the transmitting means to a terminal and displaying it on the terminal;

[1895] 10. The system of claim 1, comprising:

[1896] (Claim 3)

[1897] The system of claim 1, wherein the filtering means filters the request data and the generated text using regular expressions or a predefined list of inappropriate words. [Explanation of symbols]

[1898] 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. a means for receiving request data from a terminal by a receiving means; means for detecting and modifying or replacing inappropriate words or phrases in the request data using filtering means; means for generating text based on the requests filtered by the generating means; means for re-detecting and correcting or replacing inappropriate words or phrases in the text produced by the re-filtering means; means for transmitting the generated text filtered by the transmitting means to the terminal; A system including:

2. a receiving means for receiving request data from a terminal and filtering the request data; means for generating generated text based on the request data filtered by the generating means; means for transmitting the text generated by the transmitting means to the terminal; The system of claim 1 , comprising:

3. 2. The system of claim 1, wherein the filtering means filters the request data and the generated text using regular expressions and a predefined list of inappropriate words.

Images