System

Abstract

A system is provided.SOLUTION: A system, comprising: means for receiving a text input by a user; means for analyzing the received text to detect portions containing slander; means for generating a warning message and a suggestion based on the detected portions containing slander; and means for displaying the generated warning message and suggestion to the user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] The number of slanderous comments on the Internet is increasing, causing many cases of psychological damage to celebrities and young people. In particular, malicious behavior that takes advantage of the anonymity of the Internet has become a serious problem, leading to bullying and suicide. For this reason, there is a need for a system that can effectively detect slanderous comments on the Internet and reduce problematic expressions. [Means for solving the problem]

[0005] The present invention provides a system that receives text entered by a user, analyzes the text using a natural language processing engine, and detects any abusive or defamatory content. It also includes a function that generates warning messages and suggestions based on the detected abusive content and displays them to the user. The system also analyzes the text again after the user has completed posting, converting any abusive content into more neutral language, and then stores and distributes the converted text. This effectively reduces online abusive content and aims to improve the health of the online environment.

[0006] "User" means an individual or legal entity who uses the system to input and post text.

[0007] The "receiving means" is a combination of hardware and software that provides the function of incorporating user-entered text into the system.

[0008] "Means for analysis" refers to the function of analyzing received text using a natural language processing engine or the like to detect specific patterns or terms.

[0009] "Abuse detection measures" are algorithms and processes that, based on the results of the analysis, identify insulting or offensive language in a text.

[0010] A "warning message" is a notification generated by the system to inform the user that the text they entered contains abusive language.

[0011] "Suggestion" is a function in which the system provides suggestions for rephrasing defamatory language into more appropriate and safer language.

[0012] "Harsh expression" refers to the safe and considerate use of language to convey one's intentions without hurting others.

[0013] "Means for storing and distributing" refers to a function for storing the converted text in a database and providing the text to other users or systems as needed.

[0014] A "natural language processing engine" is a software tool and algorithm used to analyze the content of text and detect specific patterns and phrases. [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] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[0037] System Overview

[0038] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[0039] Program processing

[0040] 1. User pre-post detection process

[0041] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[0042] Next, the server generates a warning message based on the detection results and makes a suggestion such as, "This word is negative. Why not change it to 'You may not be trying hard enough.'" This suggestion and warning message are sent to the terminal, which displays it to the user. The user can then revise or reconsider the sentence based on the suggestion.

[0043] 2. Post-submission content transformation process

[0044] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[0045] The server generates the converted text, "You should try a little harder," and stores it in the database. It also displays the converted post to the user and other related users as needed. In this way, defamatory posts are detected and corrected before and after the fact, providing a safe and healthy online environment.

[0046] Specific examples

[0047] 1. Example 1: A concrete example of a user pre-post detection process

[0048] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." The server then sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[0049] 2. Example 2: A concrete example of the post-submission content transformation process

[0050] The user posts "You're useless." The device sends the post to the server, which again detects "useless." The server converts this part to "try a little harder," generating the final sentence, "You try a little harder." The converted sentence is stored in a database and displayed to the user and other relevant users as needed.

[0051] As a result, the present invention can effectively detect and correct online slander and provide a healthy online environment.

[0052] The processing flow will be explained below.

[0053] User pre-post detection process

[0054] Step 1:

[0055] The user enters the content to post.

[0056] The user types "You're really incompetent" into a text box in a chat app.

[0057] Step 2:

[0058] The terminal sends the input text to the server.

[0059] The terminal sends the input text data to the server as a POST request.

[0060] Step 3:

[0061] The server passes the received text to a natural language processing engine.

[0062] The server passes the received text "You are truly incompetent" to the NLP engine.

[0063] Step 4:

[0064] The server uses a natural language processing engine to detect slander.

[0065] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[0066] Step 5:

[0067] The server generates warning messages and suggestions based on the detection results.

[0068] For the word "incompetent," the server creates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[0069] Step 6:

[0070] The server generates warning messages and suggestions and sends them to the device.

[0071] The server sends warning messages and suggestions in JSON format to the device.

[0072] Step 7:

[0073] The terminal displays a warning message and a suggestion to the user.

[0074] The terminal displays the received message to the user in a pop-up window or dialog box.

[0075] Post-submission content transformation process

[0076] Step 1:

[0077] The user completes the post.

[0078] The user presses the "Submit" button to complete the submission.

[0079] Step 2:

[0080] The device sends the posted text to the server.

[0081] The device sends the message "You're useless" to the server.

[0082] Step 3:

[0083] The server passes the posted text back to the natural language processing engine.

[0084] The server passes the received text "You're useless" to the NLP engine.

[0085] Step 4:

[0086] The server detects the slander again.

[0087] The server recognizes the word "useless" as a slur and identifies that part.

[0088] Step 5:

[0089] The server converts abusive comments into harmless language.

[0090] The server translates "useless" into "try harder."

[0091] Step 6:

[0092] The server generates the converted text and stores it in a database.

[0093] The server saves the revised sentence, "You should try a little harder," in the database.

[0094] Step 7:

[0095] The server delivers the converted text as needed.

[0096] The server displays the modified post to the user and other interested users.

[0097] This allows for effective detection and correction of defamatory comments both before and after users post.

[0098] Example 1

[0099] 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."

[0100] On modern online platforms, abusive language can easily spread, negatively impacting communication between users. Therefore, there is a need for a system that can proactively detect abusive posts, warn users, and prompt them to change their language to more appropriate ones. However, current systems struggle to detect and convert abusive language in sufficient real-time, often resulting in users leaving abusive language in their posts after posting. Furthermore, the lack of natural suggestions for converting to more appropriate language makes it difficult for users to edit their posts in a way that is more acceptable to them.

[0101] 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.

[0102] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text using a natural language processing engine to detect portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for displaying the generated warning message and suggestion to the user, and means for correcting the text entered by the user based on the user's reaction. As a result, when a user enters a text containing slander, a warning and appropriate suggestions are displayed in real time, allowing the user to immediately correct the text and realize safe and healthy online communication.

[0103] "Means for receiving text entered by a user" is a function for sending text entered by a user into a chat app or SNS app to a server.

[0104] A "natural language processing engine" is a software technology for understanding, interpreting, and manipulating human language, including analyzing and classifying text and detecting defamation.

[0105] "Abusive language detection means" is an analytical function that identifies offensive or negative language in a text.

[0106] The "means for generating warning messages and suggestions" is a function for warning the user about detected abusive language and suggesting appropriate alternative expressions.

[0107] The "means for displaying to the user" is a function for displaying warning messages and suggestions on the user's terminal in real time.

[0108] The "means for correcting text entered by the user based on the user's response" is a function that allows the user to correct the text entered in accordance with the warnings and suggestions received by the user.

[0109] The "means for storing and distributing" is a function for storing the final corrected text in a database and distributing it to relevant users as necessary.

[0110] To specifically implement the present invention, the following system configuration and processing procedures are required.

[0111] System Overview

[0112] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[0113] Specific hardware and software configurations

[0114] server

[0115] The server receives the user's input and uses a natural language processing engine (e.g., SpaCy, Google® Cloud Natural Language API) to detect and analyze defamatory content.

[0116] The server generates warning messages and suggestions based on the detection results and sends them to the user's terminal.

[0117] Terminal

[0118] The device sends text entered by the user in a chat app or SNS app to the server.

[0119] The terminal displays the warning messages and suggestions returned by the server to the user.

[0120] The terminal transmits the text corrected by the user to the server again.

[0121] User

[0122] A user inputs text into a chat app or a social networking app.

[0123] The user corrects the text based on the displayed warning message and suggestions.

[0124] System operation example

[0125] User pre-post detection process

[0126] 1. The user types "You are truly incompetent." The device sends this sentence to the server.

[0127] 2. The server analyzes the received text using a natural language processing engine and determines that the word "incompetent" is defamatory.

[0128] 3. The server generates a warning message and suggestion saying, "This word is negative. Why not change it to 'You may not be trying hard enough'?" and sends it to the terminal.

[0129] 4. The terminal displays a warning message and a suggestion to the user.

[0130] 5. The user takes the suggestion into consideration and revises the sentence to "Perhaps you are not trying hard enough."

[0131] Post-submission content transformation process

[0132] 1. A user posts "You're useless." The device sends the post to the server.

[0133] 2. The server analyzes the received text again using a natural language processing engine and determines that the word "useless" is defamatory.

[0134] 3. The server converts this to "You should try a little harder" and stores the final generated sentence in the database.

[0135] 4. The converted text is displayed to the user and other relevant users as needed.

[0136] Through these processes, the system can effectively detect and correct online defamation and provide a safe and healthy online environment.

[0137] Here is an example of a prompt for the generative AI model:

[0138] Prompt Sentence Examples

[0139] Please change the following sentences to non-libelous language:

[0140] "You're really incompetent."

[0141]

[0142] Please change the following sentences to non-libelous language:

[0143] "You're useless."

[0144] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0145] User pre-post detection process

[0146] Step 1:

[0147] The user inputs a sentence.

[0148] Input: A user types "You're so incompetent" into a chat or social media app.

[0149] Action: The user types a sentence on the device's keyboard.

[0150] Output: The input sentence "You are truly incompetent."

[0151] Step 2:

[0152] The terminal sends the input text to the server.

[0153] Input: The sentence "You are truly incompetent" typed by the user.

[0154] Action: The device sends this text to the server via the network.

[0155] Output: The sentence "You are truly incompetent" received by the server.

[0156] Step 3:

[0157] The server passes the received text to a natural language processing engine for analysis.

[0158] Input: The sentence "You are truly incompetent" received by the server.

[0159] How it works: The server passes this sentence to a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) for analysis. At the same time, it detects the slanderous word "incompetent."

[0160] Output: The word "incompetent" is determined to be libelous.

[0161] Step 4:

[0162] The server generates warning messages and suggestions based on the slander detected.

[0163] Input: The word "incompetent" which was determined to be libelous.

[0164] What it does: The server verifies that the word "incompetent" is negative and generates a warning message: "This word is negative." and a suggestion: "Maybe you're not trying hard enough?"

[0165] Output: Warning messages and suggestions.

[0166] Step 5:

[0167] The server sends warning messages and suggestions to the device.

[0168] Input: Generated warning messages and suggestions.

[0169] Operation: The server sends a warning message and suggestions to the user's device.

[0170] Output: A warning message and a suggestion will be displayed on the user's terminal.

[0171] Step 6:

[0172] The terminal displays a warning message and a suggestion to the user.

[0173] Input: Warning messages and suggestions sent by the server.

[0174] Behavior: Displays a warning message and suggestion in a pop-up or message display area on the device screen.

[0175] Output: The user sees a warning message and a suggestion.

[0176] Step 7:

[0177] The user corrects the text.

[0178] Input: Warning messages and suggestions.

[0179] Action: Based on the suggestion, the user amends the original sentence "You are really incompetent" to "Maybe you're not trying hard enough."

[0180] Output: Corrected sentence: "Maybe you're not trying hard enough."

[0181] Post-submission content transformation process

[0182] Step 1:

[0183] The user completes the post.

[0184] Input: Corrected sentence: "Maybe you're not trying hard enough."

[0185] Action: The user clicks the submit button and confirms the text.

[0186] Output: The confirmed post text "Maybe you're not trying hard enough."

[0187] Step 2:

[0188] The device sends the posted text to the server.

[0189] Input: Confirmed post text: "Maybe you're not trying hard enough."

[0190] Action: The device sends the confirmed text to the server.

[0191] Output: The sentence received by the server: "Maybe you're not trying hard enough."

[0192] Step 3:

[0193] The server then analyzes the text again using a natural language processing engine.

[0194] Input: The sentence "You may not be trying hard enough" received by the server.

[0195] What happens: The server passes this sentence to a natural language processing engine for analysis, and verifies that it does not contain any slander.

[0196] Output: After analysis, it is determined that the sentence "Maybe you're not trying hard enough" does not contain any slander.

[0197] Step 4:

[0198] The server stores the final converted text in a database.

[0199] Input: Analysis result shows that the sentence "You may not be trying hard enough" does not contain any slander.

[0200] What it does: The server saves the sentence in a database based on the analysis results.

[0201] Output: The sentence "Maybe you're not trying hard enough" stored in the database.

[0202] Step 5:

[0203] If necessary, the converted text is displayed to the relevant users.

[0204] Input: The sentence stored in the database: "Maybe you're not trying hard enough."

[0205] How it works: The system will display the new text to relevant users (e.g., users who receive the post on the subject) as needed.

[0206] Output: The converted sentence "Maybe you're not trying hard enough" displayed on the relevant user's device.

[0207] The above are the specific steps in the pre-post detection process and post-post content conversion process, which will effectively detect and correct online slander and provide a safe and healthy online environment.

[0208] (Application example 1)

[0209] 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."

[0210] Defamation on the Internet has a significant negative impact on users' mental health and can worsen the atmosphere of the entire community. In particular, in content distribution services, where many users post comments in real time, monitoring and correcting them is difficult. Conventional methods have difficulty detecting and appropriately correcting defamation in real time, making it difficult to provide a safe and healthy online environment. To solve this problem, a system that utilizes natural language processing technology and generative AI models to detect and correct user comments in real time is needed.

[0211] 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.

[0212] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text and detecting portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for generating the generated warning message and suggestion using a natural language generation model, and means for displaying the generated warning message and suggestion to the user. This makes it possible to monitor user comments in real time, detect slander, and immediately suggest corrections, thereby providing a constructive comment environment.

[0213] The "means for receiving text entered by a user" refers to a technology or device for transmitting text information entered by a user through a device to a server and acquiring that text information.

[0214] "Means for analyzing received text and detecting defamatory or libelous parts" refers to technologies or processes that use natural language processing engines or similar to analyze text data and identify negative or offensive words or expressions contained therein.

[0215] "Means for generating warning messages and suggestions based on the detected defamatory content" refers to technology that generates messages to warn users and suggestions showing more appropriate and constructive examples of expression based on the detection results.

[0216] The "means for displaying the generated warning message and suggestions to the user" refers to a technology or interface for displaying the generated message and suggestions on the screen of the user's device.

[0217] "Means for generating using a natural language generation model" refers to a technology for generating warning messages and suggestions in natural language using a generative AI model (e.g., GPT-3 (registered trademark) or T5, etc.).

[0218] "Means of re-analyzing the text that the user has posted and detecting any parts that contain defamatory comments" refers to a technology that re-analyzes the text that the user has finally posted and identifies any defamatory comments that remain within it.

[0219] "Means for converting detected abusive content into less offensive language" refers to technologies or algorithms that convert detected negative or offensive words or phrases into more neutral or positive language.

[0220] "Means for storing and distributing the converted text" refers to the technology used to store the corrected text on a server and distribute it to relevant users and platforms as needed.

[0221] The present invention is a system that receives text entered by a user in real time, analyzes it, and detects and corrects slander. Specific embodiments of the present invention will be described below.

[0222] System Overview

[0223] This system consists of three elements: a server, a terminal, and a user.

[0224] server:

[0225] The server uses a natural language processing engine (e.g., a T5 model using the transformers library) to analyze the text sent by the user. It detects abusive language in the text and generates warning messages and suggestions based on the detected language. The suggestions are generated by a natural language generation model and suggest appropriate ways of expressing the text to the user.

[0226] Device:

[0227] The terminal is a device (smartphone or computer) where the user inputs text. When the user inputs text using a chat app or SNS app, the text is sent to the server in real time. The terminal also displays warning messages and suggestions received from the server.

[0228] User:

[0229] The user is the subject of inputting text and receiving warning messages and suggestions from the system. The user can then correct the text based on the suggestions and submit it again.

[0230] Program processing

[0231] The server first receives the text entered by the user. It then analyzes the text using a natural language processing engine (for example, a T5 model using the transformers library) to detect parts that contain slander. Based on the detected parts, it generates warning messages and suggestions. These suggestions are created in natural Japanese sentences using a generative AI model.

[0232] The server sends the generated warning message and suggestions to the user's device, where the user can view it. If the user corrects the text as suggested, the corrected text can be sent back to the server. The server receives it, analyzes it again to ensure it does not contain any slander, and then stores and distributes it.

[0233] Specific examples

[0234] For example, if a user enters the sentence "You are really incompetent," the server detects the word "incompetent" as a slander. The server then generates a suggestion saying, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?" The following is an example of a prompt:

[0235] Your comment is negative. Try rephrasing it this way: Maybe you're not trying hard enough.

[0236] The suggestion is displayed on the terminal, and the user can refer to it and correct it by saying, "Maybe you're not trying hard enough." The server then analyzes the corrected sentence again, and after confirming that there are no problems, stores and distributes the sentence.

[0237] In this way, the system can monitor user comments in real time, detect and correct defamatory comments, and provide a healthy online environment.

[0238] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0239] Step 1:

[0240] A user inputs text using a chat app or SNS app. The input text is sent in real time from the device to a server. The server receives the text data sent by the user. The input is the text entered by the user, and the output is a confirmation that the text has been received.

[0241] Step 2:

[0242] The server passes the received text data to a natural language processing engine to detect abusive content. In this step, it filters out negative keywords such as "incompetent" or "useless." The input is a sentence entered by the user, and the output is a list of abusive keywords and phrases.

[0243] Step 3:

[0244] The server generates warning messages and suggestions based on the detected abusive comments. It uses a generative AI model (such as the T5 model) to create warning messages and suggestions in natural-sounding sentences. The input is a list of abusive comments and the original sentences, and the output is the generated warning messages and suggestions.

[0245] Step 4:

[0246] The server sends the generated warning message and suggestions to the terminal, which displays the message and suggestions to the user. The input is the warning message and suggestions sent from the server, and the output is displayed on the user's device screen.

[0247] Step 5:

[0248] The user then reconsiders whether to revise the sentence based on the displayed suggestions. In this step, the user inputs new sentences based on the suggestions received from the server. The inputs are the warning message and the suggestions, and the output is the revised sentence.

[0249] Step 6:

[0250] The user then sends the revised text back to the server. The server then receives this new text and uses a natural language processing engine to detect the abusive text again. The input is the revised text, and the output is the detection results.

[0251] Step 7:

[0252] The server checks that the text is free of slander and then stores the final text in a database and distributes it as needed. This step involves storing the final text and distributing it to relevant users and platforms. The input is the final, corrected text, and the output is a safe text stored in a database and distributed.

[0253] By following these steps, the system can detect defamatory comments in real time and prompt appropriate corrections.

[0254] 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.

[0255] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[0256] System Overview

[0257] This system receives and analyzes text entered by users, detects abusive language, and recognizes the user's emotions to provide warning messages and suggest appropriate expressions. This system combines a natural language processing engine and an emotion engine to detect abusive language and respond appropriately based on the user's emotions. It also has the ability to convert abusive language into harmless language, which it then stores and distributes.

[0258] Program processing

[0259] 1. User pre-post detection process

[0260] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[0261] The server then analyzes the user's emotions using an emotion engine, which recognizes their emotional state based on the tone and word choice of the text they type. For example, if the server determines that the user is annoyed, it will adjust warning messages and suggestions accordingly.

[0262] Based on the detection results and sentiment analysis results, the server generates a warning message and appropriate suggestions, such as "This word is negative. Why not change it to 'You may not be trying hard enough'?" This suggestion and warning message are sent to the device, which then displays it to the user. The user can then revise or reconsider the sentence based on the suggestions.

[0263] 2. Post-submission content transformation process

[0264] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[0265] The server also uses an emotion engine to analyze the user's emotions related to the post and generates an optimal, non-offensive expression based on this. The server generates the converted sentence, "You should try a little harder," and stores it in a database. The server also displays the converted post content to the user and other related users as needed.

[0266] Specific examples

[0267] 1. Example 1: A concrete example of a user pre-post detection process

[0268] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." Next, the server recognizes the user's frustration based on the emotion engine, and sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[0269] 2. Example 2: A concrete example of the post-submission content transformation process

[0270] The user posts "You're useless." The device sends the post to the server, which again recognizes "useless" and converts it to "try a little harder." It also uses an emotion engine to confirm the user's emotions and selects an appropriate expression. The server generates the converted sentence "try a little harder" and saves it in the database. The converted sentence is then displayed as needed.

[0271] As a result, the present invention can detect and correct online slander while taking into consideration the user's feelings, thereby providing a healthy online environment.

[0272] The processing flow will be explained below.

[0273] User pre-post detection process

[0274] Step 1:

[0275] The user enters the content to post.

[0276] The user types "You're really incompetent" into a text box in a chat app.

[0277] Step 2:

[0278] The terminal sends the input text to the server.

[0279] The terminal sends the text data entered by the user to the server as a POST request.

[0280] Step 3:

[0281] The server passes the received text to a natural language processing engine.

[0282] The server passes the received text "You are truly incompetent" to a natural language processing engine, which analyzes the sentence.

[0283] Step 4:

[0284] The server uses a natural language processing engine to detect slander.

[0285] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[0286] Step 5:

[0287] The server analyzes the user's emotions using an emotion engine.

[0288] The server uses an emotion engine to recognize the emotion of irritation based on the tone and word choice of the text entered by the user.

[0289] Step 6:

[0290] The server generates warning messages and suggestions based on the detection results and sentiment analysis results.

[0291] For the word "incompetent," the server generates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[0292] Step 7:

[0293] The server generates warning messages and suggestions and sends them to the device.

[0294] The server sends the generated warning message and suggestions to the device in JSON format.

[0295] Step 8:

[0296] The terminal displays a warning message and a suggestion to the user.

[0297] The terminal displays the received warning messages and suggestions to the user in pop-up windows or dialog boxes.

[0298] Post-submission content transformation process

[0299] Step 1:

[0300] The user completes the post.

[0301] The user presses the "Submit" button to complete the submission.

[0302] Step 2:

[0303] The device sends the posted text to the server.

[0304] The device sends the message "You're useless" to the server.

[0305] Step 3:

[0306] The server passes the posted text back to the natural language processing engine.

[0307] The server passes the received text, "You're useless," to a natural language processing engine for further analysis.

[0308] Step 4:

[0309] The server detects the slander again.

[0310] The server recognizes the word "useless" as a slur and identifies that part.

[0311] Step 5:

[0312] The server uses an emotion engine to analyze the user's emotions regarding the posted content.

[0313] The server uses an emotion engine to ascertain the user's emotional state associated with the post.

[0314] Step 6:

[0315] The server converts abusive comments into harmless language.

[0316] The server converts "useless" into "let's try a little harder" and selects the appropriate expression.

[0317] Step 7:

[0318] The server generates the converted text and stores it in a database.

[0319] The server stores the generated sentence "You should try a little harder" in a database.

[0320] Step 8:

[0321] The server delivers the converted text as needed.

[0322] The server displays the converted post to the user and other interested users.

[0323] As a result, the present invention can effectively detect and correct slanderous comments while taking into account the user's feelings during the pre-post and post-post processes.

[0324] Example 2

[0325] 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."

[0326] The increase in slander and negative comments on the Internet has made it difficult to maintain a healthy online environment. Users often unconsciously use negative language, and these comments are often posted without being edited. Furthermore, even for content that has already been posted, there will be users who will suffer harm unless slander is reduced. A system that can solve these problems is needed.

[0327] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for receiving a sentence entered by a user, a means for analyzing the received sentence using a natural language processing engine and detecting a portion containing slander, a means for generating a warning message and a suggestion using an emotion engine based on the detected portion containing slander, and a means for displaying the generated warning message and suggestion to the user. This allows the user to detect slander before posting a sentence and make appropriate corrections. Furthermore, by re-detecting slanderous portions in content after posting and converting them into non-savage expressions, the health of the online environment can be maintained.

[0328] "User" means an individual or organization that uses the system to input and post text.

[0329] "Text" refers to text data entered by a user and sent to the system.

[0330] A "natural language processing engine" is an algorithm or software that analyzes text data to recognize slander and specific patterns.

[0331] "Slander" refers to negative language that hurts or speaks ill of others.

[0332] An "emotion engine" is an algorithm or software that analyzes and identifies a user's emotions from the tone and content of a sentence.

[0333] A "warning message" is a message that notifies the user that the text they have entered contains slander and urges them to correct it.

[0334] A "suggestion" is a message containing alternative expressions to correct the sentence entered by the user.

[0335] "Terminal" refers to the device (e.g., smartphone, tablet, or PC) that a user uses to input text and send it to the system.

[0336] A "server" is a computer system that receives data sent from a terminal and analyzes and processes it.

[0337] "Posting" means the act of publishing text entered by a user on an online platform.

[0338] "Database" is an information management system for storing posted texts and generated messages.

[0339] The following describes in detail the mode for carrying out the present invention. The present invention is a system that receives text entered by a user, detects and corrects slanderous comments, and further recognizes the user's emotions to suggest appropriate warning messages and expressions. This system is composed of a server, a terminal, and a user.

[0340] System Overview

[0341] When a user enters text using a device (e.g., smartphone, tablet, or PC), the text is sent to the server. The server analyzes the received text using a natural language processing engine (e.g., Python library NLTK or spaCy) and an emotion engine (e.g., IBM Watson (registered trademark) sentiment analysis API) to detect abusive content and recognize the user's emotions. It then generates appropriate warning messages and suggested expressions, which are displayed to the user via their device. After the user completes their post, the server again detects and analyzes the text, converting the abusive content into more neutral expressions and storing them in a database.

[0342] Specific examples

[0343] Example 1: A concrete example of a user pre-post detection process

[0344] The user types "You're really incompetent." The device sends this sentence to the server. The server uses a natural language processing engine to detect the slanderous expression "incompetent." It then uses an emotion engine to recognize that the user is frustrated. The server generates a suggestion saying, "This phrase is negative. Why not change it to 'Maybe you're not trying hard enough?'" and sends it to the device. The device displays this message to the user, who then modifies the sentence based on the suggestion.

[0345] Example 2: A concrete example of the post-submission content transformation process

[0346] A user posts "You're useless." The device sends the post to the server. The server again uses a natural language processing engine to detect the defamatory expression "useless." It then uses an emotion engine to analyze the user's emotions again and generates a sentence that converts "useless" to "try harder." The converted sentence is stored in a database and displayed to other relevant users as needed.

[0347] Prompt Sentence Examples

[0348] 1. Example prompts for generative AI models

[0349] "Check whether the sentence entered by the user contains slanderous content, and perform sentiment analysis to provide a warning message and suggested corrections. For example, provide appropriate corrections for the sentence, 'You are really incompetent.'"

[0350] 2. Example prompts for generative AI models

[0351] "Detect abusive language in user-submitted sentences and generate appropriate suggestions for changes. For example, change 'You're useless' to a more gentle expression."

[0352] As a result, the present invention can effectively detect and correct online slander while taking into consideration the user's feelings, and provide a healthy online environment.

[0353] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0354] Step 1: User enters text

[0355] A user inputs a sentence into a chat app or social networking app. For example, the user inputs "You are really incompetent." This input sentence becomes the input data for the next processing step.

[0356] Step 2: The device sends the text to the server

[0357] The terminal transmits the text entered by the user to the server in real time. At this time, the text is transmitted as text data. The input is the user's text, and the output is the text data transmitted to the server.

[0358] Step 3: The server analyzes the text using a natural language processing engine

[0359] The server passes the received text to a natural language processing engine (e.g., Python library NLTK or spaCy) to detect abusive words and expressions. The input is the text data sent to the server, and the output is a list of abusive words and expressions. For example, the detection result includes the word "incompetent."

[0360] Step 4: The server analyzes the user's emotions using the emotion engine.

[0361] The server passes the received text to an emotion engine (e.g., IBM Watson's emotion analysis API) to analyze the user's emotions. The input is the text data sent to the server, and the output is the identification result of the user's emotional state (e.g., irritation, anger). For example, the analysis result may be that the user is "irritated."

[0362] Step 5: The server generates a warning message and a suggestion

[0363] The server generates a warning message and suggestions based on the results of defamation detection and sentiment analysis. The input is the results of defamation detection and sentiment analysis, and the output is a warning message and suggestions. For example, it generates a message such as, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?"

[0364] Step 6: The device displays a warning message and a suggestion to the user.

[0365] The terminal displays the warning messages and suggestions sent from the server to the user. The input is the warning messages and suggestions sent from the server, and the output is a visual display to the user, who can then correct the text based on this.

[0366] Step 7: User submits revised text

[0367] The user refers to the warning message and suggestions, inputs the corrected text (e.g., "You may not have tried hard enough"), and submits again. The input is the corrected text data, and the output is the submitted text data.

[0368] Step 8: The device sends the corrected text to the server

[0369] The terminal sends the corrected text back to the server. The input is the corrected text data, and the output is the text data that is sent again to the server.

[0370] Step 9: The server again detects the slander and transforms it if necessary.

[0371] The server then passes the corrected text back to the natural language processing engine, which re-detects any abusive content. It then uses an emotion engine to generate a converted text if necessary. The input is the corrected text data, and the output is the final converted text. For example, converting "You're useless" to "Let's try a little harder."

[0372] Step 10: The server saves the converted text in a database and displays it.

[0373] The server saves the generated converted text in a database and displays it to other users as needed. The input is the final converted text data, and the output is saved in a database and displayed to other users.

[0374] Through the processing of each step, users' writing is corrected to include healthy expressions that do not contain slander, thereby maintaining a healthy online environment.

[0375] (Application example 2)

[0376] 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."

[0377] With the increase in online communication, abusive comments have become a problem in live chats and social media. In particular, live stream chats involve the exchange of large volumes of messages in real time, making it difficult to detect and respond to abusive comments in a timely manner. Furthermore, if appropriate feedback is not provided in response to the user's emotions, this can lead to further problems. The present invention aims to solve these problems and provide a healthy online communication environment.

[0378] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving a text entered by a user, means for analyzing the received text and detecting a portion including slander, means for generating a warning message and a suggestion based on the detected portion including slander, means for displaying the generated warning message and suggestion to the user, means for analyzing the user's emotions, and means for adjusting the warning message and the suggestion based on the results of the emotion analysis. This makes it possible to detect slander in real time in live chat and provide appropriate feedback according to the user's emotions.

[0379] "Means for receiving text entered by the user" refers to a function that allows the server to receive text messages entered by the user in live chat or on social networking sites in real time.

[0380] "Means for analyzing received text and detecting parts containing defamatory language" refers to a function that uses natural language processing technology to analyze received text messages and identify defamatory language or inappropriate expressions.

[0381] The "means for generating warning messages and suggestions" is a function that automatically generates warnings and suggestions for improvement for users based on detected abusive content.

[0382] The "means for displaying the generated warning message and suggestion to the user" is a function for displaying the generated warning message and suggestion on the user terminal.

[0383] The "means for analyzing user emotions" is a function that analyzes text messages and determines the user's emotional state (e.g., anger, joy, sadness, etc.).

[0384] The "means for adjusting warning messages and suggestions based on the results of sentiment analysis" is a function for adjusting the contents of warning messages and suggestions for improvement based on the results of sentiment analysis of the user.

[0385] "Means of re-analyzing text that has been posted by the user and detecting parts that contain slander" refers to a function that re-analyzes messages that have already been posted and detects slander.

[0386] "Means for converting into harmless expressions" is a function that converts detected abusive expressions into appropriate and mild expressions.

[0387] The "means for saving and distributing converted text" is a function for saving converted text in a database and distributing it to other users as needed.

[0388] "Means for detecting abusive comments in real time when a message is sent to live chat and issuing warnings and suggestions" refers to a function that detects abusive comments immediately after a user sends a message in live chat and immediately issues a warning and suggests improvements.

[0389] The "means for displaying the results of detection and suggestions on the user's screen" is a function for displaying the detection of slander and the subsequent suggestions on the user's screen in real time.

[0390] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention, taking a system for detecting defamation and providing feedback based on user sentiment in a live chat platform as a concrete example.

[0391] Overall system configuration

[0392] The system consists of several hardware and software components: the user's device (smartphone, PC, tablet, etc.), a server, and the necessary natural language processing and sentiment analysis engines.

[0393] The user's device runs the live chat platform's client software, which sends messages entered by the user to the server, which then analyzes the received messages in real time, detects slanderous comments, and generates warning messages and suggestions for improvement based on the user's sentiment.

[0394] Hardware and software used

[0395] 1. Hardware: Smartphones, PCs, Tablets

[0396] 2. Software: Live chat platform, Google Cloud Natural Language API, IBM Watson Tone Analyzer

[0397] Data Processing and Computation

[0398] First, a user enters a message into the live chat and presses the send button, which sends the message to the server. The server then uses the Google Cloud Natural Language API to analyze the received message and detect any abusive comments. At the same time, it uses IBM Watson Tone Analyzer to analyze the user's sentiment contained in the message.

[0399] If abusive comments are detected, the server generates an appropriate warning message and suggestions for improvement based on the sentiment analysis results, which are then sent back to the user's device for them to review and decide whether to correct them.

[0400] Even if the user does not correct the original abusive message, the server will detect the posted message again, automatically convert it to a more neutral expression, store it in the database, and, in some cases, display the converted message to other users.

[0401] Specific examples

[0402] Example 1: Real-time slander detection and suggestions

[0403] A user types "You're really incompetent" into a live chat. The server uses the Google Cloud Natural Language API to detect the keyword "incompetent" in real time, and IBM Watson Tone Analyzer detects the user's frustration. It displays a warning saying "This expression is negative," and suggests an improvement to the user, saying "Maybe you're not trying hard enough."

[0404] Example 2: Post-post conversion of slander

[0405] If a user posts "You're useless" in the live chat, the server will again detect "useless" and perform sentiment analysis using IBM Watson Tone Analyzer. Depending on the results, it will convert it into a more harmless expression such as "Try harder," store it in the database, and display the converted message as needed.

[0406] Prompt Sentence Examples

[0407] "Detect user messages and analyze their sentiment. Based on that, detect slander and provide an appropriate response to the user.\nInput: "You're really incompetent"\nOutput: "This expression is negative. How about changing it to 'Maybe you're not trying hard enough'?""

[0408] As described above, the system of the present invention analyzes the content and emotions posted by users in real time and provides appropriate warnings and suggestions for improvement, thereby realizing a healthy online communication environment.

[0409] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0410] Step 1:

[0411] The user enters a message into the live chat and presses the send button. At this stage, the text message entered by the user is sent from the terminal to the server. The input is the message entered by the user, and the message content is transmitted to the server.

[0412] Step 2:

[0413] The server passes the received message to a natural language processing engine for analysis. Specifically, it uses the Google Cloud Natural Language API to analyze the message and detect abusive words and phrases. The input is the user's message, and the output is the results of the abusive language detection.

[0414] Step 3:

[0415] Next, the server performs sentiment analysis on the messages where abusive content was detected. It uses IBM Watson Tone Analyzer to analyze the emotional state of the message. The input is the user's message and the abusive content detection results, and the output is the analysis of the user's emotional state.

[0416] Step 4:

[0417] If a slanderous remark is detected and the user's sentiment is determined to be negative, the server generates an appropriate warning message and improvement suggestions. For example, it creates content such as, "This expression is negative. Why not change it to, 'Maybe you're not trying hard enough.'" The input is the detection result of the slanderous remark and the sentiment analysis result, and the output is the generated warning message and improvement suggestions.

[0418] Step 5:

[0419] The generated warning message and improvement suggestions are sent from the server to the user's terminal. The terminal displays them to the user in real time, allowing the user to modify or reconsider the message. The input is the warning message and improvement suggestions, and the output is the screen displayed on the user's terminal.

[0420] Step 6:

[0421] Once the user has finally confirmed sending the message, it is sent to the server again. The server then passes the message to the natural language processing engine again to detect abusive content. The input is the message that was finally sent, and the output is the result of whether abusive content was detected again.

[0422] Step 7:

[0423] The server converts the detected abusive content into a more neutral expression. For example, it automatically converts a message such as "You're useless" into "Try harder." The input is a message containing abusive content, and the output is the converted, more neutral expression.

[0424] Step 8:

[0425] The converted message is stored in a database and can be displayed to other users as needed. The input is the converted message, and the output is the data stored in the database and the screen displayed.

[0426] Through these steps, the system detects defamatory comments in real time and provides feedback based on the user's emotions, helping to create a healthy online communication environment.

[0427] 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.

[0428] 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.

[0429] 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.

[0430] [Second embodiment]

[0431] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0432] 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.

[0433] 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).

[0434] 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.

[0435] 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.

[0436] 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).

[0437] 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.

[0438] 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.

[0439] 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.

[0440] 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.

[0441] 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.

[0442] 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."

[0443] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[0444] System Overview

[0445] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[0446] Program processing

[0447] 1. User pre-post detection process

[0448] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[0449] Next, the server generates a warning message based on the detection results and makes a suggestion such as, "This word is negative. Why not change it to 'You may not be trying hard enough.'" This suggestion and warning message are sent to the terminal, which displays it to the user. The user can then revise or reconsider the sentence based on the suggestion.

[0450] 2. Post-submission content transformation process

[0451] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[0452] The server generates the converted text, "You should try a little harder," and stores it in the database. It also displays the converted post to the user and other related users as needed. In this way, defamatory posts are detected and corrected before and after the fact, providing a safe and healthy online environment.

[0453] Specific examples

[0454] 1. Example 1: A concrete example of a user pre-post detection process

[0455] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." The server then sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[0456] 2. Example 2: A concrete example of the post-submission content transformation process

[0457] The user posts "You're useless." The device sends the post to the server, which again detects "useless." The server converts this part to "try a little harder," generating the final sentence, "You try a little harder." The converted sentence is stored in a database and displayed to the user and other relevant users as needed.

[0458] As a result, the present invention can effectively detect and correct online slander and provide a healthy online environment.

[0459] The processing flow will be explained below.

[0460] User pre-post detection process

[0461] Step 1:

[0462] The user enters the content to post.

[0463] The user types "You're really incompetent" into a text box in a chat app.

[0464] Step 2:

[0465] The terminal sends the input text to the server.

[0466] The terminal sends the input text data to the server as a POST request.

[0467] Step 3:

[0468] The server passes the received text to a natural language processing engine.

[0469] The server passes the received text "You are truly incompetent" to the NLP engine.

[0470] Step 4:

[0471] The server uses a natural language processing engine to detect slander.

[0472] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[0473] Step 5:

[0474] The server generates warning messages and suggestions based on the detection results.

[0475] For the word "incompetent," the server creates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[0476] Step 6:

[0477] The server generates warning messages and suggestions and sends them to the device.

[0478] The server sends warning messages and suggestions in JSON format to the device.

[0479] Step 7:

[0480] The terminal displays a warning message and a suggestion to the user.

[0481] The terminal displays the received message to the user in a pop-up window or dialog box.

[0482] Post-submission content transformation process

[0483] Step 1:

[0484] The user completes the post.

[0485] The user presses the "Submit" button to complete the submission.

[0486] Step 2:

[0487] The device sends the posted text to the server.

[0488] The device sends the message "You're useless" to the server.

[0489] Step 3:

[0490] The server passes the posted text back to the natural language processing engine.

[0491] The server passes the received text "You're useless" to the NLP engine.

[0492] Step 4:

[0493] The server detects the slander again.

[0494] The server recognizes the word "useless" as a slur and identifies that part.

[0495] Step 5:

[0496] The server converts abusive comments into harmless language.

[0497] The server translates "useless" into "try harder."

[0498] Step 6:

[0499] The server generates the converted text and stores it in a database.

[0500] The server saves the revised sentence, "You should try a little harder," in the database.

[0501] Step 7:

[0502] The server delivers the converted text as needed.

[0503] The server displays the modified post to the user and other interested users.

[0504] This allows for effective detection and correction of defamatory comments both before and after users post.

[0505] Example 1

[0506] 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."

[0507] On modern online platforms, abusive language can easily spread, negatively impacting communication between users. Therefore, there is a need for a system that can proactively detect abusive posts, warn users, and prompt them to change their language to more appropriate ones. However, current systems struggle to detect and convert abusive language in sufficient real-time, often resulting in users leaving abusive language in their posts after posting. Furthermore, the lack of natural suggestions for converting to more appropriate language makes it difficult for users to edit their posts in a way that is more acceptable to them.

[0508] 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.

[0509] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text using a natural language processing engine to detect portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for displaying the generated warning message and suggestion to the user, and means for correcting the text entered by the user based on the user's reaction. As a result, when a user enters a text containing slander, a warning and appropriate suggestions are displayed in real time, allowing the user to immediately correct the text and realize safe and healthy online communication.

[0510] "Means for receiving text entered by a user" is a function for sending text entered by a user into a chat app or SNS app to a server.

[0511] A "natural language processing engine" is a software technology for understanding, interpreting, and manipulating human language, including analyzing and classifying text and detecting defamation.

[0512] "Abusive language detection means" is an analytical function that identifies offensive or negative language in a text.

[0513] The "means for generating warning messages and suggestions" is a function for warning the user about detected abusive language and suggesting appropriate alternative expressions.

[0514] The "means for displaying to the user" is a function for displaying warning messages and suggestions on the user's terminal in real time.

[0515] The "means for correcting text entered by the user based on the user's response" is a function that allows the user to correct the text entered in accordance with the warnings and suggestions received by the user.

[0516] The "means for storing and distributing" is a function for storing the final corrected text in a database and distributing it to relevant users as necessary.

[0517] To specifically implement the present invention, the following system configuration and processing procedures are required.

[0518] System Overview

[0519] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[0520] Specific hardware and software configurations

[0521] server

[0522] The server receives user input and uses a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) to detect and analyze defamatory content.

[0523] The server generates warning messages and suggestions based on the detection results and sends them to the user's terminal.

[0524] Terminal

[0525] The device sends text entered by the user in a chat app or SNS app to the server.

[0526] The terminal displays the warning messages and suggestions returned by the server to the user.

[0527] The terminal transmits the text corrected by the user to the server again.

[0528] User

[0529] A user inputs text into a chat app or a social networking app.

[0530] The user corrects the text based on the displayed warning message and suggestions.

[0531] System operation example

[0532] User pre-post detection process

[0533] 1. The user types "You are truly incompetent." The device sends this sentence to the server.

[0534] 2. The server analyzes the received text using a natural language processing engine and determines that the word "incompetent" is defamatory.

[0535] 3. The server generates a warning message and suggestion saying, "This word is negative. Why not change it to 'You may not be trying hard enough'?" and sends it to the terminal.

[0536] 4. The terminal displays a warning message and a suggestion to the user.

[0537] 5. The user takes the suggestion into consideration and revises the sentence to "Perhaps you are not trying hard enough."

[0538] Post-submission content transformation process

[0539] 1. A user posts "You're useless." The device sends the post to the server.

[0540] 2. The server analyzes the received text again using a natural language processing engine and determines that the word "useless" is defamatory.

[0541] 3. The server converts this to "You should try a little harder" and stores the final generated sentence in the database.

[0542] 4. The converted text is displayed to the user and other relevant users as needed.

[0543] Through these processes, the system can effectively detect and correct online defamation and provide a safe and healthy online environment.

[0544] Here is an example of a prompt for the generative AI model:

[0545] Prompt Sentence Examples

[0546] Please change the following sentences to non-libelous language:

[0547] "You're really incompetent."

[0548]

[0549] Please change the following sentences to non-libelous language:

[0550] "You're useless."

[0551] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0552] User pre-post detection process

[0553] Step 1:

[0554] The user inputs a sentence.

[0555] Input: A user types "You're so incompetent" into a chat or social media app.

[0556] Action: The user types a sentence on the device's keyboard.

[0557] Output: The input sentence "You are truly incompetent."

[0558] Step 2:

[0559] The terminal sends the input text to the server.

[0560] Input: The sentence "You are truly incompetent" typed by the user.

[0561] Action: The device sends this text to the server via the network.

[0562] Output: The sentence "You are truly incompetent" received by the server.

[0563] Step 3:

[0564] The server passes the received text to a natural language processing engine for analysis.

[0565] Input: The sentence "You are truly incompetent" received by the server.

[0566] How it works: The server passes this sentence to a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) for analysis. At the same time, it detects the slanderous word "incompetent."

[0567] Output: The word "incompetent" is determined to be libelous.

[0568] Step 4:

[0569] The server generates warning messages and suggestions based on the slander detected.

[0570] Input: The word "incompetent" which was determined to be libelous.

[0571] What it does: The server verifies that the word "incompetent" is negative and generates a warning message: "This word is negative." and a suggestion: "Maybe you're not trying hard enough?"

[0572] Output: Warning messages and suggestions.

[0573] Step 5:

[0574] The server sends warning messages and suggestions to the device.

[0575] Input: Generated warning messages and suggestions.

[0576] Operation: The server sends a warning message and suggestions to the user's device.

[0577] Output: A warning message and a suggestion will be displayed on the user's terminal.

[0578] Step 6:

[0579] The terminal displays a warning message and a suggestion to the user.

[0580] Input: Warning messages and suggestions sent by the server.

[0581] Behavior: Displays a warning message and suggestion in a pop-up or message display area on the device screen.

[0582] Output: The user sees a warning message and a suggestion.

[0583] Step 7:

[0584] The user corrects the text.

[0585] Input: Warning messages and suggestions.

[0586] Action: Based on the suggestion, the user amends the original sentence "You are really incompetent" to "Maybe you're not trying hard enough."

[0587] Output: Corrected sentence: "Maybe you're not trying hard enough."

[0588] Post-submission content transformation process

[0589] Step 1:

[0590] The user completes the post.

[0591] Input: Corrected sentence: "Maybe you're not trying hard enough."

[0592] Action: The user clicks the submit button and confirms the text.

[0593] Output: The confirmed post text "Maybe you're not trying hard enough."

[0594] Step 2:

[0595] The device sends the posted text to the server.

[0596] Input: Confirmed post text: "Maybe you're not trying hard enough."

[0597] Action: The device sends the confirmed text to the server.

[0598] Output: The sentence received by the server: "Maybe you're not trying hard enough."

[0599] Step 3:

[0600] The server then analyzes the text again using a natural language processing engine.

[0601] Input: The sentence "You may not be trying hard enough" received by the server.

[0602] What happens: The server passes this sentence to a natural language processing engine for analysis, and verifies that it does not contain any slander.

[0603] Output: After analysis, it is determined that the sentence "Maybe you're not trying hard enough" does not contain any slander.

[0604] Step 4:

[0605] The server stores the final converted text in a database.

[0606] Input: Analysis result shows that the sentence "You may not be trying hard enough" does not contain any slander.

[0607] What it does: The server saves the sentence in a database based on the analysis results.

[0608] Output: The sentence "Maybe you're not trying hard enough" stored in the database.

[0609] Step 5:

[0610] If necessary, the converted text is displayed to the relevant users.

[0611] Input: The sentence stored in the database: "Maybe you're not trying hard enough."

[0612] How it works: The system will display the new text to relevant users (e.g., users who receive the post on the subject) as needed.

[0613] Output: The converted sentence "Maybe you're not trying hard enough" displayed on the relevant user's device.

[0614] The above are the specific steps in the pre-post detection process and post-post content conversion process, which will effectively detect and correct online slander and provide a safe and healthy online environment.

[0615] (Application example 1)

[0616] 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."

[0617] Defamation on the Internet has a significant negative impact on users' mental health and can worsen the atmosphere of the entire community. In particular, in content distribution services, where many users post comments in real time, monitoring and correcting them is difficult. Conventional methods have difficulty detecting and appropriately correcting defamation in real time, making it difficult to provide a safe and healthy online environment. To solve this problem, a system that utilizes natural language processing technology and generative AI models to detect and correct user comments in real time is needed.

[0618] 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.

[0619] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text and detecting portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for generating the generated warning message and suggestion using a natural language generation model, and means for displaying the generated warning message and suggestion to the user. This makes it possible to monitor user comments in real time, detect slander, and immediately suggest corrections, thereby providing a constructive comment environment.

[0620] The "means for receiving text entered by a user" refers to a technology or device for transmitting text information entered by a user through a device to a server and acquiring that text information.

[0621] "Means for analyzing received text and detecting defamatory or libelous parts" refers to technologies or processes that use natural language processing engines or similar to analyze text data and identify negative or offensive words or expressions contained therein.

[0622] "Means for generating warning messages and suggestions based on the detected defamatory content" refers to technology that generates messages to warn users and suggestions showing more appropriate and constructive examples of expression based on the detection results.

[0623] The "means for displaying the generated warning message and suggestions to the user" refers to a technology or interface for displaying the generated message and suggestions on the screen of the user's device.

[0624] "Means for generation using a natural language generation model" refers to technology for generating warning messages and suggestions in natural language using a generative AI model (e.g., GPT-3 or T5).

[0625] "Means of re-analyzing the text that the user has posted and detecting any parts that contain defamatory comments" refers to a technology that re-analyzes the text that the user has finally posted and identifies any defamatory comments that remain within it.

[0626] "Means for converting detected abusive content into less offensive language" refers to technologies or algorithms that convert detected negative or offensive words or phrases into more neutral or positive language.

[0627] "Means for storing and distributing the converted text" refers to the technology used to store the corrected text on a server and distribute it to relevant users and platforms as needed.

[0628] The present invention is a system that receives text entered by a user in real time, analyzes it, and detects and corrects slander. Specific embodiments of the present invention will be described below.

[0629] System Overview

[0630] This system consists of three elements: a server, a terminal, and a user.

[0631] server:

[0632] The server uses a natural language processing engine (e.g., a T5 model using the transformers library) to analyze the text sent by the user. It detects abusive language in the text and generates warning messages and suggestions based on the detected language. The suggestions are generated by a natural language generation model and suggest appropriate ways of expressing the text to the user.

[0633] Device:

[0634] The terminal is a device (smartphone or computer) where the user inputs text. When the user inputs text using a chat app or SNS app, the text is sent to the server in real time. The terminal also displays warning messages and suggestions received from the server.

[0635] User:

[0636] The user is the subject of inputting text and receiving warning messages and suggestions from the system. The user can then correct the text based on the suggestions and submit it again.

[0637] Program processing

[0638] The server first receives the text entered by the user. It then analyzes the text using a natural language processing engine (for example, a T5 model using the transformers library) to detect parts that contain slander. Based on the detected parts, it generates warning messages and suggestions. These suggestions are created in natural Japanese sentences using a generative AI model.

[0639] The server sends the generated warning message and suggestions to the user's device, where the user can view it. If the user corrects the text as suggested, the corrected text can be sent back to the server. The server receives it, analyzes it again to ensure it does not contain any slander, and then stores and distributes it.

[0640] Specific examples

[0641] For example, if a user enters the sentence "You are really incompetent," the server detects the word "incompetent" as a slander. The server then generates a suggestion saying, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?" The following is an example of a prompt:

[0642] Your comment is negative. Try rephrasing it this way: Maybe you're not trying hard enough.

[0643] The suggestion is displayed on the terminal, and the user can refer to it and correct it by saying, "Maybe you're not trying hard enough." The server then analyzes the corrected sentence again, and after confirming that there are no problems, stores and distributes the sentence.

[0644] In this way, the system can monitor user comments in real time, detect and correct defamatory comments, and provide a healthy online environment.

[0645] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0646] Step 1:

[0647] A user inputs text using a chat app or SNS app. The input text is sent in real time from the device to a server. The server receives the text data sent by the user. The input is the text entered by the user, and the output is a confirmation that the text has been received.

[0648] Step 2:

[0649] The server passes the received text data to a natural language processing engine to detect abusive content. In this step, it filters out negative keywords such as "incompetent" or "useless." The input is a sentence entered by the user, and the output is a list of abusive keywords and phrases.

[0650] Step 3:

[0651] The server generates warning messages and suggestions based on the detected abusive comments. It uses a generative AI model (such as the T5 model) to create warning messages and suggestions in natural-sounding sentences. The input is a list of abusive comments and the original sentences, and the output is the generated warning messages and suggestions.

[0652] Step 4:

[0653] The server sends the generated warning message and suggestions to the terminal, which displays the message and suggestions to the user. The input is the warning message and suggestions sent from the server, and the output is displayed on the user's device screen.

[0654] Step 5:

[0655] The user then reconsiders whether to revise the sentence based on the displayed suggestions. In this step, the user inputs new sentences based on the suggestions received from the server. The inputs are the warning message and the suggestions, and the output is the revised sentence.

[0656] Step 6:

[0657] The user then sends the revised text back to the server. The server then receives this new text and uses a natural language processing engine to detect the abusive text again. The input is the revised text, and the output is the detection results.

[0658] Step 7:

[0659] The server checks that the text is free of slander and then stores the final text in a database and distributes it as needed. This step involves storing the final text and distributing it to relevant users and platforms. The input is the final, corrected text, and the output is a safe text stored in a database and distributed.

[0660] By following these steps, the system can detect defamatory comments in real time and prompt appropriate corrections.

[0661] 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.

[0662] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[0663] System Overview

[0664] This system receives and analyzes text entered by users, detects abusive language, and recognizes the user's emotions to provide warning messages and suggest appropriate expressions. This system combines a natural language processing engine and an emotion engine to detect abusive language and respond appropriately based on the user's emotions. It also has the ability to convert abusive language into harmless language, which it then stores and distributes.

[0665] Program processing

[0666] 1. User pre-post detection process

[0667] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[0668] The server then analyzes the user's emotions using an emotion engine, which recognizes their emotional state based on the tone and word choice of the text they type. For example, if the server determines that the user is annoyed, it will adjust warning messages and suggestions accordingly.

[0669] Based on the detection results and sentiment analysis results, the server generates a warning message and appropriate suggestions, such as "This word is negative. Why not change it to 'You may not be trying hard enough'?" This suggestion and warning message are sent to the device, which then displays it to the user. The user can then revise or reconsider the sentence based on the suggestions.

[0670] 2. Post-submission content transformation process

[0671] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[0672] The server also uses an emotion engine to analyze the user's emotions related to the post and generates an optimal, non-offensive expression based on this. The server generates the converted sentence, "You should try a little harder," and stores it in a database. The server also displays the converted post content to the user and other related users as needed.

[0673] Specific examples

[0674] 1. Example 1: A concrete example of a user pre-post detection process

[0675] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." Next, the server recognizes the user's frustration based on the emotion engine, and sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[0676] 2. Example 2: A concrete example of the post-submission content transformation process

[0677] The user posts "You're useless." The device sends the post to the server, which again recognizes "useless" and converts it to "try a little harder." It also uses an emotion engine to confirm the user's emotions and selects an appropriate expression. The server generates the converted sentence "try a little harder" and saves it in the database. The converted sentence is then displayed as needed.

[0678] As a result, the present invention can detect and correct online slander while taking into consideration the user's feelings, thereby providing a healthy online environment.

[0679] The processing flow will be explained below.

[0680] User pre-post detection process

[0681] Step 1:

[0682] The user enters the content to post.

[0683] The user types "You're really incompetent" into a text box in a chat app.

[0684] Step 2:

[0685] The terminal sends the input text to the server.

[0686] The terminal sends the text data entered by the user to the server as a POST request.

[0687] Step 3:

[0688] The server passes the received text to a natural language processing engine.

[0689] The server passes the received text "You are truly incompetent" to a natural language processing engine, which analyzes the sentence.

[0690] Step 4:

[0691] The server uses a natural language processing engine to detect slander.

[0692] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[0693] Step 5:

[0694] The server analyzes the user's emotions using an emotion engine.

[0695] The server uses an emotion engine to recognize the emotion of irritation based on the tone and word choice of the text entered by the user.

[0696] Step 6:

[0697] The server generates warning messages and suggestions based on the detection results and sentiment analysis results.

[0698] For the word "incompetent," the server generates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[0699] Step 7:

[0700] The server generates warning messages and suggestions and sends them to the device.

[0701] The server sends the generated warning message and suggestions to the device in JSON format.

[0702] Step 8:

[0703] The terminal displays a warning message and a suggestion to the user.

[0704] The terminal displays the received warning messages and suggestions to the user in pop-up windows or dialog boxes.

[0705] Post-submission content transformation process

[0706] Step 1:

[0707] The user completes the post.

[0708] The user presses the "Submit" button to complete the submission.

[0709] Step 2:

[0710] The device sends the posted text to the server.

[0711] The device sends the message "You're useless" to the server.

[0712] Step 3:

[0713] The server passes the posted text back to the natural language processing engine.

[0714] The server passes the received text, "You're useless," to a natural language processing engine for further analysis.

[0715] Step 4:

[0716] The server detects the slander again.

[0717] The server recognizes the word "useless" as a slur and identifies that part.

[0718] Step 5:

[0719] The server uses an emotion engine to analyze the user's emotions regarding the posted content.

[0720] The server uses an emotion engine to ascertain the user's emotional state associated with the post.

[0721] Step 6:

[0722] The server converts abusive comments into harmless language.

[0723] The server converts "useless" into "let's try a little harder" and selects the appropriate expression.

[0724] Step 7:

[0725] The server generates the converted text and stores it in a database.

[0726] The server stores the generated sentence "You should try a little harder" in a database.

[0727] Step 8:

[0728] The server delivers the converted text as needed.

[0729] The server displays the converted post to the user and other interested users.

[0730] As a result, the present invention can effectively detect and correct slanderous comments while taking into account the user's feelings during the pre-post and post-post processes.

[0731] Example 2

[0732] 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."

[0733] The increase in slander and negative comments on the Internet has made it difficult to maintain a healthy online environment. Users often unconsciously use negative language, and these comments are often posted without being edited. Furthermore, even for content that has already been posted, there will be users who will suffer harm unless slander is reduced. A system that can solve these problems is needed.

[0734] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for receiving a sentence entered by a user, a means for analyzing the received sentence using a natural language processing engine and detecting a portion containing slander, a means for generating a warning message and a suggestion using an emotion engine based on the detected portion containing slander, and a means for displaying the generated warning message and suggestion to the user. This allows the user to detect slander before posting a sentence and make appropriate corrections. Furthermore, by re-detecting slanderous portions in content after posting and converting them into non-savage expressions, the health of the online environment can be maintained.

[0735] "User" means an individual or organization that uses the system to input and post text.

[0736] "Text" refers to text data entered by a user and sent to the system.

[0737] A "natural language processing engine" is an algorithm or software that analyzes text data to recognize slander and specific patterns.

[0738] "Slander" refers to negative language that hurts or speaks ill of others.

[0739] An "emotion engine" is an algorithm or software that analyzes and identifies a user's emotions from the tone and content of a sentence.

[0740] A "warning message" is a message that notifies the user that the text they have entered contains slander and urges them to correct it.

[0741] A "suggestion" is a message containing alternative expressions to correct the sentence entered by the user.

[0742] "Terminal" refers to the device (e.g., smartphone, tablet, or PC) that a user uses to input text and send it to the system.

[0743] A "server" is a computer system that receives data sent from a terminal and analyzes and processes it.

[0744] "Posting" means the act of publishing text entered by a user on an online platform.

[0745] "Database" is an information management system for storing posted texts and generated messages.

[0746] The following describes in detail the mode for carrying out the present invention. The present invention is a system that receives text entered by a user, detects and corrects slanderous comments, and further recognizes the user's emotions to suggest appropriate warning messages and expressions. This system is composed of a server, a terminal, and a user.

[0747] System Overview

[0748] When a user enters text using a device (e.g., smartphone, tablet, or PC), the text is sent to the server. The server analyzes the received text using a natural language processing engine (e.g., Python library NLTK or spaCy) and an emotion engine (e.g., IBM Watson's emotion analysis API) to detect abusive content and recognize the user's emotions. It then generates appropriate warning messages and suggested expressions, which are displayed to the user via their device. After the user has completed posting, the server again detects and analyzes the text, converting the abusive content into more neutral expressions and storing them in a database.

[0749] Specific examples

[0750] Example 1: A concrete example of a user pre-post detection process

[0751] The user types "You're really incompetent." The device sends this sentence to the server. The server uses a natural language processing engine to detect the slanderous expression "incompetent." It then uses an emotion engine to recognize that the user is frustrated. The server generates a suggestion saying, "This phrase is negative. Why not change it to 'Maybe you're not trying hard enough?'" and sends it to the device. The device displays this message to the user, who then modifies the sentence based on the suggestion.

[0752] Example 2: A concrete example of the post-submission content transformation process

[0753] A user posts "You're useless." The device sends the post to the server. The server again uses a natural language processing engine to detect the defamatory expression "useless." It then uses an emotion engine to analyze the user's emotions again and generates a sentence that converts "useless" to "try harder." The converted sentence is stored in a database and displayed to other relevant users as needed.

[0754] Prompt Sentence Examples

[0755] 1. Example prompts for generative AI models

[0756] "Check whether the sentence entered by the user contains slanderous content, and perform sentiment analysis to provide a warning message and suggested corrections. For example, provide appropriate corrections for the sentence, 'You are really incompetent.'"

[0757] 2. Example prompts for generative AI models

[0758] "Detect abusive language in user-submitted sentences and generate appropriate suggestions for changes. For example, change 'You're useless' to a more gentle expression."

[0759] As a result, the present invention can effectively detect and correct online slander while taking into consideration the user's feelings, and provide a healthy online environment.

[0760] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0761] Step 1: User enters text

[0762] A user inputs a sentence into a chat app or social networking app. For example, the user inputs "You are really incompetent." This input sentence becomes the input data for the next processing step.

[0763] Step 2: The device sends the text to the server

[0764] The terminal transmits the text entered by the user to the server in real time. At this time, the text is transmitted as text data. The input is the user's text, and the output is the text data transmitted to the server.

[0765] Step 3: The server analyzes the text using a natural language processing engine

[0766] The server passes the received text to a natural language processing engine (e.g., Python library NLTK or spaCy) to detect abusive words and expressions. The input is the text data sent to the server, and the output is a list of abusive words and expressions. For example, the detection result includes the word "incompetent."

[0767] Step 4: The server analyzes the user's emotions using the emotion engine.

[0768] The server passes the received text to an emotion engine (e.g., IBM Watson's emotion analysis API) to analyze the user's emotions. The input is the text data sent to the server, and the output is the identification result of the user's emotional state (e.g., irritation, anger). For example, the analysis result may be that the user is "irritated."

[0769] Step 5: The server generates a warning message and a suggestion

[0770] The server generates a warning message and suggestions based on the results of defamation detection and sentiment analysis. The input is the results of defamation detection and sentiment analysis, and the output is a warning message and suggestions. For example, it generates a message such as, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?"

[0771] Step 6: The device displays a warning message and a suggestion to the user.

[0772] The terminal displays the warning messages and suggestions sent from the server to the user. The input is the warning messages and suggestions sent from the server, and the output is a visual display to the user, who can then correct the text based on this.

[0773] Step 7: User submits revised text

[0774] The user refers to the warning message and suggestions, inputs the corrected text (e.g., "You may not have tried hard enough"), and submits again. The input is the corrected text data, and the output is the submitted text data.

[0775] Step 8: The device sends the corrected text to the server

[0776] The terminal sends the corrected text back to the server. The input is the corrected text data, and the output is the text data that is sent again to the server.

[0777] Step 9: The server again detects the slander and transforms it if necessary.

[0778] The server then passes the corrected text back to the natural language processing engine, which re-detects any abusive content. It then uses an emotion engine to generate a converted text if necessary. The input is the corrected text data, and the output is the final converted text. For example, converting "You're useless" to "Let's try a little harder."

[0779] Step 10: The server saves the converted text in a database and displays it.

[0780] The server saves the generated converted text in a database and displays it to other users as needed. The input is the final converted text data, and the output is saved in a database and displayed to other users.

[0781] Through the processing of each step, users' writing is corrected to include healthy expressions that do not contain slander, thereby maintaining a healthy online environment.

[0782] (Application example 2)

[0783] 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."

[0784] With the increase in online communication, abusive comments have become a problem in live chats and social media. In particular, live stream chats involve the exchange of large volumes of messages in real time, making it difficult to detect and respond to abusive comments in a timely manner. Furthermore, if appropriate feedback is not provided in response to the user's emotions, this can lead to further problems. The present invention aims to solve these problems and provide a healthy online communication environment.

[0785] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving a text entered by a user, means for analyzing the received text and detecting a portion including slander, means for generating a warning message and a suggestion based on the detected portion including slander, means for displaying the generated warning message and suggestion to the user, means for analyzing the user's emotions, and means for adjusting the warning message and the suggestion based on the results of the emotion analysis. This makes it possible to detect slander in real time in live chat and provide appropriate feedback according to the user's emotions.

[0786] "Means for receiving text entered by the user" refers to a function that allows the server to receive text messages entered by the user in live chat or on social networking sites in real time.

[0787] "Means for analyzing received text and detecting parts containing defamatory language" refers to a function that uses natural language processing technology to analyze received text messages and identify defamatory language or inappropriate expressions.

[0788] The "means for generating warning messages and suggestions" is a function that automatically generates warnings and suggestions for improvement for users based on detected abusive content.

[0789] The "means for displaying the generated warning message and suggestion to the user" is a function for displaying the generated warning message and suggestion on the user terminal.

[0790] The "means for analyzing user emotions" is a function that analyzes text messages and determines the user's emotional state (e.g., anger, joy, sadness, etc.).

[0791] The "means for adjusting warning messages and suggestions based on the results of sentiment analysis" is a function for adjusting the contents of warning messages and suggestions for improvement based on the results of sentiment analysis of the user.

[0792] "Means of re-analyzing text that has been posted by the user and detecting parts that contain slander" refers to a function that re-analyzes messages that have already been posted and detects slander.

[0793] "Means for converting into harmless expressions" is a function that converts detected abusive expressions into appropriate and mild expressions.

[0794] The "means for saving and distributing converted text" is a function for saving converted text in a database and distributing it to other users as needed.

[0795] "Means for detecting abusive comments in real time when a message is sent to live chat and issuing warnings and suggestions" refers to a function that detects abusive comments immediately after a user sends a message in live chat and immediately issues a warning and suggests improvements.

[0796] The "means for displaying the results of detection and suggestions on the user's screen" is a function for displaying the detection of slander and the subsequent suggestions on the user's screen in real time.

[0797] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention, taking a system for detecting defamation and providing feedback based on user sentiment in a live chat platform as a concrete example.

[0798] Overall system configuration

[0799] The system consists of several hardware and software components: the user's device (smartphone, PC, tablet, etc.), a server, and the necessary natural language processing and sentiment analysis engines.

[0800] The user's device runs the live chat platform's client software, which sends messages entered by the user to the server, which then analyzes the received messages in real time, detects slanderous comments, and generates warning messages and suggestions for improvement based on the user's sentiment.

[0801] Hardware and software used

[0802] 1. Hardware: Smartphones, PCs, Tablets

[0803] 2. Software: Live chat platform, Google Cloud Natural Language API, IBM Watson Tone Analyzer

[0804] Data Processing and Computation

[0805] First, a user enters a message into the live chat and presses the send button, which sends the message to the server. The server then uses the Google Cloud Natural Language API to analyze the received message and detect any abusive comments. At the same time, it uses IBM Watson Tone Analyzer to analyze the user's sentiment contained in the message.

[0806] If abusive comments are detected, the server generates an appropriate warning message and suggestions for improvement based on the sentiment analysis results, which are then sent back to the user's device for them to review and decide whether to correct them.

[0807] Even if the user does not correct the original abusive message, the server will detect the posted message again, automatically convert it to a more neutral expression, store it in the database, and, in some cases, display the converted message to other users.

[0808] Specific examples

[0809] Example 1: Real-time slander detection and suggestions

[0810] A user types "You're really incompetent" into a live chat. The server uses the Google Cloud Natural Language API to detect the keyword "incompetent" in real time, and IBM Watson Tone Analyzer detects the user's frustration. It displays a warning saying "This expression is negative," and suggests an improvement to the user, saying "Maybe you're not trying hard enough."

[0811] Example 2: Post-post conversion of slander

[0812] If a user posts "You're useless" in the live chat, the server will again detect "useless" and perform sentiment analysis using IBM Watson Tone Analyzer. Depending on the results, it will convert it into a more harmless expression such as "Try harder," store it in the database, and display the converted message as needed.

[0813] Prompt Sentence Examples

[0814] "Detect user messages and analyze their sentiment. Based on that, detect slander and provide an appropriate response to the user.\nInput: "You're really incompetent"\nOutput: "This expression is negative. How about changing it to 'Maybe you're not trying hard enough'?""

[0815] As described above, the system of the present invention analyzes the content and emotions posted by users in real time and provides appropriate warnings and suggestions for improvement, thereby realizing a healthy online communication environment.

[0816] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0817] Step 1:

[0818] The user enters a message into the live chat and presses the send button. At this stage, the text message entered by the user is sent from the terminal to the server. The input is the message entered by the user, and the message content is transmitted to the server.

[0819] Step 2:

[0820] The server passes the received message to a natural language processing engine for analysis. Specifically, it uses the Google Cloud Natural Language API to analyze the message and detect abusive words and phrases. The input is the user's message, and the output is the results of the abusive language detection.

[0821] Step 3:

[0822] Next, the server performs sentiment analysis on the messages where abusive content was detected. It uses IBM Watson Tone Analyzer to analyze the emotional state of the message. The input is the user's message and the abusive content detection results, and the output is the analysis of the user's emotional state.

[0823] Step 4:

[0824] If a slanderous remark is detected and the user's sentiment is determined to be negative, the server generates an appropriate warning message and improvement suggestions. For example, it creates content such as, "This expression is negative. Why not change it to, 'Maybe you're not trying hard enough.'" The input is the detection result of the slanderous remark and the sentiment analysis result, and the output is the generated warning message and improvement suggestions.

[0825] Step 5:

[0826] The generated warning message and improvement suggestions are sent from the server to the user's terminal. The terminal displays them to the user in real time, allowing the user to modify or reconsider the message. The input is the warning message and improvement suggestions, and the output is the screen displayed on the user's terminal.

[0827] Step 6:

[0828] Once the user has finally confirmed sending the message, it is sent to the server again. The server then passes the message to the natural language processing engine again to detect abusive content. The input is the message that was finally sent, and the output is the result of whether abusive content was detected again.

[0829] Step 7:

[0830] The server converts the detected abusive content into a more neutral expression. For example, it automatically converts a message such as "You're useless" into "Try harder." The input is a message containing abusive content, and the output is the converted, more neutral expression.

[0831] Step 8:

[0832] The converted message is stored in a database and can be displayed to other users as needed. The input is the converted message, and the output is the data stored in the database and the screen displayed.

[0833] Through these steps, the system detects defamatory comments in real time and provides feedback based on the user's emotions, helping to create a healthy online communication environment.

[0834] 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.

[0835] 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.

[0836] 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.

[0837] [Third embodiment]

[0838] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0839] 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.

[0840] 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).

[0841] 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.

[0842] 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.

[0843] 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).

[0844] 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.

[0845] 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.

[0846] 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.

[0847] 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.

[0848] 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.

[0849] 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."

[0850] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[0851] System Overview

[0852] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[0853] Program processing

[0854] 1. User pre-post detection process

[0855] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[0856] Next, the server generates a warning message based on the detection results and makes a suggestion such as, "This word is negative. Why not change it to 'You may not be trying hard enough.'" This suggestion and warning message are sent to the terminal, which displays it to the user. The user can then revise or reconsider the sentence based on the suggestion.

[0857] 2. Post-submission content transformation process

[0858] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[0859] The server generates the converted text, "You should try a little harder," and stores it in the database. It also displays the converted post to the user and other related users as needed. In this way, defamatory posts are detected and corrected before and after the fact, providing a safe and healthy online environment.

[0860] Specific examples

[0861] 1. Example 1: A concrete example of a user pre-post detection process

[0862] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." The server then sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[0863] 2. Example 2: A concrete example of the post-submission content transformation process

[0864] The user posts "You're useless." The device sends the post to the server, which again detects "useless." The server converts this part to "try a little harder," generating the final sentence, "You try a little harder." The converted sentence is stored in a database and displayed to the user and other relevant users as needed.

[0865] As a result, the present invention can effectively detect and correct online slander and provide a healthy online environment.

[0866] The processing flow will be explained below.

[0867] User pre-post detection process

[0868] Step 1:

[0869] The user enters the content to post.

[0870] The user types "You're really incompetent" into a text box in a chat app.

[0871] Step 2:

[0872] The terminal sends the input text to the server.

[0873] The terminal sends the input text data to the server as a POST request.

[0874] Step 3:

[0875] The server passes the received text to a natural language processing engine.

[0876] The server passes the received text "You are truly incompetent" to the NLP engine.

[0877] Step 4:

[0878] The server uses a natural language processing engine to detect slander.

[0879] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[0880] Step 5:

[0881] The server generates warning messages and suggestions based on the detection results.

[0882] For the word "incompetent," the server creates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[0883] Step 6:

[0884] The server generates warning messages and suggestions and sends them to the device.

[0885] The server sends warning messages and suggestions in JSON format to the device.

[0886] Step 7:

[0887] The terminal displays a warning message and a suggestion to the user.

[0888] The terminal displays the received message to the user in a pop-up window or dialog box.

[0889] Post-submission content transformation process

[0890] Step 1:

[0891] The user completes the post.

[0892] The user presses the "Submit" button to complete the submission.

[0893] Step 2:

[0894] The device sends the posted text to the server.

[0895] The device sends the message "You're useless" to the server.

[0896] Step 3:

[0897] The server passes the posted text back to the natural language processing engine.

[0898] The server passes the received text "You're useless" to the NLP engine.

[0899] Step 4:

[0900] The server detects the slander again.

[0901] The server recognizes the word "useless" as a slur and identifies that part.

[0902] Step 5:

[0903] The server converts abusive comments into harmless language.

[0904] The server translates "useless" into "try harder."

[0905] Step 6:

[0906] The server generates the converted text and stores it in a database.

[0907] The server saves the revised sentence, "You should try a little harder," in the database.

[0908] Step 7:

[0909] The server delivers the converted text as needed.

[0910] The server displays the modified post to the user and other interested users.

[0911] This allows for effective detection and correction of defamatory comments both before and after users post.

[0912] Example 1

[0913] 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."

[0914] On modern online platforms, abusive language can easily spread, negatively impacting communication between users. Therefore, there is a need for a system that can proactively detect abusive posts, warn users, and prompt them to change their language to more appropriate ones. However, current systems struggle to detect and convert abusive language in sufficient real-time, often resulting in users leaving abusive language in their posts after posting. Furthermore, the lack of natural suggestions for converting to more appropriate language makes it difficult for users to edit their posts in a way that is more acceptable to them.

[0915] 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.

[0916] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text using a natural language processing engine to detect portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for displaying the generated warning message and suggestion to the user, and means for correcting the text entered by the user based on the user's reaction. As a result, when a user enters a text containing slander, a warning and appropriate suggestions are displayed in real time, allowing the user to immediately correct the text and realize safe and healthy online communication.

[0917] "Means for receiving text entered by a user" is a function for sending text entered by a user into a chat app or SNS app to a server.

[0918] A "natural language processing engine" is a software technology for understanding, interpreting, and manipulating human language, including analyzing and classifying text and detecting defamation.

[0919] "Abusive language detection means" is an analytical function that identifies offensive or negative language in a text.

[0920] The "means for generating warning messages and suggestions" is a function for warning the user about detected abusive language and suggesting appropriate alternative expressions.

[0921] The "means for displaying to the user" is a function for displaying warning messages and suggestions on the user's terminal in real time.

[0922] The "means for correcting text entered by the user based on the user's response" is a function that allows the user to correct the text entered in accordance with the warnings and suggestions received by the user.

[0923] The "means for storing and distributing" is a function for storing the final corrected text in a database and distributing it to relevant users as necessary.

[0924] To specifically implement the present invention, the following system configuration and processing procedures are required.

[0925] System Overview

[0926] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[0927] Specific hardware and software configurations

[0928] server

[0929] The server receives user input and uses a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) to detect and analyze defamatory content.

[0930] The server generates warning messages and suggestions based on the detection results and sends them to the user's terminal.

[0931] Terminal

[0932] The device sends text entered by the user in a chat app or SNS app to the server.

[0933] The terminal displays the warning messages and suggestions returned by the server to the user.

[0934] The terminal transmits the text corrected by the user to the server again.

[0935] User

[0936] A user inputs text into a chat app or a social networking app.

[0937] The user corrects the text based on the displayed warning message and suggestions.

[0938] System operation example

[0939] User pre-post detection process

[0940] 1. The user types "You are truly incompetent." The device sends this sentence to the server.

[0941] 2. The server analyzes the received text using a natural language processing engine and determines that the word "incompetent" is defamatory.

[0942] 3. The server generates a warning message and suggestion saying, "This word is negative. Why not change it to 'You may not be trying hard enough'?" and sends it to the terminal.

[0943] 4. The terminal displays a warning message and a suggestion to the user.

[0944] 5. The user takes the suggestion into consideration and revises the sentence to "Perhaps you are not trying hard enough."

[0945] Post-submission content transformation process

[0946] 1. A user posts "You're useless." The device sends the post to the server.

[0947] 2. The server analyzes the received text again using a natural language processing engine and determines that the word "useless" is defamatory.

[0948] 3. The server converts this to "You should try a little harder" and stores the final generated sentence in the database.

[0949] 4. The converted text is displayed to the user and other relevant users as needed.

[0950] Through these processes, the system can effectively detect and correct online defamation and provide a safe and healthy online environment.

[0951] Here is an example of a prompt for the generative AI model:

[0952] Prompt Sentence Examples

[0953] Please change the following sentences to non-libelous language:

[0954] "You're really incompetent."

[0955]

[0956] Please change the following sentences to non-libelous language:

[0957] "You're useless."

[0958] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0959] User pre-post detection process

[0960] Step 1:

[0961] The user inputs a sentence.

[0962] Input: A user types "You're so incompetent" into a chat or social media app.

[0963] Action: The user types a sentence on the device's keyboard.

[0964] Output: The input sentence "You are truly incompetent."

[0965] Step 2:

[0966] The terminal sends the input text to the server.

[0967] Input: The sentence "You are truly incompetent" typed by the user.

[0968] Action: The device sends this text to the server via the network.

[0969] Output: The sentence "You are truly incompetent" received by the server.

[0970] Step 3:

[0971] The server passes the received text to a natural language processing engine for analysis.

[0972] Input: The sentence "You are truly incompetent" received by the server.

[0973] How it works: The server passes this sentence to a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) for analysis. At the same time, it detects the slanderous word "incompetent."

[0974] Output: The word "incompetent" is determined to be libelous.

[0975] Step 4:

[0976] The server generates warning messages and suggestions based on the slander detected.

[0977] Input: The word "incompetent" which was determined to be libelous.

[0978] What it does: The server verifies that the word "incompetent" is negative and generates a warning message: "This word is negative." and a suggestion: "Maybe you're not trying hard enough?"

[0979] Output: Warning messages and suggestions.

[0980] Step 5:

[0981] The server sends warning messages and suggestions to the device.

[0982] Input: Generated warning messages and suggestions.

[0983] Operation: The server sends a warning message and suggestions to the user's device.

[0984] Output: A warning message and a suggestion will be displayed on the user's terminal.

[0985] Step 6:

[0986] The terminal displays a warning message and a suggestion to the user.

[0987] Input: Warning messages and suggestions sent by the server.

[0988] Behavior: Displays a warning message and suggestion in a pop-up or message display area on the device screen.

[0989] Output: The user sees a warning message and a suggestion.

[0990] Step 7:

[0991] The user corrects the text.

[0992] Input: Warning messages and suggestions.

[0993] Action: Based on the suggestion, the user amends the original sentence "You are really incompetent" to "Maybe you're not trying hard enough."

[0994] Output: Corrected sentence: "Maybe you're not trying hard enough."

[0995] Post-submission content transformation process

[0996] Step 1:

[0997] The user completes the post.

[0998] Input: Corrected sentence: "Maybe you're not trying hard enough."

[0999] Action: The user clicks the submit button and confirms the text.

[1000] Output: The confirmed post text "Maybe you're not trying hard enough."

[1001] Step 2:

[1002] The device sends the posted text to the server.

[1003] Input: Confirmed post text: "Maybe you're not trying hard enough."

[1004] Action: The device sends the confirmed text to the server.

[1005] Output: The sentence received by the server: "Maybe you're not trying hard enough."

[1006] Step 3:

[1007] The server then analyzes the text again using a natural language processing engine.

[1008] Input: The sentence "You may not be trying hard enough" received by the server.

[1009] What happens: The server passes this sentence to a natural language processing engine for analysis, and verifies that it does not contain any slander.

[1010] Output: After analysis, it is determined that the sentence "Maybe you're not trying hard enough" does not contain any slander.

[1011] Step 4:

[1012] The server stores the final converted text in a database.

[1013] Input: Analysis result shows that the sentence "You may not be trying hard enough" does not contain any slander.

[1014] What it does: The server saves the sentence in a database based on the analysis results.

[1015] Output: The sentence "Maybe you're not trying hard enough" stored in the database.

[1016] Step 5:

[1017] If necessary, the converted text is displayed to the relevant users.

[1018] Input: The sentence stored in the database: "Maybe you're not trying hard enough."

[1019] How it works: The system will display the new text to relevant users (e.g., users who receive the post on the subject) as needed.

[1020] Output: The converted sentence "Maybe you're not trying hard enough" displayed on the relevant user's device.

[1021] The above are the specific steps in the pre-post detection process and post-post content conversion process, which will effectively detect and correct online slander and provide a safe and healthy online environment.

[1022] (Application example 1)

[1023] 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."

[1024] Defamation on the Internet has a significant negative impact on users' mental health and can worsen the atmosphere of the entire community. In particular, in content distribution services, where many users post comments in real time, monitoring and correcting them is difficult. Conventional methods have difficulty detecting and appropriately correcting defamation in real time, making it difficult to provide a safe and healthy online environment. To solve this problem, a system that utilizes natural language processing technology and generative AI models to detect and correct user comments in real time is needed.

[1025] 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.

[1026] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text and detecting portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for generating the generated warning message and suggestion using a natural language generation model, and means for displaying the generated warning message and suggestion to the user. This makes it possible to monitor user comments in real time, detect slander, and immediately suggest corrections, thereby providing a constructive comment environment.

[1027] The "means for receiving text entered by a user" refers to a technology or device for transmitting text information entered by a user through a device to a server and acquiring that text information.

[1028] "Means for analyzing received text and detecting defamatory or libelous parts" refers to technologies or processes that use natural language processing engines or similar to analyze text data and identify negative or offensive words or expressions contained therein.

[1029] "Means for generating warning messages and suggestions based on the detected defamatory content" refers to technology that generates messages to warn users and suggestions showing more appropriate and constructive examples of expression based on the detection results.

[1030] The "means for displaying the generated warning message and suggestions to the user" refers to a technology or interface for displaying the generated message and suggestions on the screen of the user's device.

[1031] "Means for generation using a natural language generation model" refers to technology for generating warning messages and suggestions in natural language using a generative AI model (e.g., GPT-3 or T5).

[1032] "Means of re-analyzing the text that the user has posted and detecting any parts that contain defamatory comments" refers to a technology that re-analyzes the text that the user has finally posted and identifies any defamatory comments that remain within it.

[1033] "Means for converting detected abusive content into less offensive language" refers to technologies or algorithms that convert detected negative or offensive words or phrases into more neutral or positive language.

[1034] "Means for storing and distributing the converted text" refers to the technology used to store the corrected text on a server and distribute it to relevant users and platforms as needed.

[1035] The present invention is a system that receives text entered by a user in real time, analyzes it, and detects and corrects slander. Specific embodiments of the present invention will be described below.

[1036] System Overview

[1037] This system consists of three elements: a server, a terminal, and a user.

[1038] server:

[1039] The server uses a natural language processing engine (e.g., a T5 model using the transformers library) to analyze the text sent by the user. It detects abusive language in the text and generates warning messages and suggestions based on the detected language. The suggestions are generated by a natural language generation model and suggest appropriate ways of expressing the text to the user.

[1040] Device:

[1041] The terminal is a device (smartphone or computer) where the user inputs text. When the user inputs text using a chat app or SNS app, the text is sent to the server in real time. The terminal also displays warning messages and suggestions received from the server.

[1042] User:

[1043] The user is the subject of inputting text and receiving warning messages and suggestions from the system. The user can then correct the text based on the suggestions and submit it again.

[1044] Program processing

[1045] The server first receives the text entered by the user. It then analyzes the text using a natural language processing engine (for example, a T5 model using the transformers library) to detect parts that contain slander. Based on the detected parts, it generates warning messages and suggestions. These suggestions are created in natural Japanese sentences using a generative AI model.

[1046] The server sends the generated warning message and suggestions to the user's device, where the user can view it. If the user corrects the text as suggested, the corrected text can be sent back to the server. The server receives it, analyzes it again to ensure it does not contain any slander, and then stores and distributes it.

[1047] Specific examples

[1048] For example, if a user enters the sentence "You are really incompetent," the server detects the word "incompetent" as a slander. The server then generates a suggestion saying, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?" The following is an example of a prompt:

[1049] Your comment is negative. Try rephrasing it this way: Maybe you're not trying hard enough.

[1050] The suggestion is displayed on the terminal, and the user can refer to it and correct it by saying, "Maybe you're not trying hard enough." The server then analyzes the corrected sentence again, and after confirming that there are no problems, stores and distributes the sentence.

[1051] In this way, the system can monitor user comments in real time, detect and correct defamatory comments, and provide a healthy online environment.

[1052] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1053] Step 1:

[1054] A user inputs text using a chat app or SNS app. The input text is sent in real time from the device to a server. The server receives the text data sent by the user. The input is the text entered by the user, and the output is a confirmation that the text has been received.

[1055] Step 2:

[1056] The server passes the received text data to a natural language processing engine to detect abusive content. In this step, it filters out negative keywords such as "incompetent" or "useless." The input is a sentence entered by the user, and the output is a list of abusive keywords and phrases.

[1057] Step 3:

[1058] The server generates warning messages and suggestions based on the detected abusive comments. It uses a generative AI model (such as the T5 model) to create warning messages and suggestions in natural-sounding sentences. The input is a list of abusive comments and the original sentences, and the output is the generated warning messages and suggestions.

[1059] Step 4:

[1060] The server sends the generated warning message and suggestions to the terminal, which displays the message and suggestions to the user. The input is the warning message and suggestions sent from the server, and the output is displayed on the user's device screen.

[1061] Step 5:

[1062] The user then reconsiders whether to revise the sentence based on the displayed suggestions. In this step, the user inputs new sentences based on the suggestions received from the server. The inputs are the warning message and the suggestions, and the output is the revised sentence.

[1063] Step 6:

[1064] The user then sends the revised text back to the server. The server then receives this new text and uses a natural language processing engine to detect the abusive text again. The input is the revised text, and the output is the detection results.

[1065] Step 7:

[1066] The server checks that the text is free of slander and then stores the final text in a database and distributes it as needed. This step involves storing the final text and distributing it to relevant users and platforms. The input is the final, corrected text, and the output is a safe text stored in a database and distributed.

[1067] By following these steps, the system can detect defamatory comments in real time and prompt appropriate corrections.

[1068] 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.

[1069] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[1070] System Overview

[1071] This system receives and analyzes text entered by users, detects abusive language, and recognizes the user's emotions to provide warning messages and suggest appropriate expressions. This system combines a natural language processing engine and an emotion engine to detect abusive language and respond appropriately based on the user's emotions. It also has the ability to convert abusive language into harmless language, which it then stores and distributes.

[1072] Program processing

[1073] 1. User pre-post detection process

[1074] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[1075] The server then analyzes the user's emotions using an emotion engine, which recognizes their emotional state based on the tone and word choice of the text they type. For example, if the server determines that the user is annoyed, it will adjust warning messages and suggestions accordingly.

[1076] Based on the detection results and sentiment analysis results, the server generates a warning message and appropriate suggestions, such as "This word is negative. Why not change it to 'You may not be trying hard enough'?" This suggestion and warning message are sent to the device, which then displays it to the user. The user can then revise or reconsider the sentence based on the suggestions.

[1077] 2. Post-submission content transformation process

[1078] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[1079] The server also uses an emotion engine to analyze the user's emotions related to the post and generates an optimal, non-offensive expression based on this. The server generates the converted sentence, "You should try a little harder," and stores it in a database. The server also displays the converted post content to the user and other related users as needed.

[1080] Specific examples

[1081] 1. Example 1: A concrete example of a user pre-post detection process

[1082] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." Next, the server recognizes the user's frustration based on the emotion engine, and sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[1083] 2. Example 2: A concrete example of the post-submission content transformation process

[1084] The user posts "You're useless." The device sends the post to the server, which again recognizes "useless" and converts it to "try a little harder." It also uses an emotion engine to confirm the user's emotions and selects an appropriate expression. The server generates the converted sentence "try a little harder" and saves it in the database. The converted sentence is then displayed as needed.

[1085] As a result, the present invention can detect and correct online slander while taking into consideration the user's feelings, thereby providing a healthy online environment.

[1086] The processing flow will be explained below.

[1087] User pre-post detection process

[1088] Step 1:

[1089] The user enters the content to post.

[1090] The user types "You're really incompetent" into a text box in a chat app.

[1091] Step 2:

[1092] The terminal sends the input text to the server.

[1093] The terminal sends the text data entered by the user to the server as a POST request.

[1094] Step 3:

[1095] The server passes the received text to a natural language processing engine.

[1096] The server passes the received text "You are truly incompetent" to a natural language processing engine, which analyzes the sentence.

[1097] Step 4:

[1098] The server uses a natural language processing engine to detect slander.

[1099] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[1100] Step 5:

[1101] The server analyzes the user's emotions using an emotion engine.

[1102] The server uses an emotion engine to recognize the emotion of irritation based on the tone and word choice of the text entered by the user.

[1103] Step 6:

[1104] The server generates warning messages and suggestions based on the detection results and sentiment analysis results.

[1105] For the word "incompetent," the server generates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[1106] Step 7:

[1107] The server generates warning messages and suggestions and sends them to the device.

[1108] The server sends the generated warning message and suggestions to the device in JSON format.

[1109] Step 8:

[1110] The terminal displays a warning message and a suggestion to the user.

[1111] The terminal displays the received warning messages and suggestions to the user in pop-up windows or dialog boxes.

[1112] Post-submission content transformation process

[1113] Step 1:

[1114] The user completes the post.

[1115] The user presses the "Submit" button to complete the submission.

[1116] Step 2:

[1117] The device sends the posted text to the server.

[1118] The device sends the message "You're useless" to the server.

[1119] Step 3:

[1120] The server passes the posted text back to the natural language processing engine.

[1121] The server passes the received text, "You're useless," to a natural language processing engine for further analysis.

[1122] Step 4:

[1123] The server detects the slander again.

[1124] The server recognizes the word "useless" as a slur and identifies that part.

[1125] Step 5:

[1126] The server uses an emotion engine to analyze the user's emotions regarding the posted content.

[1127] The server uses an emotion engine to ascertain the user's emotional state associated with the post.

[1128] Step 6:

[1129] The server converts abusive comments into harmless language.

[1130] The server converts "useless" into "let's try a little harder" and selects the appropriate expression.

[1131] Step 7:

[1132] The server generates the converted text and stores it in a database.

[1133] The server stores the generated sentence "You should try a little harder" in a database.

[1134] Step 8:

[1135] The server delivers the converted text as needed.

[1136] The server displays the converted post to the user and other interested users.

[1137] As a result, the present invention can effectively detect and correct slanderous comments while taking into account the user's feelings during the pre-post and post-post processes.

[1138] Example 2

[1139] 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."

[1140] The increase in slander and negative comments on the Internet has made it difficult to maintain a healthy online environment. Users often unconsciously use negative language, and these comments are often posted without being edited. Furthermore, even for content that has already been posted, there will be users who will suffer harm unless slander is reduced. A system that can solve these problems is needed.

[1141] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for receiving a sentence entered by a user, a means for analyzing the received sentence using a natural language processing engine and detecting a portion containing slander, a means for generating a warning message and a suggestion using an emotion engine based on the detected portion containing slander, and a means for displaying the generated warning message and suggestion to the user. This allows the user to detect slander before posting a sentence and make appropriate corrections. Furthermore, by re-detecting slanderous portions in content after posting and converting them into non-savage expressions, the health of the online environment can be maintained.

[1142] "User" means an individual or organization that uses the system to input and post text.

[1143] "Text" refers to text data entered by a user and sent to the system.

[1144] A "natural language processing engine" is an algorithm or software that analyzes text data to recognize slander and specific patterns.

[1145] "Slander" refers to negative language that hurts or speaks ill of others.

[1146] An "emotion engine" is an algorithm or software that analyzes and identifies a user's emotions from the tone and content of a sentence.

[1147] A "warning message" is a message that notifies the user that the text they have entered contains slander and urges them to correct it.

[1148] A "suggestion" is a message containing alternative expressions to correct the sentence entered by the user.

[1149] "Terminal" refers to the device (e.g., smartphone, tablet, or PC) that a user uses to input text and send it to the system.

[1150] A "server" is a computer system that receives data sent from a terminal and analyzes and processes it.

[1151] "Posting" means the act of publishing text entered by a user on an online platform.

[1152] "Database" is an information management system for storing posted texts and generated messages.

[1153] The following describes in detail the mode for carrying out the present invention. The present invention is a system that receives text entered by a user, detects and corrects slanderous comments, and further recognizes the user's emotions to suggest appropriate warning messages and expressions. This system is composed of a server, a terminal, and a user.

[1154] System Overview

[1155] When a user enters text using a device (e.g., smartphone, tablet, or PC), the text is sent to the server. The server analyzes the received text using a natural language processing engine (e.g., Python library NLTK or spaCy) and an emotion engine (e.g., IBM Watson's emotion analysis API) to detect abusive content and recognize the user's emotions. It then generates appropriate warning messages and suggested expressions, which are displayed to the user via their device. After the user has completed posting, the server again detects and analyzes the text, converting the abusive content into more neutral expressions and storing them in a database.

[1156] Specific examples

[1157] Example 1: A concrete example of a user pre-post detection process

[1158] The user types "You're really incompetent." The device sends this sentence to the server. The server uses a natural language processing engine to detect the slanderous expression "incompetent." It then uses an emotion engine to recognize that the user is frustrated. The server generates a suggestion saying, "This phrase is negative. Why not change it to 'Maybe you're not trying hard enough?'" and sends it to the device. The device displays this message to the user, who then modifies the sentence based on the suggestion.

[1159] Example 2: A concrete example of the post-submission content transformation process

[1160] A user posts "You're useless." The device sends the post to the server. The server again uses a natural language processing engine to detect the defamatory expression "useless." It then uses an emotion engine to analyze the user's emotions again and generates a sentence that converts "useless" to "try harder." The converted sentence is stored in a database and displayed to other relevant users as needed.

[1161] Prompt Sentence Examples

[1162] 1. Example prompts for generative AI models

[1163] "Check whether the sentence entered by the user contains slanderous content, and perform sentiment analysis to provide a warning message and suggested corrections. For example, provide appropriate corrections for the sentence, 'You are really incompetent.'"

[1164] 2. Example prompts for generative AI models

[1165] "Detect abusive language in user-submitted sentences and generate appropriate suggestions for changes. For example, change 'You're useless' to a more gentle expression."

[1166] As a result, the present invention can effectively detect and correct online slander while taking into consideration the user's feelings, and provide a healthy online environment.

[1167] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1168] Step 1: User enters text

[1169] A user inputs a sentence into a chat app or social networking app. For example, the user inputs "You are really incompetent." This input sentence becomes the input data for the next processing step.

[1170] Step 2: The device sends the text to the server

[1171] The terminal transmits the text entered by the user to the server in real time. At this time, the text is transmitted as text data. The input is the user's text, and the output is the text data transmitted to the server.

[1172] Step 3: The server analyzes the text using a natural language processing engine

[1173] The server passes the received text to a natural language processing engine (e.g., Python library NLTK or spaCy) to detect abusive words and expressions. The input is the text data sent to the server, and the output is a list of abusive words and expressions. For example, the detection result includes the word "incompetent."

[1174] Step 4: The server analyzes the user's emotions using the emotion engine.

[1175] The server passes the received text to an emotion engine (e.g., IBM Watson's emotion analysis API) to analyze the user's emotions. The input is the text data sent to the server, and the output is the identification result of the user's emotional state (e.g., irritation, anger). For example, the analysis result may be that the user is "irritated."

[1176] Step 5: The server generates a warning message and a suggestion

[1177] The server generates a warning message and suggestions based on the results of defamation detection and sentiment analysis. The input is the results of defamation detection and sentiment analysis, and the output is a warning message and suggestions. For example, it generates a message such as, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?"

[1178] Step 6: The device displays a warning message and a suggestion to the user.

[1179] The terminal displays the warning messages and suggestions sent from the server to the user. The input is the warning messages and suggestions sent from the server, and the output is a visual display to the user, who can then correct the text based on this.

[1180] Step 7: User submits revised text

[1181] The user refers to the warning message and suggestions, inputs the corrected text (e.g., "You may not have tried hard enough"), and submits again. The input is the corrected text data, and the output is the submitted text data.

[1182] Step 8: The device sends the corrected text to the server

[1183] The terminal sends the corrected text back to the server. The input is the corrected text data, and the output is the text data that is sent again to the server.

[1184] Step 9: The server again detects the slander and transforms it if necessary.

[1185] The server then passes the corrected text back to the natural language processing engine, which re-detects any abusive content. It then uses an emotion engine to generate a converted text if necessary. The input is the corrected text data, and the output is the final converted text. For example, converting "You're useless" to "Let's try a little harder."

[1186] Step 10: The server saves the converted text in a database and displays it.

[1187] The server saves the generated converted text in a database and displays it to other users as needed. The input is the final converted text data, and the output is saved in a database and displayed to other users.

[1188] Through the processing of each step, users' writing is corrected to include healthy expressions that do not contain slander, thereby maintaining a healthy online environment.

[1189] (Application example 2)

[1190] 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."

[1191] With the increase in online communication, abusive comments have become a problem in live chats and social media. In particular, live stream chats involve the exchange of large volumes of messages in real time, making it difficult to detect and respond to abusive comments in a timely manner. Furthermore, if appropriate feedback is not provided in response to the user's emotions, this can lead to further problems. The present invention aims to solve these problems and provide a healthy online communication environment.

[1192] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving a text entered by a user, means for analyzing the received text and detecting a portion including slander, means for generating a warning message and a suggestion based on the detected portion including slander, means for displaying the generated warning message and suggestion to the user, means for analyzing the user's emotions, and means for adjusting the warning message and the suggestion based on the results of the emotion analysis. This makes it possible to detect slander in real time in live chat and provide appropriate feedback according to the user's emotions.

[1193] "Means for receiving text entered by the user" refers to a function that allows the server to receive text messages entered by the user in live chat or on social networking sites in real time.

[1194] "Means for analyzing received text and detecting parts containing defamatory language" refers to a function that uses natural language processing technology to analyze received text messages and identify defamatory language or inappropriate expressions.

[1195] The "means for generating warning messages and suggestions" is a function that automatically generates warnings and suggestions for improvement for users based on detected abusive content.

[1196] The "means for displaying the generated warning message and suggestion to the user" is a function for displaying the generated warning message and suggestion on the user terminal.

[1197] The "means for analyzing user emotions" is a function that analyzes text messages and determines the user's emotional state (e.g., anger, joy, sadness, etc.).

[1198] The "means for adjusting warning messages and suggestions based on the results of sentiment analysis" is a function for adjusting the contents of warning messages and suggestions for improvement based on the results of sentiment analysis of the user.

[1199] "Means of re-analyzing text that has been posted by the user and detecting parts that contain slander" refers to a function that re-analyzes messages that have already been posted and detects slander.

[1200] "Means for converting into harmless expressions" is a function that converts detected abusive expressions into appropriate and mild expressions.

[1201] The "means for saving and distributing converted text" is a function for saving converted text in a database and distributing it to other users as needed.

[1202] "Means for detecting abusive comments in real time when a message is sent to live chat and issuing warnings and suggestions" refers to a function that detects abusive comments immediately after a user sends a message in live chat and immediately issues a warning and suggests improvements.

[1203] The "means for displaying the results of detection and suggestions on the user's screen" is a function for displaying the detection of slander and the subsequent suggestions on the user's screen in real time.

[1204] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention, taking a system for detecting defamation and providing feedback based on user sentiment in a live chat platform as a concrete example.

[1205] Overall system configuration

[1206] The system consists of several hardware and software components: the user's device (smartphone, PC, tablet, etc.), a server, and the necessary natural language processing and sentiment analysis engines.

[1207] The user's device runs the live chat platform's client software, which sends messages entered by the user to the server, which then analyzes the received messages in real time, detects slanderous comments, and generates warning messages and suggestions for improvement based on the user's sentiment.

[1208] Hardware and software used

[1209] 1. Hardware: Smartphones, PCs, Tablets

[1210] 2. Software: Live chat platform, Google Cloud Natural Language API, IBM Watson Tone Analyzer

[1211] Data Processing and Computation

[1212] First, a user enters a message into the live chat and presses the send button, which sends the message to the server. The server then uses the Google Cloud Natural Language API to analyze the received message and detect any abusive comments. At the same time, it uses IBM Watson Tone Analyzer to analyze the user's sentiment contained in the message.

[1213] If abusive comments are detected, the server generates an appropriate warning message and suggestions for improvement based on the sentiment analysis results, which are then sent back to the user's device for them to review and decide whether to correct them.

[1214] Even if the user does not correct the original abusive message, the server will detect the posted message again, automatically convert it to a more neutral expression, store it in the database, and, in some cases, display the converted message to other users.

[1215] Specific examples

[1216] Example 1: Real-time slander detection and suggestions

[1217] A user types "You're really incompetent" into a live chat. The server uses the Google Cloud Natural Language API to detect the keyword "incompetent" in real time, and IBM Watson Tone Analyzer detects the user's frustration. It displays a warning saying "This expression is negative," and suggests an improvement to the user, saying "Maybe you're not trying hard enough."

[1218] Example 2: Post-post conversion of slander

[1219] If a user posts "You're useless" in the live chat, the server will again detect "useless" and perform sentiment analysis using IBM Watson Tone Analyzer. Depending on the results, it will convert it into a more harmless expression such as "Try harder," store it in the database, and display the converted message as needed.

[1220] Prompt Sentence Examples

[1221] "Detect user messages and analyze their sentiment. Based on that, detect slander and provide an appropriate response to the user.\nInput: "You're really incompetent"\nOutput: "This expression is negative. How about changing it to 'Maybe you're not trying hard enough'?""

[1222] As described above, the system of the present invention analyzes the content and emotions posted by users in real time and provides appropriate warnings and suggestions for improvement, thereby realizing a healthy online communication environment.

[1223] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1224] Step 1:

[1225] The user enters a message into the live chat and presses the send button. At this stage, the text message entered by the user is sent from the terminal to the server. The input is the message entered by the user, and the message content is transmitted to the server.

[1226] Step 2:

[1227] The server passes the received message to a natural language processing engine for analysis. Specifically, it uses the Google Cloud Natural Language API to analyze the message and detect abusive words and phrases. The input is the user's message, and the output is the results of the abusive language detection.

[1228] Step 3:

[1229] Next, the server performs sentiment analysis on the messages where abusive content was detected. It uses IBM Watson Tone Analyzer to analyze the emotional state of the message. The input is the user's message and the abusive content detection results, and the output is the analysis of the user's emotional state.

[1230] Step 4:

[1231] If a slanderous remark is detected and the user's sentiment is determined to be negative, the server generates an appropriate warning message and improvement suggestions. For example, it creates content such as, "This expression is negative. Why not change it to, 'Maybe you're not trying hard enough.'" The input is the detection result of the slanderous remark and the sentiment analysis result, and the output is the generated warning message and improvement suggestions.

[1232] Step 5:

[1233] The generated warning message and improvement suggestions are sent from the server to the user's terminal. The terminal displays them to the user in real time, allowing the user to modify or reconsider the message. The input is the warning message and improvement suggestions, and the output is the screen displayed on the user's terminal.

[1234] Step 6:

[1235] Once the user has finally confirmed sending the message, it is sent to the server again. The server then passes the message to the natural language processing engine again to detect abusive content. The input is the message that was finally sent, and the output is the result of whether abusive content was detected again.

[1236] Step 7:

[1237] The server converts the detected abusive content into a more neutral expression. For example, it automatically converts a message such as "You're useless" into "Try harder." The input is a message containing abusive content, and the output is the converted, more neutral expression.

[1238] Step 8:

[1239] The converted message is stored in a database and can be displayed to other users as needed. The input is the converted message, and the output is the data stored in the database and the screen displayed.

[1240] Through these steps, the system detects defamatory comments in real time and provides feedback based on the user's emotions, helping to create a healthy online communication environment.

[1241] 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.

[1242] 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.

[1243] 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.

[1244] [Fourth embodiment]

[1245] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1246] 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.

[1247] 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).

[1248] 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.

[1249] 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.

[1250] 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).

[1251] 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.

[1252] 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.

[1253] 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.

[1254] 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.

[1255] 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.

[1256] 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.

[1257] 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."

[1258] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[1259] System Overview

[1260] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[1261] Program processing

[1262] 1. User pre-post detection process

[1263] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[1264] Next, the server generates a warning message based on the detection results and makes a suggestion such as, "This word is negative. Why not change it to 'You may not be trying hard enough.'" This suggestion and warning message are sent to the terminal, which displays it to the user. The user can then revise or reconsider the sentence based on the suggestion.

[1265] 2. Post-submission content transformation process

[1266] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[1267] The server generates the converted text, "You should try a little harder," and stores it in the database. It also displays the converted post to the user and other related users as needed. In this way, defamatory posts are detected and corrected before and after the fact, providing a safe and healthy online environment.

[1268] Specific examples

[1269] 1. Example 1: A concrete example of a user pre-post detection process

[1270] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." The server then sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[1271] 2. Example 2: A concrete example of the post-submission content transformation process

[1272] The user posts "You're useless." The device sends the post to the server, which again detects "useless." The server converts this part to "try a little harder," generating the final sentence, "You try a little harder." The converted sentence is stored in a database and displayed to the user and other relevant users as needed.

[1273] As a result, the present invention can effectively detect and correct online slander and provide a healthy online environment.

[1274] The processing flow will be explained below.

[1275] User pre-post detection process

[1276] Step 1:

[1277] The user enters the content to post.

[1278] The user types "You're really incompetent" into a text box in a chat app.

[1279] Step 2:

[1280] The terminal sends the input text to the server.

[1281] The terminal sends the input text data to the server as a POST request.

[1282] Step 3:

[1283] The server passes the received text to a natural language processing engine.

[1284] The server passes the received text "You are truly incompetent" to the NLP engine.

[1285] Step 4:

[1286] The server uses a natural language processing engine to detect slander.

[1287] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[1288] Step 5:

[1289] The server generates warning messages and suggestions based on the detection results.

[1290] For the word "incompetent," the server creates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[1291] Step 6:

[1292] The server generates warning messages and suggestions and sends them to the device.

[1293] The server sends warning messages and suggestions in JSON format to the device.

[1294] Step 7:

[1295] The terminal displays a warning message and a suggestion to the user.

[1296] The terminal displays the received message to the user in a pop-up window or dialog box.

[1297] Post-submission content transformation process

[1298] Step 1:

[1299] The user completes the post.

[1300] The user presses the "Submit" button to complete the submission.

[1301] Step 2:

[1302] The device sends the posted text to the server.

[1303] The device sends the message "You're useless" to the server.

[1304] Step 3:

[1305] The server passes the posted text back to the natural language processing engine.

[1306] The server passes the received text "You're useless" to the NLP engine.

[1307] Step 4:

[1308] The server detects the slander again.

[1309] The server recognizes the word "useless" as a slur and identifies that part.

[1310] Step 5:

[1311] The server converts abusive comments into harmless language.

[1312] The server translates "useless" into "try harder."

[1313] Step 6:

[1314] The server generates the converted text and stores it in a database.

[1315] The server saves the revised sentence, "You should try a little harder," in the database.

[1316] Step 7:

[1317] The server delivers the converted text as needed.

[1318] The server displays the modified post to the user and other interested users.

[1319] This allows for effective detection and correction of defamatory comments both before and after users post.

[1320] Example 1

[1321] 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."

[1322] On modern online platforms, abusive language can easily spread, negatively impacting communication between users. Therefore, there is a need for a system that can proactively detect abusive posts, warn users, and prompt them to change their language to more appropriate ones. However, current systems struggle to detect and convert abusive language in sufficient real-time, often resulting in users leaving abusive language in their posts after posting. Furthermore, the lack of natural suggestions for converting to more appropriate language makes it difficult for users to edit their posts in a way that is more acceptable to them.

[1323] 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.

[1324] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text using a natural language processing engine to detect portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for displaying the generated warning message and suggestion to the user, and means for correcting the text entered by the user based on the user's reaction. As a result, when a user enters a text containing slander, a warning and appropriate suggestions are displayed in real time, allowing the user to immediately correct the text and realize safe and healthy online communication.

[1325] "Means for receiving text entered by a user" is a function for sending text entered by a user into a chat app or SNS app to a server.

[1326] A "natural language processing engine" is a software technology for understanding, interpreting, and manipulating human language, including analyzing and classifying text and detecting defamation.

[1327] "Abusive language detection means" is an analytical function that identifies offensive or negative language in a text.

[1328] The "means for generating warning messages and suggestions" is a function for warning the user about detected abusive language and suggesting appropriate alternative expressions.

[1329] The "means for displaying to the user" is a function for displaying warning messages and suggestions on the user's terminal in real time.

[1330] The "means for correcting text entered by the user based on the user's response" is a function that allows the user to correct the text entered in accordance with the warnings and suggestions received by the user.

[1331] The "means for storing and distributing" is a function for storing the final corrected text in a database and distributing it to relevant users as necessary.

[1332] To specifically implement the present invention, the following system configuration and processing procedures are required.

[1333] System Overview

[1334] This invention is a system that detects slanderous comments online and converts them into more neutral expressions. The system receives text entered by the user and analyzes it using a natural language processing engine. If slanderous comments are detected, a warning message is displayed and a suggestion is made to change the comment to an appropriate one. The system also includes a function to save the posted content so that it can be reviewed and revised later.

[1335] Specific hardware and software configurations

[1336] server

[1337] The server receives user input and uses a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) to detect and analyze defamatory content.

[1338] The server generates warning messages and suggestions based on the detection results and sends them to the user's terminal.

[1339] Terminal

[1340] The device sends text entered by the user in a chat app or SNS app to the server.

[1341] The terminal displays the warning messages and suggestions returned by the server to the user.

[1342] The terminal transmits the text corrected by the user to the server again.

[1343] User

[1344] A user inputs text into a chat app or a social networking app.

[1345] The user corrects the text based on the displayed warning message and suggestions.

[1346] System operation example

[1347] User pre-post detection process

[1348] 1. The user types "You are truly incompetent." The device sends this sentence to the server.

[1349] 2. The server analyzes the received text using a natural language processing engine and determines that the word "incompetent" is defamatory.

[1350] 3. The server generates a warning message and suggestion saying, "This word is negative. Why not change it to 'You may not be trying hard enough'?" and sends it to the terminal.

[1351] 4. The terminal displays a warning message and a suggestion to the user.

[1352] 5. The user takes the suggestion into consideration and revises the sentence to "Perhaps you are not trying hard enough."

[1353] Post-submission content transformation process

[1354] 1. A user posts "You're useless." The device sends the post to the server.

[1355] 2. The server analyzes the received text again using a natural language processing engine and determines that the word "useless" is defamatory.

[1356] 3. The server converts this to "You should try a little harder" and stores the final generated sentence in the database.

[1357] 4. The converted text is displayed to the user and other relevant users as needed.

[1358] Through these processes, the system can effectively detect and correct online defamation and provide a safe and healthy online environment.

[1359] Here is an example of a prompt for the generative AI model:

[1360] Prompt Sentence Examples

[1361] Please change the following sentences to non-libelous language:

[1362] "You're really incompetent."

[1363]

[1364] Please change the following sentences to non-libelous language:

[1365] "You're useless."

[1366] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1367] User pre-post detection process

[1368] Step 1:

[1369] The user inputs a sentence.

[1370] Input: A user types "You're so incompetent" into a chat or social media app.

[1371] Action: The user types a sentence on the device's keyboard.

[1372] Output: The input sentence "You are truly incompetent."

[1373] Step 2:

[1374] The terminal sends the input text to the server.

[1375] Input: The sentence "You are truly incompetent" typed by the user.

[1376] Action: The device sends this text to the server via the network.

[1377] Output: The sentence "You are truly incompetent" received by the server.

[1378] Step 3:

[1379] The server passes the received text to a natural language processing engine for analysis.

[1380] Input: The sentence "You are truly incompetent" received by the server.

[1381] How it works: The server passes this sentence to a natural language processing engine (e.g., SpaCy, Google Cloud Natural Language API) for analysis. At the same time, it detects the slanderous word "incompetent."

[1382] Output: The word "incompetent" is determined to be libelous.

[1383] Step 4:

[1384] The server generates warning messages and suggestions based on the slander detected.

[1385] Input: The word "incompetent" which was determined to be libelous.

[1386] What it does: The server verifies that the word "incompetent" is negative and generates a warning message: "This word is negative." and a suggestion: "Maybe you're not trying hard enough?"

[1387] Output: Warning messages and suggestions.

[1388] Step 5:

[1389] The server sends warning messages and suggestions to the device.

[1390] Input: Generated warning messages and suggestions.

[1391] Operation: The server sends a warning message and suggestions to the user's device.

[1392] Output: A warning message and a suggestion will be displayed on the user's terminal.

[1393] Step 6:

[1394] The terminal displays a warning message and a suggestion to the user.

[1395] Input: Warning messages and suggestions sent by the server.

[1396] Behavior: Displays a warning message and suggestion in a pop-up or message display area on the device screen.

[1397] Output: The user sees a warning message and a suggestion.

[1398] Step 7:

[1399] The user corrects the text.

[1400] Input: Warning messages and suggestions.

[1401] Action: Based on the suggestion, the user amends the original sentence "You are really incompetent" to "Maybe you're not trying hard enough."

[1402] Output: Corrected sentence: "Maybe you're not trying hard enough."

[1403] Post-submission content transformation process

[1404] Step 1:

[1405] The user completes the post.

[1406] Input: Corrected sentence: "Maybe you're not trying hard enough."

[1407] Action: The user clicks the submit button and confirms the text.

[1408] Output: The confirmed post text "Maybe you're not trying hard enough."

[1409] Step 2:

[1410] The device sends the posted text to the server.

[1411] Input: Confirmed post text: "Maybe you're not trying hard enough."

[1412] Action: The device sends the confirmed text to the server.

[1413] Output: The sentence received by the server: "Maybe you're not trying hard enough."

[1414] Step 3:

[1415] The server then analyzes the text again using a natural language processing engine.

[1416] Input: The sentence "You may not be trying hard enough" received by the server.

[1417] What happens: The server passes this sentence to a natural language processing engine for analysis, and verifies that it does not contain any slander.

[1418] Output: After analysis, it is determined that the sentence "Maybe you're not trying hard enough" does not contain any slander.

[1419] Step 4:

[1420] The server stores the final converted text in a database.

[1421] Input: Analysis result shows that the sentence "You may not be trying hard enough" does not contain any slander.

[1422] What it does: The server saves the sentence in a database based on the analysis results.

[1423] Output: The sentence "Maybe you're not trying hard enough" stored in the database.

[1424] Step 5:

[1425] If necessary, the converted text is displayed to the relevant users.

[1426] Input: The sentence stored in the database: "Maybe you're not trying hard enough."

[1427] How it works: The system will display the new text to relevant users (e.g., users who receive the post on the subject) as needed.

[1428] Output: The converted sentence "Maybe you're not trying hard enough" displayed on the relevant user's device.

[1429] The above are the specific steps in the pre-post detection process and post-post content conversion process, which will effectively detect and correct online slander and provide a safe and healthy online environment.

[1430] (Application example 1)

[1431] 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."

[1432] Defamation on the Internet has a significant negative impact on users' mental health and can worsen the atmosphere of the entire community. In particular, in content distribution services, where many users post comments in real time, monitoring and correcting them is difficult. Conventional methods have difficulty detecting and appropriately correcting defamation in real time, making it difficult to provide a safe and healthy online environment. To solve this problem, a system that utilizes natural language processing technology and generative AI models to detect and correct user comments in real time is needed.

[1433] 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.

[1434] In this invention, the server includes means for receiving text entered by a user, means for analyzing the received text and detecting portions containing slander, means for generating a warning message and a suggestion based on the detected portions containing slander, means for generating the generated warning message and suggestion using a natural language generation model, and means for displaying the generated warning message and suggestion to the user. This makes it possible to monitor user comments in real time, detect slander, and immediately suggest corrections, thereby providing a constructive comment environment.

[1435] The "means for receiving text entered by a user" refers to a technology or device for transmitting text information entered by a user through a device to a server and acquiring that text information.

[1436] "Means for analyzing received text and detecting defamatory or libelous parts" refers to technologies or processes that use natural language processing engines or similar to analyze text data and identify negative or offensive words or expressions contained therein.

[1437] "Means for generating warning messages and suggestions based on the detected defamatory content" refers to technology that generates messages to warn users and suggestions showing more appropriate and constructive examples of expression based on the detection results.

[1438] The "means for displaying the generated warning message and suggestions to the user" refers to a technology or interface for displaying the generated message and suggestions on the screen of the user's device.

[1439] "Means for generation using a natural language generation model" refers to technology for generating warning messages and suggestions in natural language using a generative AI model (e.g., GPT-3 or T5).

[1440] "Means of re-analyzing the text that the user has posted and detecting any parts that contain defamatory comments" refers to a technology that re-analyzes the text that the user has finally posted and identifies any defamatory comments that remain within it.

[1441] "Means for converting detected abusive content into less offensive language" refers to technologies or algorithms that convert detected negative or offensive words or phrases into more neutral or positive language.

[1442] "Means for storing and distributing the converted text" refers to the technology used to store the corrected text on a server and distribute it to relevant users and platforms as needed.

[1443] The present invention is a system that receives text entered by a user in real time, analyzes it, and detects and corrects slander. Specific embodiments of the present invention will be described below.

[1444] System Overview

[1445] This system consists of three elements: a server, a terminal, and a user.

[1446] server:

[1447] The server uses a natural language processing engine (e.g., a T5 model using the transformers library) to analyze the text sent by the user. It detects abusive language in the text and generates warning messages and suggestions based on the detected language. The suggestions are generated by a natural language generation model and suggest appropriate ways of expressing the text to the user.

[1448] Device:

[1449] The terminal is a device (smartphone or computer) where the user inputs text. When the user inputs text using a chat app or SNS app, the text is sent to the server in real time. The terminal also displays warning messages and suggestions received from the server.

[1450] User:

[1451] The user is the subject of inputting text and receiving warning messages and suggestions from the system. The user can then correct the text based on the suggestions and submit it again.

[1452] Program processing

[1453] The server first receives the text entered by the user. It then analyzes the text using a natural language processing engine (for example, a T5 model using the transformers library) to detect parts that contain slander. Based on the detected parts, it generates warning messages and suggestions. These suggestions are created in natural Japanese sentences using a generative AI model.

[1454] The server sends the generated warning message and suggestions to the user's device, where the user can view it. If the user corrects the text as suggested, the corrected text can be sent back to the server. The server receives it, analyzes it again to ensure it does not contain any slander, and then stores and distributes it.

[1455] Specific examples

[1456] For example, if a user enters the sentence "You are really incompetent," the server detects the word "incompetent" as a slander. The server then generates a suggestion saying, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?" The following is an example of a prompt:

[1457] Your comment is negative. Try rephrasing it this way: Maybe you're not trying hard enough.

[1458] The suggestion is displayed on the terminal, and the user can refer to it and correct it by saying, "Maybe you're not trying hard enough." The server then analyzes the corrected sentence again, and after confirming that there are no problems, stores and distributes the sentence.

[1459] In this way, the system can monitor user comments in real time, detect and correct defamatory comments, and provide a healthy online environment.

[1460] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1461] Step 1:

[1462] A user inputs text using a chat app or SNS app. The input text is sent in real time from the device to a server. The server receives the text data sent by the user. The input is the text entered by the user, and the output is a confirmation that the text has been received.

[1463] Step 2:

[1464] The server passes the received text data to a natural language processing engine to detect abusive content. In this step, it filters out negative keywords such as "incompetent" or "useless." The input is a sentence entered by the user, and the output is a list of abusive keywords and phrases.

[1465] Step 3:

[1466] The server generates warning messages and suggestions based on the detected abusive comments. It uses a generative AI model (such as the T5 model) to create warning messages and suggestions in natural-sounding sentences. The input is a list of abusive comments and the original sentences, and the output is the generated warning messages and suggestions.

[1467] Step 4:

[1468] The server sends the generated warning message and suggestions to the terminal, which displays the message and suggestions to the user. The input is the warning message and suggestions sent from the server, and the output is displayed on the user's device screen.

[1469] Step 5:

[1470] The user then reconsiders whether to revise the sentence based on the displayed suggestions. In this step, the user inputs new sentences based on the suggestions received from the server. The inputs are the warning message and the suggestions, and the output is the revised sentence.

[1471] Step 6:

[1472] The user then sends the revised text back to the server. The server then receives this new text and uses a natural language processing engine to detect the abusive text again. The input is the revised text, and the output is the detection results.

[1473] Step 7:

[1474] The server checks that the text is free of slander and then stores the final text in a database and distributes it as needed. This step involves storing the final text and distributing it to relevant users and platforms. The input is the final, corrected text, and the output is a safe text stored in a database and distributed.

[1475] By following these steps, the system can detect defamatory comments in real time and prompt appropriate corrections.

[1476] 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.

[1477] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention.

[1478] System Overview

[1479] This system receives and analyzes text entered by users, detects abusive language, and recognizes the user's emotions to provide warning messages and suggest appropriate expressions. This system combines a natural language processing engine and an emotion engine to detect abusive language and respond appropriately based on the user's emotions. It also has the ability to convert abusive language into harmless language, which it then stores and distributes.

[1480] Program processing

[1481] 1. User pre-post detection process

[1482] When a user types a sentence using a chat app or social networking app, the device sends the sentence in real time to a server. The server then passes the received sentence to a natural language processing engine to detect words and expressions that could be considered defamatory. For example, if a user types "You're really incompetent," the server will detect the word "incompetent" and determine that it is defamatory.

[1483] The server then analyzes the user's emotions using an emotion engine, which recognizes their emotional state based on the tone and word choice of the text they type. For example, if the server determines that the user is annoyed, it will adjust warning messages and suggestions accordingly.

[1484] Based on the detection results and sentiment analysis results, the server generates a warning message and appropriate suggestions, such as "This word is negative. Why not change it to 'You may not be trying hard enough'?" This suggestion and warning message are sent to the device, which then displays it to the user. The user can then revise or reconsider the sentence based on the suggestions.

[1485] 2. Post-submission content transformation process

[1486] Once the user has finished posting, the device sends the text back to the server, which then passes it through a natural language processing engine to detect any abusive content. For example, if a user posts "You're useless," the server will detect "useless" and convert it to "Try harder."

[1487] The server also uses an emotion engine to analyze the user's emotions related to the post and generates an optimal, non-offensive expression based on this. The server generates the converted sentence, "You should try a little harder," and stores it in a database. The server also displays the converted post content to the user and other related users as needed.

[1488] Specific examples

[1489] 1. Example 1: A concrete example of a user pre-post detection process

[1490] The user types "You are really incompetent." The device sends this sentence to the server, which detects "incompetent." Next, the server recognizes the user's frustration based on the emotion engine, and sends a suggestion to the device saying, "This sentence is negative. Why not change it to 'Maybe you're not trying hard enough?'" and displays it to the user. The user uses the suggestion to correct the sentence to "Maybe you're not trying hard enough."

[1491] 2. Example 2: A concrete example of the post-submission content transformation process

[1492] The user posts "You're useless." The device sends the post to the server, which again recognizes "useless" and converts it to "try a little harder." It also uses an emotion engine to confirm the user's emotions and selects an appropriate expression. The server generates the converted sentence "try a little harder" and saves it in the database. The converted sentence is then displayed as needed.

[1493] As a result, the present invention can detect and correct online slander while taking into consideration the user's feelings, thereby providing a healthy online environment.

[1494] The processing flow will be explained below.

[1495] User pre-post detection process

[1496] Step 1:

[1497] The user enters the content to post.

[1498] The user types "You're really incompetent" into a text box in a chat app.

[1499] Step 2:

[1500] The terminal sends the input text to the server.

[1501] The terminal sends the text data entered by the user to the server as a POST request.

[1502] Step 3:

[1503] The server passes the received text to a natural language processing engine.

[1504] The server passes the received text "You are truly incompetent" to a natural language processing engine, which analyzes the sentence.

[1505] Step 4:

[1506] The server uses a natural language processing engine to detect slander.

[1507] The server recognizes the word "incompetent" as a slur and identifies its location and content.

[1508] Step 5:

[1509] The server analyzes the user's emotions using an emotion engine.

[1510] The server uses an emotion engine to recognize the emotion of irritation based on the tone and word choice of the text entered by the user.

[1511] Step 6:

[1512] The server generates warning messages and suggestions based on the detection results and sentiment analysis results.

[1513] For the word "incompetent," the server generates a warning and suggestion: "This word is negative. Why not change it to 'Maybe you're not trying hard enough?'"

[1514] Step 7:

[1515] The server generates warning messages and suggestions and sends them to the device.

[1516] The server sends the generated warning message and suggestions to the device in JSON format.

[1517] Step 8:

[1518] The terminal displays a warning message and a suggestion to the user.

[1519] The terminal displays the received warning messages and suggestions to the user in pop-up windows or dialog boxes.

[1520] Post-submission content transformation process

[1521] Step 1:

[1522] The user completes the post.

[1523] The user presses the "Submit" button to complete the submission.

[1524] Step 2:

[1525] The device sends the posted text to the server.

[1526] The device sends the message "You're useless" to the server.

[1527] Step 3:

[1528] The server passes the posted text back to the natural language processing engine.

[1529] The server passes the received text, "You're useless," to a natural language processing engine for further analysis.

[1530] Step 4:

[1531] The server detects the slander again.

[1532] The server recognizes the word "useless" as a slur and identifies that part.

[1533] Step 5:

[1534] The server uses an emotion engine to analyze the user's emotions regarding the posted content.

[1535] The server uses an emotion engine to ascertain the user's emotional state associated with the post.

[1536] Step 6:

[1537] The server converts abusive comments into harmless language.

[1538] The server converts "useless" into "let's try a little harder" and selects the appropriate expression.

[1539] Step 7:

[1540] The server generates the converted text and stores it in a database.

[1541] The server stores the generated sentence "You should try a little harder" in a database.

[1542] Step 8:

[1543] The server delivers the converted text as needed.

[1544] The server displays the converted post to the user and other interested users.

[1545] As a result, the present invention can effectively detect and correct slanderous comments while taking into account the user's feelings during the pre-post and post-post processes.

[1546] Example 2

[1547] 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."

[1548] The increase in slander and negative comments on the Internet has made it difficult to maintain a healthy online environment. Users often unconsciously use negative language, and these comments are often posted without being edited. Furthermore, even for content that has already been posted, there will be users who will suffer harm unless slander is reduced. A system that can solve these problems is needed.

[1549] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for receiving a sentence entered by a user, a means for analyzing the received sentence using a natural language processing engine and detecting a portion containing slander, a means for generating a warning message and a suggestion using an emotion engine based on the detected portion containing slander, and a means for displaying the generated warning message and suggestion to the user. This allows the user to detect slander before posting a sentence and make appropriate corrections. Furthermore, by re-detecting slanderous portions in content after posting and converting them into non-savage expressions, the health of the online environment can be maintained.

[1550] "User" means an individual or organization that uses the system to input and post text.

[1551] "Text" refers to text data entered by a user and sent to the system.

[1552] A "natural language processing engine" is an algorithm or software that analyzes text data to recognize slander and specific patterns.

[1553] "Slander" refers to negative language that hurts or speaks ill of others.

[1554] An "emotion engine" is an algorithm or software that analyzes and identifies a user's emotions from the tone and content of a sentence.

[1555] A "warning message" is a message that notifies the user that the text they have entered contains slander and urges them to correct it.

[1556] A "suggestion" is a message containing alternative expressions to correct the sentence entered by the user.

[1557] "Terminal" refers to the device (e.g., smartphone, tablet, or PC) that a user uses to input text and send it to the system.

[1558] A "server" is a computer system that receives data sent from a terminal and analyzes and processes it.

[1559] "Posting" means the act of publishing text entered by a user on an online platform.

[1560] "Database" is an information management system for storing posted texts and generated messages.

[1561] The following describes in detail the mode for carrying out the present invention. The present invention is a system that receives text entered by a user, detects and corrects slanderous comments, and further recognizes the user's emotions to suggest appropriate warning messages and expressions. This system is composed of a server, a terminal, and a user.

[1562] System Overview

[1563] When a user enters text using a device (e.g., smartphone, tablet, or PC), the text is sent to the server. The server analyzes the received text using a natural language processing engine (e.g., Python library NLTK or spaCy) and an emotion engine (e.g., IBM Watson's emotion analysis API) to detect abusive content and recognize the user's emotions. It then generates appropriate warning messages and suggested expressions, which are displayed to the user via their device. After the user has completed posting, the server again detects and analyzes the text, converting the abusive content into more neutral expressions and storing them in a database.

[1564] Specific examples

[1565] Example 1: A concrete example of a user pre-post detection process

[1566] The user types "You're really incompetent." The device sends this sentence to the server. The server uses a natural language processing engine to detect the slanderous expression "incompetent." It then uses an emotion engine to recognize that the user is frustrated. The server generates a suggestion saying, "This phrase is negative. Why not change it to 'Maybe you're not trying hard enough?'" and sends it to the device. The device displays this message to the user, who then modifies the sentence based on the suggestion.

[1567] Example 2: A concrete example of the post-submission content transformation process

[1568] A user posts "You're useless." The device sends the post to the server. The server again uses a natural language processing engine to detect the defamatory expression "useless." It then uses an emotion engine to analyze the user's emotions again and generates a sentence that converts "useless" to "try harder." The converted sentence is stored in a database and displayed to other relevant users as needed.

[1569] Prompt Sentence Examples

[1570] 1. Example prompts for generative AI models

[1571] "Check whether the sentence entered by the user contains slanderous content, and perform sentiment analysis to provide a warning message and suggested corrections. For example, provide appropriate corrections for the sentence, 'You are really incompetent.'"

[1572] 2. Example prompts for generative AI models

[1573] "Detect abusive language in user-submitted sentences and generate appropriate suggestions for changes. For example, change 'You're useless' to a more gentle expression."

[1574] As a result, the present invention can effectively detect and correct online slander while taking into consideration the user's feelings, and provide a healthy online environment.

[1575] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1576] Step 1: User enters text

[1577] A user inputs a sentence into a chat app or social networking app. For example, the user inputs "You are really incompetent." This input sentence becomes the input data for the next processing step.

[1578] Step 2: The device sends the text to the server

[1579] The terminal transmits the text entered by the user to the server in real time. At this time, the text is transmitted as text data. The input is the user's text, and the output is the text data transmitted to the server.

[1580] Step 3: The server analyzes the text using a natural language processing engine

[1581] The server passes the received text to a natural language processing engine (e.g., Python library NLTK or spaCy) to detect abusive words and expressions. The input is the text data sent to the server, and the output is a list of abusive words and expressions. For example, the detection result includes the word "incompetent."

[1582] Step 4: The server analyzes the user's emotions using the emotion engine.

[1583] The server passes the received text to an emotion engine (e.g., IBM Watson's emotion analysis API) to analyze the user's emotions. The input is the text data sent to the server, and the output is the identification result of the user's emotional state (e.g., irritation, anger). For example, the analysis result may be that the user is "irritated."

[1584] Step 5: The server generates a warning message and a suggestion

[1585] The server generates a warning message and suggestions based on the results of defamation detection and sentiment analysis. The input is the results of defamation detection and sentiment analysis, and the output is a warning message and suggestions. For example, it generates a message such as, "This word is negative. Why not change it to 'Maybe you're not trying hard enough'?"

[1586] Step 6: The device displays a warning message and a suggestion to the user.

[1587] The terminal displays the warning messages and suggestions sent from the server to the user. The input is the warning messages and suggestions sent from the server, and the output is a visual display to the user, who can then correct the text based on this.

[1588] Step 7: User submits revised text

[1589] The user refers to the warning message and suggestions, inputs the corrected text (e.g., "You may not have tried hard enough"), and submits again. The input is the corrected text data, and the output is the submitted text data.

[1590] Step 8: The device sends the corrected text to the server

[1591] The terminal sends the corrected text back to the server. The input is the corrected text data, and the output is the text data that is sent again to the server.

[1592] Step 9: The server again detects the slander and transforms it if necessary.

[1593] The server then passes the corrected text back to the natural language processing engine, which re-detects any abusive content. It then uses an emotion engine to generate a converted text if necessary. The input is the corrected text data, and the output is the final converted text. For example, converting "You're useless" to "Let's try a little harder."

[1594] Step 10: The server saves the converted text in a database and displays it.

[1595] The server saves the generated converted text in a database and displays it to other users as needed. The input is the final converted text data, and the output is saved in a database and displayed to other users.

[1596] Through the processing of each step, users' writing is corrected to include healthy expressions that do not contain slander, thereby maintaining a healthy online environment.

[1597] (Application example 2)

[1598] 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."

[1599] With the increase in online communication, abusive comments have become a problem in live chats and social media. In particular, live stream chats involve the exchange of large volumes of messages in real time, making it difficult to detect and respond to abusive comments in a timely manner. Furthermore, if appropriate feedback is not provided in response to the user's emotions, this can lead to further problems. The present invention aims to solve these problems and provide a healthy online communication environment.

[1600] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for receiving a text entered by a user, means for analyzing the received text and detecting a portion including slander, means for generating a warning message and a suggestion based on the detected portion including slander, means for displaying the generated warning message and suggestion to the user, means for analyzing the user's emotions, and means for adjusting the warning message and the suggestion based on the results of the emotion analysis. This makes it possible to detect slander in real time in live chat and provide appropriate feedback according to the user's emotions.

[1601] "Means for receiving text entered by the user" refers to a function that allows the server to receive text messages entered by the user in live chat or on social networking sites in real time.

[1602] "Means for analyzing received text and detecting parts containing defamatory language" refers to a function that uses natural language processing technology to analyze received text messages and identify defamatory language or inappropriate expressions.

[1603] The "means for generating warning messages and suggestions" is a function that automatically generates warnings and suggestions for improvement for users based on detected abusive content.

[1604] The "means for displaying the generated warning message and suggestion to the user" is a function for displaying the generated warning message and suggestion on the user terminal.

[1605] The "means for analyzing user emotions" is a function that analyzes text messages and determines the user's emotional state (e.g., anger, joy, sadness, etc.).

[1606] The "means for adjusting warning messages and suggestions based on the results of sentiment analysis" is a function for adjusting the contents of warning messages and suggestions for improvement based on the results of sentiment analysis of the user.

[1607] "Means of re-analyzing text that has been posted by the user and detecting parts that contain slander" refers to a function that re-analyzes messages that have already been posted and detects slander.

[1608] "Means for converting into harmless expressions" is a function that converts detected abusive expressions into appropriate and mild expressions.

[1609] The "means for saving and distributing converted text" is a function for saving converted text in a database and distributing it to other users as needed.

[1610] "Means for detecting abusive comments in real time when a message is sent to live chat and issuing warnings and suggestions" refers to a function that detects abusive comments immediately after a user sends a message in live chat and immediately issues a warning and suggests improvements.

[1611] The "means for displaying the results of detection and suggestions on the user's screen" is a function for displaying the detection of slander and the subsequent suggestions on the user's screen in real time.

[1612] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes an embodiment of the present invention, taking a system for detecting defamation and providing feedback based on user sentiment in a live chat platform as a concrete example.

[1613] Overall system configuration

[1614] The system consists of several hardware and software components: the user's device (smartphone, PC, tablet, etc.), a server, and the necessary natural language processing and sentiment analysis engines.

[1615] The user's device runs the live chat platform's client software, which sends messages entered by the user to the server, which then analyzes the received messages in real time, detects slanderous comments, and generates warning messages and suggestions for improvement based on the user's sentiment.

[1616] Hardware and software used

[1617] 1. Hardware: Smartphones, PCs, Tablets

[1618] 2. Software: Live chat platform, Google Cloud Natural Language API, IBM Watson Tone Analyzer

[1619] Data Processing and Computation

[1620] First, a user enters a message into the live chat and presses the send button, which sends the message to the server. The server then uses the Google Cloud Natural Language API to analyze the received message and detect any abusive comments. At the same time, it uses IBM Watson Tone Analyzer to analyze the user's sentiment contained in the message.

[1621] If abusive comments are detected, the server generates an appropriate warning message and suggestions for improvement based on the sentiment analysis results, which are then sent back to the user's device for them to review and decide whether to correct them.

[1622] Even if the user does not correct the original abusive message, the server will detect the posted message again, automatically convert it to a more neutral expression, store it in the database, and, in some cases, display the converted message to other users.

[1623] Specific examples

[1624] Example 1: Real-time slander detection and suggestions

[1625] A user types "You're really incompetent" into a live chat. The server uses the Google Cloud Natural Language API to detect the keyword "incompetent" in real time, and IBM Watson Tone Analyzer detects the user's frustration. It displays a warning saying "This expression is negative," and suggests an improvement to the user, saying "Maybe you're not trying hard enough."

[1626] Example 2: Post-post conversion of slander

[1627] If a user posts "You're useless" in the live chat, the server will again detect "useless" and perform sentiment analysis using IBM Watson Tone Analyzer. Depending on the results, it will convert it into a more harmless expression such as "Try harder," store it in the database, and display the converted message as needed.

[1628] Prompt Sentence Examples

[1629] "Detect user messages and analyze their sentiment. Based on that, detect slander and provide an appropriate response to the user.\nInput: "You're really incompetent"\nOutput: "This expression is negative. How about changing it to 'Maybe you're not trying hard enough'?""

[1630] As described above, the system of the present invention analyzes the content and emotions posted by users in real time and provides appropriate warnings and suggestions for improvement, thereby realizing a healthy online communication environment.

[1631] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1632] Step 1:

[1633] The user enters a message into the live chat and presses the send button. At this stage, the text message entered by the user is sent from the terminal to the server. The input is the message entered by the user, and the message content is transmitted to the server.

[1634] Step 2:

[1635] The server passes the received message to a natural language processing engine for analysis. Specifically, it uses the Google Cloud Natural Language API to analyze the message and detect abusive words and phrases. The input is the user's message, and the output is the results of the abusive language detection.

[1636] Step 3:

[1637] Next, the server performs sentiment analysis on the messages where abusive content was detected. It uses IBM Watson Tone Analyzer to analyze the emotional state of the message. The input is the user's message and the abusive content detection results, and the output is the analysis of the user's emotional state.

[1638] Step 4:

[1639] If a slanderous remark is detected and the user's sentiment is determined to be negative, the server generates an appropriate warning message and improvement suggestions. For example, it creates content such as, "This expression is negative. Why not change it to, 'Maybe you're not trying hard enough.'" The input is the detection result of the slanderous remark and the sentiment analysis result, and the output is the generated warning message and improvement suggestions.

[1640] Step 5:

[1641] The generated warning message and improvement suggestions are sent from the server to the user's terminal. The terminal displays them to the user in real time, allowing the user to modify or reconsider the message. The input is the warning message and improvement suggestions, and the output is the screen displayed on the user's terminal.

[1642] Step 6:

[1643] Once the user has finally confirmed sending the message, it is sent to the server again. The server then passes the message to the natural language processing engine again to detect abusive content. The input is the message that was finally sent, and the output is the result of whether abusive content was detected again.

[1644] Step 7:

[1645] The server converts the detected abusive content into a more neutral expression. For example, it automatically converts a message such as "You're useless" into "Try harder." The input is a message containing abusive content, and the output is the converted, more neutral expression.

[1646] Step 8:

[1647] The converted message is stored in a database and can be displayed to other users as needed. The input is the converted message, and the output is the data stored in the database and the screen displayed.

[1648] Through these steps, the system detects defamatory comments in real time and provides feedback based on the user's emotions, helping to create a healthy online communication environment.

[1649] 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.

[1650] 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.

[1651] 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.

[1652] 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.

[1653] 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.

[1654] 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.

[1655] 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).

[1656] 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.

[1657] 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."

[1658] 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.

[1659] 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).

[1660] 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.

[1661] 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.

[1662] 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.

[1663] 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.

[1664] 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.

[1665] 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.

[1666] 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.

[1667] 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.

[1668] 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.

[1669] 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.

[1670] The following is further disclosed regarding the above embodiment.

[1671] (Claim 1)

[1672] means for receiving text entered by a user;

[1673] means for analyzing the received text and detecting any libelous or slanderous content;

[1674] means for generating a warning message and a suggestion based on the detected abusive portion;

[1675] means for displaying the generated warning message and suggestions to a user;

[1676] A system including:

[1677] (Claim 2)

[1678] A means for re-analyzing the text posted by the user and detecting any defamatory content;

[1679] means for converting the detected slanderous portion into a non-savory expression;

[1680] means for storing and distributing the converted text;

[1681] The system of claim 1 further comprising:

[1682] (Claim 3)

[1683] The analysis means detects slander by utilizing a natural language processing engine.

[1684] 10. The system of claim 1.

[1685] "Example 1"

[1686] (Claim 1)

[1687] means for receiving text entered by a user;

[1688] A means for analyzing the received text using a natural language processing engine to detect parts containing slander;

[1689] means for generating a warning message and a suggestion based on the detected abusive portion;

[1690] means for displaying the generated warning message and suggestions to a user;

[1691] a means for correcting the sentence input by the user based on the user's response;

[1692] A system including:

[1693] (Claim 2)

[1694] A means for re-analyzing the text posted by the user and detecting any defamatory content using a natural language processing engine;

[1695] means for converting the detected slanderous portion into a non-savory expression;

[1696] means for storing and distributing the converted text;

[1697] The system of claim 1 further comprising:

[1698] (Claim 3)

[1699] The generated warning messages and suggestions are displayed to the user in real time, allowing the user to immediately correct the text.

[1700] 10. The system of claim 1.

[1701] "Application Example 1"

[1702] (Claim 1)

[1703] means for receiving text entered by a user;

[1704] means for analyzing the received text and detecting any libelous or slanderous content;

[1705] means for generating a warning message and a suggestion based on the detected abusive portion;

[1706] means for displaying the generated warning message and suggestions to a user;

[1707] means for generating the generated warning message and suggestions using a natural language generation model;

[1708] A system including:

[1709] (Claim 2)

[1710] A means for re-analyzing the text posted by the user and detecting any defamatory content;

[1711] means for converting the detected slanderous portion into a non-savory expression;

[1712] means for storing and distributing the converted text;

[1713] means for performing said transformation using a natural language generation model;

[1714] The system of claim 1 further comprising:

[1715] (Claim 3)

[1716] The analysis means detects slander by utilizing a natural language processing engine.

[1717] 10. The system of claim 1.

[1718] "Example 2: Combining Emotion Engines"

[1719] (Claim 1)

[1720] means for receiving text entered by a user;

[1721] A means for analyzing the received text using a natural language processing engine to detect parts containing slander;

[1722] A means for generating a warning message and a suggestion using an emotion engine based on the detected abusive content;

[1723] means for displaying the generated warning message and suggestions to a user;

[1724] A system including:

[1725] (Claim 2)

[1726] A means for re-analyzing the text posted by the user and detecting any defamatory content;

[1727] means for converting the detected slanderous portion into a non-savory expression;

[1728] means for storing and distributing the converted text;

[1729] The system of claim 1 further comprising:

[1730] (Claim 3)

[1731] The emotion engine is used to analyze the user's emotional state.

[1732] 10. The system of claim 1.

[1733] "Application example 2 when combining emotion engines"

[1734] (Claim 1)

[1735] means for receiving text entered by a user;

[1736] means for analyzing the received text and detecting any libelous or slanderous content;

[1737] means for generating a warning message and a suggestion based on the detected abusive portion;

[1738] means for displaying the generated warning message and suggestions to a user;

[1739] means for analyzing user sentiment;

[1740] means for tailoring warning messages and suggestions based on the results of said sentiment analysis;

[1741] A system including:

[1742] (Claim 2)

[1743] A means for re-analyzing the text posted by the user and detecting any defamatory content;

[1744] means for converting the detected slanderous portion into a non-savory expression;

[1745] means for storing and distributing the converted text;

[1746] A means for detecting abusive content and providing warnings and suggestions in real time when users send messages to live chat;

[1747] means for displaying the results of said detection and suggestions on a user's screen;

[1748] means for detecting the converted message again and displaying it appropriately;

[1749] The system of claim 1 further comprising:

[1750] (Claim 3)

[1751] The analysis means detects slander by utilizing a natural language processing engine.

[1752] 10. The system of claim 1. [Explanation of symbols]

[1753] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. means for receiving text entered by a user; means for analyzing the received text and detecting any libelous or slanderous content; means for generating a warning message and a suggestion based on the detected abusive portion; means for displaying the generated warning message and suggestions to a user; A system including:

2. A means for re-analyzing the text posted by the user and detecting any defamatory content; means for converting the detected slanderous portion into a non-savory expression; means for storing and distributing the converted text; The system of claim 1 further comprising:

3. The analysis means detects slander by utilizing a natural language processing engine. The system of claim 1 .

Images