System

Abstract

A system is provided.SOLUTION: A system, comprising: means for obtaining credentials of a user and collecting historical post data of a plurality of federated social networking services; means for preprocessing the collected historical post data and performing topic modeling to extract key topics; means for obtaining newly posted data in real-time and comparing topics of the new posts with the historical data to evaluate consistency; means for displaying a warning to the user and suggesting an optimal expression if the consistency is low; and means for evaluating a burning risk of the new post based on historical burning cases and notifying a risk value.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] In modern society, social networking services (SNS) have become widespread, allowing anyone to easily disseminate information. However, there is a risk of inconsistency between past and present statements, which can lead to a backlash. Such backlash can cause serious problems, particularly for companies that employ celebrities or influencers, such as large damages and a decline in brand image. The purpose of this invention is to provide a system that evaluates the consistency between a user's past and present SNS statements and provides advance warnings and suggestions regarding the risk of backlash. [Means for solving the problem]

[0005] The present invention is a system that includes means for acquiring user authentication information and collecting past posting data from multiple linked SNSs, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, and means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value. This makes it possible to maintain consistency between the user's past and present comments and prevent potential uproar risks.

[0006] "User credentials" refers to the authentication data a user needs to access a system, typically a username, password, or OAuth token.

[0007] "Social networking services (SNS)" refers to all online services that enable users to share personal and corporate information and interact with other users.

[0008] "Past posting data" refers to all posting content, including text, images, audio, etc., that users have previously sent via social media, as well as their metadata.

[0009] "Means of collection" refers to the software and hardware functions for obtaining past posting data via SNS API and storing it in a database.

[0010] "Preprocessing" refers to the steps involved in converting raw data into an analyzable form, and includes tasks such as tokenization, morphological analysis, and stop-word removal.

[0011] "Topic modeling" is an analytical technique for extracting hidden topics from large amounts of text data.

[0012] "Latent Dirichlet Allocation (LDA)" is a topic modeling method, a probabilistic model that generates topic distributions within documents and word distributions within topics.

[0013] "Newly posted data" refers to content such as text, images, and audio that users intend to post on SNS.

[0014] "Real-time capture" refers to the process by which the system captures data instantly the moment a new post is made.

[0015] "Assessing consistency" means determining whether the topic of a new post is consistent with previous posts.

[0016] "Displaying a warning" means displaying a message to alert the user when consistency is low or risk is high.

[0017] "Suggesting optimal expressions" means suggesting revisions so that a user's new post does not contradict previous posts or reduces the risk of a backlash.

[0018] "Flame cases" refer to specific posts or incidents that have caused significant controversy or criticism on social media in the past.

[0019] "Notifying risk value" means evaluating the risk of a new post causing a controversy and notifying the user of the result as a number or index. [Brief explanation of the drawings]

[0020] [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

[0021] 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.

[0022] First, the terms used in the following description will be explained.

[0023] 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).

[0024] 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.

[0025] 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.

[0026] 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.

[0027] 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."

[0028] [First embodiment]

[0029] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0030] 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.

[0031] 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).

[0032] 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.

[0033] 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.

[0034] 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.

[0035] 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.

[0036] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0037] 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.

[0038] 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.

[0039] 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.

[0040] 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."

[0041] MODE FOR CARRYING OUT THE INVENTION

[0042] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. This system has the following components:

[0043] (1. Data collection and integration)

[0044] User authentication and SNS integration

[0045] Users log in to the system and link their various SNS accounts, which then stores their authentication information (username, password, OAuth token, etc.) in the system.

[0046] Social Media Data Extraction

[0047] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database and used for later analysis.

[0048] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms.

[0049] (2. Speech analysis and consistency check)

[0050] Data Preprocessing

[0051] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[0052] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[0053] Topic Modeling

[0054] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[0055] Example: If a user's past posts have primarily been about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[0056] (3. Real-time warning)

[0057] Submit a new post

[0058] When a user posts a new message, the message is sent to the server in real time.

[0059] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to a server.

[0060] Consistency Check

[0061] The server preprocesses the text of new posts and compares them with topic models to assess consistency.

[0062] Example: Detect when the topic of a new post "I bought a new car!" contradicts a previous post that said "I won't buy a car because it's bad for the environment."

[0063] Displaying warnings and suggestions

[0064] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[0065] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[0066] (4. Risk Assessment)

[0067] Risk Assessment

[0068] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0069] Example: If a past post containing "car" is deemed high risk, the system will determine that "this post is high risk" and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[0070] (5. Final confirmation and submission)

[0071] Confirm or cancel your post

[0072] Users can choose to modify their post based on the suggestions and warnings provided, or cancel the post.

[0073] Example: A user corrects the post to "I bought a new, green car" and confirms the post.

[0074] Through these steps, the present invention can maintain consistency between a user's past and present comments and prevent potential risks of flame wars.

[0075] The processing flow will be explained below.

[0076] Step 1:

[0077] Users log in to the platform and link their various social networking service (SNS) accounts using authentication protocols such as OAuth.

[0078] Step 2:

[0079] The server uses the authentication information provided by the user to collect past posting data via the SNS API, including the post content, date and time, and metadata.

[0080] Step 3:

[0081] The server stores the collected data in a database for later analysis.

[0082] Step 4:

[0083] The server periodically scans the contents of the database to see if any newly collected submitted data has been collected.

[0084] Step 5:

[0085] The server pre-processes the newly collected submission data, which includes text tokenization, morphological analysis, and stop-word removal.

[0086] Step 6:

[0087] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[0088] Step 7:

[0089] When a user attempts to post a new message, the user's device sends the message to the server in real time.

[0090] Step 8:

[0091] The server receives the content of new posts and performs pre-processing, which includes tokenization and morphological analysis.

[0092] Step 9:

[0093] The server compares the topic of each new post with past posts to assess consistency.

[0094] Step 10:

[0095] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading expressions and suggest optimal expressions.

[0096] Step 11:

[0097] The server calculates a risk value for new posts based on past examples of controversy. This risk assessment takes into account the probability that similar posts have caused controversy in the past and the frequency of negative reactions to a particular topic.

[0098] Step 12:

[0099] The server notifies the user of the results of the risk assessment and warns them if the risk is high.

[0100] Step 13:

[0101] Based on the feedback from the server, the user can choose to modify the content of the new post or to confirm the post as is.

[0102] Step 14:

[0103] The user's device sends the final post content to the SNS, completing the post.

[0104] Example 1

[0105] 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."

[0106] There is a need for a system that maintains the consistency of new posts with past posting data on social networking services (SNS) and prevents potential flame war risks. However, conventional SNS management systems do not check the consistency of new posts or evaluate the flame war risk in real time, which can result in users posting inappropriate content. Furthermore, if a user posts content that contradicts past posts, this can damage the user's credibility and reputation. To solve these issues, a system is needed that checks the consistency of past posting data and evaluates the flame war risk in real time when a user posts a new post, and provides appropriate warnings and suggestions.

[0107] 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.

[0108] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked SNSs, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topic of the new post with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, and means for allowing the user to modify or cancel the new post based on the warning or suggestion received. This allows users to maintain the consistency of their new posts while preventing potential uproar risks.

[0109] "Authentication information" refers to information used to verify the identity of a user, and includes a username, password, OAuth token, and the like.

[0110] A "social networking service (SNS)" is an online service that allows users to interact with and share information with other users.

[0111] "Posted data" is a general term for content such as text, images, and videos posted by users on SNS.

[0112] "Preprocessing" refers to the cleaning and formatting of collected posted data to make it analyzable, such as tokenization, morphological analysis, and removal of stop words.

[0113] "Topic modeling" is a machine learning technique for extracting major topics and themes from collected text data, and commonly used techniques include Latent Dirichlet Allocation (LDA).

[0114] "Acquiring in real time" refers to the process of immediately sending the content of a post to a server for analysis when a user posts a new post.

[0115] "Consistency" is an evaluation criterion that indicates whether a new post is consistent with previous posts.

[0116] "Displaying a warning" means that if there is a problem with the content of a new post made by a user, the content will be notified to the user and a warning will be issued.

[0117] "Flame risk" refers to the risk that a new post will receive a negative reaction from ordinary users and be criticized or attacked online.

[0118] The "risk value" is an index that quantitatively indicates the degree of risk of a controversy.

[0119] "Cancel" means that the user cancels a new post.

[0120] A "database system" is a structured data storage system for storing and managing collected data.

[0121] The present invention is a system that collects past posting data from social networking services (SNS) that users connect to, and performs consistency and risk assessment of new posts. This system is composed of several major hardware and software components, each of which works in conjunction to achieve the objectives of the invention.

[0122] 1. Data collection and integration

[0123] User authentication and SNS integration

[0124] First, a user logs in to the system and connects their social media account. At this time, the system stores the user's authentication information (username, password, OAuth token, etc.). Authentication is performed using the API of the specific social media platform (e.g., Twitter or Instagram).

[0125] Social Media Data Extraction

[0126] The server uses the API of each SNS to collect past posting data from the linked SNSs. The collected data is stored in a centralized database. API calls can be implemented using programming languages ​​such as Python and Node.js.

[0127] Example: When a user connects Twitter and Instagram, the server uses the Twitter API and Instagram API to retrieve past tweets and Instagram posts and save them in a database.

[0128] 2. Speech analysis and consistency check

[0129] Data Preprocessing

[0130] The server preprocesses the collected historical post data using natural language processing (NLP) techniques, such as tokenization, morphological analysis, and stop word removal. For NLP, Python libraries such as NLTK and spacy can be used.

[0131] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[0132] Topic Modeling

[0133] The server extracts major topics from users' past posts using topic modeling techniques such as Latent Dirichlet Allocation (LDA), which can be implemented using the scikit-learn or gensim libraries.

[0134] Example: If a user's past posts are mainly about "environmental issues," topics such as "environment," "eco," and "climate change" will be extracted.

[0135] 3. Real-time alerts

[0136] Submit a new post

[0137] When a user posts a new post, the content is sent to the server in real time. When the user clicks the post button, the content is sent to the server as an asynchronous request.

[0138] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to the server and first analyzed by the system.

[0139] Consistency Check

[0140] The server preprocesses the text of new posts and compares it with the topic model to evaluate consistency. Based on the comparison, it determines whether the new post is inconsistent with previous posts.

[0141] Example: If the topic of a new post "I bought a new car!" contradicts a previous post "I won't buy a car because it's bad for the environment," the contradiction will be detected.

[0142] Displaying warnings and suggestions

[0143] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[0144] Example: A warning message appears saying, "This statement contradicts a previous statement. Would you say, 'I bought a new, eco-friendly car'?"

[0145] 4. Risk Assessment

[0146] Risk Assessment

[0147] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the result. The risk assessment can use a machine learning model using past data.

[0148] Example: If a post containing the keyword "car" has caused controversy in the past, a notification will be sent to any new posts saying, "This post is at high risk of causing controversy. Please reconsider."

[0149] 5. Final check and submission

[0150] Confirm or cancel your post

[0151] Based on the suggestions and warnings provided, users can choose to either correct their post or cancel it. The corrected post is then sent back to the server for a final check.

[0152] Example: If the user changes the post to "I bought a new, eco-friendly car" and confirms the post, the post is sent to the server again. If the user wants to cancel the post, they can press the cancel button to abort the post process.

[0153] Examples of prompt statements

[0154] As an example of user input, the following prompt sentence can be input to the generative AI model.

[0155] Example prompt:

[0156] "Link your Twitter and Instagram accounts to collect past posts. Use that data to evaluate new posts in real time for consistency and risk of outrage, and implement a system that provides warnings and suggestions to users as needed."

[0157] This allows users to maintain consistency in new posts and prevent potential flame wars.

[0158] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0159] Step 1:

[0160] A user logs in to the system and connects their social media account. At this time, they enter their authentication information (username, password, OAuth token, etc.). The server receives this authentication information and, once authenticated, stores it internally in the system.

[0161] Input: The credentials entered by the user into the system.

[0162] Output: Authenticated user information and linked SNS account information are saved in the system.

[0163] Specific operation: The user enters their username and password on the login screen, clicks the login button, and then clicks the SNS link button to complete OAuth authentication on the authentication screen of each SNS.

[0164] Step 2:

[0165] The server uses the API of each SNS to collect past posting data from the SNSs that the user has linked. At this time, the server sends authentication information to the API of each SNS and stores the posting data in a centralized database.

[0166] Input: API key, user credentials.

[0167] Output: Past posting data is saved in the database.

[0168] Specific operation: The server periodically calls the API of each SNS, retrieves past posting data (for example, tweets and Instagram posts from the past year), and stores it in a database.

[0169] Step 3:

[0170] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[0171] Input: Past posting data stored in the database.

[0172] Output: Preprocessed submission data is stored in a database for analysis.

[0173] What it does: The server uses Python's NLTK library to tokenize the text "I bought a new car!" into "new," "car," and "buy." It also removes stop words (such as "!").

[0174] Step 4:

[0175] The server uses the preprocessed data to extract key topics using topic modeling techniques such as Latent Dirichlet Allocation (LDA).

[0176] Input: Preprocessed submission data.

[0177] Output: The extracted topic information is stored in the topic model database.

[0178] What it does: The server uses the scikit-learn library to apply an LDA model to extract topics such as "environment," "eco," and "climate change."

[0179] Step 5:

[0180] When a user posts a new post, the device sends the post content to the server in real time. The device sends an asynchronous request when the user clicks the post button.

[0181] Input: The user's new post.

[0182] Output: The post is sent to the server and saved.

[0183] Specific action: A user types "I bought a new car!" into the Instagram posting screen and presses the post button.

[0184] Step 6:

[0185] The server preprocesses the content of new posts and compares them with the topic model to assess consistency, ensuring that the topic of the new post matches previous data.

[0186] Input: New post content.

[0187] Output: Topic comparison results. Coherence evaluation results are generated.

[0188] What happens: The server preprocesses the text "I bought a new car!" and determines that it is inconsistent with past posts such as "I won't buy a car because it's bad for the environment."

[0189] Step 7:

[0190] If the server determines that the consistency is low, it will send a warning message to the user's device and suggest the most appropriate expression.

[0191] Input: Consistency assessment results.

[0192] Output: A warning message and a suggestion are displayed on the user's terminal.

[0193] Specific operation: The server sends a warning message to the user's terminal saying, "This statement contradicts a previous statement. Would you like to say, 'I bought a new, environmentally friendly car'?"

[0194] Step 8:

[0195] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the results.

[0196] Input: New post content, past flame war data.

[0197] Output: The result of the flame risk assessment is notified to the user.

[0198] Specific operation: The server uses a machine learning model to send a notification saying, "This post has a high risk of causing a backlash. Please reconsider," based on past examples of controversy surrounding the keyword "car."

[0199] Step 9:

[0200] Based on the suggestions and warnings provided, users can choose to modify or cancel their post, after which the modified post is sent back to the server for a final review.

[0201] Input: User's revised submission or cancellation information.

[0202] Output: The revised submission is confirmed or the submission is cancelled.

[0203] Specific actions: The user corrects the post to "I bought a new, eco-friendly car" and confirms the post, or presses the cancel button without pressing the post button.

[0204] Through these steps, the system can maintain consistency in new posts by users and prevent potential flame wars.

[0205] (Application example 1)

[0206] 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."

[0207] Conventional SNS posting systems lack a means to evaluate whether the content of a user's new post is consistent with past posts, making it difficult to prevent the risk of a flame war. Furthermore, there is a lack of a systematic method to determine whether a new post poses a flame war risk. This can lead to users unintentionally causing a flame war and damaging the credibility of individuals and companies.

[0208] 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.

[0209] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk value of the new post based on past risk cases and notifying the user of the risk value, and means for inputting prompt sentences into the generative AI model to assist in the consistency evaluation. This makes it possible to maintain consistency in the content newly posted by users and prevent potential flame wars.

[0210] "User authentication information" refers to the identifiers and authentication data required for a user to access the system, including, for example, a username, password, OAuth token, etc.

[0211] "Integrated Social Networking Service" refers to a social networking platform that allows users to connect their social networking account to the system and enable data collection.

[0212] "Past posting data" refers to all text, images, videos, and related metadata that a user has posted to the linked SNS platform.

[0213] "Preprocessing" refers to a series of data cleansing operations performed to convert collected data into an analyzable format, including tokenization, morphological analysis, and stop word removal.

[0214] "Topic modeling" is a statistical method for extracting major topics or themes from collected text data, using algorithms such as Latent Dirichlet Allocation (LDA).

[0215] "Acquisition in real time" means that newly posted data is sent to the system the moment a user attempts to post it, and analysis begins immediately.

[0216] "Evaluating consistency" means determining whether the subject matter and expressions of newly posted content are consistent with topics extracted from past posting data.

[0217] A "warning" is a message that is displayed to the user when a contradiction is detected as a result of the consistency evaluation, and provides specific instructions and suggestions for improvement.

[0218] "Evaluating the risk value" means quantitatively calculating the risk factors contained in a new post based on past risk cases and notifying the user of the results.

[0219] A "generative AI model" is an artificial intelligence model for natural language generation and text analysis, and refers to advanced models such as GPT.

[0220] A "prompt" is an instruction or question input to a generative AI model, which serves as a trigger for the model to generate an appropriate response or analysis result.

[0221] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. The system includes the following components:

[0222] First, the user's authentication information is obtained. The user connects their social media accounts (e.g., multiple social media platforms) to the system. Specifically, authentication information such as username, password, and OAuth token are stored on the server. The server then collects the user's past posting data from the connected social media platforms via API. This data is stored in a centralized database and used for later analysis.

[0223] The collected data is preprocessed on the server. This preprocessing includes data cleansing operations, such as removing URLs, mentions, and hashtags, and merging consecutive spaces. This preprocessed data is then subjected to topic modeling. The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. This makes it possible to clarify trends in users' interests and concerns.

[0224] Next, newly posted data is sent to the server in real time. When a user attempts to create a new post, the content of the post is immediately sent to the server. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. If the result of this consistency evaluation indicates low consistency, a warning is displayed on the user's device. For example, a warning message such as "This post contradicts previous statements" is displayed. Specific alternatives are also provided, such as "I bought a new, environmentally friendly car."

[0225] Additionally, the server can input prompts to the generative AI model to assist in consistency assessment, such as "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If not, please tell me why."

[0226] This makes it possible to maintain consistency in the content of new posts by users and prevent potential risks of online outrage. For example, if a user tries to create a post saying "I bought a new car!", the system will detect that the content of that post is inconsistent with a previous post saying "I won't buy a car because it's bad for the environment," and generate a warning.

[0227] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0228] Step 1:

[0229] User authentication and SNS integration

[0230] The user logs in to the system and connects various SNS accounts. This causes the user's authentication information (username, password, OAuth token, etc.) to be stored on the server. The input is the user's authentication information and SNS account information, and the output is the authentication information being saved. Specifically, the authentication information provided by the user is used to authenticate via the SNS API and establish the connection.

[0231] Step 2:

[0232] Social media data collection

[0233] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database. The input is authenticated SNS account information, and the output is the collected past posting data. Specifically, the server sends an API request and stores the posting data returned from the SNS in the database.

[0234] Step 3:

[0235] Data Preprocessing

[0236] The server preprocesses the collected past posting data. This includes removing URLs, mentions, and hashtags, and merging consecutive spaces. The input is the collected posting data, and the output is preprocessed text data. Specifically, the data is cleansed using regular expressions and text analysis libraries.

[0237] Step 4:

[0238] Topic Modeling

[0239] The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. The input is preprocessed text data, and the output is a topic distribution. Specifically, the preprocessed data is vectorized using CountVectorizer, and then the topics are extracted by training an LDA model.

[0240] Step 5:

[0241] Submitting and analyzing new posts

[0242] When a user attempts to create a new post, its contents are sent to the server in real time. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. The input is the text of the new post, and the output is the result of the consistency evaluation. Specifically, the server preprocesses the new post to calculate a topic distribution and compares it with past topic distributions.

[0243] Step 6:

[0244] Displaying warnings and suggestions

[0245] If the server determines that the consistency is low, it displays a warning message on the user's device and offers alternatives. The input is the consistency evaluation result, and the output is a warning message and alternatives. Specifically, a warning message pops up on the user's device and suggests expressions such as "I bought a new, environmentally friendly car."

[0246] Step 7:

[0247] Risk Assessment and Notification

[0248] The server calculates the risk value of a new post based on past risk case data and notifies the user of the result. The input is the text of the new post and past risk case data, and the output is a risk value and a notification message. Specifically, the server uses the new post as input and inputs a prompt sentence into the risk assessment model, generating a notification such as "This post is high risk. Please reconsider."

[0249] Step 8:

[0250] Using generative AI models

[0251] The server inputs a prompt sentence into the generative AI model to assist in consistency evaluation. The input is the prompt sentence, and the output is the analysis result of the generative AI model. A specific prompt sentence is used: "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If it does not match, please tell me the reason."

[0252] 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.

[0253] MODE FOR CARRYING OUT THE INVENTION

[0254] This invention is a system that precisely evaluates the consistency and risk of flaming of new posts on SNS by combining an emotion engine that recognizes user emotions. This system has the following components.

[0255] (1. Data collection and integration)

[0256] User authentication and SNS integration

[0257] Users log in to the system and link their social networking service (SNS) accounts, which allows the system to store their authentication information.

[0258] Social Media Data Extraction

[0259] The server collects users' past posting data from linked SNSs via the SNS API and stores it in a database.

[0260] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms and stores them centrally.

[0261] (2. Speech analysis and consistency check)

[0262] Data Preprocessing

[0263] The server pre-processes the collected historical posting data, which includes text tokenization, morphological analysis, and stop word removal.

[0264] Example: The text "I bought a new car!" is tokenized and split into the words "new," "car," and "buy."

[0265] Topic Modeling

[0266] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[0267] Example: If a user's past posts are about "environmental issues," the topic model might include "environment," "eco," "climate change," etc.

[0268] (3. Emotion Engine Integration)

[0269] Emotion recognition

[0270] The server uses an emotion engine to recognize emotions from users' past posting data and store them in a database.

[0271] Example: A past post such as "I'm so happy today!" is perceived as a positive emotion, while a post such as "I'm sad" is perceived as a negative emotion.

[0272] Emotional tone analysis of new posts

[0273] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[0274] Example: When a user posts on Instagram, "I bought a new car!", that positive emotional tone is also sent to the server.

[0275] (4. Real-time warning)

[0276] Consistency Assessment

[0277] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[0278] Example: Detecting a mismatch in positive emotional tone between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment."

[0279] Displaying warnings and suggestions

[0280] If the server determines that the consistency is low, it will display a warning and suggest the most appropriate expression on the user's device.

[0281] Example: "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[0282] (5. Risk Assessment)

[0283] Flaming risk assessment

[0284] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0285] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[0286] (6. Final confirmation and submission)

[0287] Confirm or cancel your post

[0288] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[0289] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[0290] Through these steps, the present invention can maintain consistency between a user's past and present comments and emotions, and prevent potential risks of flame wars.

[0291] The processing flow will be explained below.

[0292] Step 1:

[0293] Users log in to the system and link their social networking service (SNS) accounts. This linking is done using an authentication protocol such as OAuth, and the authentication information is stored in the system.

[0294] Step 2:

[0295] The server uses the user's authentication information to collect past posting data via the SNS API, including the content, date and time of the post, and metadata.

[0296] Step 3:

[0297] The server stores the collected data in a database for later analysis.

[0298] Step 4:

[0299] The server pre-processes the historical submission data stored in the database, including text tokenization, morphological analysis, and stop word removal.

[0300] Example: "I bought a new car!" is tokenized as "new," "car," and "buy."

[0301] Step 5:

[0302] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[0303] Example: If a user's past posts are about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[0304] Step 6:

[0305] The server uses an emotion engine to recognize emotions from users' past posts, using text analysis to classify emotional states as positive, negative, etc.

[0306] Example: "I'm so happy today!" is perceived as positive, while "I'm sad" is perceived as negative.

[0307] Step 7:

[0308] The server stores the emotion recognition results in a database.

[0309] Step 8:

[0310] When a user attempts to post a new message, the user's device sends the message's content and emotional tone to the server in real time.

[0311] Example: When a user posts "I bought a new car!" on Instagram, the content and positive emotional tone are sent to the server.

[0312] Step 9:

[0313] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[0314] Example: Detect a contradiction between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment" in terms of emotional tone.

[0315] Step 10:

[0316] If the server determines that the consistency is low, it displays a warning message on the user's device and suggests the most appropriate expression.

[0317] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[0318] Step 11:

[0319] The server calculates the risk of a new post becoming a hot topic based on past examples of such hot topics, and notifies the user of the results.

[0320] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[0321] Step 12:

[0322] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post.

[0323] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[0324] Step 13:

[0325] The user's device sends the final post content to the SNS, completing the post.

[0326] Example 2

[0327] 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."

[0328] In modern social networking services (SNS), when users post a new message, it is difficult to judge whether the message is consistent with past comments or whether it poses a risk of causing a backlash. This can lead to misunderstandings due to inconsistencies with the user's past posts, or inappropriate comments that can cause a backlash, which can have a negative impact on the user.

[0329] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for acquiring user authentication information and collecting past post data from multiple linked social networking services; means for preprocessing the collected past post data and performing topic modeling to extract major topics; means for recognizing emotions from the collected past post data and saving the emotional data; means for transmitting new posts by users to the server in real time and acquiring the content and emotional tone of the new post; means for preprocessing the newly posted data and comparing the topic and emotional tone of the new post with past data to evaluate consistency; means for displaying a warning on the user's terminal and suggesting optimal expressions if consistency is low; means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value; and means for the user to modify or cancel the content and emotional tone of the new post based on feedback from the server. This makes it possible to prevent the risk of a new post causing an uproar while maintaining consistency with the user's past comments.

[0330] "User authentication information" refers to information such as an ID and password required for a user to log in to a system.

[0331] "Social networking service" is a general term for services that provide a platform for users to interact with other users online.

[0332] "Posted data" refers to information such as messages, images, and videos posted by users on social networking services.

[0333] "Preprocessing" refers to processes such as tokenization, morphological analysis, and removal of stop words that are performed to make the collected data easier to analyze.

[0334] "Topic modeling" is a statistical method for extracting latent topics (themes) from large amounts of text data.

[0335] "Emotion recognition" is a technology that identifies the emotions (positive, negative, etc.) contained in text from text data.

[0336] "Emotion data" is information about the emotions indicated by the text data, obtained through emotion recognition.

[0337] "Real-time" refers to data and information being processed immediately.

[0338] "Emotional tone" refers to the emotional atmosphere or nuance of a particular text or message.

[0339] "Consistency assessment" is the process of assessing whether the content of a new post is consistent with previous posts and whether the emotional tone is consistent.

[0340] "Warning display" refers to displaying a message to alert the user.

[0341] "Expression suggestions" refers to presenting appropriate expressions for a user's new posts.

[0342] "Flame risk" is the likelihood that a particular post will provoke a large number of negative reactions.

[0343] "Feedback" refers to information provided to users about analysis and evaluation results from the system.

[0344] MODE FOR CARRYING OUT THE INVENTION

[0345] The present invention provides a system that recognizes user emotions and more precisely evaluates the consistency and risk of a new post on a social networking site. This system is constructed through the operation of a server, a terminal, and a user. Specific embodiments of the system are described in detail below.

[0346] 1. User authentication and SNS integration

[0347] Users log in to the system by entering their ID and password, and then link their various social media accounts (e.g., Twitter, Instagram) to the system, which then securely stores their authentication information on the server.

[0348] 2. Collection and storage of social media data

[0349] The server uses APIs such as Twitter API and Instagram API to collect past posting data from linked social media accounts. This data is stored in a database and used for subsequent processing.

[0350] 3. Data Preprocessing

[0351] The server preprocesses the collected data, which includes text tokenization, morphological analysis, stopword removal, etc. For example, a post saying "I bought a new car!" is broken down into "new," "car," and "buy."

[0352] 4. Topic Modeling

[0353] The server uses a Latent Dirichlet Allocation (LDA) model to extract key topics from the preprocessed data, for example, "environment," "eco," and "climate change" from posts about "environmental issues."

[0354] 5. Recognizing and storing emotions

[0355] The server analyzes the collected post data using an emotion engine (e.g., VADER), recognizes the emotion (e.g., positive or negative) of each post, and stores it in a database. For example, a post saying, "I'm really happy today!" is recognized as a positive emotion.

[0356] 6. Emotional Tone Analysis of New Posts

[0357] When a user posts a new message, the user's device sends the message's text and emotional tone to the server in real time. The server preprocesses the text and emotional tone of the new message and temporarily stores them in a database.

[0358] 7. Consistency Assessment

[0359] The server compares the text and emotional tone of new posts with past data to assess consistency. If the content of a new post contradicts previous posts, for example, "I bought a new car!", it will determine that this contradicts the previous post, "I won't buy a car because it's bad for the environment."

[0360] 8. Displaying warnings and suggestions

[0361] If the server determines that the statement is inconsistent, it displays a warning message on the user's device and suggests the most appropriate expression. For example, it displays a warning such as, "This statement contradicts previous statements and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[0362] 9. Risk Assessment

[0363] The server calculates the risk value of new posts based on past data on flaming cases and notifies the user of the result. For example, if a new post containing "car" is judged to have a high risk of flaming, the server sends a notification saying, "This post is high risk. Please reconsider."

[0364] 10. Confirming or cancelling a submission

[0365] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether. The user can modify the content and emotional tone to "I bought a new, eco-friendly car," and finally confirm the post after confirming the emotional tone.

[0366] Examples and prompts

[0367] As a specific example of operation, a user links their Twitter and Instagram accounts to the system and collects past posting data from these accounts. When a user posts a new message such as "I bought a new car!", if that message contradicts a previous message such as "I won't buy a car because it's bad for the environment," the system displays a warning saying, "This statement contradicts your previous message and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[0368] Examples of input prompts for a generative AI model might include the following:

[0369] "Design a system that analyzes past posts on Twitter and Instagram and evaluates the consistency and sentiment of new posts by users."

[0370] "Create a system that incorporates an emotion engine that recognizes user emotions and evaluates the risk of social media posts causing controversy."

[0371] These techniques make it possible to maintain consistency between a user's past and present comments and emotions, and prevent potential risks of online outrage.

[0372] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0373] Program processing flow

[0374] Step 1:

[0375] Users log in to the system by entering their ID and password.

[0376] Input: User ID, Password

[0377] Output: Feedback of authentication success or failure

[0378] Specific operation: The user enters information into the login form and presses the "Login" button. The system authenticates the user ID and password, and if successful, displays the dashboard.

[0379] Step 2:

[0380] Users use the system's interface to link their Twitter and Instagram accounts to the system.

[0381] Input: User's social media account

[0382] Output: Notification of successful SNS account linking

[0383] Specific operation: The user clicks the link button and completes authentication on the SNS authentication screen that appears. The authentication token is saved on the server.

[0384] Step 3:

[0385] The server uses the Twitter API and Instagram API to collect users' past posting data and store it in a database.

[0386] Input: SNS account authentication token

[0387] Output: Collected past posting data

[0388] Specific operation: The server calls the Twitter API to retrieve tweets within the specified period, and similarly calls the Instagram API to retrieve past post data.

[0389] Step 4:

[0390] The server pre-processes the collected data, which includes text tokenization, morphological analysis, and stop-word removal.

[0391] Input: Collected submission data

[0392] Output: Preprocessed data (tokenized, morphologically analyzed, stopwords removed)

[0393] Specific behavior: Breaks down a post like "I bought a new car!" into "new," "car," and "buy," and removes meaningless words.

[0394] Step 5:

[0395] The server uses a Latent Dirichlet Allocation (LDA) model to extract major topics from the preprocessed data.

[0396] Input: Preprocessed data

[0397] Output: Extracted topics (e.g. "environment", "eco", "climate change")

[0398] What it does: Runs an LDA model to identify key topics from a user's past posts.

[0399] Step 6:

[0400] The server analyzes the collected posting data using an emotion engine, recognizes the emotion of each post, and stores it in a database.

[0401] Input: Collected submission data

[0402] Output: Recognized emotion data (positive, negative, etc.)

[0403] Specific behavior: Using an emotion engine such as VADER, we recognize "I'm so happy today!" as a positive emotion and "I'm sad" as a negative emotion.

[0404] Step 7:

[0405] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[0406] Input: New post text and emotional tone

[0407] Output: New post data and emotional tone sent to the server

[0408] What happens: A user posts on Instagram, "I bought a new car!" and sends that positive emotional tone to the server.

[0409] Step 8:

[0410] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[0411] Input: New post text and emotional tone

[0412] Output: Result of consistency assessment (e.g., consistent, inconsistent)

[0413] Specific behavior: Determine whether the new post "I bought a new car!" contradicts the previous post "I won't buy a car because it's bad for the environment."

[0414] Step 9:

[0415] If the server determines that the consistency is low, it displays a warning message and suggests the most appropriate expression on the user's device.

[0416] Input: Results of consistency assessment

[0417] Output: Warning message and suggested expression

[0418] What to do: Display a warning that says, "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[0419] Step 10:

[0420] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0421] Input: Results of consistency assessment and content of new submission

[0422] Output: Flame risk assessment notification

[0423] What it does: If a new post containing the word "car" has previously sparked outrage, we'll send a notification saying, "This post is high risk. Please reconsider."

[0424] Step 11:

[0425] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[0426] Input: Feedback

[0427] Output: The revised post or the cancellation of the post

[0428] What happens: The user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[0429] (Application example 2)

[0430] 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."

[0431] In recent years, as social networking services (SNS) have become more widespread, users are increasingly facing the risk of their posts sparking outrage. In particular, if a user's emotions are not consistent with their previous posts, it is easy for misunderstandings and resentment to arise, resulting in a higher risk of outrage. Furthermore, as the number of posts on content distribution services increases, there is a demand for systems that can appropriately evaluate emotions and consistency. However, existing systems lack sufficient emotional analysis and outrage risk assessment, making it difficult for users to receive appropriate feedback and warnings. Providing an effective system to improve this situation is a challenge.

[0432] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0433] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, means for analyzing the emotional tone of the new post in real time using sentiment analysis, and means for the user to confirm, modify, or cancel the content and emotional tone of the post in the content distribution service. This makes it possible to maintain consistency between the user's past and present comments and emotions, prevent potential uproars, and provide appropriate feedback in real time.

[0434] "User Credentials" means the unique identifying information required for a user to access a system.

[0435] A "social networking service" is an online platform that enables users to interact with and share content with other users over the Internet.

[0436] "Previously posted data" refers to digital content such as text, images, and videos that a user previously posted on a social networking service.

[0437] "Preprocessing" refers to a series of processes, such as text normalization, tokenization, morphological analysis, and stop word removal, that are carried out to convert collected data into a format that is easier to analyze.

[0438] "Topic modeling" is a statistical method for automatically extracting themes and topics from large amounts of text data.

[0439] "Latent Dirichlet Allocation (LDA)" is a machine learning technique for understanding the topic structure hidden in documents, and models topics based on the probability of occurrence of each word.

[0440] "Acquisition in real time" means that the data is collected and processed immediately the moment a new post is made by a user.

[0441] "Emotion analysis" is a technology that recognizes a user's emotional state (positive, negative, neutral, etc.) from text data.

[0442] "Consistency assessment" is the process of comparing and determining whether a user's new posts are consistent with previous posts in terms of content and emotional tone.

[0443] "Display warning" means that when low consistency is detected, the problem and its solution are presented to the user.

[0444] "Optimal expression suggestions" means providing specific expressions to modify a user's new posts to ensure consistency with past posts and appropriately reflect the emotional tone.

[0445] "Flame risk assessment" is the process of predicting the likelihood that a new post will provoke criticism or resentment from other users if it contains inappropriate content or sentiment.

[0446] "Notifying risk value" means quantifying the possibility of a new post causing a backlash and informing the user.

[0447] A "content distribution service" is an internet-based service that allows users to create, share, and view digital content such as text, images, and videos.

[0448] "Means for checking, modifying, and canceling post content and emotional tone" refers to a function that allows users to check the content of new posts and the accompanying emotional tone in real time on the user interface, and to modify or cancel posts as necessary.

[0449] MODE FOR CARRYING OUT THE INVENTION

[0450] System configuration and program overview

[0451] The system acquires user authentication information, collects and analyzes past posting data from multiple linked social networking services (SNS), evaluates the emotional tone and consistency of new posts in real time, and notifies users of the risk of a controversy.The main software and hardware used include Django, Tweepy, Instagram Graph API, NLTK, Gensim, VaderSentiment, Scikit-learn, React Native, and PostgreSQL.

[0452] Data collection and integration

[0453] When a user logs in to the system and connects to a social media account, the server obtains their authentication information. This information is used to retrieve past posting data from the connected social media platform. Specifically, Twitter posting data is collected using Tweepy, and Instagram posting data is collected using the Instagram Graph API. The collected data is stored in a PostgreSQL database.

[0454] Data Preprocessing and Topic Modeling

[0455] The server converts the collected raw data into a format that is easy to analyze. This includes text tokenization, morphological analysis, and stop word removal. Gensim then performs topic modeling using Latent Dirichlet Allocation (LDA) on the preprocessed data to extract major topics. For example, a post saying "I bought a new car!" is broken down into words such as "new," "car," and "buy," and the major topics extracted are "car" and "shopping."

[0456] Sentiment analysis and real-time evaluation

[0457] When a new post is created on a user's device, its text and emotional tone are sent to the server in real time. VaderSentiment is used to analyze the emotional tone of the new post and the results are stored in a database. This makes it possible to evaluate the consistency between past posts and new posts. For example, if a user tries to post "I bought a new car!", the positive emotion is analyzed.

[0458] Displaying warnings and suggestions

[0459] The server compares new posts with past data to evaluate their consistency, and if the consistency is low, it displays a real-time warning on the user's device. It also suggests optimal expressions. For example, if a new post, "I bought a new car!", contradicts a past post, "Cars are bad for the environment," it suggests expressions like, "I bought a new, environmentally friendly car."

[0460] Flaming risk assessment and notification

[0461] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the risk value. Scikit-learn is used to compare past posts with new posts and calculate the risk value. For example, if a post containing the word "car" has become a hot topic in the past, a notification will be sent to the user stating that "this post is high risk."

[0462] Final check and submission

[0463] Based on the feedback from this system, users can choose to modify their new post or cancel the post altogether. This confirmation, modification, and cancellation functionality is provided on the user interface using React Native. This allows users to receive appropriate feedback and post safely while avoiding the risk of a social media outcry.

[0464] Specific examples

[0465] Examples:

[0466] User post: "The environment in urban areas has been deteriorating recently. New narrow roads are to blame."

[0467] Warning / Suggestion: This statement contradicts a previous post that said, "Urban development is a good thing." Would you say, "I've noticed an increase in narrow roads in urban areas recently"?

[0468] Prompt statement:

[0469] markdown

[0470] A user is attempting to post content about environmental issues. The post states, "The environment in urban areas has been deteriorating recently. New, narrow roads are to blame." Please assess the consistency and controversy risk of this post and previous posts and provide appropriate warnings and suggestions.

[0471] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0472] Step 1:

[0473] A user logs in to the system and links their social media account. In this step, the user enters their authentication information, and the server establishes a link with the social media account through the API. Specifically, the user provides their Twitter and Instagram authentication information via the API, and the server receives it and begins collecting past posting data from each platform. The input is the user's authentication information, and the output is the linked social media account information.

[0474] Step 2:

[0475] The server collects past posting data from linked SNSs and stores it in a database. The collected data includes content such as text, images, and videos. Specifically, it uses Tweepy to collect Twitter tweet data and Instagram posts using the Instagram Graph API. The input is post data obtained through the SNS API, and the output is past posting data stored in the database.

[0476] Step 3:

[0477] The server preprocesses the collected post data. In this step, processes such as text tokenization, morphological analysis, and stop word removal are performed. For example, a post "I bought a new car!" is split into words such as "new," "car," and "bought." Specifically, this preprocessing is performed using the NLTK library. The input is the post data stored in the database, and the output is the preprocessed text data.

[0478] Step 4:

[0479] The server performs topic modeling using the preprocessed data. In this step, Gensim is used to extract major topics using Latent Dirichlet Allocation (LDA). For example, from posts related to "environmental issues," topics such as "climate change," "recycling," and "energy" are extracted. The input is the preprocessed text data, and the output is the extracted topic information.

[0480] Step 5:

[0481] Newly posted data is acquired in real time from the user's device and sentiment analysis is performed. When a new post is entered by a user, its text and emotional tone are sent to the server in real time, and the emotional tone is analyzed using VaderSentiment. For example, the sentence "I'm so happy today!" is analyzed as a positive sentiment. The input is the user's new post data, and the output is the analyzed emotional tone.

[0482] Step 6:

[0483] The server compares the topic and emotional tone of new posts acquired in real time with past post data to evaluate consistency. Specifically, it compares the topic information and emotional tone of new posts with the past database and calculates the degree of match. The input is the topic information and emotional tone of new posts, and the output is the consistency evaluation result.

[0484] Step 7:

[0485] If the consistency is low, the server will warn the user and suggest the most appropriate expression. For example, if a user posts "Cars are bad for the environment" while also posting "I bought a new car!", the server will display a warning saying, "This statement contradicts previous statements. Would you like to express it as 'I bought a new, environmentally friendly car'?" The input is the consistency evaluation result and the topic information of the new post, and the output is a warning and suggestion to be displayed to the user.

[0486] Step 8:

[0487] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user's device of the risk value. Specifically, it uses Scikit-learn to compare past hot topics with the new post and calculate the risk value. The input is the topic information of the new post and past hot topics, and the output is the hot topic risk value.

[0488] Step 9:

[0489] Based on the feedback from the server, the user can modify the content and emotional tone of the new post, or cancel the post altogether. Specifically, using React Native, the user can confirm, modify, or cancel the post from the user interface. The input is warnings and suggestions from the server, and the output is the modified post content, the cancellation of the post, or the post content as is.

[0490] 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.

[0491] 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.

[0492] 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.

[0493] [Second embodiment]

[0494] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0495] 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.

[0496] 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).

[0497] 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.

[0498] 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.

[0499] 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).

[0500] 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. 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.

[0501] 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.

[0502] 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.

[0503] 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.

[0504] 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.

[0505] 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."

[0506] MODE FOR CARRYING OUT THE INVENTION

[0507] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. This system has the following components:

[0508] (1. Data collection and integration)

[0509] User authentication and SNS integration

[0510] Users log in to the system and link their various SNS accounts, which then stores their authentication information (username, password, OAuth token, etc.) in the system.

[0511] Social Media Data Extraction

[0512] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database and used for later analysis.

[0513] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms.

[0514] (2. Speech analysis and consistency check)

[0515] Data Preprocessing

[0516] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[0517] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[0518] Topic Modeling

[0519] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[0520] Example: If a user's past posts have primarily been about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[0521] (3. Real-time warning)

[0522] Submit a new post

[0523] When a user posts a new message, the message is sent to the server in real time.

[0524] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to a server.

[0525] Consistency Check

[0526] The server preprocesses the text of new posts and compares them with topic models to assess consistency.

[0527] Example: Detect when the topic of a new post "I bought a new car!" contradicts a previous post that said "I won't buy a car because it's bad for the environment."

[0528] Displaying warnings and suggestions

[0529] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[0530] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[0531] (4. Risk Assessment)

[0532] Risk Assessment

[0533] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0534] Example: If a past post containing "car" is deemed high risk, the system will determine that "this post is high risk" and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[0535] (5. Final confirmation and submission)

[0536] Confirm or cancel your post

[0537] Users can choose to modify their post based on the suggestions and warnings provided, or cancel the post.

[0538] Example: A user corrects the post to "I bought a new, green car" and confirms the post.

[0539] Through these steps, the present invention can maintain consistency between a user's past and present comments and prevent potential risks of flame wars.

[0540] The processing flow will be explained below.

[0541] Step 1:

[0542] Users log in to the platform and link their various social networking service (SNS) accounts using authentication protocols such as OAuth.

[0543] Step 2:

[0544] The server uses the authentication information provided by the user to collect past posting data via the SNS API, including the post content, date and time, and metadata.

[0545] Step 3:

[0546] The server stores the collected data in a database for later analysis.

[0547] Step 4:

[0548] The server periodically scans the contents of the database to see if any newly collected submitted data has been collected.

[0549] Step 5:

[0550] The server pre-processes the newly collected submission data, which includes text tokenization, morphological analysis, and stop-word removal.

[0551] Step 6:

[0552] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[0553] Step 7:

[0554] When a user attempts to post a new message, the user's device sends the message to the server in real time.

[0555] Step 8:

[0556] The server receives the content of new posts and performs pre-processing, which includes tokenization and morphological analysis.

[0557] Step 9:

[0558] The server compares the topic of each new post with past posts to assess consistency.

[0559] Step 10:

[0560] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading expressions and suggest optimal expressions.

[0561] Step 11:

[0562] The server calculates a risk value for new posts based on past examples of controversy. This risk assessment takes into account the probability that similar posts have caused controversy in the past and the frequency of negative reactions to a particular topic.

[0563] Step 12:

[0564] The server notifies the user of the results of the risk assessment and warns them if the risk is high.

[0565] Step 13:

[0566] Based on the feedback from the server, the user can choose to modify the content of the new post or to confirm the post as is.

[0567] Step 14:

[0568] The user's device sends the final post content to the SNS, completing the post.

[0569] Example 1

[0570] 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."

[0571] There is a need for a system that maintains the consistency of new posts with past posting data on social networking services (SNS) and prevents potential flame war risks. However, conventional SNS management systems do not check the consistency of new posts or evaluate the flame war risk in real time, which can result in users posting inappropriate content. Furthermore, if a user posts content that contradicts past posts, this can damage the user's credibility and reputation. To solve these issues, a system is needed that checks the consistency of past posting data and evaluates the flame war risk in real time when a user posts a new post, and provides appropriate warnings and suggestions.

[0572] 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.

[0573] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked SNSs, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topic of the new post with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, and means for allowing the user to modify or cancel the new post based on the warning or suggestion received. This allows users to maintain the consistency of their new posts while preventing potential uproar risks.

[0574] "Authentication information" refers to information used to verify the identity of a user, and includes a username, password, OAuth token, and the like.

[0575] A "social networking service (SNS)" is an online service that allows users to interact with and share information with other users.

[0576] "Posted data" is a general term for content such as text, images, and videos posted by users on SNS.

[0577] "Preprocessing" refers to the cleaning and formatting of collected posted data to make it analyzable, such as tokenization, morphological analysis, and removal of stop words.

[0578] "Topic modeling" is a machine learning technique for extracting major topics and themes from collected text data, and commonly used techniques include Latent Dirichlet Allocation (LDA).

[0579] "Acquiring in real time" refers to the process of immediately sending the content of a post to a server for analysis when a user posts a new post.

[0580] "Consistency" is an evaluation criterion that indicates whether a new post is consistent with previous posts.

[0581] "Displaying a warning" means that if there is a problem with the content of a new post made by a user, the content will be notified to the user and a warning will be issued.

[0582] "Flame risk" refers to the risk that a new post will receive a negative reaction from ordinary users and be criticized or attacked online.

[0583] The "risk value" is an index that quantitatively indicates the degree of risk of a controversy.

[0584] "Cancel" means that the user cancels a new post.

[0585] A "database system" is a structured data storage system for storing and managing collected data.

[0586] The present invention is a system that collects past posting data from social networking services (SNS) that users connect to, and performs consistency and risk assessment of new posts. This system is composed of several major hardware and software components, each of which works in conjunction to achieve the objectives of the invention.

[0587] 1. Data collection and integration

[0588] User authentication and SNS integration

[0589] First, a user logs in to the system and connects their social media account. At this time, the system stores the user's authentication information (username, password, OAuth token, etc.). Authentication is performed using the API of the specific social media platform (e.g., Twitter or Instagram).

[0590] Social Media Data Extraction

[0591] The server uses the API of each SNS to collect past posting data from the linked SNSs. The collected data is stored in a centralized database. API calls can be implemented using programming languages ​​such as Python and Node.js.

[0592] Example: When a user connects Twitter and Instagram, the server uses the Twitter API and Instagram API to retrieve past tweets and Instagram posts and save them in a database.

[0593] 2. Speech analysis and consistency check

[0594] Data Preprocessing

[0595] The server preprocesses the collected historical post data using natural language processing (NLP) techniques, such as tokenization, morphological analysis, and stop word removal. For NLP, Python libraries such as NLTK and spacy can be used.

[0596] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[0597] Topic Modeling

[0598] The server extracts major topics from users' past posts using topic modeling techniques such as Latent Dirichlet Allocation (LDA), which can be implemented using the scikit-learn or gensim libraries.

[0599] Example: If a user's past posts are mainly about "environmental issues," topics such as "environment," "eco," and "climate change" will be extracted.

[0600] 3. Real-time alerts

[0601] Submit a new post

[0602] When a user posts a new post, the content is sent to the server in real time. When the user clicks the post button, the content is sent to the server as an asynchronous request.

[0603] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to the server and first analyzed by the system.

[0604] Consistency Check

[0605] The server preprocesses the text of new posts and compares it with the topic model to evaluate consistency. Based on the comparison, it determines whether the new post is inconsistent with previous posts.

[0606] Example: If the topic of a new post "I bought a new car!" contradicts a previous post "I won't buy a car because it's bad for the environment," the contradiction will be detected.

[0607] Displaying warnings and suggestions

[0608] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[0609] Example: A warning message appears saying, "This statement contradicts a previous statement. Would you say, 'I bought a new, eco-friendly car'?"

[0610] 4. Risk Assessment

[0611] Risk Assessment

[0612] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the result. The risk assessment can use a machine learning model using past data.

[0613] Example: If a post containing the keyword "car" has caused controversy in the past, a notification will be sent to any new posts saying, "This post is at high risk of causing controversy. Please reconsider."

[0614] 5. Final check and submission

[0615] Confirm or cancel your post

[0616] Based on the suggestions and warnings provided, users can choose to either correct their post or cancel it. The corrected post is then sent back to the server for a final check.

[0617] Example: If the user changes the post to "I bought a new, eco-friendly car" and confirms the post, the post is sent to the server again. If the user wants to cancel the post, they can press the cancel button to abort the post process.

[0618] Examples of prompt statements

[0619] As an example of user input, the following prompt sentence can be input to the generative AI model.

[0620] Example prompt:

[0621] "Link your Twitter and Instagram accounts to collect past posts. Use that data to evaluate new posts in real time for consistency and risk of outrage, and implement a system that provides warnings and suggestions to users as needed."

[0622] This allows users to maintain consistency in new posts and prevent potential flame wars.

[0623] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0624] Step 1:

[0625] A user logs in to the system and connects their social media account. At this time, they enter their authentication information (username, password, OAuth token, etc.). The server receives this authentication information and, once authenticated, stores it internally in the system.

[0626] Input: The credentials entered by the user into the system.

[0627] Output: Authenticated user information and linked SNS account information are saved in the system.

[0628] Specific operation: The user enters their username and password on the login screen, clicks the login button, and then clicks the SNS link button to complete OAuth authentication on the authentication screen of each SNS.

[0629] Step 2:

[0630] The server uses the API of each SNS to collect past posting data from the SNSs that the user has linked. At this time, the server sends authentication information to the API of each SNS and stores the posting data in a centralized database.

[0631] Input: API key, user credentials.

[0632] Output: Past posting data is saved in the database.

[0633] Specific operation: The server periodically calls the API of each SNS, retrieves past posting data (for example, tweets and Instagram posts from the past year), and stores it in a database.

[0634] Step 3:

[0635] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[0636] Input: Past posting data stored in the database.

[0637] Output: Preprocessed submission data is stored in a database for analysis.

[0638] What it does: The server uses Python's NLTK library to tokenize the text "I bought a new car!" into "new," "car," and "buy." It also removes stop words (such as "!").

[0639] Step 4:

[0640] The server uses the preprocessed data to extract key topics using topic modeling techniques such as Latent Dirichlet Allocation (LDA).

[0641] Input: Preprocessed submission data.

[0642] Output: The extracted topic information is stored in the topic model database.

[0643] What it does: The server uses the scikit-learn library to apply an LDA model to extract topics such as "environment," "eco," and "climate change."

[0644] Step 5:

[0645] When a user posts a new post, the device sends the post content to the server in real time. The device sends an asynchronous request when the user clicks the post button.

[0646] Input: The user's new post.

[0647] Output: The post is sent to the server and saved.

[0648] Specific action: A user types "I bought a new car!" into the Instagram posting screen and presses the post button.

[0649] Step 6:

[0650] The server preprocesses the content of new posts and compares them with the topic model to assess consistency, ensuring that the topic of the new post matches previous data.

[0651] Input: New post content.

[0652] Output: Topic comparison results. Coherence evaluation results are generated.

[0653] What happens: The server preprocesses the text "I bought a new car!" and determines that it is inconsistent with past posts such as "I won't buy a car because it's bad for the environment."

[0654] Step 7:

[0655] If the server determines that the consistency is low, it will send a warning message to the user's device and suggest the most appropriate expression.

[0656] Input: Consistency assessment results.

[0657] Output: A warning message and a suggestion are displayed on the user's terminal.

[0658] Specific operation: The server sends a warning message to the user's terminal saying, "This statement contradicts a previous statement. Would you like to say, 'I bought a new, environmentally friendly car'?"

[0659] Step 8:

[0660] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the results.

[0661] Input: New post content, past flame war data.

[0662] Output: The result of the flame risk assessment is notified to the user.

[0663] Specific operation: The server uses a machine learning model to send a notification saying, "This post has a high risk of causing a backlash. Please reconsider," based on past examples of controversy surrounding the keyword "car."

[0664] Step 9:

[0665] Based on the suggestions and warnings provided, users can choose to modify or cancel their post, after which the modified post is sent back to the server for a final review.

[0666] Input: User's revised submission or cancellation information.

[0667] Output: The revised submission is confirmed or the submission is cancelled.

[0668] Specific actions: The user corrects the post to "I bought a new, eco-friendly car" and confirms the post, or presses the cancel button without pressing the post button.

[0669] Through these steps, the system can maintain consistency in new posts by users and prevent potential flame wars.

[0670] (Application example 1)

[0671] 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."

[0672] Conventional SNS posting systems lack a means to evaluate whether the content of a user's new post is consistent with past posts, making it difficult to prevent the risk of a flame war. Furthermore, there is a lack of a systematic method to determine whether a new post poses a flame war risk. This can lead to users unintentionally causing a flame war and damaging the credibility of individuals and companies.

[0673] 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.

[0674] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk value of the new post based on past risk cases and notifying the user of the risk value, and means for inputting prompt sentences into the generative AI model to assist in the consistency evaluation. This makes it possible to maintain consistency in the content newly posted by users and prevent potential flame wars.

[0675] "User authentication information" refers to the identifiers and authentication data required for a user to access the system, including, for example, a username, password, OAuth token, etc.

[0676] "Integrated Social Networking Service" refers to a social networking platform that allows users to connect their social networking account to the system and enable data collection.

[0677] "Past posting data" refers to all text, images, videos, and related metadata that a user has posted to the linked SNS platform.

[0678] "Preprocessing" refers to a series of data cleansing operations performed to convert collected data into an analyzable format, including tokenization, morphological analysis, and stop word removal.

[0679] "Topic modeling" is a statistical method for extracting major topics or themes from collected text data, using algorithms such as Latent Dirichlet Allocation (LDA).

[0680] "Acquisition in real time" means that newly posted data is sent to the system the moment a user attempts to post it, and analysis begins immediately.

[0681] "Evaluating consistency" means determining whether the subject matter and expressions of newly posted content are consistent with topics extracted from past posting data.

[0682] A "warning" is a message that is displayed to the user when a contradiction is detected as a result of the consistency evaluation, and provides specific instructions and suggestions for improvement.

[0683] "Evaluating the risk value" means quantitatively calculating the risk factors contained in a new post based on past risk cases and notifying the user of the results.

[0684] A "generative AI model" is an artificial intelligence model for natural language generation and text analysis, and refers to advanced models such as GPT.

[0685] A "prompt" is an instruction or question input to a generative AI model, which serves as a trigger for the model to generate an appropriate response or analysis result.

[0686] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. The system includes the following components:

[0687] First, the user's authentication information is obtained. The user connects their social media accounts (e.g., multiple social media platforms) to the system. Specifically, authentication information such as username, password, and OAuth token are stored on the server. The server then collects the user's past posting data from the connected social media platforms via API. This data is stored in a centralized database and used for later analysis.

[0688] The collected data is preprocessed on the server. This preprocessing includes data cleansing operations, such as removing URLs, mentions, and hashtags, and merging consecutive spaces. This preprocessed data is then subjected to topic modeling. The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. This makes it possible to clarify trends in users' interests and concerns.

[0689] Next, newly posted data is sent to the server in real time. When a user attempts to create a new post, the content of the post is immediately sent to the server. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. If the result of this consistency evaluation indicates low consistency, a warning is displayed on the user's device. For example, a warning message such as "This post contradicts previous statements" is displayed. Specific alternatives are also provided, such as "I bought a new, environmentally friendly car."

[0690] Additionally, the server can input prompts to the generative AI model to assist in consistency assessment, such as "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If not, please tell me why."

[0691] This makes it possible to maintain consistency in the content of new posts by users and prevent potential risks of online outrage. For example, if a user tries to create a post saying "I bought a new car!", the system will detect that the content of that post is inconsistent with a previous post saying "I won't buy a car because it's bad for the environment," and generate a warning.

[0692] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0693] Step 1:

[0694] User authentication and SNS integration

[0695] The user logs in to the system and connects various SNS accounts. This causes the user's authentication information (username, password, OAuth token, etc.) to be stored on the server. The input is the user's authentication information and SNS account information, and the output is the authentication information being saved. Specifically, the authentication information provided by the user is used to authenticate via the SNS API and establish the connection.

[0696] Step 2:

[0697] Social media data collection

[0698] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database. The input is authenticated SNS account information, and the output is the collected past posting data. Specifically, the server sends an API request and stores the posting data returned from the SNS in the database.

[0699] Step 3:

[0700] Data Preprocessing

[0701] The server preprocesses the collected past posting data. This includes removing URLs, mentions, and hashtags, and merging consecutive spaces. The input is the collected posting data, and the output is preprocessed text data. Specifically, the data is cleansed using regular expressions and text analysis libraries.

[0702] Step 4:

[0703] Topic Modeling

[0704] The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. The input is preprocessed text data, and the output is a topic distribution. Specifically, the preprocessed data is vectorized using CountVectorizer, and then the topics are extracted by training an LDA model.

[0705] Step 5:

[0706] Submitting and analyzing new posts

[0707] When a user attempts to create a new post, its contents are sent to the server in real time. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. The input is the text of the new post, and the output is the result of the consistency evaluation. Specifically, the server preprocesses the new post to calculate a topic distribution and compares it with past topic distributions.

[0708] Step 6:

[0709] Displaying warnings and suggestions

[0710] If the server determines that the consistency is low, it displays a warning message on the user's device and offers alternatives. The input is the consistency evaluation result, and the output is a warning message and alternatives. Specifically, a warning message pops up on the user's device and suggests expressions such as "I bought a new, environmentally friendly car."

[0711] Step 7:

[0712] Risk Assessment and Notification

[0713] The server calculates the risk value of a new post based on past risk case data and notifies the user of the result. The input is the text of the new post and past risk case data, and the output is a risk value and a notification message. Specifically, the server uses the new post as input and inputs a prompt sentence into the risk assessment model, generating a notification such as "This post is high risk. Please reconsider."

[0714] Step 8:

[0715] Using generative AI models

[0716] The server inputs a prompt sentence into the generative AI model to assist in consistency evaluation. The input is the prompt sentence, and the output is the analysis result of the generative AI model. A specific prompt sentence is used: "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If it does not match, please tell me the reason."

[0717] 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.

[0718] MODE FOR CARRYING OUT THE INVENTION

[0719] This invention is a system that precisely evaluates the consistency and risk of flaming of new posts on SNS by combining an emotion engine that recognizes user emotions. This system has the following components.

[0720] (1. Data collection and integration)

[0721] User authentication and SNS integration

[0722] Users log in to the system and link their social networking service (SNS) accounts, which allows the system to store their authentication information.

[0723] Social Media Data Extraction

[0724] The server collects users' past posting data from linked SNSs via the SNS API and stores it in a database.

[0725] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms and stores them centrally.

[0726] (2. Speech analysis and consistency check)

[0727] Data Preprocessing

[0728] The server pre-processes the collected historical posting data, which includes text tokenization, morphological analysis, and stop word removal.

[0729] Example: The text "I bought a new car!" is tokenized and split into the words "new," "car," and "buy."

[0730] Topic Modeling

[0731] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[0732] Example: If a user's past posts are about "environmental issues," the topic model might include "environment," "eco," "climate change," etc.

[0733] (3. Emotion Engine Integration)

[0734] Emotion recognition

[0735] The server uses an emotion engine to recognize emotions from users' past posting data and store them in a database.

[0736] Example: A past post such as "I'm so happy today!" is perceived as a positive emotion, while a post such as "I'm sad" is perceived as a negative emotion.

[0737] Emotional tone analysis of new posts

[0738] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[0739] Example: When a user posts on Instagram, "I bought a new car!", that positive emotional tone is also sent to the server.

[0740] (4. Real-time warning)

[0741] Consistency Assessment

[0742] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[0743] Example: Detecting a mismatch in positive emotional tone between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment."

[0744] Displaying warnings and suggestions

[0745] If the server determines that the consistency is low, it will display a warning and suggest the most appropriate expression on the user's device.

[0746] Example: "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[0747] (5. Risk Assessment)

[0748] Flaming risk assessment

[0749] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0750] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[0751] (6. Final confirmation and submission)

[0752] Confirm or cancel your post

[0753] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[0754] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[0755] Through these steps, the present invention can maintain consistency between a user's past and present comments and emotions, and prevent potential risks of flame wars.

[0756] The processing flow will be explained below.

[0757] Step 1:

[0758] Users log in to the system and link their social networking service (SNS) accounts. This linking is done using an authentication protocol such as OAuth, and the authentication information is stored in the system.

[0759] Step 2:

[0760] The server uses the user's authentication information to collect past posting data via the SNS API, including the content, date and time of the post, and metadata.

[0761] Step 3:

[0762] The server stores the collected data in a database for later analysis.

[0763] Step 4:

[0764] The server pre-processes the historical submission data stored in the database, including text tokenization, morphological analysis, and stop word removal.

[0765] Example: "I bought a new car!" is tokenized as "new," "car," and "buy."

[0766] Step 5:

[0767] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[0768] Example: If a user's past posts are about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[0769] Step 6:

[0770] The server uses an emotion engine to recognize emotions from users' past posts, using text analysis to classify emotional states as positive, negative, etc.

[0771] Example: "I'm so happy today!" is perceived as positive, while "I'm sad" is perceived as negative.

[0772] Step 7:

[0773] The server stores the emotion recognition results in a database.

[0774] Step 8:

[0775] When a user attempts to post a new message, the user's device sends the message's content and emotional tone to the server in real time.

[0776] Example: When a user posts "I bought a new car!" on Instagram, the content and positive emotional tone are sent to the server.

[0777] Step 9:

[0778] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[0779] Example: Detect a contradiction between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment" in terms of emotional tone.

[0780] Step 10:

[0781] If the server determines that the consistency is low, it displays a warning message on the user's device and suggests the most appropriate expression.

[0782] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[0783] Step 11:

[0784] The server calculates the risk of a new post becoming a hot topic based on past examples of such hot topics, and notifies the user of the results.

[0785] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[0786] Step 12:

[0787] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post.

[0788] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[0789] Step 13:

[0790] The user's device sends the final post content to the SNS, completing the post.

[0791] Example 2

[0792] 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."

[0793] In modern social networking services (SNS), when users post a new message, it is difficult to judge whether the message is consistent with past comments or whether it poses a risk of causing a backlash. This can lead to misunderstandings due to inconsistencies with the user's past posts, or inappropriate comments that can cause a backlash, which can have a negative impact on the user.

[0794] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for acquiring user authentication information and collecting past post data from multiple linked social networking services; means for preprocessing the collected past post data and performing topic modeling to extract major topics; means for recognizing emotions from the collected past post data and saving the emotional data; means for transmitting new posts by users to the server in real time and acquiring the content and emotional tone of the new post; means for preprocessing the newly posted data and comparing the topic and emotional tone of the new post with past data to evaluate consistency; means for displaying a warning on the user's terminal and suggesting optimal expressions if consistency is low; means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value; and means for the user to modify or cancel the content and emotional tone of the new post based on feedback from the server. This makes it possible to prevent the risk of a new post causing an uproar while maintaining consistency with the user's past comments.

[0795] "User authentication information" refers to information such as an ID and password required for a user to log in to a system.

[0796] "Social networking service" is a general term for services that provide a platform for users to interact with other users online.

[0797] "Posted data" refers to information such as messages, images, and videos posted by users on social networking services.

[0798] "Preprocessing" refers to processes such as tokenization, morphological analysis, and removal of stop words that are performed to make the collected data easier to analyze.

[0799] "Topic modeling" is a statistical method for extracting latent topics (themes) from large amounts of text data.

[0800] "Emotion recognition" is a technology that identifies the emotions (positive, negative, etc.) contained in text from text data.

[0801] "Emotion data" is information about the emotions indicated by the text data, obtained through emotion recognition.

[0802] "Real-time" refers to data and information being processed immediately.

[0803] "Emotional tone" refers to the emotional atmosphere or nuance of a particular text or message.

[0804] "Consistency assessment" is the process of assessing whether the content of a new post is consistent with previous posts and whether the emotional tone is consistent.

[0805] "Warning display" refers to displaying a message to alert the user.

[0806] "Expression suggestions" refers to presenting appropriate expressions for a user's new posts.

[0807] "Flame risk" is the likelihood that a particular post will provoke a large number of negative reactions.

[0808] "Feedback" refers to information provided to users about analysis and evaluation results from the system.

[0809] MODE FOR CARRYING OUT THE INVENTION

[0810] The present invention provides a system that recognizes user emotions and more precisely evaluates the consistency and risk of a new post on a social networking site. This system is constructed through the operation of a server, a terminal, and a user. Specific embodiments of the system are described in detail below.

[0811] 1. User authentication and SNS integration

[0812] Users log in to the system by entering their ID and password, and then link their various social media accounts (e.g., Twitter, Instagram) to the system, which then securely stores their authentication information on the server.

[0813] 2. Collection and storage of social media data

[0814] The server uses APIs such as Twitter API and Instagram API to collect past posting data from linked social media accounts. This data is stored in a database and used for subsequent processing.

[0815] 3. Data Preprocessing

[0816] The server preprocesses the collected data, which includes text tokenization, morphological analysis, stopword removal, etc. For example, a post saying "I bought a new car!" is broken down into "new," "car," and "buy."

[0817] 4. Topic Modeling

[0818] The server uses a Latent Dirichlet Allocation (LDA) model to extract key topics from the preprocessed data, for example, "environment," "eco," and "climate change" from posts about "environmental issues."

[0819] 5. Recognizing and storing emotions

[0820] The server analyzes the collected post data using an emotion engine (e.g., VADER), recognizes the emotion (e.g., positive or negative) of each post, and stores it in a database. For example, a post saying, "I'm really happy today!" is recognized as a positive emotion.

[0821] 6. Emotional Tone Analysis of New Posts

[0822] When a user posts a new message, the user's device sends the message's text and emotional tone to the server in real time. The server preprocesses the text and emotional tone of the new message and temporarily stores them in a database.

[0823] 7. Consistency Assessment

[0824] The server compares the text and emotional tone of new posts with past data to assess consistency. If the content of a new post contradicts previous posts, for example, "I bought a new car!", it will determine that this contradicts the previous post, "I won't buy a car because it's bad for the environment."

[0825] 8. Displaying warnings and suggestions

[0826] If the server determines that the statement is inconsistent, it displays a warning message on the user's device and suggests the most appropriate expression. For example, it displays a warning such as, "This statement contradicts previous statements and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[0827] 9. Risk Assessment

[0828] The server calculates the risk value of new posts based on past data on flaming cases and notifies the user of the result. For example, if a new post containing "car" is judged to have a high risk of flaming, the server sends a notification saying, "This post is high risk. Please reconsider."

[0829] 10. Confirming or cancelling a submission

[0830] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether. The user can modify the content and emotional tone to "I bought a new, eco-friendly car," and finally confirm the post after confirming the emotional tone.

[0831] Examples and prompts

[0832] As a specific example of operation, a user links their Twitter and Instagram accounts to the system and collects past posting data from these accounts. When a user posts a new message such as "I bought a new car!", if that message contradicts a previous message such as "I won't buy a car because it's bad for the environment," the system displays a warning saying, "This statement contradicts your previous message and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[0833] Examples of input prompts for a generative AI model might include the following:

[0834] "Design a system that analyzes past posts on Twitter and Instagram and evaluates the consistency and sentiment of new posts by users."

[0835] "Create a system that incorporates an emotion engine that recognizes user emotions and evaluates the risk of social media posts causing controversy."

[0836] These techniques make it possible to maintain consistency between a user's past and present comments and emotions, and prevent potential risks of online outrage.

[0837] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0838] Program processing flow

[0839] Step 1:

[0840] Users log in to the system by entering their ID and password.

[0841] Input: User ID, Password

[0842] Output: Feedback of authentication success or failure

[0843] Specific operation: The user enters information into the login form and presses the "Login" button. The system authenticates the user ID and password, and if successful, displays the dashboard.

[0844] Step 2:

[0845] Users use the system's interface to link their Twitter and Instagram accounts to the system.

[0846] Input: User's social media account

[0847] Output: Notification of successful SNS account linking

[0848] Specific operation: The user clicks the link button and completes authentication on the SNS authentication screen that appears. The authentication token is saved on the server.

[0849] Step 3:

[0850] The server uses the Twitter API and Instagram API to collect users' past posting data and store it in a database.

[0851] Input: SNS account authentication token

[0852] Output: Collected past posting data

[0853] Specific operation: The server calls the Twitter API to retrieve tweets within the specified period, and similarly calls the Instagram API to retrieve past post data.

[0854] Step 4:

[0855] The server pre-processes the collected data, which includes text tokenization, morphological analysis, and stop-word removal.

[0856] Input: Collected submission data

[0857] Output: Preprocessed data (tokenized, morphologically analyzed, stopwords removed)

[0858] Specific behavior: Breaks down a post like "I bought a new car!" into "new," "car," and "buy," and removes meaningless words.

[0859] Step 5:

[0860] The server uses a Latent Dirichlet Allocation (LDA) model to extract major topics from the preprocessed data.

[0861] Input: Preprocessed data

[0862] Output: Extracted topics (e.g. "environment", "eco", "climate change")

[0863] What it does: Runs an LDA model to identify key topics from a user's past posts.

[0864] Step 6:

[0865] The server analyzes the collected posting data using an emotion engine, recognizes the emotion of each post, and stores it in a database.

[0866] Input: Collected submission data

[0867] Output: Recognized emotion data (positive, negative, etc.)

[0868] Specific behavior: Using an emotion engine such as VADER, we recognize "I'm so happy today!" as a positive emotion and "I'm sad" as a negative emotion.

[0869] Step 7:

[0870] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[0871] Input: New post text and emotional tone

[0872] Output: New post data and emotional tone sent to the server

[0873] What happens: A user posts on Instagram, "I bought a new car!" and sends that positive emotional tone to the server.

[0874] Step 8:

[0875] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[0876] Input: New post text and emotional tone

[0877] Output: Result of consistency assessment (e.g., consistent, inconsistent)

[0878] Specific behavior: Determine whether the new post "I bought a new car!" contradicts the previous post "I won't buy a car because it's bad for the environment."

[0879] Step 9:

[0880] If the server determines that the consistency is low, it displays a warning message and suggests the most appropriate expression on the user's device.

[0881] Input: Results of consistency assessment

[0882] Output: Warning message and suggested expression

[0883] What to do: Display a warning that says, "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[0884] Step 10:

[0885] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0886] Input: Results of consistency assessment and content of new submission

[0887] Output: Flame risk assessment notification

[0888] What it does: If a new post containing the word "car" has previously sparked outrage, we'll send a notification saying, "This post is high risk. Please reconsider."

[0889] Step 11:

[0890] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[0891] Input: Feedback

[0892] Output: The revised post or the cancellation of the post

[0893] What happens: The user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[0894] (Application example 2)

[0895] 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."

[0896] In recent years, as social networking services (SNS) have become more widespread, users are increasingly facing the risk of their posts sparking outrage. In particular, if a user's emotions are not consistent with their previous posts, it is easy for misunderstandings and resentment to arise, resulting in a higher risk of outrage. Furthermore, as the number of posts on content distribution services increases, there is a demand for systems that can appropriately evaluate emotions and consistency. However, existing systems lack sufficient emotional analysis and outrage risk assessment, making it difficult for users to receive appropriate feedback and warnings. Providing an effective system to improve this situation is a challenge.

[0897] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0898] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, means for analyzing the emotional tone of the new post in real time using sentiment analysis, and means for the user to confirm, modify, or cancel the content and emotional tone of the post in the content distribution service. This makes it possible to maintain consistency between the user's past and present comments and emotions, prevent potential uproars, and provide appropriate feedback in real time.

[0899] "User Credentials" means the unique identifying information required for a user to access a system.

[0900] A "social networking service" is an online platform that enables users to interact with and share content with other users over the Internet.

[0901] "Previously posted data" refers to digital content such as text, images, and videos that a user previously posted on a social networking service.

[0902] "Preprocessing" refers to a series of processes, such as text normalization, tokenization, morphological analysis, and stop word removal, that are carried out to convert collected data into a format that is easier to analyze.

[0903] "Topic modeling" is a statistical method for automatically extracting themes and topics from large amounts of text data.

[0904] "Latent Dirichlet Allocation (LDA)" is a machine learning technique for understanding the topic structure hidden in documents, and models topics based on the probability of occurrence of each word.

[0905] "Acquisition in real time" means that the data is collected and processed immediately the moment a new post is made by a user.

[0906] "Emotion analysis" is a technology that recognizes a user's emotional state (positive, negative, neutral, etc.) from text data.

[0907] "Consistency assessment" is the process of comparing and determining whether a user's new posts are consistent with previous posts in terms of content and emotional tone.

[0908] "Display warning" means that when low consistency is detected, the problem and its solution are presented to the user.

[0909] "Optimal expression suggestions" means providing specific expressions to modify a user's new posts to ensure consistency with past posts and appropriately reflect the emotional tone.

[0910] "Flame risk assessment" is the process of predicting the likelihood that a new post will provoke criticism or resentment from other users if it contains inappropriate content or sentiment.

[0911] "Notifying risk value" means quantifying the possibility of a new post causing a backlash and informing the user.

[0912] A "content distribution service" is an internet-based service that allows users to create, share, and view digital content such as text, images, and videos.

[0913] "Means for checking, modifying, and canceling post content and emotional tone" refers to a function that allows users to check the content of new posts and the accompanying emotional tone in real time on the user interface, and to modify or cancel posts as necessary.

[0914] MODE FOR CARRYING OUT THE INVENTION

[0915] System configuration and program overview

[0916] The system acquires user authentication information, collects and analyzes past posting data from multiple linked social networking services (SNS), evaluates the emotional tone and consistency of new posts in real time, and notifies users of the risk of a controversy.The main software and hardware used include Django, Tweepy, Instagram Graph API, NLTK, Gensim, VaderSentiment, Scikit-learn, React Native, and PostgreSQL.

[0917] Data collection and integration

[0918] When a user logs in to the system and connects to a social media account, the server obtains their authentication information. This information is used to retrieve past posting data from the connected social media platform. Specifically, Twitter posting data is collected using Tweepy, and Instagram posting data is collected using the Instagram Graph API. The collected data is stored in a PostgreSQL database.

[0919] Data Preprocessing and Topic Modeling

[0920] The server converts the collected raw data into a format that is easy to analyze. This includes text tokenization, morphological analysis, and stop word removal. Gensim then performs topic modeling using Latent Dirichlet Allocation (LDA) on the preprocessed data to extract major topics. For example, a post saying "I bought a new car!" is broken down into words such as "new," "car," and "buy," and the major topics extracted are "car" and "shopping."

[0921] Sentiment analysis and real-time evaluation

[0922] When a new post is created on a user's device, its text and emotional tone are sent to the server in real time. VaderSentiment is used to analyze the emotional tone of the new post and the results are stored in a database. This makes it possible to evaluate the consistency between past posts and new posts. For example, if a user tries to post "I bought a new car!", the positive emotion is analyzed.

[0923] Displaying warnings and suggestions

[0924] The server compares new posts with past data to evaluate their consistency, and if the consistency is low, it displays a real-time warning on the user's device. It also suggests optimal expressions. For example, if a new post, "I bought a new car!", contradicts a past post, "Cars are bad for the environment," it suggests expressions like, "I bought a new, environmentally friendly car."

[0925] Flaming risk assessment and notification

[0926] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the risk value. Scikit-learn is used to compare past posts with new posts and calculate the risk value. For example, if a post containing the word "car" has become a hot topic in the past, a notification will be sent to the user stating that "this post is high risk."

[0927] Final check and submission

[0928] Based on the feedback from this system, users can choose to modify their new post or cancel the post altogether. This confirmation, modification, and cancellation functionality is provided on the user interface using React Native. This allows users to receive appropriate feedback and post safely while avoiding the risk of a social media outcry.

[0929] Specific examples

[0930] Examples:

[0931] User post: "The environment in urban areas has been deteriorating recently. New narrow roads are to blame."

[0932] Warning / Suggestion: This statement contradicts a previous post that said, "Urban development is a good thing." Would you say, "I've noticed an increase in narrow roads in urban areas recently"?

[0933] Prompt statement:

[0934] markdown

[0935] A user is attempting to post content about environmental issues. The post states, "The environment in urban areas has been deteriorating recently. New, narrow roads are to blame." Please assess the consistency and controversy risk of this post and previous posts and provide appropriate warnings and suggestions.

[0936] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0937] Step 1:

[0938] A user logs in to the system and links their social media account. In this step, the user enters their authentication information, and the server establishes a link with the social media account through the API. Specifically, the user provides their Twitter and Instagram authentication information via the API, and the server receives it and begins collecting past posting data from each platform. The input is the user's authentication information, and the output is the linked social media account information.

[0939] Step 2:

[0940] The server collects past posting data from linked SNSs and stores it in a database. The collected data includes content such as text, images, and videos. Specifically, it uses Tweepy to collect Twitter tweet data and Instagram posts using the Instagram Graph API. The input is post data obtained through the SNS API, and the output is past posting data stored in the database.

[0941] Step 3:

[0942] The server preprocesses the collected post data. In this step, processes such as text tokenization, morphological analysis, and stop word removal are performed. For example, a post "I bought a new car!" is split into words such as "new," "car," and "bought." Specifically, this preprocessing is performed using the NLTK library. The input is the post data stored in the database, and the output is the preprocessed text data.

[0943] Step 4:

[0944] The server performs topic modeling using the preprocessed data. In this step, Gensim is used to extract major topics using Latent Dirichlet Allocation (LDA). For example, from posts related to "environmental issues," topics such as "climate change," "recycling," and "energy" are extracted. The input is the preprocessed text data, and the output is the extracted topic information.

[0945] Step 5:

[0946] Newly posted data is acquired in real time from the user's device and sentiment analysis is performed. When a new post is entered by a user, its text and emotional tone are sent to the server in real time, and the emotional tone is analyzed using VaderSentiment. For example, the sentence "I'm so happy today!" is analyzed as a positive sentiment. The input is the user's new post data, and the output is the analyzed emotional tone.

[0947] Step 6:

[0948] The server compares the topic and emotional tone of new posts acquired in real time with past post data to evaluate consistency. Specifically, it compares the topic information and emotional tone of new posts with the past database and calculates the degree of match. The input is the topic information and emotional tone of new posts, and the output is the consistency evaluation result.

[0949] Step 7:

[0950] If the consistency is low, the server will warn the user and suggest the most appropriate expression. For example, if a user posts "Cars are bad for the environment" while also posting "I bought a new car!", the server will display a warning saying, "This statement contradicts previous statements. Would you like to express it as 'I bought a new, environmentally friendly car'?" The input is the consistency evaluation result and the topic information of the new post, and the output is a warning and suggestion to be displayed to the user.

[0951] Step 8:

[0952] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user's device of the risk value. Specifically, it uses Scikit-learn to compare past hot topics with the new post and calculate the risk value. The input is the topic information of the new post and past hot topics, and the output is the hot topic risk value.

[0953] Step 9:

[0954] Based on the feedback from the server, the user can modify the content and emotional tone of the new post, or cancel the post altogether. Specifically, using React Native, the user can confirm, modify, or cancel the post from the user interface. The input is warnings and suggestions from the server, and the output is the modified post content, the cancellation of the post, or the post content as is.

[0955] 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.

[0956] 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.

[0957] 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.

[0958] [Third embodiment]

[0959] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0960] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0961] 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).

[0962] 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.

[0963] 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.

[0964] 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).

[0965] 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. 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.

[0966] 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.

[0967] 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.

[0968] 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.

[0969] 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.

[0970] 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."

[0971] MODE FOR CARRYING OUT THE INVENTION

[0972] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. This system has the following components:

[0973] (1. Data collection and integration)

[0974] User authentication and SNS integration

[0975] Users log in to the system and link their various SNS accounts, which then stores their authentication information (username, password, OAuth token, etc.) in the system.

[0976] Social Media Data Extraction

[0977] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database and used for later analysis.

[0978] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms.

[0979] (2. Speech analysis and consistency check)

[0980] Data Preprocessing

[0981] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[0982] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[0983] Topic Modeling

[0984] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[0985] Example: If a user's past posts have primarily been about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[0986] (3. Real-time warning)

[0987] Submit a new post

[0988] When a user posts a new message, the message is sent to the server in real time.

[0989] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to a server.

[0990] Consistency Check

[0991] The server preprocesses the text of new posts and compares them with topic models to assess consistency.

[0992] Example: Detect when the topic of a new post "I bought a new car!" contradicts a previous post that said "I won't buy a car because it's bad for the environment."

[0993] Displaying warnings and suggestions

[0994] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[0995] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[0996] (4. Risk Assessment)

[0997] Risk Assessment

[0998] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[0999] Example: If a past post containing "car" is deemed high risk, the system will determine that "this post is high risk" and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[1000] (5. Final confirmation and submission)

[1001] Confirm or cancel your post

[1002] Users can choose to modify their post based on the suggestions and warnings provided, or cancel the post.

[1003] Example: A user corrects the post to "I bought a new, green car" and confirms the post.

[1004] Through these steps, the present invention can maintain consistency between a user's past and present comments and prevent potential risks of flame wars.

[1005] The processing flow will be explained below.

[1006] Step 1:

[1007] Users log in to the platform and link their various social networking service (SNS) accounts using authentication protocols such as OAuth.

[1008] Step 2:

[1009] The server uses the authentication information provided by the user to collect past posting data via the SNS API, including the post content, date and time, and metadata.

[1010] Step 3:

[1011] The server stores the collected data in a database for later analysis.

[1012] Step 4:

[1013] The server periodically scans the contents of the database to see if any newly collected submitted data has been collected.

[1014] Step 5:

[1015] The server pre-processes the newly collected submission data, which includes text tokenization, morphological analysis, and stop-word removal.

[1016] Step 6:

[1017] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[1018] Step 7:

[1019] When a user attempts to post a new message, the user's device sends the message to the server in real time.

[1020] Step 8:

[1021] The server receives the content of new posts and performs pre-processing, which includes tokenization and morphological analysis.

[1022] Step 9:

[1023] The server compares the topic of each new post with past posts to assess consistency.

[1024] Step 10:

[1025] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading expressions and suggest optimal expressions.

[1026] Step 11:

[1027] The server calculates a risk value for new posts based on past examples of controversy. This risk assessment takes into account the probability that similar posts have caused controversy in the past and the frequency of negative reactions to a particular topic.

[1028] Step 12:

[1029] The server notifies the user of the results of the risk assessment and warns them if the risk is high.

[1030] Step 13:

[1031] Based on the feedback from the server, the user can choose to modify the content of the new post or to confirm the post as is.

[1032] Step 14:

[1033] The user's device sends the final post content to the SNS, completing the post.

[1034] Example 1

[1035] 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."

[1036] There is a need for a system that maintains the consistency of new posts with past posting data on social networking services (SNS) and prevents potential flame war risks. However, conventional SNS management systems do not check the consistency of new posts or evaluate the flame war risk in real time, which can result in users posting inappropriate content. Furthermore, if a user posts content that contradicts past posts, this can damage the user's credibility and reputation. To solve these issues, a system is needed that checks the consistency of past posting data and evaluates the flame war risk in real time when a user posts a new post, and provides appropriate warnings and suggestions.

[1037] 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.

[1038] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked SNSs, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topic of the new post with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, and means for allowing the user to modify or cancel the new post based on the warning or suggestion received. This allows users to maintain the consistency of their new posts while preventing potential uproar risks.

[1039] "Authentication information" refers to information used to verify the identity of a user, and includes a username, password, OAuth token, and the like.

[1040] A "social networking service (SNS)" is an online service that allows users to interact with and share information with other users.

[1041] "Posted data" is a general term for content such as text, images, and videos posted by users on SNS.

[1042] "Preprocessing" refers to the cleaning and formatting of collected posted data to make it analyzable, such as tokenization, morphological analysis, and removal of stop words.

[1043] "Topic modeling" is a machine learning technique for extracting major topics and themes from collected text data, and commonly used techniques include Latent Dirichlet Allocation (LDA).

[1044] "Acquiring in real time" refers to the process of immediately sending the content of a post to a server for analysis when a user posts a new post.

[1045] "Consistency" is an evaluation criterion that indicates whether a new post is consistent with previous posts.

[1046] "Displaying a warning" means that if there is a problem with the content of a new post made by a user, the content will be notified to the user and a warning will be issued.

[1047] "Flame risk" refers to the risk that a new post will receive a negative reaction from ordinary users and be criticized or attacked online.

[1048] The "risk value" is an index that quantitatively indicates the degree of risk of a controversy.

[1049] "Cancel" means that the user cancels a new post.

[1050] A "database system" is a structured data storage system for storing and managing collected data.

[1051] The present invention is a system that collects past posting data from social networking services (SNS) that users connect to, and performs consistency and risk assessment of new posts. This system is composed of several major hardware and software components, each of which works in conjunction to achieve the objectives of the invention.

[1052] 1. Data collection and integration

[1053] User authentication and SNS integration

[1054] First, a user logs in to the system and connects their social media account. At this time, the system stores the user's authentication information (username, password, OAuth token, etc.). Authentication is performed using the API of the specific social media platform (e.g., Twitter or Instagram).

[1055] Social Media Data Extraction

[1056] The server uses the API of each SNS to collect past posting data from the linked SNSs. The collected data is stored in a centralized database. API calls can be implemented using programming languages ​​such as Python and Node.js.

[1057] Example: When a user connects Twitter and Instagram, the server uses the Twitter API and Instagram API to retrieve past tweets and Instagram posts and save them in a database.

[1058] 2. Speech analysis and consistency check

[1059] Data Preprocessing

[1060] The server preprocesses the collected historical post data using natural language processing (NLP) techniques, such as tokenization, morphological analysis, and stop word removal. For NLP, Python libraries such as NLTK and spacy can be used.

[1061] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[1062] Topic Modeling

[1063] The server extracts major topics from users' past posts using topic modeling techniques such as Latent Dirichlet Allocation (LDA), which can be implemented using the scikit-learn or gensim libraries.

[1064] Example: If a user's past posts are mainly about "environmental issues," topics such as "environment," "eco," and "climate change" will be extracted.

[1065] 3. Real-time alerts

[1066] Submit a new post

[1067] When a user posts a new post, the content is sent to the server in real time. When the user clicks the post button, the content is sent to the server as an asynchronous request.

[1068] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to the server and first analyzed by the system.

[1069] Consistency Check

[1070] The server preprocesses the text of new posts and compares it with the topic model to evaluate consistency. Based on the comparison, it determines whether the new post is inconsistent with previous posts.

[1071] Example: If the topic of a new post "I bought a new car!" contradicts a previous post "I won't buy a car because it's bad for the environment," the contradiction will be detected.

[1072] Displaying warnings and suggestions

[1073] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[1074] Example: A warning message appears saying, "This statement contradicts a previous statement. Would you say, 'I bought a new, eco-friendly car'?"

[1075] 4. Risk Assessment

[1076] Risk Assessment

[1077] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the result. The risk assessment can use a machine learning model using past data.

[1078] Example: If a post containing the keyword "car" has caused controversy in the past, a notification will be sent to any new posts saying, "This post is at high risk of causing controversy. Please reconsider."

[1079] 5. Final check and submission

[1080] Confirm or cancel your post

[1081] Based on the suggestions and warnings provided, users can choose to either correct their post or cancel it. The corrected post is then sent back to the server for a final check.

[1082] Example: If the user changes the post to "I bought a new, eco-friendly car" and confirms the post, the post is sent to the server again. If the user wants to cancel the post, they can press the cancel button to abort the post process.

[1083] Examples of prompt statements

[1084] As an example of user input, the following prompt sentence can be input to the generative AI model.

[1085] Example prompt:

[1086] "Link your Twitter and Instagram accounts to collect past posts. Use that data to evaluate new posts in real time for consistency and risk of outrage, and implement a system that provides warnings and suggestions to users as needed."

[1087] This allows users to maintain consistency in new posts and prevent potential flame wars.

[1088] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1089] Step 1:

[1090] A user logs in to the system and connects their social media account. At this time, they enter their authentication information (username, password, OAuth token, etc.). The server receives this authentication information and, once authenticated, stores it internally in the system.

[1091] Input: The credentials entered by the user into the system.

[1092] Output: Authenticated user information and linked SNS account information are saved in the system.

[1093] Specific operation: The user enters their username and password on the login screen, clicks the login button, and then clicks the SNS link button to complete OAuth authentication on the authentication screen of each SNS.

[1094] Step 2:

[1095] The server uses the API of each SNS to collect past posting data from the SNSs that the user has linked. At this time, the server sends authentication information to the API of each SNS and stores the posting data in a centralized database.

[1096] Input: API key, user credentials.

[1097] Output: Past posting data is saved in the database.

[1098] Specific operation: The server periodically calls the API of each SNS, retrieves past posting data (for example, tweets and Instagram posts from the past year), and stores it in a database.

[1099] Step 3:

[1100] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[1101] Input: Past posting data stored in the database.

[1102] Output: Preprocessed submission data is stored in a database for analysis.

[1103] What it does: The server uses Python's NLTK library to tokenize the text "I bought a new car!" into "new," "car," and "buy." It also removes stop words (such as "!").

[1104] Step 4:

[1105] The server uses the preprocessed data to extract key topics using topic modeling techniques such as Latent Dirichlet Allocation (LDA).

[1106] Input: Preprocessed submission data.

[1107] Output: The extracted topic information is stored in the topic model database.

[1108] What it does: The server uses the scikit-learn library to apply an LDA model to extract topics such as "environment," "eco," and "climate change."

[1109] Step 5:

[1110] When a user posts a new post, the device sends the post content to the server in real time. The device sends an asynchronous request when the user clicks the post button.

[1111] Input: The user's new post.

[1112] Output: The post is sent to the server and saved.

[1113] Specific action: A user types "I bought a new car!" into the Instagram posting screen and presses the post button.

[1114] Step 6:

[1115] The server preprocesses the content of new posts and compares them with the topic model to assess consistency, ensuring that the topic of the new post matches previous data.

[1116] Input: New post content.

[1117] Output: Topic comparison results. Coherence evaluation results are generated.

[1118] What happens: The server preprocesses the text "I bought a new car!" and determines that it is inconsistent with past posts such as "I won't buy a car because it's bad for the environment."

[1119] Step 7:

[1120] If the server determines that the consistency is low, it will send a warning message to the user's device and suggest the most appropriate expression.

[1121] Input: Consistency assessment results.

[1122] Output: A warning message and a suggestion are displayed on the user's terminal.

[1123] Specific operation: The server sends a warning message to the user's terminal saying, "This statement contradicts a previous statement. Would you like to say, 'I bought a new, environmentally friendly car'?"

[1124] Step 8:

[1125] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the results.

[1126] Input: New post content, past flame war data.

[1127] Output: The result of the flame risk assessment is notified to the user.

[1128] Specific operation: The server uses a machine learning model to send a notification saying, "This post has a high risk of causing a backlash. Please reconsider," based on past examples of controversy surrounding the keyword "car."

[1129] Step 9:

[1130] Based on the suggestions and warnings provided, users can choose to modify or cancel their post, after which the modified post is sent back to the server for a final review.

[1131] Input: User's revised submission or cancellation information.

[1132] Output: The revised submission is confirmed or the submission is cancelled.

[1133] Specific actions: The user corrects the post to "I bought a new, eco-friendly car" and confirms the post, or presses the cancel button without pressing the post button.

[1134] Through these steps, the system can maintain consistency in new posts by users and prevent potential flame wars.

[1135] (Application example 1)

[1136] 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."

[1137] Conventional SNS posting systems lack a means to evaluate whether the content of a user's new post is consistent with past posts, making it difficult to prevent the risk of a flame war. Furthermore, there is a lack of a systematic method to determine whether a new post poses a flame war risk. This can lead to users unintentionally causing a flame war and damaging the credibility of individuals and companies.

[1138] 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.

[1139] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk value of the new post based on past risk cases and notifying the user of the risk value, and means for inputting prompt sentences into the generative AI model to assist in the consistency evaluation. This makes it possible to maintain consistency in the content newly posted by users and prevent potential flame wars.

[1140] "User authentication information" refers to the identifiers and authentication data required for a user to access the system, including, for example, a username, password, OAuth token, etc.

[1141] "Integrated Social Networking Service" refers to a social networking platform that allows users to connect their social networking account to the system and enable data collection.

[1142] "Past posting data" refers to all text, images, videos, and related metadata that a user has posted to the linked SNS platform.

[1143] "Preprocessing" refers to a series of data cleansing operations performed to convert collected data into an analyzable format, including tokenization, morphological analysis, and stop word removal.

[1144] "Topic modeling" is a statistical method for extracting major topics or themes from collected text data, using algorithms such as Latent Dirichlet Allocation (LDA).

[1145] "Acquisition in real time" means that newly posted data is sent to the system the moment a user attempts to post it, and analysis begins immediately.

[1146] "Evaluating consistency" means determining whether the subject matter and expressions of newly posted content are consistent with topics extracted from past posting data.

[1147] A "warning" is a message that is displayed to the user when a contradiction is detected as a result of the consistency evaluation, and provides specific instructions and suggestions for improvement.

[1148] "Evaluating the risk value" means quantitatively calculating the risk factors contained in a new post based on past risk cases and notifying the user of the results.

[1149] A "generative AI model" is an artificial intelligence model for natural language generation and text analysis, and refers to advanced models such as GPT.

[1150] A "prompt" is an instruction or question input to a generative AI model, which serves as a trigger for the model to generate an appropriate response or analysis result.

[1151] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. The system includes the following components:

[1152] First, the user's authentication information is obtained. The user connects their social media accounts (e.g., multiple social media platforms) to the system. Specifically, authentication information such as username, password, and OAuth token are stored on the server. The server then collects the user's past posting data from the connected social media platforms via API. This data is stored in a centralized database and used for later analysis.

[1153] The collected data is preprocessed on the server. This preprocessing includes data cleansing operations, such as removing URLs, mentions, and hashtags, and merging consecutive spaces. This preprocessed data is then subjected to topic modeling. The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. This makes it possible to clarify trends in users' interests and concerns.

[1154] Next, newly posted data is sent to the server in real time. When a user attempts to create a new post, the content of the post is immediately sent to the server. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. If the result of this consistency evaluation indicates low consistency, a warning is displayed on the user's device. For example, a warning message such as "This post contradicts previous statements" is displayed. Specific alternatives are also provided, such as "I bought a new, environmentally friendly car."

[1155] Additionally, the server can input prompts to the generative AI model to assist in consistency assessment, such as "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If not, please tell me why."

[1156] This makes it possible to maintain consistency in the content of new posts by users and prevent potential risks of online outrage. For example, if a user tries to create a post saying "I bought a new car!", the system will detect that the content of that post is inconsistent with a previous post saying "I won't buy a car because it's bad for the environment," and generate a warning.

[1157] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1158] Step 1:

[1159] User authentication and SNS integration

[1160] The user logs in to the system and connects various SNS accounts. This causes the user's authentication information (username, password, OAuth token, etc.) to be stored on the server. The input is the user's authentication information and SNS account information, and the output is the authentication information being saved. Specifically, the authentication information provided by the user is used to authenticate via the SNS API and establish the connection.

[1161] Step 2:

[1162] Social media data collection

[1163] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database. The input is authenticated SNS account information, and the output is the collected past posting data. Specifically, the server sends an API request and stores the posting data returned from the SNS in the database.

[1164] Step 3:

[1165] Data Preprocessing

[1166] The server preprocesses the collected past posting data. This includes removing URLs, mentions, and hashtags, and merging consecutive spaces. The input is the collected posting data, and the output is preprocessed text data. Specifically, the data is cleansed using regular expressions and text analysis libraries.

[1167] Step 4:

[1168] Topic Modeling

[1169] The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. The input is preprocessed text data, and the output is a topic distribution. Specifically, the preprocessed data is vectorized using CountVectorizer, and then the topics are extracted by training an LDA model.

[1170] Step 5:

[1171] Submitting and analyzing new posts

[1172] When a user attempts to create a new post, its contents are sent to the server in real time. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. The input is the text of the new post, and the output is the result of the consistency evaluation. Specifically, the server preprocesses the new post to calculate a topic distribution and compares it with past topic distributions.

[1173] Step 6:

[1174] Displaying warnings and suggestions

[1175] If the server determines that the consistency is low, it displays a warning message on the user's device and offers alternatives. The input is the consistency evaluation result, and the output is a warning message and alternatives. Specifically, a warning message pops up on the user's device and suggests expressions such as "I bought a new, environmentally friendly car."

[1176] Step 7:

[1177] Risk Assessment and Notification

[1178] The server calculates the risk value of a new post based on past risk case data and notifies the user of the result. The input is the text of the new post and past risk case data, and the output is a risk value and a notification message. Specifically, the server uses the new post as input and inputs a prompt sentence into the risk assessment model, generating a notification such as "This post is high risk. Please reconsider."

[1179] Step 8:

[1180] Using generative AI models

[1181] The server inputs a prompt sentence into the generative AI model to assist in consistency evaluation. The input is the prompt sentence, and the output is the analysis result of the generative AI model. A specific prompt sentence is used: "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If it does not match, please tell me the reason."

[1182] 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.

[1183] MODE FOR CARRYING OUT THE INVENTION

[1184] This invention is a system that precisely evaluates the consistency and risk of flaming of new posts on SNS by combining an emotion engine that recognizes user emotions. This system has the following components.

[1185] (1. Data collection and integration)

[1186] User authentication and SNS integration

[1187] Users log in to the system and link their social networking service (SNS) accounts, which allows the system to store their authentication information.

[1188] Social Media Data Extraction

[1189] The server collects users' past posting data from linked SNSs via the SNS API and stores it in a database.

[1190] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms and stores them centrally.

[1191] (2. Speech analysis and consistency check)

[1192] Data Preprocessing

[1193] The server pre-processes the collected historical posting data, which includes text tokenization, morphological analysis, and stop word removal.

[1194] Example: The text "I bought a new car!" is tokenized and split into the words "new," "car," and "buy."

[1195] Topic Modeling

[1196] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[1197] Example: If a user's past posts are about "environmental issues," the topic model might include "environment," "eco," "climate change," etc.

[1198] (3. Emotion Engine Integration)

[1199] Emotion recognition

[1200] The server uses an emotion engine to recognize emotions from users' past posting data and store them in a database.

[1201] Example: A past post such as "I'm so happy today!" is perceived as a positive emotion, while a post such as "I'm sad" is perceived as a negative emotion.

[1202] Emotional tone analysis of new posts

[1203] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[1204] Example: When a user posts on Instagram, "I bought a new car!", that positive emotional tone is also sent to the server.

[1205] (4. Real-time warning)

[1206] Consistency Assessment

[1207] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[1208] Example: Detecting a mismatch in positive emotional tone between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment."

[1209] Displaying warnings and suggestions

[1210] If the server determines that the consistency is low, it will display a warning and suggest the most appropriate expression on the user's device.

[1211] Example: "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[1212] (5. Risk Assessment)

[1213] Flaming risk assessment

[1214] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[1215] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[1216] (6. Final confirmation and submission)

[1217] Confirm or cancel your post

[1218] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[1219] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[1220] Through these steps, the present invention can maintain consistency between a user's past and present comments and emotions, and prevent potential risks of flame wars.

[1221] The processing flow will be explained below.

[1222] Step 1:

[1223] Users log in to the system and link their social networking service (SNS) accounts. This linking is done using an authentication protocol such as OAuth, and the authentication information is stored in the system.

[1224] Step 2:

[1225] The server uses the user's authentication information to collect past posting data via the SNS API, including the content, date and time of the post, and metadata.

[1226] Step 3:

[1227] The server stores the collected data in a database for later analysis.

[1228] Step 4:

[1229] The server pre-processes the historical submission data stored in the database, including text tokenization, morphological analysis, and stop word removal.

[1230] Example: "I bought a new car!" is tokenized as "new," "car," and "buy."

[1231] Step 5:

[1232] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[1233] Example: If a user's past posts are about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[1234] Step 6:

[1235] The server uses an emotion engine to recognize emotions from users' past posts, using text analysis to classify emotional states as positive, negative, etc.

[1236] Example: "I'm so happy today!" is perceived as positive, while "I'm sad" is perceived as negative.

[1237] Step 7:

[1238] The server stores the emotion recognition results in a database.

[1239] Step 8:

[1240] When a user attempts to post a new message, the user's device sends the message's content and emotional tone to the server in real time.

[1241] Example: When a user posts "I bought a new car!" on Instagram, the content and positive emotional tone are sent to the server.

[1242] Step 9:

[1243] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[1244] Example: Detect a contradiction between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment" in terms of emotional tone.

[1245] Step 10:

[1246] If the server determines that the consistency is low, it displays a warning message on the user's device and suggests the most appropriate expression.

[1247] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[1248] Step 11:

[1249] The server calculates the risk of a new post becoming a hot topic based on past examples of such hot topics, and notifies the user of the results.

[1250] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[1251] Step 12:

[1252] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post.

[1253] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[1254] Step 13:

[1255] The user's device sends the final post content to the SNS, completing the post.

[1256] Example 2

[1257] 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."

[1258] In modern social networking services (SNS), when users post a new message, it is difficult to judge whether the message is consistent with past comments or whether it poses a risk of causing a backlash. This can lead to misunderstandings due to inconsistencies with the user's past posts, or inappropriate comments that can cause a backlash, which can have a negative impact on the user.

[1259] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for acquiring user authentication information and collecting past post data from multiple linked social networking services; means for preprocessing the collected past post data and performing topic modeling to extract major topics; means for recognizing emotions from the collected past post data and saving the emotional data; means for transmitting new posts by users to the server in real time and acquiring the content and emotional tone of the new post; means for preprocessing the newly posted data and comparing the topic and emotional tone of the new post with past data to evaluate consistency; means for displaying a warning on the user's terminal and suggesting optimal expressions if consistency is low; means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value; and means for the user to modify or cancel the content and emotional tone of the new post based on feedback from the server. This makes it possible to prevent the risk of a new post causing an uproar while maintaining consistency with the user's past comments.

[1260] "User authentication information" refers to information such as an ID and password required for a user to log in to a system.

[1261] "Social networking service" is a general term for services that provide a platform for users to interact with other users online.

[1262] "Posted data" refers to information such as messages, images, and videos posted by users on social networking services.

[1263] "Preprocessing" refers to processes such as tokenization, morphological analysis, and removal of stop words that are performed to make the collected data easier to analyze.

[1264] "Topic modeling" is a statistical method for extracting latent topics (themes) from large amounts of text data.

[1265] "Emotion recognition" is a technology that identifies the emotions (positive, negative, etc.) contained in text from text data.

[1266] "Emotion data" is information about the emotions indicated by the text data, obtained through emotion recognition.

[1267] "Real-time" refers to data and information being processed immediately.

[1268] "Emotional tone" refers to the emotional atmosphere or nuance of a particular text or message.

[1269] "Consistency assessment" is the process of assessing whether the content of a new post is consistent with previous posts and whether the emotional tone is consistent.

[1270] "Warning display" refers to displaying a message to alert the user.

[1271] "Expression suggestions" refers to presenting appropriate expressions for a user's new posts.

[1272] "Flame risk" is the likelihood that a particular post will provoke a large number of negative reactions.

[1273] "Feedback" refers to information provided to users about analysis and evaluation results from the system.

[1274] MODE FOR CARRYING OUT THE INVENTION

[1275] The present invention provides a system that recognizes user emotions and more precisely evaluates the consistency and risk of a new post on a social networking site. This system is constructed through the operation of a server, a terminal, and a user. Specific embodiments of the system are described in detail below.

[1276] 1. User authentication and SNS integration

[1277] Users log in to the system by entering their ID and password, and then link their various social media accounts (e.g., Twitter, Instagram) to the system, which then securely stores their authentication information on the server.

[1278] 2. Collection and storage of social media data

[1279] The server uses APIs such as Twitter API and Instagram API to collect past posting data from linked social media accounts. This data is stored in a database and used for subsequent processing.

[1280] 3. Data Preprocessing

[1281] The server preprocesses the collected data, which includes text tokenization, morphological analysis, stopword removal, etc. For example, a post saying "I bought a new car!" is broken down into "new," "car," and "buy."

[1282] 4. Topic Modeling

[1283] The server uses a Latent Dirichlet Allocation (LDA) model to extract key topics from the preprocessed data, for example, "environment," "eco," and "climate change" from posts about "environmental issues."

[1284] 5. Recognizing and storing emotions

[1285] The server analyzes the collected post data using an emotion engine (e.g., VADER), recognizes the emotion (e.g., positive or negative) of each post, and stores it in a database. For example, a post saying, "I'm really happy today!" is recognized as a positive emotion.

[1286] 6. Emotional Tone Analysis of New Posts

[1287] When a user posts a new message, the user's device sends the message's text and emotional tone to the server in real time. The server preprocesses the text and emotional tone of the new message and temporarily stores them in a database.

[1288] 7. Consistency Assessment

[1289] The server compares the text and emotional tone of new posts with past data to assess consistency. If the content of a new post contradicts previous posts, for example, "I bought a new car!", it will determine that this contradicts the previous post, "I won't buy a car because it's bad for the environment."

[1290] 8. Displaying warnings and suggestions

[1291] If the server determines that the statement is inconsistent, it displays a warning message on the user's device and suggests the most appropriate expression. For example, it displays a warning such as, "This statement contradicts previous statements and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[1292] 9. Risk Assessment

[1293] The server calculates the risk value of new posts based on past data on flaming cases and notifies the user of the result. For example, if a new post containing "car" is judged to have a high risk of flaming, the server sends a notification saying, "This post is high risk. Please reconsider."

[1294] 10. Confirming or Cancelling a Post

[1295] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether. The user can modify the content and emotional tone to "I bought a new, eco-friendly car," and finally confirm the post after confirming the emotional tone.

[1296] Examples and prompts

[1297] As a specific example of operation, a user links their Twitter and Instagram accounts to the system and collects past posting data from these accounts. When a user posts a new message such as "I bought a new car!", if that message contradicts a previous message such as "I won't buy a car because it's bad for the environment," the system displays a warning saying, "This statement contradicts your previous message and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[1298] Examples of input prompts for a generative AI model might include the following:

[1299] "Design a system that analyzes past posts on Twitter and Instagram and evaluates the consistency and sentiment of new posts by users."

[1300] "Create a system that incorporates an emotion engine that recognizes user emotions and evaluates the risk of social media posts causing controversy."

[1301] These techniques make it possible to maintain consistency between a user's past and present comments and emotions, and prevent potential risks of online outrage.

[1302] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1303] Program processing flow

[1304] Step 1:

[1305] Users log in to the system by entering their ID and password.

[1306] Input: User ID, Password

[1307] Output: Feedback of authentication success or failure

[1308] Specific operation: The user enters information into the login form and presses the "Login" button. The system authenticates the user ID and password, and if successful, displays the dashboard.

[1309] Step 2:

[1310] Users use the system's interface to link their Twitter and Instagram accounts to the system.

[1311] Input: User's social media account

[1312] Output: Notification of successful SNS account linking

[1313] Specific operation: The user clicks the link button and completes authentication on the SNS authentication screen that appears. The authentication token is saved on the server.

[1314] Step 3:

[1315] The server uses the Twitter API and Instagram API to collect users' past posting data and store it in a database.

[1316] Input: SNS account authentication token

[1317] Output: Collected past posting data

[1318] Specific operation: The server calls the Twitter API to retrieve tweets within the specified period, and similarly calls the Instagram API to retrieve past post data.

[1319] Step 4:

[1320] The server pre-processes the collected data, which includes text tokenization, morphological analysis, and stop-word removal.

[1321] Input: Collected submission data

[1322] Output: Preprocessed data (tokenized, morphologically analyzed, stopwords removed)

[1323] Specific behavior: Breaks down a post like "I bought a new car!" into "new," "car," and "buy," and removes meaningless words.

[1324] Step 5:

[1325] The server uses a Latent Dirichlet Allocation (LDA) model to extract major topics from the preprocessed data.

[1326] Input: Preprocessed data

[1327] Output: Extracted topics (e.g. "environment", "eco", "climate change")

[1328] What it does: Runs an LDA model to identify key topics from a user's past posts.

[1329] Step 6:

[1330] The server analyzes the collected posting data using an emotion engine, recognizes the emotion of each post, and stores it in a database.

[1331] Input: Collected submission data

[1332] Output: Recognized emotion data (positive, negative, etc.)

[1333] Specific behavior: Using an emotion engine such as VADER, we recognize "I'm so happy today!" as a positive emotion and "I'm sad" as a negative emotion.

[1334] Step 7:

[1335] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[1336] Input: New post text and emotional tone

[1337] Output: New post data and emotional tone sent to the server

[1338] What happens: A user posts on Instagram, "I bought a new car!" and sends that positive emotional tone to the server.

[1339] Step 8:

[1340] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[1341] Input: New post text and emotional tone

[1342] Output: Result of consistency assessment (e.g., consistent, inconsistent)

[1343] Specific behavior: Determine whether the new post "I bought a new car!" contradicts the previous post "I won't buy a car because it's bad for the environment."

[1344] Step 9:

[1345] If the server determines that the consistency is low, it displays a warning message and suggests the most appropriate expression on the user's device.

[1346] Input: Results of consistency assessment

[1347] Output: Warning message and suggested expression

[1348] What to do: Display a warning that says, "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[1349] Step 10:

[1350] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[1351] Input: Results of consistency assessment and content of new submission

[1352] Output: Flame risk assessment notification

[1353] What it does: If a new post containing the word "car" has previously sparked outrage, we'll send a notification saying, "This post is high risk. Please reconsider."

[1354] Step 11:

[1355] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[1356] Input: Feedback

[1357] Output: The revised post or the cancellation of the post

[1358] What happens: The user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[1359] (Application example 2)

[1360] 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."

[1361] In recent years, as social networking services (SNS) have become more widespread, users are increasingly facing the risk of their posts sparking outrage. In particular, if a user's emotions are not consistent with their previous posts, it is easy for misunderstandings and resentment to arise, resulting in a higher risk of outrage. Furthermore, as the number of posts on content distribution services increases, there is a demand for systems that can appropriately evaluate emotions and consistency. However, existing systems lack sufficient emotional analysis and outrage risk assessment, making it difficult for users to receive appropriate feedback and warnings. Providing an effective system to improve this situation is a challenge.

[1362] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1363] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, means for analyzing the emotional tone of the new post in real time using sentiment analysis, and means for the user to confirm, modify, or cancel the content and emotional tone of the post in the content distribution service. This makes it possible to maintain consistency between the user's past and present comments and emotions, prevent potential uproars, and provide appropriate feedback in real time.

[1364] "User Credentials" means the unique identifying information required for a user to access a system.

[1365] A "social networking service" is an online platform that enables users to interact with and share content with other users over the Internet.

[1366] "Previously posted data" refers to digital content such as text, images, and videos that a user previously posted on a social networking service.

[1367] "Preprocessing" refers to a series of processes, such as text normalization, tokenization, morphological analysis, and stop word removal, that are carried out to convert collected data into a format that is easier to analyze.

[1368] "Topic modeling" is a statistical method for automatically extracting themes and topics from large amounts of text data.

[1369] "Latent Dirichlet Allocation (LDA)" is a machine learning technique for understanding the topic structure hidden in documents, and models topics based on the probability of occurrence of each word.

[1370] "Acquisition in real time" means that the data is collected and processed immediately the moment a new post is made by a user.

[1371] "Emotion analysis" is a technology that recognizes a user's emotional state (positive, negative, neutral, etc.) from text data.

[1372] "Consistency assessment" is the process of comparing and determining whether a user's new posts are consistent with previous posts in terms of content and emotional tone.

[1373] "Display warning" means that when low consistency is detected, the problem and its solution are presented to the user.

[1374] "Optimal expression suggestions" means providing specific expressions to modify a user's new posts to ensure consistency with past posts and appropriately reflect the emotional tone.

[1375] "Flame risk assessment" is the process of predicting the likelihood that a new post will provoke criticism or resentment from other users if it contains inappropriate content or sentiment.

[1376] "Notifying risk value" means quantifying the possibility of a new post causing a backlash and informing the user.

[1377] A "content distribution service" is an internet-based service that allows users to create, share, and view digital content such as text, images, and videos.

[1378] "Means for checking, modifying, and canceling post content and emotional tone" refers to a function that allows users to check the content of new posts and the accompanying emotional tone in real time on the user interface, and to modify or cancel posts as necessary.

[1379] MODE FOR CARRYING OUT THE INVENTION

[1380] System configuration and program overview

[1381] The system acquires user authentication information, collects and analyzes past posting data from multiple linked social networking services (SNS), evaluates the emotional tone and consistency of new posts in real time, and notifies users of the risk of a controversy.The main software and hardware used include Django, Tweepy, Instagram Graph API, NLTK, Gensim, VaderSentiment, Scikit-learn, React Native, and PostgreSQL.

[1382] Data collection and integration

[1383] When a user logs in to the system and connects to a social media account, the server obtains their authentication information. This information is used to retrieve past posting data from the connected social media platform. Specifically, Twitter posting data is collected using Tweepy, and Instagram posting data is collected using the Instagram Graph API. The collected data is stored in a PostgreSQL database.

[1384] Data Preprocessing and Topic Modeling

[1385] The server converts the collected raw data into a format that is easy to analyze. This includes text tokenization, morphological analysis, and stop word removal. Gensim then performs topic modeling using Latent Dirichlet Allocation (LDA) on the preprocessed data to extract major topics. For example, a post saying "I bought a new car!" is broken down into words such as "new," "car," and "buy," and the major topics extracted are "car" and "shopping."

[1386] Sentiment analysis and real-time evaluation

[1387] When a new post is created on a user's device, its text and emotional tone are sent to the server in real time. VaderSentiment is used to analyze the emotional tone of the new post and the results are stored in a database. This makes it possible to evaluate the consistency between past posts and new posts. For example, if a user tries to post "I bought a new car!", the positive emotion is analyzed.

[1388] Displaying warnings and suggestions

[1389] The server compares new posts with past data to evaluate their consistency, and if the consistency is low, it displays a real-time warning on the user's device. It also suggests optimal expressions. For example, if a new post, "I bought a new car!", contradicts a past post, "Cars are bad for the environment," it suggests expressions like, "I bought a new, environmentally friendly car."

[1390] Flaming risk assessment and notification

[1391] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the risk value. Scikit-learn is used to compare past posts with new posts and calculate the risk value. For example, if a post containing the word "car" has become a hot topic in the past, a notification will be sent to the user stating that "this post is high risk."

[1392] Final check and submission

[1393] Based on the feedback from this system, users can choose to modify their new post or cancel the post altogether. This confirmation, modification, and cancellation functionality is provided on the user interface using React Native. This allows users to receive appropriate feedback and post safely while avoiding the risk of a social media outcry.

[1394] Specific examples

[1395] Examples:

[1396] User post: "The environment in urban areas has been deteriorating recently. New narrow roads are to blame."

[1397] Warning / Suggestion: This statement contradicts a previous post that said, "Urban development is a good thing." Would you say, "I've noticed an increase in narrow roads in urban areas recently"?

[1398] Prompt statement:

[1399] markdown

[1400] A user is attempting to post content about environmental issues. The post states, "The environment in urban areas has been deteriorating recently. New, narrow roads are to blame." Please assess the consistency and controversy risk of this post and previous posts and provide appropriate warnings and suggestions.

[1401] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1402] Step 1:

[1403] A user logs in to the system and links their social media account. In this step, the user enters their authentication information, and the server establishes a link with the social media account through the API. Specifically, the user provides their Twitter and Instagram authentication information via the API, and the server receives it and begins collecting past posting data from each platform. The input is the user's authentication information, and the output is the linked social media account information.

[1404] Step 2:

[1405] The server collects past posting data from linked SNSs and stores it in a database. The collected data includes content such as text, images, and videos. Specifically, it uses Tweepy to collect Twitter tweet data and Instagram posts using the Instagram Graph API. The input is post data obtained through the SNS API, and the output is past posting data stored in the database.

[1406] Step 3:

[1407] The server preprocesses the collected post data. In this step, processes such as text tokenization, morphological analysis, and stop word removal are performed. For example, a post "I bought a new car!" is split into words such as "new," "car," and "bought." Specifically, this preprocessing is performed using the NLTK library. The input is the post data stored in the database, and the output is the preprocessed text data.

[1408] Step 4:

[1409] The server performs topic modeling using the preprocessed data. In this step, Gensim is used to extract major topics using Latent Dirichlet Allocation (LDA). For example, from posts related to "environmental issues," topics such as "climate change," "recycling," and "energy" are extracted. The input is the preprocessed text data, and the output is the extracted topic information.

[1410] Step 5:

[1411] Newly posted data is acquired in real time from the user's device and sentiment analysis is performed. When a new post is entered by a user, its text and emotional tone are sent to the server in real time, and the emotional tone is analyzed using VaderSentiment. For example, the sentence "I'm so happy today!" is analyzed as a positive sentiment. The input is the user's new post data, and the output is the analyzed emotional tone.

[1412] Step 6:

[1413] The server compares the topic and emotional tone of new posts acquired in real time with past post data to evaluate consistency. Specifically, it compares the topic information and emotional tone of new posts with the past database and calculates the degree of match. The input is the topic information and emotional tone of new posts, and the output is the consistency evaluation result.

[1414] Step 7:

[1415] If the consistency is low, the server will warn the user and suggest the most appropriate expression. For example, if a user posts "Cars are bad for the environment" while also posting "I bought a new car!", the server will display a warning saying, "This statement contradicts previous statements. Would you like to express it as 'I bought a new, environmentally friendly car'?" The input is the consistency evaluation result and the topic information of the new post, and the output is a warning and suggestion to be displayed to the user.

[1416] Step 8:

[1417] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user's device of the risk value. Specifically, it uses Scikit-learn to compare past hot topics with the new post and calculate the risk value. The input is the topic information of the new post and past hot topics, and the output is the hot topic risk value.

[1418] Step 9:

[1419] Based on the feedback from the server, the user can modify the content and emotional tone of the new post, or cancel the post altogether. Specifically, using React Native, the user can confirm, modify, or cancel the post from the user interface. The input is warnings and suggestions from the server, and the output is the modified post content, the cancellation of the post, or the post content as is.

[1420] 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.

[1421] 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.

[1422] 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.

[1423] [Fourth embodiment]

[1424] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1425] 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.

[1426] 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).

[1427] 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.

[1428] 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.

[1429] 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).

[1430] 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. 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.

[1431] 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.

[1432] 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.

[1433] 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.

[1434] 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.

[1435] 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.

[1436] 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."

[1437] MODE FOR CARRYING OUT THE INVENTION

[1438] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. This system has the following components:

[1439] (1. Data collection and integration)

[1440] User authentication and SNS integration

[1441] Users log in to the system and link their various SNS accounts, which then stores their authentication information (username, password, OAuth token, etc.) in the system.

[1442] Social Media Data Extraction

[1443] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database and used for later analysis.

[1444] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms.

[1445] (2. Speech analysis and consistency check)

[1446] Data Preprocessing

[1447] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[1448] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[1449] Topic Modeling

[1450] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[1451] Example: If a user's past posts have primarily been about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[1452] (3. Real-time warning)

[1453] Submit a new post

[1454] When a user posts a new message, the message is sent to the server in real time.

[1455] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to a server.

[1456] Consistency Check

[1457] The server preprocesses the text of new posts and compares them with topic models to assess consistency.

[1458] Example: Detect when the topic of a new post "I bought a new car!" contradicts a previous post that said "I won't buy a car because it's bad for the environment."

[1459] Displaying warnings and suggestions

[1460] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[1461] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[1462] (4. Risk Assessment)

[1463] Risk Assessment

[1464] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[1465] Example: If a past post containing "car" is deemed high risk, the system will determine that "this post is high risk" and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[1466] (5. Final confirmation and submission)

[1467] Confirm or cancel your post

[1468] Users can choose to modify their post based on the suggestions and warnings provided, or cancel the post.

[1469] Example: A user corrects the post to "I bought a new, green car" and confirms the post.

[1470] Through these steps, the present invention can maintain consistency between a user's past and present comments and prevent potential risks of flame wars.

[1471] The processing flow will be explained below.

[1472] Step 1:

[1473] Users log in to the platform and link their various social networking service (SNS) accounts using authentication protocols such as OAuth.

[1474] Step 2:

[1475] The server uses the authentication information provided by the user to collect past posting data via the SNS API, including the post content, date and time, and metadata.

[1476] Step 3:

[1477] The server stores the collected data in a database for later analysis.

[1478] Step 4:

[1479] The server periodically scans the contents of the database to see if any newly collected submitted data has been collected.

[1480] Step 5:

[1481] The server pre-processes the newly collected submission data, which includes text tokenization, morphological analysis, and stop-word removal.

[1482] Step 6:

[1483] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[1484] Step 7:

[1485] When a user attempts to post a new message, the user's device sends the message to the server in real time.

[1486] Step 8:

[1487] The server receives the content of new posts and performs pre-processing, which includes tokenization and morphological analysis.

[1488] Step 9:

[1489] The server compares the topic of each new post with past posts to assess consistency.

[1490] Step 10:

[1491] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading expressions and suggest optimal expressions.

[1492] Step 11:

[1493] The server calculates a risk value for new posts based on past examples of controversy. This risk assessment takes into account the probability that similar posts have caused controversy in the past and the frequency of negative reactions to a particular topic.

[1494] Step 12:

[1495] The server notifies the user of the results of the risk assessment and warns them if the risk is high.

[1496] Step 13:

[1497] Based on the feedback from the server, the user can choose to modify the content of the new post or to confirm the post as is.

[1498] Step 14:

[1499] The user's device sends the final post content to the SNS, completing the post.

[1500] Example 1

[1501] 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."

[1502] There is a need for a system that maintains the consistency of new posts with past posting data on social networking services (SNS) and prevents potential flame war risks. However, conventional SNS management systems do not check the consistency of new posts or evaluate the flame war risk in real time, which can result in users posting inappropriate content. Furthermore, if a user posts content that contradicts past posts, this can damage the user's credibility and reputation. To solve these issues, a system is needed that checks the consistency of past posting data and evaluates the flame war risk in real time when a user posts a new post, and provides appropriate warnings and suggestions.

[1503] 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.

[1504] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked SNSs, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topic of the new post with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, and means for allowing the user to modify or cancel the new post based on the warning or suggestion received. This allows users to maintain the consistency of their new posts while preventing potential uproar risks.

[1505] "Authentication information" refers to information used to verify the identity of a user, and includes a username, password, OAuth token, and the like.

[1506] A "social networking service (SNS)" is an online service that allows users to interact with and share information with other users.

[1507] "Posted data" is a general term for content such as text, images, and videos posted by users on SNS.

[1508] "Preprocessing" refers to the cleaning and formatting of collected posted data to make it analyzable, such as tokenization, morphological analysis, and removal of stop words.

[1509] "Topic modeling" is a machine learning technique for extracting major topics and themes from collected text data, and commonly used techniques include Latent Dirichlet Allocation (LDA).

[1510] "Acquiring in real time" refers to the process of immediately sending the content of a post to a server for analysis when a user posts a new post.

[1511] "Consistency" is an evaluation criterion that indicates whether a new post is consistent with previous posts.

[1512] "Displaying a warning" means that if there is a problem with the content of a new post made by a user, the content will be notified to the user and a warning will be issued.

[1513] "Flame risk" refers to the risk that a new post will receive a negative reaction from ordinary users and be criticized or attacked online.

[1514] The "risk value" is an index that quantitatively indicates the degree of risk of a controversy.

[1515] "Cancel" means that the user cancels a new post.

[1516] A "database system" is a structured data storage system for storing and managing collected data.

[1517] The present invention is a system that collects past posting data from social networking services (SNS) that users connect to, and performs consistency and risk assessment of new posts. This system is composed of several major hardware and software components, each of which works in conjunction to achieve the objectives of the invention.

[1518] 1. Data collection and integration

[1519] User authentication and SNS integration

[1520] First, a user logs in to the system and connects their social media account. At this time, the system stores the user's authentication information (username, password, OAuth token, etc.). Authentication is performed using the API of the specific social media platform (e.g., Twitter or Instagram).

[1521] Social Media Data Extraction

[1522] The server uses the API of each SNS to collect past posting data from the linked SNSs. The collected data is stored in a centralized database. API calls can be implemented using programming languages ​​such as Python and Node.js.

[1523] Example: When a user connects Twitter and Instagram, the server uses the Twitter API and Instagram API to retrieve past tweets and Instagram posts and save them in a database.

[1524] 2. Speech analysis and consistency check

[1525] Data Preprocessing

[1526] The server preprocesses the collected historical post data using natural language processing (NLP) techniques, such as tokenization, morphological analysis, and stop word removal. For NLP, Python libraries such as NLTK and spacy can be used.

[1527] Example: The text "I bought a new car!" is tokenized as "new", "car", and "buy" and saved with stop words removed.

[1528] Topic Modeling

[1529] The server extracts major topics from users' past posts using topic modeling techniques such as Latent Dirichlet Allocation (LDA), which can be implemented using the scikit-learn or gensim libraries.

[1530] Example: If a user's past posts are mainly about "environmental issues," topics such as "environment," "eco," and "climate change" will be extracted.

[1531] 3. Real-time alerts

[1532] Submit a new post

[1533] When a user posts a new post, the content is sent to the server in real time. When the user clicks the post button, the content is sent to the server as an asynchronous request.

[1534] Example: When a user posts on Instagram, "I bought a new car!", the post is sent to the server and first analyzed by the system.

[1535] Consistency Check

[1536] The server preprocesses the text of new posts and compares it with the topic model to evaluate consistency. Based on the comparison, it determines whether the new post is inconsistent with previous posts.

[1537] Example: If the topic of a new post "I bought a new car!" contradicts a previous post "I won't buy a car because it's bad for the environment," the contradiction will be detected.

[1538] Displaying warnings and suggestions

[1539] If the server determines that the content is inconsistent, it will display a warning message on the user's device. It will also point out potentially misleading content and offer alternatives.

[1540] Example: A warning message appears saying, "This statement contradicts a previous statement. Would you say, 'I bought a new, eco-friendly car'?"

[1541] 4. Risk Assessment

[1542] Risk Assessment

[1543] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the result. The risk assessment can use a machine learning model using past data.

[1544] Example: If a post containing the keyword "car" has caused controversy in the past, a notification will be sent to any new posts saying, "This post is at high risk of causing controversy. Please reconsider."

[1545] 5. Final check and submission

[1546] Confirm or cancel your post

[1547] Based on the suggestions and warnings provided, users can choose to either correct their post or cancel it. The corrected post is then sent back to the server for a final check.

[1548] Example: If the user changes the post to "I bought a new, eco-friendly car" and confirms the post, the post is sent to the server again. If the user wants to cancel the post, they can press the cancel button to abort the post process.

[1549] Examples of prompt statements

[1550] As an example of user input, the following prompt sentence can be input to the generative AI model.

[1551] Example prompt:

[1552] "Link your Twitter and Instagram accounts to collect past posts. Use that data to evaluate new posts in real time for consistency and risk of outrage, and implement a system that provides warnings and suggestions to users as needed."

[1553] This allows users to maintain consistency in new posts and prevent potential flame wars.

[1554] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1555] Step 1:

[1556] A user logs in to the system and connects their social media account. At this time, they enter their authentication information (username, password, OAuth token, etc.). The server receives this authentication information and, once authenticated, stores it internally in the system.

[1557] Input: The credentials entered by the user into the system.

[1558] Output: Authenticated user information and linked SNS account information are saved in the system.

[1559] Specific operation: The user enters their username and password on the login screen, clicks the login button, and then clicks the SNS link button to complete OAuth authentication on the authentication screen of each SNS.

[1560] Step 2:

[1561] The server uses the API of each SNS to collect past posting data from the SNSs that the user has linked. At this time, the server sends authentication information to the API of each SNS and stores the posting data in a centralized database.

[1562] Input: API key, user credentials.

[1563] Output: Past posting data is saved in the database.

[1564] Specific operation: The server periodically calls the API of each SNS, retrieves past posting data (for example, tweets and Instagram posts from the past year), and stores it in a database.

[1565] Step 3:

[1566] The server pre-processes the collected historical posting data, which includes tokenization, morphological analysis, and stop word removal.

[1567] Input: Past posting data stored in the database.

[1568] Output: Preprocessed submission data is stored in a database for analysis.

[1569] What it does: The server uses Python's NLTK library to tokenize the text "I bought a new car!" into "new," "car," and "buy." It also removes stop words (such as "!").

[1570] Step 4:

[1571] The server uses the preprocessed data to extract key topics using topic modeling techniques such as Latent Dirichlet Allocation (LDA).

[1572] Input: Preprocessed submission data.

[1573] Output: The extracted topic information is stored in the topic model database.

[1574] What it does: The server uses the scikit-learn library to apply an LDA model to extract topics such as "environment," "eco," and "climate change."

[1575] Step 5:

[1576] When a user posts a new post, the device sends the post content to the server in real time. The device sends an asynchronous request when the user clicks the post button.

[1577] Input: The user's new post.

[1578] Output: The post is sent to the server and saved.

[1579] Specific action: A user types "I bought a new car!" into the Instagram posting screen and presses the post button.

[1580] Step 6:

[1581] The server preprocesses the content of new posts and compares them with the topic model to assess consistency, ensuring that the topic of the new post matches previous data.

[1582] Input: New post content.

[1583] Output: Topic comparison results. Coherence evaluation results are generated.

[1584] What happens: The server preprocesses the text "I bought a new car!" and determines that it is inconsistent with past posts such as "I won't buy a car because it's bad for the environment."

[1585] Step 7:

[1586] If the server determines that the consistency is low, it will send a warning message to the user's device and suggest the most appropriate expression.

[1587] Input: Consistency assessment results.

[1588] Output: A warning message and a suggestion are displayed on the user's terminal.

[1589] Specific operation: The server sends a warning message to the user's terminal saying, "This statement contradicts a previous statement. Would you like to say, 'I bought a new, environmentally friendly car'?"

[1590] Step 8:

[1591] The server calculates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the results.

[1592] Input: New post content, past flame war data.

[1593] Output: The result of the flame risk assessment is notified to the user.

[1594] Specific operation: The server uses a machine learning model to send a notification saying, "This post has a high risk of causing a backlash. Please reconsider," based on past examples of controversy surrounding the keyword "car."

[1595] Step 9:

[1596] Based on the suggestions and warnings provided, users can choose to modify or cancel their post, after which the modified post is sent back to the server for a final review.

[1597] Input: User's revised submission or cancellation information.

[1598] Output: The revised submission is confirmed or the submission is cancelled.

[1599] Specific actions: The user corrects the post to "I bought a new, eco-friendly car" and confirms the post, or presses the cancel button without pressing the post button.

[1600] Through these steps, the system can maintain consistency in new posts by users and prevent potential flame wars.

[1601] (Application example 1)

[1602] 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."

[1603] Conventional SNS posting systems lack a means to evaluate whether the content of a user's new post is consistent with past posts, making it difficult to prevent the risk of a flame war. Furthermore, there is a lack of a systematic method to determine whether a new post poses a flame war risk. This can lead to users unintentionally causing a flame war and damaging the credibility of individuals and companies.

[1604] 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.

[1605] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk value of the new post based on past risk cases and notifying the user of the risk value, and means for inputting prompt sentences into the generative AI model to assist in the consistency evaluation. This makes it possible to maintain consistency in the content newly posted by users and prevent potential flame wars.

[1606] "User authentication information" refers to the identifiers and authentication data required for a user to access the system, including, for example, a username, password, OAuth token, etc.

[1607] "Integrated Social Networking Service" refers to a social networking platform that allows users to connect their social networking account to the system and enable data collection.

[1608] "Past posting data" refers to all text, images, videos, and related metadata that a user has posted to the linked SNS platform.

[1609] "Preprocessing" refers to a series of data cleansing operations performed to convert collected data into an analyzable format, including tokenization, morphological analysis, and stop word removal.

[1610] "Topic modeling" is a statistical method for extracting major topics or themes from collected text data, using algorithms such as Latent Dirichlet Allocation (LDA).

[1611] "Acquisition in real time" means that newly posted data is sent to the system the moment a user attempts to post it, and analysis begins immediately.

[1612] "Evaluating consistency" means determining whether the subject matter and expressions of newly posted content are consistent with topics extracted from past posting data.

[1613] A "warning" is a message that is displayed to the user when a contradiction is detected as a result of the consistency evaluation, and provides specific instructions and suggestions for improvement.

[1614] "Evaluating the risk value" means quantitatively calculating the risk factors contained in a new post based on past risk cases and notifying the user of the results.

[1615] A "generative AI model" is an artificial intelligence model for natural language generation and text analysis, and refers to advanced models such as GPT.

[1616] A "prompt" is an instruction or question input to a generative AI model, which serves as a trigger for the model to generate an appropriate response or analysis result.

[1617] The present invention is a system that collects data on past posts from social networking services (SNS) that users connect to, and performs consistency and risk assessment on new posts. The system includes the following components:

[1618] First, the user's authentication information is obtained. The user connects their social media accounts (e.g., multiple social media platforms) to the system. Specifically, authentication information such as username, password, and OAuth token are stored on the server. The server then collects the user's past posting data from the connected social media platforms via API. This data is stored in a centralized database and used for later analysis.

[1619] The collected data is preprocessed on the server. This preprocessing includes data cleansing operations, such as removing URLs, mentions, and hashtags, and merging consecutive spaces. This preprocessed data is then subjected to topic modeling. The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. This makes it possible to clarify trends in users' interests and concerns.

[1620] Next, newly posted data is sent to the server in real time. When a user attempts to create a new post, the content of the post is immediately sent to the server. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. If the result of this consistency evaluation indicates low consistency, a warning is displayed on the user's device. For example, a warning message such as "This post contradicts previous statements" is displayed. Specific alternatives are also provided, such as "I bought a new, environmentally friendly car."

[1621] Additionally, the server can input prompts to the generative AI model to assist in consistency assessment, such as "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If not, please tell me why."

[1622] This makes it possible to maintain consistency in the content of new posts by users and prevent potential risks of online outrage. For example, if a user tries to create a post saying "I bought a new car!", the system will detect that the content of that post is inconsistent with a previous post saying "I won't buy a car because it's bad for the environment," and generate a warning.

[1623] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1624] Step 1:

[1625] User authentication and SNS integration

[1626] The user logs in to the system and connects various SNS accounts. This causes the user's authentication information (username, password, OAuth token, etc.) to be stored on the server. The input is the user's authentication information and SNS account information, and the output is the authentication information being saved. Specifically, the authentication information provided by the user is used to authenticate via the SNS API and establish the connection.

[1627] Step 2:

[1628] Social media data collection

[1629] The server collects users' past posting data from linked SNSs via the SNS API. This data is stored in a centralized database. The input is authenticated SNS account information, and the output is the collected past posting data. Specifically, the server sends an API request and stores the posting data returned from the SNS in the database.

[1630] Step 3:

[1631] Data Preprocessing

[1632] The server preprocesses the collected past posting data. This includes removing URLs, mentions, and hashtags, and merging consecutive spaces. The input is the collected posting data, and the output is preprocessed text data. Specifically, the data is cleansed using regular expressions and text analysis libraries.

[1633] Step 4:

[1634] Topic Modeling

[1635] The server uses Latent Dirichlet Allocation (LDA) to extract major topics from users' past posting data. The input is preprocessed text data, and the output is a topic distribution. Specifically, the preprocessed data is vectorized using CountVectorizer, and then the topics are extracted by training an LDA model.

[1636] Step 5:

[1637] Submitting and analyzing new posts

[1638] When a user attempts to create a new post, its contents are sent to the server in real time. The server preprocesses the text of the new post and compares it with the topic model to evaluate its consistency. The input is the text of the new post, and the output is the result of the consistency evaluation. Specifically, the server preprocesses the new post to calculate a topic distribution and compares it with past topic distributions.

[1639] Step 6:

[1640] Displaying warnings and suggestions

[1641] If the server determines that the consistency is low, it displays a warning message on the user's device and offers alternatives. The input is the consistency evaluation result, and the output is a warning message and alternatives. Specifically, a warning message pops up on the user's device and suggests expressions such as "I bought a new, environmentally friendly car."

[1642] Step 7:

[1643] Risk Assessment and Notification

[1644] The server calculates the risk value of a new post based on past risk case data and notifies the user of the result. The input is the text of the new post and past risk case data, and the output is a risk value and a notification message. Specifically, the server uses the new post as input and inputs a prompt sentence into the risk assessment model, generating a notification such as "This post is high risk. Please reconsider."

[1645] Step 8:

[1646] Using generative AI models

[1647] The server inputs a prompt sentence into the generative AI model to assist in consistency evaluation. The input is the prompt sentence, and the output is the analysis result of the generative AI model. A specific prompt sentence is used: "Extract the topic distribution from my past Twitter posts and check whether my new post 'I bought a new car!' matches a specific topic. If it does not match, please tell me the reason."

[1648] 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.

[1649] MODE FOR CARRYING OUT THE INVENTION

[1650] This invention is a system that precisely evaluates the consistency and risk of flaming of new posts on SNS by combining an emotion engine that recognizes user emotions. This system has the following components.

[1651] (1. Data collection and integration)

[1652] User authentication and SNS integration

[1653] Users log in to the system and link their social networking service (SNS) accounts, which allows the system to store their authentication information.

[1654] Social Media Data Extraction

[1655] The server collects users' past posting data from linked SNSs via the SNS API and stores it in a database.

[1656] Example: When a user connects Twitter and Instagram, the server retrieves past tweets and Instagram posts from these platforms and stores them centrally.

[1657] (2. Speech analysis and consistency check)

[1658] Data Preprocessing

[1659] The server pre-processes the collected historical posting data, which includes text tokenization, morphological analysis, and stop word removal.

[1660] Example: The text "I bought a new car!" is tokenized and split into the words "new," "car," and "buy."

[1661] Topic Modeling

[1662] The server uses topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract major topics from users' past posts.

[1663] Example: If a user's past posts are about "environmental issues," the topic model might include "environment," "eco," "climate change," etc.

[1664] (3. Emotion Engine Integration)

[1665] Emotion recognition

[1666] The server uses an emotion engine to recognize emotions from users' past posting data and store them in a database.

[1667] Example: A past post such as "I'm so happy today!" is perceived as a positive emotion, while a post such as "I'm sad" is perceived as a negative emotion.

[1668] Emotional tone analysis of new posts

[1669] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[1670] Example: When a user posts on Instagram, "I bought a new car!", that positive emotional tone is also sent to the server.

[1671] (4. Real-time warning)

[1672] Consistency Assessment

[1673] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[1674] Example: Detecting a mismatch in positive emotional tone between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment."

[1675] Displaying warnings and suggestions

[1676] If the server determines that the consistency is low, it will display a warning and suggest the most appropriate expression on the user's device.

[1677] Example: "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[1678] (5. Risk Assessment)

[1679] Flaming risk assessment

[1680] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[1681] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[1682] (6. Final confirmation and submission)

[1683] Confirm or cancel your post

[1684] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[1685] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[1686] Through these steps, the present invention can maintain consistency between a user's past and present comments and emotions, and prevent potential risks of flame wars.

[1687] The processing flow will be explained below.

[1688] Step 1:

[1689] Users log in to the system and link their social networking service (SNS) accounts. This linking is done using an authentication protocol such as OAuth, and the authentication information is stored in the system.

[1690] Step 2:

[1691] The server uses the user's authentication information to collect past posting data via the SNS API, including the content, date and time of the post, and metadata.

[1692] Step 3:

[1693] The server stores the collected data in a database for later analysis.

[1694] Step 4:

[1695] The server pre-processes the historical submission data stored in the database, including text tokenization, morphological analysis, and stop word removal.

[1696] Example: "I bought a new car!" is tokenized as "new," "car," and "buy."

[1697] Step 5:

[1698] The server uses the preprocessed data to run topic modeling techniques such as Latent Dirichlet Allocation (LDA) to extract key topics.

[1699] Example: If a user's past posts are about "environmental issues," the topics might be "environment," "eco," "climate change," etc.

[1700] Step 6:

[1701] The server uses an emotion engine to recognize emotions from users' past posts, using text analysis to classify emotional states as positive, negative, etc.

[1702] Example: "I'm so happy today!" is perceived as positive, while "I'm sad" is perceived as negative.

[1703] Step 7:

[1704] The server stores the emotion recognition results in a database.

[1705] Step 8:

[1706] When a user attempts to post a new message, the user's device sends the message's content and emotional tone to the server in real time.

[1707] Example: When a user posts "I bought a new car!" on Instagram, the content and positive emotional tone are sent to the server.

[1708] Step 9:

[1709] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[1710] Example: Detect a contradiction between a new post "I bought a new car!" and a previous post "I won't buy a car because it's bad for the environment" in terms of emotional tone.

[1711] Step 10:

[1712] If the server determines that the consistency is low, it displays a warning message on the user's device and suggests the most appropriate expression.

[1713] Example: Display a warning that reads, "This statement contradicts a previous statement. Would you say, 'I bought a new, green car'?"

[1714] Step 11:

[1715] The server calculates the risk of a new post becoming a hot topic based on past examples of such hot topics, and notifies the user of the results.

[1716] Example: If a post containing the word "car" has caused controversy in the past, the system will determine that the post is high risk and send a notification saying, "This post has a high risk of causing controversy. Please reconsider."

[1717] Step 12:

[1718] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post.

[1719] Example: A user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[1720] Step 13:

[1721] The user's device sends the final post content to the SNS, completing the post.

[1722] Example 2

[1723] 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."

[1724] In modern social networking services (SNS), when users post a new message, it is difficult to judge whether the message is consistent with past comments or whether it poses a risk of causing a backlash. This can lead to misunderstandings due to inconsistencies with the user's past posts, or inappropriate comments that can cause a backlash, which can have a negative impact on the user.

[1725] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for acquiring user authentication information and collecting past post data from multiple linked social networking services; means for preprocessing the collected past post data and performing topic modeling to extract major topics; means for recognizing emotions from the collected past post data and saving the emotional data; means for transmitting new posts by users to the server in real time and acquiring the content and emotional tone of the new post; means for preprocessing the newly posted data and comparing the topic and emotional tone of the new post with past data to evaluate consistency; means for displaying a warning on the user's terminal and suggesting optimal expressions if consistency is low; means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value; and means for the user to modify or cancel the content and emotional tone of the new post based on feedback from the server. This makes it possible to prevent the risk of a new post causing an uproar while maintaining consistency with the user's past comments.

[1726] "User authentication information" refers to information such as an ID and password required for a user to log in to a system.

[1727] "Social networking service" is a general term for services that provide a platform for users to interact with other users online.

[1728] "Posted data" refers to information such as messages, images, and videos posted by users on social networking services.

[1729] "Preprocessing" refers to processes such as tokenization, morphological analysis, and removal of stop words that are performed to make the collected data easier to analyze.

[1730] "Topic modeling" is a statistical method for extracting latent topics (themes) from large amounts of text data.

[1731] "Emotion recognition" is a technology that identifies the emotions (positive, negative, etc.) contained in text from text data.

[1732] "Emotion data" is information about the emotions indicated by the text data, obtained through emotion recognition.

[1733] "Real-time" refers to data and information being processed immediately.

[1734] "Emotional tone" refers to the emotional atmosphere or nuance of a particular text or message.

[1735] "Consistency assessment" is the process of assessing whether the content of a new post is consistent with previous posts and whether the emotional tone is consistent.

[1736] "Warning display" refers to displaying a message to alert the user.

[1737] "Expression suggestions" refers to presenting appropriate expressions for a user's new posts.

[1738] "Flame risk" is the likelihood that a particular post will provoke a large number of negative reactions.

[1739] "Feedback" refers to information provided to users about analysis and evaluation results from the system.

[1740] MODE FOR CARRYING OUT THE INVENTION

[1741] The present invention provides a system that recognizes user emotions and more precisely evaluates the consistency and risk of a new post on a social networking site. This system is constructed through the operation of a server, a terminal, and a user. Specific embodiments of the system are described in detail below.

[1742] 1. User authentication and SNS integration

[1743] Users log in to the system by entering their ID and password, and then link their various social media accounts (e.g., Twitter, Instagram) to the system, which then securely stores their authentication information on the server.

[1744] 2. Collection and storage of social media data

[1745] The server uses APIs such as Twitter API and Instagram API to collect past posting data from linked social media accounts. This data is stored in a database and used for subsequent processing.

[1746] 3. Data Preprocessing

[1747] The server preprocesses the collected data, which includes text tokenization, morphological analysis, stopword removal, etc. For example, a post saying "I bought a new car!" is broken down into "new," "car," and "buy."

[1748] 4. Topic Modeling

[1749] The server uses a Latent Dirichlet Allocation (LDA) model to extract key topics from the preprocessed data, for example, "environment," "eco," and "climate change" from posts about "environmental issues."

[1750] 5. Recognizing and storing emotions

[1751] The server analyzes the collected post data using an emotion engine (e.g., VADER), recognizes the emotion (e.g., positive or negative) of each post, and stores it in a database. For example, a post saying, "I'm really happy today!" is recognized as a positive emotion.

[1752] 6. Emotional Tone Analysis of New Posts

[1753] When a user posts a new message, the user's device sends the message's text and emotional tone to the server in real time. The server preprocesses the text and emotional tone of the new message and temporarily stores them in a database.

[1754] 7. Consistency Assessment

[1755] The server compares the text and emotional tone of new posts with past data to assess consistency. If the content of a new post contradicts previous posts, for example, "I bought a new car!", it will determine that this contradicts the previous post, "I won't buy a car because it's bad for the environment."

[1756] 8. Displaying warnings and suggestions

[1757] If the server determines that the statement is inconsistent, it displays a warning message on the user's device and suggests the most appropriate expression. For example, it displays a warning such as, "This statement contradicts previous statements and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[1758] 9. Risk Assessment

[1759] The server calculates the risk value of new posts based on past data on flaming cases and notifies the user of the result. For example, if a new post containing "car" is judged to have a high risk of flaming, the server sends a notification saying, "This post is high risk. Please reconsider."

[1760] 10. Confirming or cancelling a submission

[1761] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether. The user can modify the content and emotional tone to "I bought a new, eco-friendly car," and finally confirm the post after confirming the emotional tone.

[1762] Examples and prompts

[1763] As a specific example of operation, a user links their Twitter and Instagram accounts to the system and collects past posting data from these accounts. When a user posts a new message such as "I bought a new car!", if that message contradicts a previous message such as "I won't buy a car because it's bad for the environment," the system displays a warning saying, "This statement contradicts your previous message and has a different emotional tone. Would you like to express it as 'I bought a new, environmentally friendly car'?"

[1764] Examples of input prompts for a generative AI model might include the following:

[1765] "Design a system that analyzes past posts on Twitter and Instagram and evaluates the consistency and sentiment of new posts by users."

[1766] "Create a system that incorporates an emotion engine that recognizes user emotions and evaluates the risk of social media posts causing controversy."

[1767] These techniques make it possible to maintain consistency between a user's past and present comments and emotions, and prevent potential risks of online outrage.

[1768] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1769] Program processing flow

[1770] Step 1:

[1771] Users log in to the system by entering their ID and password.

[1772] Input: User ID, Password

[1773] Output: Feedback of authentication success or failure

[1774] Specific operation: The user enters information into the login form and presses the "Login" button. The system authenticates the user ID and password, and if successful, displays the dashboard.

[1775] Step 2:

[1776] Users use the system's interface to link their Twitter and Instagram accounts to the system.

[1777] Input: User's social media account

[1778] Output: Notification of successful SNS account linking

[1779] Specific operation: The user clicks the link button and completes authentication on the SNS authentication screen that appears. The authentication token is saved on the server.

[1780] Step 3:

[1781] The server uses the Twitter API and Instagram API to collect users' past posting data and store it in a database.

[1782] Input: SNS account authentication token

[1783] Output: Collected past posting data

[1784] Specific operation: The server calls the Twitter API to retrieve tweets within the specified period, and similarly calls the Instagram API to retrieve past post data.

[1785] Step 4:

[1786] The server pre-processes the collected data, which includes text tokenization, morphological analysis, and stop-word removal.

[1787] Input: Collected submission data

[1788] Output: Preprocessed data (tokenized, morphologically analyzed, stopwords removed)

[1789] Specific behavior: Breaks down a post like "I bought a new car!" into "new," "car," and "buy," and removes meaningless words.

[1790] Step 5:

[1791] The server uses a Latent Dirichlet Allocation (LDA) model to extract major topics from the preprocessed data.

[1792] Input: Preprocessed data

[1793] Output: Extracted topics (e.g. "environment", "eco", "climate change")

[1794] What it does: Runs an LDA model to identify key topics from a user's past posts.

[1795] Step 6:

[1796] The server analyzes the collected posting data using an emotion engine, recognizes the emotion of each post, and stores it in a database.

[1797] Input: Collected submission data

[1798] Output: Recognized emotion data (positive, negative, etc.)

[1799] Specific behavior: Using an emotion engine such as VADER, we recognize "I'm so happy today!" as a positive emotion and "I'm sad" as a negative emotion.

[1800] Step 7:

[1801] When a user posts a new message, the message content and emotional tone are sent to the server in real time.

[1802] Input: New post text and emotional tone

[1803] Output: New post data and emotional tone sent to the server

[1804] What happens: A user posts on Instagram, "I bought a new car!" and sends that positive emotional tone to the server.

[1805] Step 8:

[1806] The server preprocesses the text and emotional tone of new posts and compares them with historical data to assess consistency.

[1807] Input: New post text and emotional tone

[1808] Output: Result of consistency assessment (e.g., consistent, inconsistent)

[1809] Specific behavior: Determine whether the new post "I bought a new car!" contradicts the previous post "I won't buy a car because it's bad for the environment."

[1810] Step 9:

[1811] If the server determines that the consistency is low, it displays a warning message and suggests the most appropriate expression on the user's device.

[1812] Input: Results of consistency assessment

[1813] Output: Warning message and suggested expression

[1814] What to do: Display a warning that says, "This statement contradicts previous statements and has a different emotional tone. Would you say, 'I bought a new, eco-friendly car'?"

[1815] Step 10:

[1816] The server calculates the risk value of new posts based on data on past flame war cases and notifies the user of the results.

[1817] Input: Results of consistency assessment and content of new submission

[1818] Output: Flame risk assessment notification

[1819] What it does: If a new post containing the word "car" has previously sparked outrage, we'll send a notification saying, "This post is high risk. Please reconsider."

[1820] Step 11:

[1821] Based on the feedback from the server, the user can choose to modify the content and emotional tone of the new post, or cancel the post altogether.

[1822] Input: Feedback

[1823] Output: The revised post or the cancellation of the post

[1824] What happens: The user amends the comment to "I bought a new, eco-friendly car," checks the emotional tone, and then confirms the post.

[1825] (Application example 2)

[1826] 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."

[1827] In recent years, as social networking services (SNS) have become more widespread, users are increasingly facing the risk of their posts sparking outrage. In particular, if a user's emotions are not consistent with their previous posts, it is easy for misunderstandings and resentment to arise, resulting in a higher risk of outrage. Furthermore, as the number of posts on content distribution services increases, there is a demand for systems that can appropriately evaluate emotions and consistency. However, existing systems lack sufficient emotional analysis and outrage risk assessment, making it difficult for users to receive appropriate feedback and warnings. Providing an effective system to improve this situation is a challenge.

[1828] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1829] In this invention, the server includes means for acquiring user authentication information and collecting past posting data from multiple linked social networking services, means for preprocessing the collected past posting data and performing topic modeling to extract major topics, means for acquiring newly posted data in real time and comparing the topics of the new posts with past data to evaluate consistency, means for displaying a warning to the user and suggesting optimal wording if consistency is low, means for evaluating the risk of a new post causing an uproar based on past examples of uproar and notifying the user of the risk value, means for analyzing the emotional tone of the new post in real time using sentiment analysis, and means for the user to confirm, modify, or cancel the content and emotional tone of the post in the content distribution service. This makes it possible to maintain consistency between the user's past and present comments and emotions, prevent potential uproars, and provide appropriate feedback in real time.

[1830] "User Credentials" means the unique identifying information required for a user to access a system.

[1831] A "social networking service" is an online platform that enables users to interact with and share content with other users over the Internet.

[1832] "Previously posted data" refers to digital content such as text, images, and videos that a user previously posted on a social networking service.

[1833] "Preprocessing" refers to a series of processes, such as text normalization, tokenization, morphological analysis, and stop word removal, that are carried out to convert collected data into a format that is easier to analyze.

[1834] "Topic modeling" is a statistical method for automatically extracting themes and topics from large amounts of text data.

[1835] "Latent Dirichlet Allocation (LDA)" is a machine learning technique for understanding the topic structure hidden in documents, and models topics based on the probability of occurrence of each word.

[1836] "Acquisition in real time" means that the data is collected and processed immediately the moment a new post is made by a user.

[1837] "Emotion analysis" is a technology that recognizes a user's emotional state (positive, negative, neutral, etc.) from text data.

[1838] "Consistency assessment" is the process of comparing and determining whether a user's new posts are consistent with previous posts in terms of content and emotional tone.

[1839] "Display warning" means that when low consistency is detected, the problem and its solution are presented to the user.

[1840] "Optimal expression suggestions" means providing specific expressions to modify a user's new posts to ensure consistency with past posts and appropriately reflect the emotional tone.

[1841] "Flame risk assessment" is the process of predicting the likelihood that a new post will provoke criticism or resentment from other users if it contains inappropriate content or sentiment.

[1842] "Notifying risk value" means quantifying the possibility of a new post causing a backlash and informing the user.

[1843] A "content distribution service" is an internet-based service that allows users to create, share, and view digital content such as text, images, and videos.

[1844] "Means for checking, modifying, and canceling post content and emotional tone" refers to a function that allows users to check the content of new posts and the accompanying emotional tone in real time on the user interface, and to modify or cancel posts as necessary.

[1845] MODE FOR CARRYING OUT THE INVENTION

[1846] System configuration and program overview

[1847] The system acquires user authentication information, collects and analyzes past posting data from multiple linked social networking services (SNS), evaluates the emotional tone and consistency of new posts in real time, and notifies users of the risk of a controversy.The main software and hardware used include Django, Tweepy, Instagram Graph API, NLTK, Gensim, VaderSentiment, Scikit-learn, React Native, and PostgreSQL.

[1848] Data collection and integration

[1849] When a user logs in to the system and connects to a social media account, the server obtains their authentication information. This information is used to retrieve past posting data from the connected social media platform. Specifically, Twitter posting data is collected using Tweepy, and Instagram posting data is collected using the Instagram Graph API. The collected data is stored in a PostgreSQL database.

[1850] Data Preprocessing and Topic Modeling

[1851] The server converts the collected raw data into a format that is easy to analyze. This includes text tokenization, morphological analysis, and stop word removal. Gensim then performs topic modeling using Latent Dirichlet Allocation (LDA) on the preprocessed data to extract major topics. For example, a post saying "I bought a new car!" is broken down into words such as "new," "car," and "buy," and the major topics extracted are "car" and "shopping."

[1852] Sentiment analysis and real-time evaluation

[1853] When a new post is created on a user's device, its text and emotional tone are sent to the server in real time. VaderSentiment is used to analyze the emotional tone of the new post and the results are stored in a database. This makes it possible to evaluate the consistency between past posts and new posts. For example, if a user tries to post "I bought a new car!", the positive emotion is analyzed.

[1854] Displaying warnings and suggestions

[1855] The server compares new posts with past data to evaluate their consistency, and if the consistency is low, it displays a real-time warning on the user's device. It also suggests optimal expressions. For example, if a new post, "I bought a new car!", contradicts a past post, "Cars are bad for the environment," it suggests expressions like, "I bought a new, environmentally friendly car."

[1856] Flaming risk assessment and notification

[1857] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user of the risk value. Scikit-learn is used to compare past posts with new posts and calculate the risk value. For example, if a post containing the word "car" has become a hot topic in the past, a notification will be sent to the user stating that "this post is high risk."

[1858] Final check and submission

[1859] Based on the feedback from this system, users can choose to modify their new post or cancel the post altogether. This confirmation, modification, and cancellation functionality is provided on the user interface using React Native. This allows users to receive appropriate feedback and post safely while avoiding the risk of a social media outcry.

[1860] Specific examples

[1861] Examples:

[1862] User post: "The environment in urban areas has been deteriorating recently. New narrow roads are to blame."

[1863] Warning / Suggestion: This statement contradicts a previous post that said, "Urban development is a good thing." Would you say, "I've noticed an increase in narrow roads in urban areas recently"?

[1864] Prompt statement:

[1865] markdown

[1866] A user is attempting to post content about environmental issues. The post states, "The environment in urban areas has been deteriorating recently. New, narrow roads are to blame." Please assess the consistency and controversy risk of this post and previous posts and provide appropriate warnings and suggestions.

[1867] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1868] Step 1:

[1869] A user logs in to the system and links their social media account. In this step, the user enters their authentication information, and the server establishes a link with the social media account through the API. Specifically, the user provides their Twitter and Instagram authentication information via the API, and the server receives it and begins collecting past posting data from each platform. The input is the user's authentication information, and the output is the linked social media account information.

[1870] Step 2:

[1871] The server collects past posting data from linked SNSs and stores it in a database. The collected data includes content such as text, images, and videos. Specifically, it uses Tweepy to collect Twitter tweet data and Instagram posts using the Instagram Graph API. The input is post data obtained through the SNS API, and the output is past posting data stored in the database.

[1872] Step 3:

[1873] The server preprocesses the collected post data. In this step, processes such as text tokenization, morphological analysis, and stop word removal are performed. For example, a post "I bought a new car!" is split into words such as "new," "car," and "bought." Specifically, this preprocessing is performed using the NLTK library. The input is the post data stored in the database, and the output is the preprocessed text data.

[1874] Step 4:

[1875] The server performs topic modeling using the preprocessed data. In this step, Gensim is used to extract major topics using Latent Dirichlet Allocation (LDA). For example, from posts related to "environmental issues," topics such as "climate change," "recycling," and "energy" are extracted. The input is the preprocessed text data, and the output is the extracted topic information.

[1876] Step 5:

[1877] Newly posted data is acquired in real time from the user's device and sentiment analysis is performed. When a new post is entered by a user, its text and emotional tone are sent to the server in real time, and the emotional tone is analyzed using VaderSentiment. For example, the sentence "I'm so happy today!" is analyzed as a positive sentiment. The input is the user's new post data, and the output is the analyzed emotional tone.

[1878] Step 6:

[1879] The server compares the topic and emotional tone of new posts acquired in real time with past post data to evaluate consistency. Specifically, it compares the topic information and emotional tone of new posts with the past database and calculates the degree of match. The input is the topic information and emotional tone of new posts, and the output is the consistency evaluation result.

[1880] Step 7:

[1881] If the consistency is low, the server will warn the user and suggest the most appropriate expression. For example, if a user posts "Cars are bad for the environment" while also posting "I bought a new car!", the server will display a warning saying, "This statement contradicts previous statements. Would you like to express it as 'I bought a new, environmentally friendly car'?" The input is the consistency evaluation result and the topic information of the new post, and the output is a warning and suggestion to be displayed to the user.

[1882] Step 8:

[1883] The server evaluates the risk of a new post becoming a hot topic based on past data on hot topics and notifies the user's device of the risk value. Specifically, it uses Scikit-learn to compare past hot topics with the new post and calculate the risk value. The input is the topic information of the new post and past hot topics, and the output is the hot topic risk value.

[1884] Step 9:

[1885] Based on the feedback from the server, the user can modify the content and emotional tone of the new post, or cancel the post altogether. Specifically, using React Native, the user can confirm, modify, or cancel the post from the user interface. The input is warnings and suggestions from the server, and the output is the modified post content, the cancellation of the post, or the post content as is.

[1886] 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.

[1887] 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.

[1888] 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.

[1889] 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.

[1890] FIG. 9 illustrates an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and behaviors arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1891] 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.

[1892] 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).

[1893] 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.

[1894] 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."

[1895] 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.

[1896] 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).

[1897] 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.

[1898] 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.

[1899] 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.

[1900] 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.

[1901] 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 design...

Claims

1. A means for acquiring user authentication information and collecting past posting data from multiple linked social networking services; A means for preprocessing the collected past posting data and performing topic modeling to extract key topics; A means of capturing newly posted data in real time and comparing the topics of new posts with past data to assess consistency; If consistency is low, a warning is displayed to the user and a means of suggesting the best possible representation is provided. A method to evaluate the risk of a new post causing an uproar based on past examples of uproars and notify users of the risk value; A system including:

2. 10. The system of claim 1, further comprising a database system for storing, preprocessing and analyzing the collected data.

3. The system of claim 1 , wherein the topic modeling uses Latent Dirichlet Allocation.

Images