System

Abstract

To provide a system.SOLUTION: A system comprises: means of recognizing an emotion of a user by using an emotion engine; means of generating a submission text on the basis of a prompt text in accordance with the emotion of the user by using a generative AI with an additional learning on the basis of a submission extracted by a predetermined condition containing at least any one of a specific keyword, a specific hash tag, and the number of reactions; and means of outputting the generated submission text.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] There is a demand for automatically generating posts that are rated as "interesting" on social media and providing new services that utilize these posts. However, the technology for automatically generating posts that people find "interesting" has not yet been fully developed. [Means for solving the problem]

[0005] This invention crawls posts that are rated as interesting on social media and uses that data for additional training of a generation AI. After additional training, the generation AI can generate sentences that people find "interesting." This makes it possible to provide new services such as virtual influencer activities and a platform where social media users can create posts that will "go viral." [Brief explanation of the drawings]

[0006] [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. 2 is a sequence diagram showing a flow of processing in the data processing system according to the first embodiment of the first form example. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Embodiment 1. [Figure 13] FIG. 10 is a sequence diagram showing a processing flow of a data processing system in a second embodiment of the second form example. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 of Embodiment 2. [Figure 15] FIG. 10 is a sequence diagram showing the flow of processing in a data processing system according to a third embodiment of the third embodiment. [Figure 16] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 3 of Embodiment 3. [Figure 17] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the first embodiment of the first form example when an emotion engine is combined. [Figure 18] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Form Example 1 when an emotion engine is combined. [Figure 19] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the second embodiment of the second form example when an emotion engine is combined. [Figure 20] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 of Form Example 2 when an emotion engine is combined. [Figure 21] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the third embodiment of the third form example when an emotion engine is combined. [Figure 22] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 3 of Form Example 3 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0007] 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.

[0008] First, the terms used in the following description will be explained.

[0009] 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, the 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), an APU (Accelerated Processing Unit), or a TPU (TENSOR PROCESSING UNIT (registered trademark)).

[0010] 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.

[0011] 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.

[0012] 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.

[0013] 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."

[0014] [First embodiment]

[0015] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0016] 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.

[0017] 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).

[0018] 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.

[0019] 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.

[0020] 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.

[0021] 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.

[0022] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0023] 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.

[0024] 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.

[0025] 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.

[0026] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0027] "Example 1"

[0028] One embodiment of the present invention is a system that includes a means for crawling posts from social media that are rated as interesting. This system selects posts based on specific keywords, hashtags, the number of reactions, etc., and crawls their text. For example, it is conceivable to crawl posts with the hashtag "interesting" or posts with a certain number of "likes."

[0029] "Example 2"

[0030] Next, it is equipped with a means for additional training of the generation AI based on the crawled posts. This AI uses the text of the crawled posts as training data to learn how to generate sentences that will be evaluated as interesting. Specifically, the text of the crawled posts is used as training data for additional training of the existing learning model.

[0031] "Example 3"

[0032] Furthermore, the generative AI, which has undergone additional learning, is equipped with a means to generate sentences that it deems interesting. This generative AI generates new sentences based on the results of its learning. For example, it can generate sentences such as "Something interesting happened today! That's funny" as posts from virtual influencers.

[0033] "Example 4"

[0034] It also provides a means for users to act as virtual influencers using generative AI, and a platform for social media users to create posts that are intended to go viral. Specifically, it publishes interesting sentences created by generative AI as posts by virtual influencers, and provides a platform for social media users to create their own posts using generative AI.

[0035] The processing flow of each embodiment will be described below.

[0036] "Example 1"

[0037] Step 1: The system selects posts from social media based on specific keywords, hashtags, number of reactions, etc. For example, it selects posts with the hashtag "funny" or posts with a certain number of "likes."

[0038] Step 2: The text of the selected posts is crawled. The crawled data is used as training data for the generation AI.

[0039] "Example 2"

[0040] Step 1: Additional training is performed on the generation AI based on the crawled posts. Specifically, additional training is performed on the existing learning model using the text of the crawled posts as training data.

[0041] Step 2: After additional learning, the generation AI generates new sentences based on the results of its learning.

[0042] Step 1: The interesting sentences created by the generative AI are published as posts by virtual influencers.

[0043] Step 2: Provide a platform for social media users to create their own posts using generative AI.

[0044] Example 1

[0045] Next, a description will be given of Example 1 of Form 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."

[0046] There is a demand for a system that can efficiently collect posts that are rated as interesting on social media and automatically generate more attractive content by training a generation AI based on that data. However, with conventional methods, the collection and selection of posts is often done manually, which takes time and effort. In addition, it is difficult to select appropriate posts to use as training data for the generation AI, which results in an issue of inconsistent quality of the generated content.

[0047] 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.

[0048] In this invention, the server includes means for inputting conditions such as specific keywords, hashtags, and the number of reactions, means for accessing the API of the SNS and searching for posts based on the conditions, means for selecting posts that match the conditions from the search results, means for saving the text of the selected posts in a database, means for providing the saved data to the user, means for additional training of the generation AI based on the crawled posts, and means for the additionally trained generation AI to generate sentences that are considered interesting. This makes it possible to efficiently collect posts that are rated as interesting on the SNS and use them as training data for the generation AI, thereby automatically generating more attractive content.

[0049] "Means for entering specific keywords, hashtags, and reaction counts" is a function that provides an interface for users to enter keywords, hashtags, and reaction counts that they specify as search conditions.

[0050] "A means of accessing an SNS's API and searching for posts based on conditions" refers to a function that uses the application programming interface (API) provided by the SNS to search for posts based on conditions specified by the user.

[0051] "Means for selecting posts that match conditions from search results" refers to a function that filters and selects posts that match conditions specified by the user from the search results obtained from the SNS API.

[0052] The "means for saving the text of the selected posts to a database" is a function for storing the text of the filtered posts in a database.

[0053] The "means for providing saved data to users" is a function that provides an interface for displaying posted information saved in the database to users.

[0054] "Means for additional training of the generative AI based on crawled posts" is a function for additional training of the generative AI model using collected post data.

[0055] "Means for a generative AI that has undergone additional training to generate sentences that are deemed interesting" is a function that enables a generative AI model that has undergone additional training to automatically generate sentences that are evaluated as interesting.

[0056] MODE FOR CARRYING OUT THE INVENTION

[0057] This invention is a system that efficiently collects posts that are rated as interesting from social media, and automatically generates more attractive content by training a generation AI based on that data. Specific embodiments of this system are described below.

[0058] 1. Program Generation

[0059] The system's programs are written using programming languages ​​such as Python® and JavaScript®, and include code to access SNS APIs and collect posts based on specific conditions.

[0060] 2. Program Processing

[0061] The server performs its processing using the following hardware and software:

[0062] Hardware: The server uses a computer system with a powerful processor and sufficient memory.

[0063] Software: The server uses a library for accessing the SNS API and a database management system (for example, MySQL (registered trademark)).

[0064] The server receives specific keywords, hashtags, and reaction count conditions entered by the user through their device and sends a request to the social networking service's API. For example, if a user specifies the hashtag "funny" and the conditions "more than 100 likes," the server searches for posts based on these conditions.

[0065] The server analyzes the search results returned by the API and selects posts that match the criteria. The selected post text is saved in a database, which stores information such as the post ID, text, hashtags, and number of reactions.

[0066] The server then further trains the generative AI based on the crawled posts. The generative AI model uses an advanced natural language processing model such as GPT-3 (registered trademark). After further training, the generative AI automatically generates sentences that it deems interesting.

[0067] 3. Examples of concrete examples and prompts

[0068] As a concrete example, consider the case where you want to use a social media API to crawl posts with the hashtag "interesting." Below is an example of a prompt for that purpose.

[0069] Examples:

[0070] The server uses the social networking site's API to search for posts with the hashtag "funny."

[0071] The server selects posts from the search results that have 100 or more likes.

[0072] The server stores the selected post text in a database.

[0073] Example prompt sentence:

[0074] Create a program that uses a social media API to search for posts with the hashtag "interesting," select posts with more than 100 likes, and save the text of those posts to a database.

[0075] In this way, the server can efficiently crawl posts that are rated as interesting from social media, store them in a database, and use them as learning data for the generative AI, which will enable the AI ​​to automatically generate more attractive content.

[0076] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0077] Program processing flow

[0078] Step 1:

[0079] Users enter conditions such as specific keywords, hashtags, and number of reactions.

[0080] The user inputs search criteria through the device interface, for example, specifying the hashtag "interesting" and the criteria "more than 100 likes."

[0081] Input: Keyword (e.g. "interesting"), reaction count condition (e.g. "100 or more likes")

[0082] Output: The search criteria is sent to the server.

[0083] Step 2:

[0084] The server accesses the SNS API and searches for posts based on the criteria.

[0085] The server accesses the SNS API based on the conditions entered by the user, sends an API request, and searches for posts with the hashtag "interesting."

[0086] Input: Search criteria (e.g. "interesting", reaction count criteria "100 or more likes")

[0087] Data processing: API request generation

[0088] Output: Search results returned from the SNS API

[0089] Step 3:

[0090] The server selects posts that match the search criteria from the search results.

[0091] The server receives the search results returned by the API and selects posts with more than 100 likes. The server extracts information such as the ID, text, and number of reactions from the selected posts.

[0092] Input: Search results from SNS API

[0093] Data processing: Filtering posts that match criteria

[0094] Output: A list of posts that match the criteria

[0095] Step 4:

[0096] The server saves the selected post text to the database.

[0097] The server stores the selected post text in a database, which stores information such as the post ID, text, hashtags, and number of reactions.

[0098] Input: A list of posts that match the criteria

[0099] Data processing: storing submitted information in a database

[0100] Output: Post information saved in the database

[0101] Step 5:

[0102] The server provides the stored data to the user.

[0103] The server provides the user with information about the posts stored in the database, and the user can view a list of the saved posts through the interface of the terminal.

[0104] Input: Post information stored in the database

[0105] Data processing: Providing posted information to users

[0106] Output: A list of posts that the user can see

[0107] Step 6:

[0108] The generation AI further learns based on the posts crawled by the server

[0109] The server uses the crawled post data to further train the generative AI model, which uses an advanced natural language processing model such as GPT-3.

[0110] Input: Post information stored in the database

[0111] Data processing: Additional training for generative AI models

[0112] Output: A trained generative AI model

[0113] Step 7:

[0114] The server performs additional training and the AI ​​generates sentences that it deems interesting.

[0115] The server uses a trained generative AI model to automatically generate sentences that it deems interesting.

[0116] Input: A trained generative AI model

[0117] Data processing: Sentence generation using generative AI

[0118] Output: Funny sentence generated

[0119] In this way, the server can efficiently crawl posts from social media that are rated as interesting, store them in a database, and use them as learning data for the generation AI, making it possible to automatically generate more attractive content.

[0120] (Application example 1)

[0121] Next, a description will be given of Application Example 1 of Embodiment 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."

[0122] Conventional crawling systems for social media posts have had difficulty efficiently collecting and providing users with "interesting" posts that interest them. Furthermore, they lacked the functionality to allow users to customize their feeds to suit their interests, creating a need for an improved user experience. Furthermore, when using generative AI to generate content, it was difficult to accurately generate the "interesting" content users were looking for.

[0123] 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.

[0124] In this invention, the server includes a means for crawling posts rated as interesting from social media, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that it deems interesting, a means for generating a feed for providing the generated sentences to users, and a setting means for users to customize the feed. This makes it possible to efficiently collect "interesting" posts that interest users and provide them to users. It also allows users to customize their feeds to suit their interests, improving the user experience. Furthermore, when generating content using a generation AI, it becomes possible to accurately generate the "interesting" content that users desire.

[0125] "SNS" is an abbreviation for Social Networking Service, an online platform that allows users to share information and interact with other users over the Internet.

[0126] "Crawling" is the process of automatically collecting information on the Internet, and is a technique for obtaining data based on specific keywords or conditions.

[0127] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new information and content based on given data.

[0128] "Additive training" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's accuracy and performance.

[0129] A "feed" is a stream or list of information provided to a user, in which selected content is displayed sequentially based on specific criteria.

[0130] "Configuration means" refers to the interface and functions that allow the user to customize the system's operation and display content.

[0131] A system for implementing this invention includes a means for crawling posts from social media that are rated as interesting, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are deemed interesting, a means for generating a feed to provide the generated sentences to users, and a setting means for users to customize the feed.

[0132] System Program

[0133] The server uses Python and social media APIs to crawl posts that are rated as interesting by social media. Specifically, it accesses the API using social media API credentials and crawls posts based on specific keywords and hashtags. The crawled posts are then filtered based on rating criteria such as the number of likes and retweets.

[0134] Next, the generation AI is trained based on the crawled posts. For the generation AI, an advanced natural language processing model such as GPT-3 is used. The training is carried out using the crawled post data to improve the accuracy and performance of the generation AI.

[0135] After additional learning, the generative AI generates sentences that it deems interesting. These generated sentences are stored on the server as feeds to be provided to users. The feeds are designed to efficiently provide content that users are interested in, and users can customize them to suit their interests.

[0136] Users can browse feeds using devices such as smartphones and PCs and customize their feeds through the settings means, which include an interface for changing the content displayed on the feed and the evaluation criteria.

[0137] Hardware and software used

[0138] Hardware: Servers, smartphones, PCs

[0139] Software: Python, social media APIs, generative AI models (e.g., GPT-3)

[0140] Specific examples

[0141] For example, if a user sets the hashtags "funny" and "hilarious" and crawls posts with more than 100 likes, the server will collect posts based on these conditions and further train the AI. The AI ​​will then generate sentences that it deems funny and display them in the user's feed.

[0142] Prompt Sentence Examples

[0143] Crawl for posts with the hashtags "funny" and "hilarious" and display posts with over 100 likes in your feed.

[0144] In this way, it is possible to efficiently collect and provide "interesting" posts that interest users. It also allows users to customize their feeds to suit their interests, improving the user experience. Furthermore, when generating content using generative AI, it is possible to accurately generate the "interesting" content that users are looking for.

[0145] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0146] Step 1:

[0147] The server crawls posts from social media sites that are rated as interesting. Specifically, it uses Python and social media APIs to collect posts based on specific keywords and hashtags. Evaluation criteria such as keywords, hashtags, and number of likes are given as input. The server searches for posts based on these conditions and outputs the collected post data.

[0148] Step 2:

[0149] The server performs additional training on the generative AI based on the crawled posts. Specifically, the collected post data is input into a generative AI model (e.g., GPT-3) to perform additional training on the model. The crawled post data is given as input, and the trained generative AI model is obtained as output. Through this process, the server improves the accuracy and performance of the generative AI.

[0150] Step 3:

[0151] The server uses a generative AI that has undergone additional training to generate sentences that it deems interesting. Specifically, it inputs a prompt sentence into the generative AI model that has undergone additional training, and outputs the generated sentence. The prompt sentence is given as input, and the generated sentence is obtained as output. The server uses this generated sentence in the next step.

[0152] Step 4:

[0153] The server generates a feed to provide the generated sentences to the user. Specifically, it converts the generated sentences into a feed format and sends it to the user's device. The generated sentences are given as input, and data in feed format is obtained as output. By providing this feed to the user, the server enables the user to efficiently browse content that interests them.

[0154] Step 5:

[0155] Users browse feeds using their devices and customize them through settings. Specifically, users view feeds using smartphones or computers and operate the settings interface to change the content and evaluation criteria of the feed. The user's customization settings are given as input, and a customized feed is obtained as output. Through this process, users can create a feed tailored to their interests.

[0156] Example 2

[0157] Next, a description will be given of Example 2 of Form 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."

[0158] Conventional generative AI models lack the training data to generate sentences that are rated as interesting, resulting in low-quality generated sentences. In addition, the evaluation methods for generated sentences are inadequate, making it difficult to accurately evaluate the performance of generative AI models.

[0159] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for crawling posts on the Internet, a means for cleansing the crawled posts, a means for additionally training the generative AI model using the cleansed text data, and a means for evaluating the additionally trained generative AI model. This enables the generative AI model to generate high-quality sentences that are evaluated as interesting.

[0160] "Means of crawling internet posts" refers to technologies and methods for automatically collecting post data from websites and social networking sites on the internet.

[0161] "Methods for cleansing crawled posts" are techniques and methods for removing unnecessary information from collected post data and normalizing text data.

[0162] "Means for additional training of generative AI models using cleansed text data" refers to techniques and methods for using preprocessed text data to add new knowledge to existing generative AI models and improve their performance.

[0163] "Means for evaluating a trained generative AI model" refers to techniques or methods for measuring the quality of generated sentences in order to evaluate the performance of a trained generative AI model.

[0164] "Means of acting as a virtual character" refers to the technology or method used to create a virtual character using a generative AI model and have that character act on the Internet.

[0165] "Means for providing a platform for internet users to create posts intended to attract attention" refers to technologies and methods that use generative AI models to provide an environment or service that enables internet users to create posts that attract the attention of many people.

[0166] MODE FOR CARRYING OUT THE INVENTION

[0167] The present invention relates to a system for crawling posts on the Internet and additionally training a generative AI model. Specific embodiments of this system are described below.

[0168] Data collection

[0169] The server uses a Python web scraping library to crawl posts on the Internet, collects post data from designated websites and social media sites, and stores it in text files or a database.

[0170] Examples:

[0171] The server retrieves the HTML using requests.get('https: / / example-blog.com') and parses it using web scraping (html.content, 'html.parser'). The post content is extracted from the parsed data and saved in a text file.

[0172] Data Preprocessing

[0173] The server cleanses the collected text data. Specifically, it removes unnecessary HTML tags, deletes special characters, normalizes the data, etc. The server preprocesses the text data using Python's regular expression library (re) and natural language processing libraries (NLTK, spaCy).

[0174] Examples:

[0175] The server removes HTML tags using re.sub('<[^<]+?>', '', text), tokenizes using nltk.word_tokenize(text), and normalizes the text using spacy.load('en_core_web_sm').

[0176] Data training

[0177] The server uses the cleansed text data to further train the generative AI model. The server uses a generative AI model such as GPT-3 or GPT-4 (registered trademark) from OpenAI (registered trademark) and trains the model using the Transformers library.

[0178] Examples:

[0179] The server loads the model and tokenizer using from transformers import GPT2LMHeadModel, GPT2Tokenizer, and starts additional training with model.train(). Use the cleansed text data as training data.

[0180] Evaluating generative AI models

[0181] The server checks the quality of the generated sentences using test data to evaluate the trained generative AI model. The server performs the evaluation using natural language generation evaluation metrics such as the BLEU score and ROUGE score.

[0182] Examples:

[0183] The server calculates the BLEU score by comparing the generated sentence with a reference sentence using from nltk.translate.bleu_score import sentence_bleu, e.g., sentence_bleu([reference], candidate).

[0184] Prompt Sentence Examples

[0185] Examples of prompts to be input to a generative AI model include:

[0186] "Tell us about your fun day."

[0187] "Please tell me a funny story."

[0188] "Tell me about something fun that happened recently."

[0189] Describe a fun time you had with friends.

[0190] In this way, the server further trains the generation AI based on the crawled posts, building a system that learns how to generate sentences that will be evaluated as interesting.

[0191] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0192] Step 1: Data collection

[0193] The server crawls posts on the Internet. As input, it receives the URL of a specified website or social media site. The server uses a Python web scraping library to collect post data from the specified URL. Specifically, it retrieves HTML using requests.get('https: / / example-blog.com') and parses it using web scraping (html.content, 'html.parser'). As output, the collected post data is saved in a text file or database.

[0194] Step 2: Data Preprocessing

[0195] The server cleanses the collected text data. As input, it receives the post data collected in step 1. The server performs tasks such as removing unnecessary HTML tags, deleting special characters, and normalizing the data. Specifically, it removes HTML tags using re.sub('<[^<]+?>', '', text) and tokenizes the text using nltk.word_tokenize(text). It then normalizes the text using spacy.load('en_core_web_sm'). The cleansed text data is obtained as output.

[0196] Step 3: Train the data

[0197] The server uses the cleansed text data to train the generative AI model. As input, it receives the cleansed text data from step 2. The server uses a generative AI model such as OpenAI's GPT-3 or GPT-4, and trains the model using the Transformers library. Specifically, it loads the model and tokenizer using from transformers import GPT2LMHeadModel, GPT2Tokenizer, and starts the additional training with model.train(). The output is the trained generative AI model.

[0198] Step 4: Evaluating the generative AI model

[0199] The server evaluates the trained generative AI model. As input, it receives the trained generative AI model from step 3 and test data. The server performs evaluation using natural language generation evaluation metrics such as the BLEU score and ROUGE score. Specifically, it compares the generated sentence with the reference sentence using from nltk.translate.bleu_score import sentence_bleu and calculates the BLEU score. The evaluation result is obtained as output.

[0200] Step 5: Generate a prompt statement

[0201] The server generates prompt sentences to be input into the generative AI model. As input, it receives requests from the user and system settings. In concrete terms, the user inputs prompt sentences such as "Tell me about a fun day you had" or "Tell me an interesting story." As output, it obtains prompt sentences to be input into the generative AI model.

[0202] (Application example 2)

[0203] Next, a description will be given of Application Example 2 of Form 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."

[0204] Conventional generative AI models lacked the training data needed to generate sentences that would be rated as interesting, making it difficult to generate the high-quality content users demand. Furthermore, they lacked the ability to instantly generate interesting sentences based on keywords or themes entered by users, making it impossible to meet user needs.

[0205] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for generating sentences that the additionally trained generation AI considers to be interesting, means for generating interesting sentences based on keywords and themes entered by the user, and means for displaying the generated sentences to the user. This makes it possible to instantly generate and provide high-quality content that users desire.

[0206] "SNS" is an abbreviation for social networking service, an online platform for users to share information and interact over the Internet.

[0207] "Crawling" is the technique of automatically collecting web pages and posts on the Internet, used by search engines and for data collection purposes.

[0208] "Generative AI" is a system that uses artificial intelligence technology to automatically generate content such as text and images.

[0209] "Retraining" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's performance.

[0210] "Keywords" are specific words or phrases entered by users that serve as instructions for the generative AI to generate content.

[0211] A "theme" is a specific topic or topic entered by the user, which serves as instructions for the generative AI when generating content.

[0212] "User" refers to an individual or organization that uses the system and generates content using generative AI.

[0213] "Content" refers to information such as text and images generated by generative AI and provided to users.

[0214] "Display" is the act of visually presenting the generated content on a user's device.

[0215] A system for implementing the present invention has the following configuration.

[0216] First, the server has a means for crawling posts that are rated as interesting from social media. Specifically, the server runs a crawling program to automatically collect post data from social media platforms on the Internet. This crawling program uses the Python requests library to obtain post data from specified URLs.

[0217] Next, the server has a means to train the generative AI based on the crawled posts. Specifically, the server trains the generative AI model (e.g., GPT-2) using the crawled post data. For this training, the server uses the transformers library, loads a tokenizer and model, tokenizes the post data, and inputs it into the model.

[0218] Furthermore, the server has a means for the generative AI, which has undergone additional training, to generate sentences that it deems interesting. Based on the keywords and themes entered by the user, the generative AI generates interesting sentences. In this generation process, the prompt entered by the user is tokenized using a tokenizer and input into the generative AI model. The generated sentences are then decoded using the tokenizer and displayed to the user.

[0219] Users access the system using a device such as a smartphone and input keywords or themes. For example, if a user inputs the prompt "funny cat stories," the server uses a generative AI model to generate sentences like the following:

[0220] Example prompt sentence:

[0221] Funny cat stories

[0222] Example of generated statement:

[0223] One day, a cat started playing the piano, and its owner, amazed at how perfectly the cat played Beethoven's "Moonlight Sonata," burst into applause.

[0224] In this way, users can instantly generate and enjoy high-quality content.The system is realized using Python programs, the requests library, the transformers library, and a generative AI model (GPT-2).

[0225] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0226] Step 1:

[0227] The server crawls posts that are rated as interesting from social media. Specifically, the server uses the requests library to retrieve post data from the API of the specified social media platform. The input is the SNS API endpoint, and the output is the retrieved post data.

[0228] Step 2:

[0229] The server further trains the generative AI model based on the crawled post data. Specifically, the server uses the transformers library to load the GPT-2 model and tokenizer and tokenize the post data. Then, it further trains the model using the tokenized data. The input is the crawled post data, and the output is the further trained generative AI model.

[0230] Step 3:

[0231] A user accesses the system using a terminal and inputs keywords and themes. Specifically, the user opens the smartphone application and enters a prompt sentence in a text box. The input is the prompt sentence entered by the user, and the output is a request to the server.

[0232] Step 4:

[0233] The server generates funny sentences based on the prompt sentence entered by the user. Specifically, the server tokenizes the prompt sentence using a tokenizer and inputs it into an additionally trained generative AI model. The model outputs the generated token sequence, which the server decodes and converts into a natural language sentence. The input is the prompt sentence entered by the user, and the output is the generated funny sentence.

[0234] Step 5:

[0235] The server displays the generated sentence on the user's device. Specifically, the server sends the generated sentence to the user's smartphone application, which then displays it on the screen. The input is the generated sentence, and the output is the sentence displayed on the user's device.

[0236] Example 3

[0237] Next, a description will be given of a third embodiment of the third embodiment. 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."

[0238] In conventional social media posting, users had to understand trends and think of appropriate content to create posts that would go viral, which took time and effort. Virtual influencers also had to constantly generate new content, which was a significant burden.

[0239] The identification process by the identification processing unit 290 of the data processing device 12 in Example 3 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for the user to input a prompt sentence, means for the terminal to send the prompt sentence to the server, means for the server to generate sentences using the generation AI model, means for the server to send the generated sentences to the terminal, means for the terminal to display the generated sentences to the user, and means for the user to post the generated sentences to SNS. This allows users to easily create posts that will go viral and efficiently carry out virtual influencer activities.

[0240] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[0241] "Crawling" is the process of automatically gathering information on the web, using specific algorithms to crawl through web pages and obtain data.

[0242] "Generative AI" is a system that uses artificial intelligence technology to generate new text and content, particularly utilizing natural language processing technology.

[0243] "Additive training" is the process of adding new data to an existing AI model and retraining it, in order to improve the model's performance and accuracy.

[0244] A "prompt" is input text that gives specific instructions or questions to the generation AI, and determines the direction of the generated content.

[0245] A "terminal" is a device such as a computer or smartphone used by a user, which communicates with a server to send and receive data.

[0246] A "server" is a computer system that processes and stores data on a network, and provides services in response to requests from client terminals.

[0247] A "virtual influencer" is someone who uses an artificially created character or avatar to act as an influential presence on social media.

[0248] A "platform" is a system that serves as the foundation for providing specific services and applications, and on which users can carry out various activities.

[0249] MODE FOR CARRYING OUT THE INVENTION

[0250] This invention is a system that crawls posts that are rated as interesting on social media and performs additional learning on a generation AI based on that data. A specific embodiment of this system is described below.

[0251] Hardware and software used

[0252] Hardware: High-performance servers (e.g., servers equipped with GPUs), user devices (e.g., smartphones, PCs)

[0253] Software: Generative AI models (e.g., OpenAI GPT-4), social media platforms (e.g., social media APIs)

[0254] System configuration

[0255] 1. Server

[0256] The server has a means to crawl posts that are rated as interesting from social media, thereby automatically collecting trending and popular posts.

[0257] The server has a means to further train the generative AI based on the crawled posts, which improves the generative AI model's ability to generate sentences based on the latest trends.

[0258] The server has a means for the generative AI to generate sentences that it thinks are interesting after additional training. The generative AI model generates new sentences based on prompts from the user.

[0259] 2. Terminal

[0260] The terminal has a means for the user to input a prompt sentence, which can be input using a dedicated application or a web interface.

[0261] The terminal has a means to send prompt text to the server, and the entered prompt text is securely sent to the server using the HTTPS protocol.

[0262] The terminal has a means for displaying the generated sentence to the user, and displays the generated sentence received from the server to the user so that the user can confirm it.

[0263] 3. Users

[0264] The user has a means to post the generated sentence to the SNS. After checking the generated sentence and making corrections as necessary, the user posts it to the SNS.

[0265] Specific examples

[0266] For example, if a user wants to create a viral post on social media, they can input the following prompt into the generative AI model:

[0267] Example prompt sentence:

[0268] "Create a funny post based on a recent trend. For example, cat videos are popular, so I'd like a funny post related to that."

[0269] Based on this prompt, the generative AI model generates the following post:

[0270] "Today's cat was jumping like a ninja! Cats are so funny!"

[0271] The server sends the generated message to the user's device, which displays it to the user. The user can then check the message and post it to the SNS.

[0272] In this way, the present invention allows users to easily create posts that will go viral, and also allows virtual influencers to carry out their activities efficiently. The flow of the identification process in the third embodiment will be described with reference to FIG.

[0273] Step 1:

[0274] The user enters a prompt statement.

[0275] The user enters a prompt using a dedicated application or web interface, for example, "Create an interesting post based on recent trends," and the prompt is sent to the device.

[0276] Step 2:

[0277] The terminal sends the prompt to the server.

[0278] The terminal sends the prompt text entered by the user to the server using the HTTPS protocol. The input data is the prompt text, and the output is a request sent to the server.

[0279] Step 3:

[0280] The server generates sentences using a generative AI model.

[0281] The server inputs the received prompt sentence into the generative AI model to generate an interesting sentence. The generative AI model generates a new sentence based on the results of prior additional learning. The input data is the prompt sentence, and the output is the generated sentence.

[0282] Step 4:

[0283] The server sends the generated statement to the terminal.

[0284] The server sends the generated sentence to the terminal using the HTTPS protocol. The input data is the generated sentence, and the output is the response sent to the terminal.

[0285] Step 5:

[0286] The terminal displays the generated statements to the user.

[0287] The terminal displays the received generated sentence in the text area of ​​the application so that the user can check it. The input data is the generated sentence, and the output is what is displayed to the user.

[0288] Step 6:

[0289] The user posts the generated sentence to a social networking site.

[0290] The user checks the generated sentences displayed, makes corrections as necessary, and then posts them to a social networking site. For example, the user posts using the Twitter API. The input data is the generated sentence, and the output is a post to the social networking site.

[0291] (Application example 3)

[0292] Next, a description will be given of Application Example 3 of Form Example 3. 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."

[0293] In conventional virtual stores, product introductions and promotions tend to be monotonous, making it difficult to attract users' interest. Furthermore, there is a lack of support for social media users to create posts that will create buzz, making effective marketing difficult. To solve these issues, a method is needed to generate interesting text using generative AI and utilize it effectively within the virtual store.

[0294] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[0295] In this invention, the server includes means for crawling posts that are rated as interesting on SNS, means for additionally training a generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for introducing and promoting products in a virtual store using the generation AI, and means for users to view posts by virtual influencers through their smart devices. This enables interesting product introductions and promotions that attract users' interest in the virtual store, and also supports SNS users in creating posts that will go viral.

[0296] "SNS" is an abbreviation for social networking service, an online platform for users to share information and interact over the Internet.

[0297] "Crawling" is a technique for automatically crawling web pages on the Internet and collecting information.

[0298] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new text and content based on learned data.

[0299] "Additive learning" is the process of adding new data to an existing learning model and re-training it.

[0300] A "virtual store" is a virtual store that exists on the Internet and is a platform where users can browse and purchase products online.

[0301] "Promotion" is a marketing activity to promote the sale of a product or service.

[0302] "Smart devices" is a general term for electronic devices that can connect to the Internet and run various applications. Specifically, it includes smartphones and head-mounted displays.

[0303] A "virtual influencer" is a computer-generated character who acts as an influential presence on social media, in virtual stores, and elsewhere.

[0304] "Posting" refers to the act of a user disclosing information or the content of that information on online platforms such as social media and blogs.

[0305] "Going viral" refers to the phenomenon in which a post spreads rapidly on social media and attracts the attention of many people.

[0306] A system for implementing this invention includes a means for crawling posts from social media that are rated as interesting, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are considered interesting, a means for using the generation AI to introduce and promote products in a virtual store, and a means for users to view posts from virtual influencers via a smart device.

[0307] The server first crawls posts from social media that are rated as interesting. This is done using technology that automatically crawls web pages on the Internet and collects information. Next, the generative AI is trained based on the crawled posts. Training is the process of adding new data to an existing learning model and retraining it.

[0308] The generative AI generates sentences that it deems entertaining based on the results of additional learning. This generative AI is implemented using OpenAI's API. The generated sentences are used to introduce and promote products in the virtual store. Specifically, virtual influencers use the generative AI to generate entertaining sentences and catchphrases, enhancing the appeal of the products.

[0309] Users can view posts from virtual influencers through smart devices (e.g., smartphones or head-mounted displays), which allows them to access detailed product information and purchase links.

[0310] As a concrete example, let's consider the case of generating a description for a smartwatch. The server inputs the following prompt text to the generation AI:

[0311] Product Name: Smartwatch

[0312] Features: Heart rate monitoring, GPS, music playback

[0313] Generate an interesting description of this product.

[0314] Based on this prompt, the AI ​​generates a fun introduction like this:

[0315] "When you wear this smartwatch, it will handle everything from heart rate to GPS to music! It's like a gadget from the future, and it will make your everyday life just like a sci-fi movie!"

[0316] The testimonials generated in this way are provided to users by the virtual influencers in the virtual store, which makes it possible to attract users' interest and effectively communicate the appeal of the products.

[0317] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0318] Step 1:

[0319] The server crawls posts that are rated as interesting on social media. It uses the URL or API endpoint of the social media site as input and collects data on posts that are rated as interesting as output. Specifically, the server launches a web crawler, patrols the social media posts, and extracts posts that have received high ratings, such as "likes" and "shares."

[0320] Step 2:

[0321] The server further trains the generative AI based on the crawled posts. The post data collected in step 1 is used as input, and the trained generative AI model is obtained as output. Specifically, the server preprocesses the collected post data and feeds it back to the generative AI model for re-training.

[0322] Step 3:

[0323] The server generates sentences that the generation AI, which has undergone additional training, thinks are interesting. As input, it uses a prompt sentence containing the product name and features, and as output, it obtains a generated interesting introduction sentence. Specifically, the server inputs the following prompt sentence into the generation AI:

[0324] Product Name: Smartwatch

[0325] Features: Heart rate monitoring, GPS, music playback

[0326] Generate an interesting description of this product.

[0327] Step 4:

[0328] The server uses the generated testimonial to introduce and promote the product in the virtual store. It uses the testimonial generated in step 3 as input and obtains promotional content to be displayed in the virtual store as output. Specifically, the server formats the generated testimonial as a post by the virtual influencer and displays it on the interface of the virtual store.

[0329] Step 5:

[0330] Users view posts by virtual influencers through their smart devices. As input, they use the URL and application of the virtual store, and as output, they obtain a screen on which they can view posts by virtual influencers. Specifically, users access the virtual store using a smartphone or head-mounted display and view posts containing generated testimonials.

[0331] 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.

[0332] "Example 1"

[0333] In one embodiment of the present invention, a sentence generated by a generation AI is combined with an emotion engine that recognizes the user's emotions. Specifically, the sentence generated by the generation AI is presented to the user, and the emotion engine analyzes the user's reaction (e.g., facial expression, tone of voice, etc.). The analysis results are used as training data for the generation AI, improving its ability to generate sentences that correspond to the user's emotions.

[0334] "Example 2"

[0335] In another embodiment of the present invention, a generation AI generates a sentence based on the user's emotion recognized by an emotion engine. Specifically, the emotion engine feeds back the user's emotion (e.g., joy or sadness) recognized by the emotion engine to the generation AI, which then generates a sentence appropriate to that emotion. For example, if the generation AI recognizes that the user is feeling joy, it generates a sentence that amplifies the user's joy.

[0336] "Example 3"

[0337] In yet another embodiment of the present invention, the activity of the virtual influencer is adjusted based on the user's emotion recognized by the emotion engine. Specifically, the content, tone, timing, etc. of the virtual influencer's posts are adjusted according to the user's emotion recognized by the emotion engine. For example, if the emotion engine recognizes that the user is feeling sad, the virtual influencer will take an action corresponding to the user's emotion, such as posting comforting content.

[0338] The processing flow of each embodiment will be described below.

[0339] "Example 1"

[0340] Step 1: The generative AI generates sentences to present to the user.

[0341] Step 2: The user reads the sentence generated by the AI.

[0342] Step 3: The emotion engine analyzes the user's reactions (e.g., facial expressions, tone of voice, etc.).

[0343] Step 4: Use the analysis results as training data for the generation AI to improve its ability to generate sentences that correspond to the user's emotions.

[0344] "Example 2"

[0345] Step 1: The emotion engine recognizes the user's emotion (e.g., joy, sadness, etc.).

[0346] Step 2: The recognized emotion is fed back to the generative AI.

[0347] Step 3: The AI ​​generates a sentence that matches the emotion. For example, if the AI ​​recognizes that the user is feeling joy, it generates a sentence that amplifies the user's joy.

[0348] "Example 3"

[0349] Step 1: The emotion engine recognizes the user's emotion.

[0350] Step 2: The content, tone, and timing of the virtual influencer's posts are adjusted based on the user's emotions as identified by the emotion engine.

[0351] Step 3: For example, if the virtual influencer recognizes that the user is feeling sad, it will take action that corresponds to the user's emotions, such as posting comforting messages.

[0352] Example 1

[0353] Next, a description will be given of Example 1 of Form 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."

[0354] With conventional crawling systems for social media posts, it was difficult to efficiently collect posts that were rated as interesting and train a generation AI based on them. Furthermore, there was a lack of a mechanism for analyzing the emotions that generated sentences evoke in users and feeding the results back to the generation AI, making it difficult to improve the accuracy of the generation AI. Furthermore, it was not possible to collect and analyze user reactions in real time to use them as training data for the generation AI.

[0355] 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.

[0356] In this invention, the server includes: means for crawling posts rated as interesting from social media; means for additionally training a generation AI based on the crawled posts; means for the additionally trained generation AI to generate sentences that are deemed interesting; means for presenting the generated sentences to users; means for collecting user responses; means including an emotion engine for analyzing the collected user responses; and means for feeding back the analysis results as learning data for the generation AI. This makes it possible to efficiently collect posts rated as interesting and improve the accuracy of the generation AI. Furthermore, by analyzing user emotions in real time and feeding back the results to the generation AI, it is possible to generate sentences that are more suitable for the user.

[0357] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[0358] "Crawling" is the process of automatically collecting information on the Internet and obtaining data based on specific criteria.

[0359] "Generative AI" refers to systems that use artificial intelligence techniques to generate new text or content, particularly in natural language processing.

[0360] "Additive training" is the process of adding new data to an existing machine learning model and retraining it, in order to improve the model's accuracy and performance.

[0361] The term "user" refers to an individual or organization that uses a system or service, and in the present invention particularly refers to a recipient of a generated sentence.

[0362] The "emotion engine" is a system that analyzes data such as the user's facial expressions and tone of voice to recognize the user's emotions.

[0363] "Feedback" is the process of using the system's output or analysis results as input again to improve the system's performance and accuracy.

[0364] A "virtual influencer" is a fictitious character created using artificial intelligence and animation technology who has influence on social media and other online platforms.

[0365] "Platform" refers to a system or environment that serves as the foundation for providing specific services or functions, and on which users can operate.

[0366] This invention is a system that crawls posts rated as interesting on social media, trains a generation AI based on those posts, presents the generated sentences to users, and analyzes user reactions to improve the accuracy of the generation AI. A specific embodiment of this system is described below.

[0367] The server accesses social media platforms and selects posts based on specific keywords, hashtags, and the number of reactions. For example, it crawls posts with the hashtag "funny" or posts with a certain number of "likes." The server then stores these posts in a database.

[0368] As a specific example, the server uses a social networking service's API to search for posts with the hashtag "funny" and crawl posts with more than 100 likes. The server also uses a social networking service's API to search for posts containing a specific keyword (e.g., "funny") and crawl posts with more than 50 comments.

[0369] Example prompt sentence:

[0370] "Using a social media API, create a program that crawls posts with interesting hashtags and retrieves posts with more than 100 likes."

[0371] The server then uses a generative AI model to generate new sentences based on the crawled posts, which are then presented to the user, who can then display the sentences or read them aloud using speech synthesis.

[0372] The user sees and hears the generated sentences. The device's camera and microphone are used to collect the user's reactions. The device then sends the collected data to an emotion engine, which analyzes the user's reactions. The results of this analysis are used as learning data for the generative AI.

[0373] As a concrete example, a generative AI generates the sentence "Today was a very fun day!" and presents it to the user. The device's camera captures the user's smile, and the microphone records the user's laughter. The emotion engine analyzes this data and recognizes that the user is feeling "joy." The results of this analysis are fed back to the generative AI, which improves its ability to generate sentences that make the user feel joyful from the next time onwards.

[0374] Example prompt sentence:

[0375] "Design a system that presents sentences generated by a generative AI to the user, collects the user's reactions using a camera and microphone, and analyzes them with an emotion engine."

[0376] In this way, the server, device, and user work together to efficiently collect posts that are rated as interesting, improving the accuracy of the generation AI. Also, by analyzing user emotions in real time and feeding the results back to the generation AI, it is possible to generate sentences that are more suitable for the user.

[0377] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0378] Step 1:

[0379] The server accesses the API of the social media platform and searches for posts based on specific keywords, hashtags, and the number of reactions. As input, it uses the API endpoint of the social media platform and search criteria (e.g., "funny" hashtag, 100 or more likes). The server receives the post data retrieved from the API and filters it. As output, it obtains a list of posts that meet the criteria. Specifically, the server sends a request to the API endpoint of the social media platform and retrieves posts that match the criteria.

[0380] Step 2:

[0381] The server stores the filtered posts in a database. As input, it uses the list of posts obtained in step 1. The server executes a SQL query to insert the post text and metadata into the database. As output, it gets the post data stored in the database. Specifically, the server stores each post's text, username, number of likes, etc. in the database.

[0382] Step 3:

[0383] The server generates new sentences using a generative AI model based on posts stored in a database. It uses the stored post data as input. The server sends a prompt to the generative AI model to generate a new sentence. The generated sentence is obtained as output. Specifically, the server sends a prompt to a generative AI model such as GPT-3 saying, "Generate a new sentence based on an interesting post."

[0384] Step 4:

[0385] The terminal presents the generated sentence to the user. It uses the sentence generated in step 3 as input. The terminal displays the generated sentence on the screen or reads it to the user using speech synthesis. As output, it obtains the sentence presented to the user. Specifically, the terminal displays the generated sentence on the display and reads it using a speech synthesis engine.

[0386] Step 5:

[0387] The device collects the user's reactions using a camera and a microphone. The inputs are the user's facial expressions and tone of voice. The device captures the user's facial expressions with the camera and records the user's voice with the microphone. The output is the collected reaction data. Specifically, the device captures the user's smile with the camera and records the user's laughter with the microphone.

[0388] Step 6:

[0389] The device sends the collected reaction data to the emotion engine and analyzes the user's emotions. The reaction data collected in step 5 is used as input. The device inputs the data into the emotion engine and recognizes the user's emotions. The output is the analysis result. Specifically, the device inputs image data and voice data into the emotion engine and recognizes the user's emotions (e.g., joy, surprise).

[0390] Step 7:

[0391] The server feeds back the emotion engine's analysis results as learning data for the generative AI. The analysis results obtained in step 6 are used as input. The server feeds back the analysis results to the generative AI model and reflects them in subsequent sentence generation. The output is an improvement in the accuracy of the generative AI model. Specifically, the server inputs the emotion engine's analysis results into the generative AI model and uses them as learning data for the model.

[0392] (Application example 1)

[0393] Next, a description will be given of Application Example 1 of Embodiment 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."

[0394] In conventional SNS, it was difficult to efficiently collect posts that were rated as interesting and generate content that would attract users' interest based on them. In addition, there was a lack of a mechanism to analyze how the generated content affected users' emotions and reflect the results in the next content generation. This made it difficult to continuously attract users' interest and lacked a feedback loop to improve the quality of the content.

[0395] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for presenting the generated sentences to the user and analyzing the user's emotional response, and means for using the analysis results as training data for the generation AI. This makes it possible to efficiently generate content that attracts the user's interest and improve the quality of the content based on the user's emotional response.

[0396] "SNS" is an abbreviation for social networking service, an online platform for users to interact with each other and share information over the Internet.

[0397] "Crawling" is the process of automatically gathering information on a website, a method of gathering posts based on specific keywords or hashtags.

[0398] "Generative AI" is a system that uses artificial intelligence technology to generate new text and content, and has the ability to generate sentences based on specific training data.

[0399] "Additive training" is the process of retraining an existing generative AI model by adding new data to it in order to improve the model's performance.

[0400] An "emotional response" is an emotional reaction a user has to the content presented to them, and is expressed through facial expressions, tone of voice, etc.

[0401] "Analysis" is the process of examining collected data in detail to find specific patterns and trends.

[0402] "Training data" is a dataset that generative AI uses to generate new content and improve the model's performance.

[0403] "Content" is a general term for information and entertainment provided to users, and includes text, images, videos, etc.

[0404] The system for implementing this invention includes a series of processes: crawling posts rated as interesting on social media, conducting additional learning by a generation AI based on those posts, presenting the generated sentences to the user, and analyzing the user's emotional response. Specific embodiments are described below.

[0405] System Configuration

[0406] The server uses web scraping technology to crawl posts that are rated as interesting from social media. Specifically, it uses the requests library and the web scraping library to collect posts that contain a specific hashtag (e.g., "interesting"). The collected posts are filtered based on the number of "likes."

[0407] Next, the server trains the generative AI based on the crawled posts. Here, we use the GPT-3 model from the transformers library. The generative AI generates new content using the crawled posts as input.

[0408] The generated content is presented on the user's device (smartphone or PC). To analyze the user's emotional response, the device's camera is used to capture facial expressions and analyze them using the cv2 (OpenCV) library. The analysis results are processed by an emotion recognition model and sent to the server as training data for the generation AI.

[0409] Hardware and software used

[0410] Hardware: Server, user device (smartphone, PC), camera

[0411] Software: requests library, web scraping library, transformers library, cv2 (OpenCV) library, emotion recognition model

[0412] Specific examples

[0413] For example, the server might input the following prompt to the generator AI:

[0414] Example prompt sentence:

[0415] Generate new content based on this post: "Something really interesting happened today! Interesting and fun"

[0416] The AI ​​then generates new content based on this prompt and presents it to the user. While the user is viewing the content, a camera captures the user's facial expressions, which are then analyzed by an emotion recognition model. The analysis results are sent to the server and reflected in the next content generation.

[0417] In this way, it is possible to efficiently generate content that attracts the user's interest and improve the quality of the content based on the user's emotional response.

[0418] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0419] Step 1:

[0420] The server crawls posts rated as interesting from social media. Specifically, it uses the requests library to access web pages containing a specific hashtag (e.g., "interesting") and then uses a web scraping library to analyze the posts. The input is the URL of the social media site, and the output is a list of posts rated as interesting.

[0421] Step 2:

[0422] The server further trains the generative AI based on the crawled posts. It uses the GPT-3 model from the transformers library and inputs the crawled posts as prompts. The input is the list of crawled posts, and the output is the generated new content.

[0423] Step 3:

[0424] The server sends the generated content to the user's terminal. The user's terminal displays the generated content. The input is the generated content, and the output is the content displayed on the user's terminal.

[0425] Step 4:

[0426] The user's device uses a camera to capture the user's facial expressions. The captured images are processed using the cv2 (OpenCV) library to analyze the facial expressions. The input is the image captured by the camera, and the output is the analyzed emotion data.

[0427] Step 5:

[0428] The user's device transmits the analyzed emotion data to the server. The input is the analyzed emotion data, and the output is the emotion data transmitted to the server.

[0429] Step 6:

[0430] The server uses the received emotion data as learning data for the generative AI, which is then reflected in the next content generation. The input is emotion data, and the output is an updated generative AI model.

[0431] Example 2

[0432] Next, a description will be given of Example 2 of Form 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."

[0433] Conventional generative AI systems have difficulty generating sentences based on user emotions, and are unable to provide appropriate sentences that reflect the user's emotions. Furthermore, there was a lack of effective ways to train generative AI using posts from social media, resulting in low-quality generated sentences. This led to a decrease in user satisfaction and limited the utility of generative AI.

[0434] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0435] In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for the generation AI to generate sentences based on the user's emotions recognized by the emotion engine, and means for the emotion engine to recognize the user's emotions. This makes it possible to generate appropriate sentences according to the user's emotions and improve user satisfaction.

[0436] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[0437] "Crawling" is the process of automatically visiting web pages on the Internet and gathering information.

[0438] "Generative AI" is a system that uses artificial intelligence technology to automatically generate content such as text and images.

[0439] "Retraining" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's performance.

[0440] An "emotion engine" is a system that recognizes emotions from a user's text, facial expressions, etc. and analyzes those emotions.

[0441] "User" refers to an individual or organization that uses the system and receives the output of the generative AI or emotion engine.

[0442] MODE FOR CARRYING OUT THE INVENTION

[0443] This invention is a system that crawls posts that are rated as interesting on social media and uses those posts for additional learning by the generation AI. Furthermore, it has a function that the emotion engine recognizes the user's emotions and the generation AI generates sentences based on those emotions.

[0444] System program generation

[0445] The server generates a program to crawl posts from social media sites, and the program uses a web scraping tool to collect posts from a specified list of URLs.

[0446] Program processing

[0447] The server saves the crawled posts as text data and formats them as training data for the generative AI. Natural language processing libraries such as NLTK and spaCy are used for text preprocessing. The server uses the formatted text data for additional training of the generative AI model. Machine learning frameworks such as TENSORFLOW (registered trademark) and PyTorch are used for this process.

[0448] The device uses software such as OpenCV and Emotion API to recognize emotions from user input and facial expressions. The emotion engine recognizes the user's emotions and feeds that information back to the server. The server inputs the emotion data received from the emotion engine into a generative AI that generates a sentence appropriate to that emotion.

[0449] Specific examples

[0450] As a concrete example, the following prompt sentence can be input to a generative AI model:

[0451] Example prompt sentence:

[0452] 1. "Generate funny jokes based on recent posts."

[0453] 2. "Generate messages that are appropriate for the user's delight."

[0454] Example 1:

[0455] If a user enters the prompt "Generate a funny joke based on recent posts," the server will format the crawled posts as learning data, and the generation AI will generate jokes like the following:

[0456] "Why did computers go to sea? The answer is, they went in search of a network!"

[0457] Example 2:

[0458] When a user inputs the prompt "Please generate a message appropriate for when the user is happy," the device analyzes the user's input text and facial expressions, and the emotion engine recognizes the user's happiness. The server inputs the emotion data received from the emotion engine into the generation AI, which then generates a message like the one below.

[0459] "That's wonderful news! We can feel your joy!"

[0460] In this way, the server and the terminal work together to provide a system that generates appropriate sentences based on crawled posts and user sentiment.

[0461] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0462] Program processing flow

[0463] Step 1: Crawl the posts

[0464] The server crawls posts that are rated as interesting by social media.

[0465] Input: List of social media URLs

[0466] What happens: The server uses a web scraping tool to collect posts from a specified list of URLs.

[0467] Output: Crawled post data

[0468] Step 2: Generate training data

[0469] The server saves the crawled posts as text data and formats them as learning data for the generation AI.

[0470] Input: Crawled post data

[0471] What it does: The server uses natural language processing libraries such as NLTK and spaCy to remove unnecessary HTML tags and special characters from the posted data and clean up the text data.

[0472] Output: Formatted text data

[0473] Step 3: Further training of the generative AI

[0474] The server uses the formatted text data to perform additional training on the generative AI model.

[0475] Input: Formatted text data

[0476] How it works: The server uses machine learning frameworks such as TensorFlow and PyTorch to input new text data into existing generative AI models and update the model's parameters.

[0477] Output: A trained generative AI model

[0478] Step 4: Emotion Recognition with the Emotion Engine

[0479] The device recognizes emotions from the user's input and facial expressions.

[0480] Input: User-entered text and camera footage

[0481] How it works: The device uses software such as OpenCV and Emotion API to analyze the user's input text and camera footage, and the emotion engine identifies the user's emotions. For example, text analysis calculates the frequency of positive words, and video analysis analyzes facial expressions.

[0482] Output: Recognized user emotion data

[0483] Step 5: Sentiment-based sentence generation

[0484] The server feeds back the user's emotions recognized by the emotion engine to the generation AI, which then generates a sentence appropriate to that emotion.

[0485] Input: Recognized user emotion data

[0486] Specific operation: The server inputs the emotion data received from the emotion engine into the generation AI, which then generates a sentence appropriate to that emotion. For example, if the server recognizes that the user is happy, the generation AI will generate a sentence such as, "That's wonderful! I can feel your joy!"

[0487] Output: Sentence generated based on sentiment

[0488] Step 6: Providing generated statements

[0489] The server provides the generated sentences to the user.

[0490] Input: Sentiment-based generated sentence

[0491] Specific operation: The server sends the generated sentence to the user's terminal, allowing the user to view the sentence.

[0492] Output: The generated text provided to the user

[0493] In this way, the server and the terminal work together to create a system that generates appropriate sentences based on crawled posts and user sentiment and provides them to the user.

[0494] (Application example 2)

[0495] Next, a description will be given of Application Example 2 of Form 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."

[0496] Conventional generative AI systems have difficulty generating content based on user emotions, and have been unable to provide content that is appropriate to the user's emotions. In addition, there is a lack of means to train the generative AI based on posts from social media, and the generation of sentences that are rated as interesting has not been sufficiently achieved.

[0497] 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.

[0498] In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for an emotion engine to recognize a user's emotion, and means for the generation AI to generate sentences based on the recognized emotion, thereby making it possible to generate and provide content that is appropriate for the user's emotion.

[0499] "SNS" is an abbreviation for Social Networking Service, an online platform for users to interact with other users over the Internet.

[0500] "Crawling" is a technique for automatically crawling web pages on the Internet and collecting information.

[0501] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new data and content based on given data.

[0502] "Additive training" is the process of adding new data to an existing learning model and retraining it.

[0503] An "emotion engine" is a system that analyzes emotions from a user's text or voice and recognizes those emotions.

[0504] "User emotion" refers to the emotional state, such as joy, sadness, or anger, that a user feels in response to a particular situation or piece of information.

[0505] "Content" means the information and entertainment provided on the Website or Application.

[0506] A "virtual influencer" is a fictitious person created using artificial intelligence and computer graphics who has influence on social media and other online platforms.

[0507] "Platform" refers to the underlying system or environment for providing specific services or applications.

[0508] The embodiments of the present invention will be described in detail below.

[0509] First, the server crawls posts that are rated as interesting from social media. The crawled posts are used for additional learning of the generative AI. The generative AI learns new data based on the crawled posts and improves its ability to generate sentences that are considered interesting.

[0510] Next, the emotion engine recognizes the user's emotion. The emotion engine analyzes emotions from the text or voice input by the user and recognizes their emotional state. For example, if the user inputs "I'm very happy today," the emotion engine recognizes "joy."

[0511] Based on the recognized emotions, the generative AI generates sentences that match the user's emotions. For example, if the user is feeling "joy," the generative AI will generate positive content that amplifies the user's joy. Specifically, it will generate sentences such as, "The weather is very nice today. A new cafe has opened."

[0512] The following hardware and software are used to realize this system. The hardware requires a server and a user device. The software requires a crawler to crawl social media, a generative AI model (e.g., GPT-2), and an emotion engine (e.g., TextBlob).

[0513] For example, if a user types, "I'm very happy today," the emotion engine will recognize "joy," and the generative AI will generate a sentence such as, "Content appropriate for when the user is feeling happy: The weather is very nice today. A new cafe has opened."

[0514] Example prompt sentence:

[0515] Content suitable for delighted users:

[0516] By inputting this prompt into the generative AI, content that matches the user's emotions can be generated.

[0517] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0518] Step 1:

[0519] The server crawls posts from social media that are rated as interesting. Specifically, it uses a crawler to collect posts from social media and selects posts that have received high ratings, such as "likes" and "shares." The input is the social media post data, and the output is a list of posts that are rated as interesting.

[0520] Step 2:

[0521] The server trains the generative AI based on the crawled posts. Specifically, the crawled posts are input into a generative AI model (e.g., GPT-2) and the model is retrained. The input is a list of crawled posts, and the output is the trained generative AI model.

[0522] Step 3:

[0523] The user terminal receives user input. Specifically, the user expresses their emotions through text or voice. The input is the user's text or voice data, and the output is passed directly to the emotion engine.

[0524] Step 4:

[0525] The server uses an emotion engine to recognize the user's emotions. Specifically, it uses an emotion analysis tool such as TextBlob to analyze emotions from the user's input. The input is the user's text or voice data, and the output is the recognized emotion (e.g., joy, sadness).

[0526] Step 5:

[0527] The server uses a generation AI to generate a sentence based on the recognized emotion. Specifically, the recognized emotion is input to the generation AI as a prompt sentence, which generates a sentence appropriate to the emotion. The input is the recognized emotion and the prompt sentence, and the output is the generated sentence.

[0528] Step 6:

[0529] The user terminal displays the generated sentence to the user. Specifically, the generated sentence is displayed on a screen so that the user can view it. The input is the generated sentence, and the output is the content displayed to the user.

[0530] Example 3

[0531] Next, a description will be given of a third embodiment of the third embodiment. 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."

[0532] In traditional social media, creating posts that will go viral requires a great deal of time and effort. Virtual influencers are also required to consistently provide fresh and interesting content. However, these tasks are extremely difficult, and there is a lack of efficient methods for doing so. Furthermore, providing appropriate content that reflects users' emotions is also a difficult challenge.

[0533] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[0534] In this invention, the server includes a means for crawling posts rated as interesting from social media, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are considered interesting, a means for a user to input a prompt sentence, a means for inputting the prompt sentence to a generation AI model, a means for the generation AI model to generate sentences, a means for publishing the generated sentences as posts by a virtual influencer, and a means for a user to confirm the generated sentences. This allows users to efficiently create posts that will "go viral," and allows virtual influencers to always provide fresh and interesting content. It also makes it possible to provide appropriate content based on user emotions.

[0535] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[0536] "Crawling" is the process of automatically visiting web pages on the Internet and gathering information.

[0537] "Generative AI" is a system that uses artificial intelligence technology to generate new data and content.

[0538] "Additive learning" is the process of adding new data to an existing learning model and re-training it.

[0539] A "prompt" is text that contains instructions or questions that are input to a generative AI model.

[0540] A "generative AI model" is a specific implementation of generative AI, an algorithm that generates new sentences based on input prompt sentences.

[0541] A "virtual influencer" is an artificially created character who acts as an influential presence on social media.

[0542] "User" means an individual or organization that uses the system.

[0543] A "post" is content such as text, images, and videos that are published on social media.

[0544] A "platform" is a system or environment that serves as the foundation for providing specific services or functions.

[0545] MODE FOR CARRYING OUT THE INVENTION

[0546] This invention is a system that crawls posts on social media that are rated as interesting, and then further trains a generative AI model to generate interesting sentences simply by the user entering a prompt sentence, which are then published as posts by virtual influencers.

[0547] Hardware and software used

[0548] Hardware:

[0549] Server equipped with a high-performance GPU

[0550] software:

[0551] Generative AI models (e.g., GPT-3)

[0552] Emotion engine (e.g. IBM Watson® Tone Analyzer)

[0553] Program processing explanation

[0554] The server crawls posts that are rated as interesting from social media. These crawled posts are used for further training of the generative AI model. After further training, the generative AI model generates new sentences based on prompt sentences entered by the user.

[0555] The user accesses the system's interface using a device (e.g., a smartphone or PC) and inputs a prompt sentence, such as "What interesting things happened today?" The server inputs this prompt sentence into the generative AI model, which then generates a new sentence.

[0556] The generated sentences are published by the server on a social networking platform as posts by the virtual influencer, which then become available for followers and other users to view. Users can then check the generated sentences using their devices.

[0557] Specific examples

[0558] Enter the prompt statement:

[0559] A user opens an app on their smartphone and types, "What interesting things happened today?"

[0560] Check the generated statements:

[0561] A user sees a post in the app that says, "Something interesting happened today! That's fun."

[0562] In this way, the server uses the generative AI model to generate interesting sentences in response to user prompts, and supports the creation of posts by virtual influencers and SNS users. The flow of the identification process in Example 3 will be described with reference to FIG. 21.

[0563] Program processing flow

[0564] Step 1:

[0565] The server crawls posts that are rated as interesting by social media.

[0566] Input: SNS URL or API endpoint

[0567] Data processing: Extract the posted content in text format and analyze the evaluation data.

[0568] Output: A list of posts that were rated interesting

[0569] Step 2:

[0570] The server further trains the generative AI model based on the crawled posts.

[0571] Input: A list of posts that were rated as interesting

[0572] Data processing: Feed the post content into the generative AI model and update the model parameters

[0573] Output: A trained generative AI model

[0574] Step 3:

[0575] The user uses the terminal to enter a prompt sentence.

[0576] Input: A prompt entered by the user (e.g., "What interesting things happened today?")

[0577] Data processing: Send the prompt to the server in text format

[0578] Output: Prompt text sent to the server

[0579] Step 4:

[0580] The server inputs the received prompt sentence into the generative AI model.

[0581] Input: The prompt text sent by the user

[0582] Data processing: Input the prompt sentence into the generative AI model to start the sentence generation process.

[0583] Output: The prompt input to the generative AI model

[0584] Step 5:

[0585] The generative AI model generates new sentences based on the prompt sentence.

[0586] Input: The prompt given to the generative AI model

[0587] Data processing: Analyze prompt sentences and generate new sentences based on trained data

[0588] Output: Generated sentence (e.g. "Something interesting happened today! It's interesting.")

[0589] Step 6:

[0590] The server publishes the generated sentences as posts of the virtual influencer.

[0591] Input: Generated statement

[0592] Data processing: Post the generated text using the API of the social media platform

[0593] Output: Virtual influencer posts published on social media

[0594] Step 7:

[0595] The user uses the terminal to review the generated sentences.

[0596] Input: Virtual influencer posts on social media platforms

[0597] Data processing: Displaying the posted content on the device interface

[0598] Output: Generated statements displayed on the terminal

[0599] In this way, the server uses a generative AI model to generate interesting sentences in response to user prompts, helping virtual influencers and social media users create posts.

[0600] (Application example 3)

[0601] Next, a description will be given of Application Example 3 of Form Example 3. 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."

[0602] Conventional virtual influencers and social media posting platforms have had difficulty providing appropriate content that reflects users' emotions. Furthermore, they lacked support for users to create posts that would create buzz, leading to a need for a method to generate effective posts.

[0603] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[0604] In this invention, the server includes means for crawling posts rated as interesting from social media, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for recognizing a user's emotions using an emotion recognition engine, means for adjusting the sentences generated by the generation AI based on the recognized emotions, means for publishing the adjusted sentences as posts by a virtual influencer, and means for providing a platform for users to create their own posts using the generation AI. This makes it possible to provide appropriate content according to user emotions and generate effective "buzzworthy" posts.

[0605] "SNS" is an abbreviation for social networking service, a platform for users to share information and interact online.

[0606] "Crawling" is a technique for automatically collecting information on the Internet, and is a means of obtaining data from specific websites and social media sites.

[0607] "Generative AI" is a type of artificial intelligence technology, an algorithm that generates new text and content based on learned data.

[0608] "Additive training" is the process of adding new data to an existing learning model and retraining it, and is a means of improving the model's performance and accuracy.

[0609] An "emotion recognition engine" is a technology that analyzes emotions from a user's text or voice and recognizes a specific emotional state.

[0610] A "virtual influencer" is a non-existent digital character who acts as an influential presence on social media and other online platforms.

[0611] A "platform" is a system or environment that serves as the foundation for providing specific services or applications.

[0612] "User" refers to an individual or organization that uses the system or service, and in this case refers to a user of the SNS.

[0613] A "post" is content such as text, images, or videos that a user makes public on a social networking site or other online platform.

[0614] "Adjustment" is the process of modifying generated content or system behavior in response to specific conditions or circumstances.

[0615] A system for implementing this invention includes means for crawling posts from social media that are rated as interesting, means for allowing a generation AI to perform additional learning based on the crawled posts, means for the generation AI to generate sentences that are deemed interesting after additional learning, means for recognizing a user's emotions using an emotion recognition engine, means for adjusting the sentences generated by the generation AI based on the recognized emotions, means for publishing the adjusted sentences as posts by a virtual influencer, and means for providing a platform for users to create their own posts using the generation AI.

[0616] System configuration

[0617] This system uses the following hardware and software:

[0618] Hardware: Servers, smartphones

[0619] Software: OpenAI API, EmotionEngine (emotion recognition engine)

[0620] Program processing explanation

[0621] The server crawls posts that are rated as interesting on social media. These crawled posts are used as additional training data for the generation AI. The generation AI then generates new, interesting sentences based on the additional training data.

[0622] When a user accesses the system via their smartphone, the emotion recognition engine analyzes the user's input text and recognizes the user's emotions. Based on the recognized emotions, the sentences generated by the generation AI are adjusted. For example, if the user is feeling sad, the generation AI will generate comforting sentences.

[0623] The adjusted statements are published as posts by virtual influencers, and users can use the platform to create their own posts using generative AI. The platform makes it easy for users to create posts that are intended to go viral.

[0624] Specific examples

[0625] For example, if a user types, "I've been feeling a bit down lately," the emotion recognition engine will recognize the user's emotion as "sadness." Based on this, the generative AI will generate a sentence such as, "I found this great product today! Cheer up!" and publish it as a post by the virtual influencer.

[0626] Prompt Sentence Examples

[0627] If the virtual influencer recognizes that the user is feeling sad, they should generate a comforting post, such as, "I found this great product today! Cheer up!"

[0628] In this way, it becomes possible to provide appropriate content that reflects the user's emotions and to generate effective "buzz" posts.

[0629] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0630] Step 1:

[0631] The server crawls posts rated as interesting from social media. It uses the social media URL or API endpoint as input and obtains the post data rated as interesting as output. Specifically, it uses the social media API to collect posts based on specific hashtags or keywords, and then selects posts with high ratings such as "likes" and "shares."

[0632] Step 2:

[0633] The server further trains the generative AI based on the crawled posts. It uses the crawled post data as input and obtains a trained generative AI model as output. Specifically, it preprocesses the collected post data and supplies it to the generative AI as text data. The generative AI learns new patterns and trends based on this data.

[0634] Step 3:

[0635] The server generates sentences that the trained generative AI finds interesting. It uses the trained generative AI model as input and obtains the generated interesting sentences as output. Specifically, it inputs a prompt sentence to the generative AI and generates new text based on it.

[0636] Step 4:

[0637] The device recognizes the user's emotions using an emotion recognition engine. It uses the user's text input as input and obtains recognized emotion data as output. Specifically, the user provides the input text to the emotion recognition engine, which then analyzes the emotion from the text.

[0638] Step 5:

[0639] The server adjusts the sentences generated by the AI ​​based on the recognized emotions. It uses the recognized emotion data and the generated sentence as input and obtains the adjusted sentence as output. Specifically, it changes the tone and content of the generated sentence according to the emotion data. For example, if the user is feeling sad, it changes the generated sentence to be comforting.

[0640] Step 6:

[0641] The server publishes the adjusted sentence as a post of the virtual influencer. Using the adjusted sentence as input, the server obtains a post published on the SNS as output. Specifically, the adjusted sentence is posted to the SNS from the virtual influencer's account.

[0642] Step 7:

[0643] Users use the platform to create their own posts using generative AI. The platform uses the user's prompt as input and gets the generated post as output. Specifically, the user enters a prompt on the platform, and the generative AI generates a post based on it.

[0644] 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.

[0645] 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> ) 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.

[0646] Another example of generative AI is Gemini (registered trademark) (Internet search engine). <url: https: gemini.google.com ?hl="ja">) are listed.

[0647] 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.

[0648] [Second embodiment]

[0649] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0650] 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.

[0651] 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).

[0652] 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.

[0653] 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.

[0654] 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).

[0655] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0656] 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.

[0657] 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.

[0658] 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.

[0659] 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.

[0660] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0661] "Example 1"

[0662] One embodiment of the present invention is a system that includes a means for crawling posts from social media that are rated as interesting. This system selects posts based on specific keywords, hashtags, the number of reactions, etc., and crawls their text. For example, it is conceivable to crawl posts with the hashtag "interesting" or posts with a certain number of "likes."

[0663] "Example 2"

[0664] Next, it is equipped with a means for additional training of the generation AI based on the crawled posts. This AI uses the text of the crawled posts as training data to learn how to generate sentences that will be evaluated as interesting. Specifically, the text of the crawled posts is used as training data for additional training of the existing learning model.

[0665] "Example 3"

[0666] Furthermore, the generative AI, which has undergone additional learning, is equipped with a means to generate sentences that it deems interesting. This generative AI generates new sentences based on the results of its learning. For example, it can generate sentences such as "Something interesting happened today! That's funny" as posts from virtual influencers.

[0667] "Example 4"

[0668] It also provides a means for users to act as virtual influencers using generative AI, and a platform for social media users to create posts that are intended to go viral. Specifically, it publishes interesting sentences created by generative AI as posts by virtual influencers, and provides a platform for social media users to create their own posts using generative AI.

[0669] The processing flow of each embodiment will be described below.

[0670] "Example 1"

[0671] Step 1: The system selects posts from social media based on specific keywords, hashtags, number of reactions, etc. For example, it selects posts with the hashtag "funny" or posts with a certain number of "likes."

[0672] Step 2: The text of the selected posts is crawled. The crawled data is used as training data for the generation AI.

[0673] "Example 2"

[0674] Step 1: Additional training is performed on the generation AI based on the crawled posts. Specifically, additional training is performed on the existing learning model using the text of the crawled posts as training data.

[0675] Step 2: After additional learning, the generation AI generates new sentences based on the results of its learning.

[0676] Step 1: The interesting sentences created by the generative AI are published as posts by virtual influencers.

[0677] Step 2: Provide a platform for social media users to create their own posts using generative AI.

[0678] Example 1

[0679] Next, a description will be given of Example 1 of Form 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."

[0680] There is a demand for a system that can efficiently collect posts that are rated as interesting on social media and automatically generate more attractive content by training a generation AI based on that data. However, with conventional methods, the collection and selection of posts is often done manually, which takes time and effort. In addition, it is difficult to select appropriate posts to use as training data for the generation AI, which results in an issue of inconsistent quality of the generated content.

[0681] 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.

[0682] In this invention, the server includes means for inputting conditions such as specific keywords, hashtags, and the number of reactions, means for accessing the API of the SNS and searching for posts based on the conditions, means for selecting posts that match the conditions from the search results, means for saving the text of the selected posts in a database, means for providing the saved data to the user, means for additional training of the generation AI based on the crawled posts, and means for the additionally trained generation AI to generate sentences that are considered interesting. This makes it possible to efficiently collect posts that are rated as interesting on the SNS and use them as training data for the generation AI, thereby automatically generating more attractive content.

[0683] "Means for entering specific keywords, hashtags, and reaction counts" is a function that provides an interface for users to enter keywords, hashtags, and reaction counts that they specify as search conditions.

[0684] "A means of accessing an SNS's API and searching for posts based on conditions" refers to a function that uses the application programming interface (API) provided by the SNS to search for posts based on conditions specified by the user.

[0685] "Means for selecting posts that match conditions from search results" refers to a function that filters and selects posts that match conditions specified by the user from the search results obtained from the SNS API.

[0686] The "means for saving the text of the selected posts to a database" is a function for storing the text of the filtered posts in a database.

[0687] The "means for providing saved data to users" is a function that provides an interface for displaying posted information saved in the database to users.

[0688] "Means for additional training of the generative AI based on crawled posts" is a function for additional training of the generative AI model using collected post data.

[0689] "Means for a generative AI that has undergone additional training to generate sentences that are deemed interesting" is a function that enables a generative AI model that has undergone additional training to automatically generate sentences that are evaluated as interesting.

[0690] MODE FOR CARRYING OUT THE INVENTION

[0691] This invention is a system that efficiently collects posts that are rated as interesting from social media, and automatically generates more attractive content by training a generation AI based on that data. Specific embodiments of this system are described below.

[0692] 1. Program Generation

[0693] The system's programs are written using programming languages ​​such as Python and JavaScript, and include code to access social networking site APIs and collect posts based on specific criteria.

[0694] 2. Program Processing

[0695] The server performs its processing using the following hardware and software:

[0696] Hardware: The server uses a computer system with a powerful processor and sufficient memory.

[0697] Software: The server uses libraries to access the SNS's API and a database management system (e.g., MySQL).

[0698] The server receives specific keywords, hashtags, and reaction count conditions entered by the user through their device and sends a request to the social networking service's API. For example, if a user specifies the hashtag "funny" and the conditions "more than 100 likes," the server searches for posts based on these conditions.

[0699] The server analyzes the search results returned by the API and selects posts that match the criteria. The selected post text is saved in a database, which stores information such as the post ID, text, hashtags, and number of reactions.

[0700] The server then further trains the generative AI based on the crawled posts. The generative AI model uses an advanced natural language processing model such as GPT-3. After further training, the generative AI automatically generates sentences that it deems interesting.

[0701] 3. Examples of concrete examples and prompts

[0702] As a concrete example, consider the case where you want to use a social media API to crawl posts with the hashtag "interesting." Below is an example of a prompt for that purpose.

[0703] Examples:

[0704] The server uses the social networking site's API to search for posts with the hashtag "funny."

[0705] The server selects posts from the search results that have 100 or more likes.

[0706] The server stores the selected post text in a database.

[0707] Example prompt sentence:

[0708] Create a program that uses a social media API to search for posts with the hashtag "interesting," select posts with more than 100 likes, and save the text of those posts to a database.

[0709] In this way, the server can efficiently crawl posts that are rated as interesting from social media, store them in a database, and use them as learning data for the generative AI, which will enable the AI ​​to automatically generate more attractive content.

[0710] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0711] Program processing flow

[0712] Step 1:

[0713] Users enter conditions such as specific keywords, hashtags, and number of reactions.

[0714] The user inputs search criteria through the device interface, for example, specifying the hashtag "interesting" and the criteria "more than 100 likes."

[0715] Input: Keyword (e.g. "interesting"), reaction count condition (e.g. "100 or more likes")

[0716] Output: The search criteria is sent to the server.

[0717] Step 2:

[0718] The server accesses the SNS API and searches for posts based on the criteria.

[0719] The server accesses the SNS API based on the conditions entered by the user, sends an API request, and searches for posts with the hashtag "interesting."

[0720] Input: Search criteria (e.g. "interesting", reaction count criteria "100 or more likes")

[0721] Data processing: API request generation

[0722] Output: Search results returned from the SNS API

[0723] Step 3:

[0724] The server selects posts that match the search criteria from the search results.

[0725] The server receives the search results returned by the API and selects posts with more than 100 likes. The server extracts information such as the ID, text, and number of reactions from the selected posts.

[0726] Input: Search results from SNS API

[0727] Data processing: Filtering posts that match criteria

[0728] Output: A list of posts that match the criteria

[0729] Step 4:

[0730] The server saves the selected post text to the database.

[0731] The server stores the selected post text in a database, which stores information such as the post ID, text, hashtags, and number of reactions.

[0732] Input: A list of posts that match the criteria

[0733] Data processing: storing submitted information in a database

[0734] Output: Post information saved in the database

[0735] Step 5:

[0736] The server provides the stored data to the user.

[0737] The server provides the user with information about the posts stored in the database, and the user can view a list of the saved posts through the interface of the terminal.

[0738] Input: Post information stored in the database

[0739] Data processing: Providing posted information to users

[0740] Output: A list of posts that the user can see

[0741] Step 6:

[0742] The generation AI further learns based on the posts crawled by the server

[0743] The server uses the crawled post data to further train the generative AI model, which uses an advanced natural language processing model such as GPT-3.

[0744] Input: Post information stored in the database

[0745] Data processing: Additional training for generative AI models

[0746] Output: A trained generative AI model

[0747] Step 7:

[0748] The server performs additional training and the AI ​​generates sentences that it deems interesting.

[0749] The server uses a trained generative AI model to automatically generate sentences that it deems interesting.

[0750] Input: A trained generative AI model

[0751] Data processing: Sentence generation using generative AI

[0752] Output: Funny sentence generated

[0753] In this way, the server can efficiently crawl posts from social media that are rated as interesting, store them in a database, and use them as learning data for the generation AI, making it possible to automatically generate more attractive content.

[0754] (Application example 1)

[0755] Next, a description will be given of Application Example 1 of Form 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."

[0756] Conventional crawling systems for social media posts have had difficulty efficiently collecting and providing users with "interesting" posts that interest them. Furthermore, they lacked the functionality to allow users to customize their feeds to suit their interests, creating a need for an improved user experience. Furthermore, when using generative AI to generate content, it was difficult to accurately generate the "interesting" content users were looking for.

[0757] 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.

[0758] In this invention, the server includes a means for crawling posts rated as interesting from social media, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that it deems interesting, a means for generating a feed for providing the generated sentences to users, and a setting means for users to customize the feed. This makes it possible to efficiently collect "interesting" posts that interest users and provide them to users. It also allows users to customize their feeds to suit their interests, improving the user experience. Furthermore, when generating content using a generation AI, it becomes possible to accurately generate the "interesting" content that users desire.

[0759] "SNS" is an abbreviation for Social Networking Service, an online platform that allows users to share information and interact with other users over the Internet.

[0760] "Crawling" is the process of automatically collecting information on the Internet, and is a technique for obtaining data based on specific keywords or conditions.

[0761] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new information and content based on given data.

[0762] "Additive training" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's accuracy and performance.

[0763] A "feed" is a stream or list of information provided to a user, in which selected content is displayed sequentially based on specific criteria.

[0764] "Configuration means" refers to the interface and functions that allow the user to customize the system's operation and display content.

[0765] A system for implementing this invention includes a means for crawling posts from social media that are rated as interesting, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are deemed interesting, a means for generating a feed to provide the generated sentences to users, and a setting means for users to customize the feed.

[0766] System Program

[0767] The server uses Python and social media APIs to crawl posts that are rated as interesting by social media. Specifically, it accesses the API using social media API credentials and crawls posts based on specific keywords and hashtags. The crawled posts are then filtered based on rating criteria such as the number of likes and retweets.

[0768] Next, the generation AI is trained based on the crawled posts. For the generation AI, an advanced natural language processing model such as GPT-3 is used. The training is carried out using the crawled post data to improve the accuracy and performance of the generation AI.

[0769] After additional learning, the generative AI generates sentences that it deems interesting. These generated sentences are stored on the server as feeds to be provided to users. The feeds are designed to efficiently provide content that users are interested in, and users can customize them to suit their interests.

[0770] Users can browse feeds using devices such as smartphones and PCs and customize their feeds through the settings means, which include an interface for changing the content displayed on the feed and the evaluation criteria.

[0771] Hardware and software used

[0772] Hardware: Servers, smartphones, PCs

[0773] Software: Python, social media APIs, generative AI models (e.g., GPT-3)

[0774] Specific examples

[0775] For example, if a user sets the hashtags "funny" and "hilarious" and crawls posts with more than 100 likes, the server will collect posts based on these conditions and further train the AI. The AI ​​will then generate sentences that it deems funny and display them in the user's feed.

[0776] Prompt Sentence Examples

[0777] Crawl for posts with the hashtags "funny" and "hilarious" and display posts with over 100 likes in your feed.

[0778] In this way, it is possible to efficiently collect and provide "interesting" posts that interest users. It also allows users to customize their feeds to suit their interests, improving the user experience. Furthermore, when generating content using generative AI, it is possible to accurately generate the "interesting" content that users are looking for.

[0779] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0780] Step 1:

[0781] The server crawls posts from social media sites that are rated as interesting. Specifically, it uses Python and social media APIs to collect posts based on specific keywords and hashtags. Evaluation criteria such as keywords, hashtags, and number of likes are given as input. The server searches for posts based on these conditions and outputs the collected post data.

[0782] Step 2:

[0783] The server performs additional training on the generative AI based on the crawled posts. Specifically, the collected post data is input into a generative AI model (e.g., GPT-3) to perform additional training on the model. The crawled post data is given as input, and the trained generative AI model is obtained as output. Through this process, the server improves the accuracy and performance of the generative AI.

[0784] Step 3:

[0785] The server uses a generative AI that has undergone additional training to generate sentences that it deems interesting. Specifically, it inputs a prompt sentence into the generative AI model that has undergone additional training, and outputs the generated sentence. The prompt sentence is given as input, and the generated sentence is obtained as output. The server uses this generated sentence in the next step.

[0786] Step 4:

[0787] The server generates a feed to provide the generated sentences to the user. Specifically, it converts the generated sentences into a feed format and sends it to the user's device. The generated sentences are given as input, and data in feed format is obtained as output. By providing this feed to the user, the server enables the user to efficiently browse content that interests them.

[0788] Step 5:

[0789] Users browse feeds using their devices and customize them through settings. Specifically, users view feeds using smartphones or computers and operate the settings interface to change the content and evaluation criteria of the feed. The user's customization settings are given as input, and a customized feed is obtained as output. Through this process, users can create a feed tailored to their interests.

[0790] Example 2

[0791] Next, a description will be given of Example 2 of Form 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."

[0792] Conventional generative AI models lack the training data to generate sentences that are rated as interesting, resulting in low-quality generated sentences. In addition, the evaluation methods for generated sentences are inadequate, making it difficult to accurately evaluate the performance of generative AI models.

[0793] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for crawling posts on the Internet, a means for cleansing the crawled posts, a means for additionally training the generative AI model using the cleansed text data, and a means for evaluating the additionally trained generative AI model. This enables the generative AI model to generate high-quality sentences that are evaluated as interesting.

[0794] "Means of crawling internet posts" refers to technologies and methods for automatically collecting post data from websites and social networking sites on the internet.

[0795] "Methods for cleansing crawled posts" are techniques and methods for removing unnecessary information from collected post data and normalizing text data.

[0796] "Means for additional training of generative AI models using cleansed text data" refers to techniques and methods for using preprocessed text data to add new knowledge to existing generative AI models and improve their performance.

[0797] "Means for evaluating a trained generative AI model" refers to techniques or methods for measuring the quality of generated sentences in order to evaluate the performance of a trained generative AI model.

[0798] "Means of acting as a virtual character" refers to the technology or method used to create a virtual character using a generative AI model and have that character act on the Internet.

[0799] "Means for providing a platform for internet users to create posts intended to attract attention" refers to technologies and methods that use generative AI models to provide an environment or service that enables internet users to create posts that attract the attention of many people.

[0800] MODE FOR CARRYING OUT THE INVENTION

[0801] The present invention relates to a system for crawling posts on the Internet and additionally training a generative AI model. Specific embodiments of this system are described below.

[0802] Data collection

[0803] The server uses a Python web scraping library to crawl posts on the Internet, collects post data from designated websites and social media sites, and stores it in text files or a database.

[0804] Examples:

[0805] The server retrieves the HTML using requests.get('https: / / example-blog.com') and parses it using web scraping (html.content, 'html.parser'). The post content is extracted from the parsed data and saved in a text file.

[0806] Data Preprocessing

[0807] The server cleanses the collected text data. Specifically, it removes unnecessary HTML tags, deletes special characters, normalizes the data, etc. The server preprocesses the text data using Python's regular expression library (re) and natural language processing libraries (NLTK, spaCy).

[0808] Examples:

[0809] The server removes HTML tags using re.sub('<[^<]+?>', '', text), tokenizes using nltk.word_tokenize(text), and normalizes the text using spacy.load('en_core_web_sm').

[0810] Data training

[0811] The server uses the cleansed text data to further train the generative AI model. The server uses generative AI models such as OpenAI's GPT-3 and GPT-4, and trains the model using the Transformers library.

[0812] Examples:

[0813] The server loads the model and tokenizer using from transformers import GPT2LMHeadModel, GPT2Tokenizer, and starts additional training with model.train(). Use the cleansed text data as training data.

[0814] Evaluating generative AI models

[0815] The server checks the quality of the generated sentences using test data to evaluate the trained generative AI model. The server performs the evaluation using natural language generation evaluation metrics such as the BLEU score and ROUGE score.

[0816] Examples:

[0817] The server calculates the BLEU score by comparing the generated sentence with a reference sentence using from nltk.translate.bleu_score import sentence_bleu, e.g., sentence_bleu([reference], candidate).

[0818] Prompt Sentence Examples

[0819] Examples of prompts to be input to a generative AI model include:

[0820] "Tell us about your fun day."

[0821] "Please tell me a funny story."

[0822] "Tell me about something fun that happened recently."

[0823] Describe a fun time you had with friends.

[0824] In this way, the server further trains the generation AI based on the crawled posts, building a system that learns how to generate sentences that will be evaluated as interesting.

[0825] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0826] Step 1: Data collection

[0827] The server crawls posts on the Internet. As input, it receives the URL of a specified website or social media site. The server uses a Python web scraping library to collect post data from the specified URL. Specifically, it retrieves HTML using requests.get('https: / / example-blog.com') and parses it using web scraping (html.content, 'html.parser'). As output, the collected post data is saved in a text file or database.

[0828] Step 2: Data Preprocessing

[0829] The server cleanses the collected text data. As input, it receives the post data collected in step 1. The server performs tasks such as removing unnecessary HTML tags, deleting special characters, and normalizing the data. Specifically, it removes HTML tags using re.sub('<[^<]+?>', '', text) and tokenizes the text using nltk.word_tokenize(text). It then normalizes the text using spacy.load('en_core_web_sm'). The cleansed text data is obtained as output.

[0830] Step 3: Train the data

[0831] The server uses the cleansed text data to train the generative AI model. As input, it receives the cleansed text data from step 2. The server uses a generative AI model such as OpenAI's GPT-3 or GPT-4, and trains the model using the Transformers library. Specifically, it loads the model and tokenizer using from transformers import GPT2LMHeadModel, GPT2Tokenizer, and starts the additional training with model.train(). The output is the trained generative AI model.

[0832] Step 4: Evaluating the generative AI model

[0833] The server evaluates the trained generative AI model. As input, it receives the trained generative AI model from step 3 and test data. The server performs evaluation using natural language generation evaluation metrics such as the BLEU score and ROUGE score. Specifically, it compares the generated sentence with the reference sentence using from nltk.translate.bleu_score import sentence_bleu and calculates the BLEU score. The evaluation result is obtained as output.

[0834] Step 5: Generate a prompt statement

[0835] The server generates prompt sentences to be input into the generative AI model. As input, it receives requests from the user and system settings. In concrete terms, the user inputs prompt sentences such as "Tell me about a fun day you had" or "Tell me an interesting story." As output, it obtains prompt sentences to be input into the generative AI model.

[0836] (Application example 2)

[0837] Next, a description will be given of Application Example 2 of Form 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."

[0838] Conventional generative AI models lacked the training data needed to generate sentences that would be rated as interesting, making it difficult to generate the high-quality content users demand. Furthermore, they lacked the ability to instantly generate interesting sentences based on keywords or themes entered by users, making it impossible to meet user needs.

[0839] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for generating sentences that the additionally trained generation AI considers to be interesting, means for generating interesting sentences based on keywords and themes entered by the user, and means for displaying the generated sentences to the user. This makes it possible to instantly generate and provide high-quality content that users desire.

[0840] "SNS" is an abbreviation for social networking service, an online platform for users to share information and interact over the Internet.

[0841] "Crawling" is the technique of automatically collecting web pages and posts on the Internet, used by search engines and for data collection purposes.

[0842] "Generative AI" is a system that uses artificial intelligence technology to automatically generate content such as text and images.

[0843] "Retraining" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's performance.

[0844] "Keywords" are specific words or phrases entered by users that serve as instructions for the generative AI to generate content.

[0845] A "theme" is a specific topic or topic entered by the user, which serves as instructions for the generative AI when generating content.

[0846] "User" refers to an individual or organization that uses the system and generates content using generative AI.

[0847] "Content" refers to information such as text and images generated by generative AI and provided to users.

[0848] "Display" is the act of visually presenting the generated content on a user's device.

[0849] A system for implementing the present invention has the following configuration.

[0850] First, the server has a means for crawling posts that are rated as interesting from social media. Specifically, the server runs a crawling program to automatically collect post data from social media platforms on the Internet. This crawling program uses the Python requests library to obtain post data from specified URLs.

[0851] Next, the server has a means to train the generative AI based on the crawled posts. Specifically, the server trains the generative AI model (e.g., GPT-2) using the crawled post data. For this training, the server uses the transformers library, loads a tokenizer and model, tokenizes the post data, and inputs it into the model.

[0852] Furthermore, the server has a means for the generative AI, which has undergone additional training, to generate sentences that it deems interesting. Based on the keywords and themes entered by the user, the generative AI generates interesting sentences. In this generation process, the prompt entered by the user is tokenized using a tokenizer and input into the generative AI model. The generated sentences are then decoded using the tokenizer and displayed to the user.

[0853] Users access the system using a device such as a smartphone and input keywords or themes. For example, if a user inputs the prompt "funny cat stories," the server uses a generative AI model to generate sentences like the following:

[0854] Example prompt sentence:

[0855] Funny cat stories

[0856] Example of generated statement:

[0857] One day, a cat started playing the piano, and its owner, amazed at how perfectly the cat played Beethoven's "Moonlight Sonata," burst into applause.

[0858] In this way, users can instantly generate and enjoy high-quality content.The system is realized using Python programs, the requests library, the transformers library, and a generative AI model (GPT-2).

[0859] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0860] Step 1:

[0861] The server crawls posts that are rated as interesting from social media. Specifically, the server uses the requests library to retrieve post data from the API of the specified social media platform. The input is the SNS API endpoint, and the output is the retrieved post data.

[0862] Step 2:

[0863] The server further trains the generative AI model based on the crawled post data. Specifically, the server uses the transformers library to load the GPT-2 model and tokenizer and tokenize the post data. Then, it further trains the model using the tokenized data. The input is the crawled post data, and the output is the further trained generative AI model.

[0864] Step 3:

[0865] A user accesses the system using a terminal and inputs keywords and themes. Specifically, the user opens the smartphone application and enters a prompt sentence in a text box. The input is the prompt sentence entered by the user, and the output is a request to the server.

[0866] Step 4:

[0867] The server generates funny sentences based on the prompt sentence entered by the user. Specifically, the server tokenizes the prompt sentence using a tokenizer and inputs it into an additionally trained generative AI model. The model outputs the generated token sequence, which the server decodes and converts into a natural language sentence. The input is the prompt sentence entered by the user, and the output is the generated funny sentence.

[0868] Step 5:

[0869] The server displays the generated sentence on the user's device. Specifically, the server sends the generated sentence to the user's smartphone application, which then displays it on the screen. The input is the generated sentence, and the output is the sentence displayed on the user's device.

[0870] Example 3

[0871] Next, a description will be given of Example 3 of Form Example 3. 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."

[0872] In conventional social media posting, users had to understand trends and think of appropriate content to create posts that would go viral, which took time and effort. Virtual influencers also had to constantly generate new content, which was a significant burden.

[0873] The identification process by the identification processing unit 290 of the data processing device 12 in Example 3 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for the user to input a prompt sentence, means for the terminal to send the prompt sentence to the server, means for the server to generate sentences using the generation AI model, means for the server to send the generated sentences to the terminal, means for the terminal to display the generated sentences to the user, and means for the user to post the generated sentences to SNS. This allows users to easily create posts that will go viral and efficiently carry out virtual influencer activities.

[0874] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[0875] "Crawling" is the process of automatically gathering information on the web, using specific algorithms to crawl through web pages and obtain data.

[0876] "Generative AI" is a system that uses artificial intelligence technology to generate new text and content, particularly utilizing natural language processing technology.

[0877] "Additive training" is the process of adding new data to an existing AI model and retraining it, in order to improve the model's performance and accuracy.

[0878] A "prompt" is input text that gives specific instructions or questions to the generation AI, and determines the direction of the generated content.

[0879] A "terminal" is a device such as a computer or smartphone used by a user, which communicates with a server to send and receive data.

[0880] A "server" is a computer system that processes and stores data on a network, and provides services in response to requests from client terminals.

[0881] A "virtual influencer" is someone who uses an artificially created character or avatar to act as an influential presence on social media.

[0882] A "platform" is a system that serves as the foundation for providing specific services and applications, and on which users can carry out various activities.

[0883] MODE FOR CARRYING OUT THE INVENTION

[0884] This invention is a system that crawls posts that are rated as interesting on social media and performs additional learning on a generation AI based on that data. A specific embodiment of this system is described below.

[0885] Hardware and software used

[0886] Hardware: High-performance servers (e.g., servers equipped with GPUs), user devices (e.g., smartphones, PCs)

[0887] Software: Generative AI models (e.g., OpenAI GPT-4), social media platforms (e.g., social media APIs)

[0888] System configuration

[0889] 1. Server

[0890] The server has a means to crawl posts that are rated as interesting from social media, thereby automatically collecting trending and popular posts.

[0891] The server has a means to further train the generative AI based on the crawled posts, which improves the generative AI model's ability to generate sentences based on the latest trends.

[0892] The server has a means for the generative AI to generate sentences that it thinks are interesting after additional training. The generative AI model generates new sentences based on prompts from the user.

[0893] 2. Terminal

[0894] The terminal has a means for the user to input a prompt sentence, which can be input using a dedicated application or a web interface.

[0895] The terminal has a means to send prompt text to the server, and the entered prompt text is securely sent to the server using the HTTPS protocol.

[0896] The terminal has a means for displaying the generated sentence to the user, and displays the generated sentence received from the server to the user so that the user can confirm it.

[0897] 3. Users

[0898] The user has a means to post the generated sentence to the SNS. After checking the generated sentence and making corrections as necessary, the user posts it to the SNS.

[0899] Specific examples

[0900] For example, if a user wants to create a viral post on social media, they can input the following prompt into the generative AI model:

[0901] Example prompt sentence:

[0902] "Create a funny post based on a recent trend. For example, cat videos are popular, so I'd like a funny post related to that."

[0903] Based on this prompt, the generative AI model generates the following post:

[0904] "Today's cat was jumping like a ninja! Cats are so funny!"

[0905] The server sends the generated message to the user's device, which displays it to the user. The user can then check the message and post it to the SNS.

[0906] In this way, the present invention allows users to easily create posts that will go viral, and also allows virtual influencers to carry out their activities efficiently. The flow of the identification process in the third embodiment will be described with reference to FIG.

[0907] Step 1:

[0908] The user enters a prompt statement.

[0909] The user enters a prompt using a dedicated application or web interface, for example, "Create an interesting post based on recent trends," and the prompt is sent to the device.

[0910] Step 2:

[0911] The terminal sends the prompt to the server.

[0912] The terminal sends the prompt text entered by the user to the server using the HTTPS protocol. The input data is the prompt text, and the output is a request sent to the server.

[0913] Step 3:

[0914] The server generates sentences using a generative AI model.

[0915] The server inputs the received prompt sentence into the generative AI model to generate an interesting sentence. The generative AI model generates a new sentence based on the results of prior additional learning. The input data is the prompt sentence, and the output is the generated sentence.

[0916] Step 4:

[0917] The server sends the generated statement to the terminal.

[0918] The server sends the generated sentence to the terminal using the HTTPS protocol. The input data is the generated sentence, and the output is the response sent to the terminal.

[0919] Step 5:

[0920] The terminal displays the generated statements to the user.

[0921] The terminal displays the received generated sentence in the text area of ​​the application so that the user can check it. The input data is the generated sentence, and the output is what is displayed to the user.

[0922] Step 6:

[0923] The user posts the generated sentence to a social networking site.

[0924] The user checks the generated sentences displayed, makes corrections as necessary, and then posts them to a social networking site. For example, the user posts using the Twitter API. The input data is the generated sentence, and the output is a post to the social networking site.

[0925] (Application example 3)

[0926] Next, a description will be given of Application Example 3 of Form Example 3. 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."

[0927] In conventional virtual stores, product introductions and promotions tend to be monotonous, making it difficult to attract users' interest. Furthermore, there is a lack of support for social media users to create posts that will create buzz, making effective marketing difficult. To solve these issues, a method is needed to generate interesting text using generative AI and utilize it effectively within the virtual store.

[0928] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[0929] In this invention, the server includes means for crawling posts that are rated as interesting on SNS, means for additionally training a generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for introducing and promoting products in a virtual store using the generation AI, and means for users to view posts by virtual influencers through their smart devices. This enables interesting product introductions and promotions that attract users' interest in the virtual store, and also supports SNS users in creating posts that will go viral.

[0930] "SNS" is an abbreviation for social networking service, an online platform for users to share information and interact over the Internet.

[0931] "Crawling" is a technique for automatically crawling web pages on the Internet and collecting information.

[0932] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new text and content based on learned data.

[0933] "Additive learning" is the process of adding new data to an existing learning model and re-training it.

[0934] A "virtual store" is a virtual store that exists on the Internet and is a platform where users can browse and purchase products online.

[0935] "Promotion" is a marketing activity to promote the sale of a product or service.

[0936] "Smart devices" is a general term for electronic devices that can connect to the Internet and run various applications. Specifically, it includes smartphones and head-mounted displays.

[0937] A "virtual influencer" is a computer-generated character who acts as an influential presence on social media, in virtual stores, and elsewhere.

[0938] "Posting" refers to the act of a user disclosing information or the content of that information on online platforms such as social media and blogs.

[0939] "Going viral" refers to the phenomenon in which a post spreads rapidly on social media and attracts the attention of many people.

[0940] A system for implementing this invention includes a means for crawling posts from social media that are rated as interesting, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are considered interesting, a means for using the generation AI to introduce and promote products in a virtual store, and a means for users to view posts from virtual influencers via a smart device.

[0941] The server first crawls posts from social media that are rated as interesting. This is done using technology that automatically crawls web pages on the Internet and collects information. Next, the generative AI is trained based on the crawled posts. Training is the process of adding new data to an existing learning model and retraining it.

[0942] The generative AI generates sentences that it deems entertaining based on the results of additional learning. This generative AI is implemented using OpenAI's API. The generated sentences are used to introduce and promote products in the virtual store. Specifically, virtual influencers use the generative AI to generate entertaining sentences and catchphrases, enhancing the appeal of the products.

[0943] Users can view posts from virtual influencers through smart devices (e.g., smartphones or head-mounted displays), which allows them to access detailed product information and purchase links.

[0944] As a concrete example, let's consider the case of generating a description for a smartwatch. The server inputs the following prompt text to the generation AI:

[0945] Product Name: Smartwatch

[0946] Features: Heart rate monitoring, GPS, music playback

[0947] Generate an interesting description of this product.

[0948] Based on this prompt, the AI ​​generates a fun introduction like this:

[0949] "When you wear this smartwatch, it will handle everything from heart rate to GPS to music! It's like a gadget from the future, and it will make your everyday life just like a sci-fi movie!"

[0950] The testimonials generated in this way are provided to users by the virtual influencers in the virtual store, which makes it possible to attract users' interest and effectively communicate the appeal of the products.

[0951] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0952] Step 1:

[0953] The server crawls posts that are rated as interesting on social media. It uses the URL or API endpoint of the social media site as input and collects data on posts that are rated as interesting as output. Specifically, the server launches a web crawler, patrols the social media posts, and extracts posts that have received high ratings, such as "likes" and "shares."

[0954] Step 2:

[0955] The server further trains the generative AI based on the crawled posts. The post data collected in step 1 is used as input, and the trained generative AI model is obtained as output. Specifically, the server preprocesses the collected post data and feeds it back to the generative AI model for re-training.

[0956] Step 3:

[0957] The server generates sentences that the generation AI, which has undergone additional training, thinks are interesting. As input, it uses a prompt sentence containing the product name and features, and as output, it obtains a generated interesting introduction sentence. Specifically, the server inputs the following prompt sentence into the generation AI:

[0958] Product Name: Smartwatch

[0959] Features: Heart rate monitoring, GPS, music playback

[0960] Generate an interesting description of this product.

[0961] Step 4:

[0962] The server uses the generated testimonial to introduce and promote the product in the virtual store. It uses the testimonial generated in step 3 as input and obtains promotional content to be displayed in the virtual store as output. Specifically, the server formats the generated testimonial as a post by the virtual influencer and displays it on the interface of the virtual store.

[0963] Step 5:

[0964] Users view posts by virtual influencers through their smart devices. As input, they use the URL and application of the virtual store, and as output, they obtain a screen on which they can view posts by virtual influencers. Specifically, users access the virtual store using a smartphone or head-mounted display and view posts containing generated testimonials.

[0965] 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.

[0966] "Example 1"

[0967] In one embodiment of the present invention, a sentence generated by a generation AI is combined with an emotion engine that recognizes the user's emotions. Specifically, the sentence generated by the generation AI is presented to the user, and the emotion engine analyzes the user's reaction (e.g., facial expression, tone of voice, etc.). The analysis results are used as training data for the generation AI, improving its ability to generate sentences that correspond to the user's emotions.

[0968] "Example 2"

[0969] In another embodiment of the present invention, a generation AI generates a sentence based on the user's emotion recognized by an emotion engine. Specifically, the emotion engine feeds back the user's emotion (e.g., joy or sadness) recognized by the emotion engine to the generation AI, which then generates a sentence appropriate to that emotion. For example, if the generation AI recognizes that the user is feeling joy, it generates a sentence that amplifies the user's joy.

[0970] "Example 3"

[0971] In yet another embodiment of the present invention, the activity of the virtual influencer is adjusted based on the user's emotion recognized by the emotion engine. Specifically, the content, tone, timing, etc. of the virtual influencer's posts are adjusted according to the user's emotion recognized by the emotion engine. For example, if the emotion engine recognizes that the user is feeling sad, the virtual influencer will take an action corresponding to the user's emotion, such as posting comforting content.

[0972] The processing flow of each embodiment will be described below.

[0973] "Example 1"

[0974] Step 1: The generative AI generates sentences to present to the user.

[0975] Step 2: The user reads the sentence generated by the AI.

[0976] Step 3: The emotion engine analyzes the user's reactions (e.g., facial expressions, tone of voice, etc.).

[0977] Step 4: Use the analysis results as training data for the generation AI to improve its ability to generate sentences that correspond to the user's emotions.

[0978] "Example 2"

[0979] Step 1: The emotion engine recognizes the user's emotion (e.g., joy, sadness, etc.).

[0980] Step 2: The recognized emotion is fed back to the generative AI.

[0981] Step 3: The AI ​​generates a sentence that matches the emotion. For example, if the AI ​​recognizes that the user is feeling joy, it generates a sentence that amplifies the user's joy.

[0982] "Example 3"

[0983] Step 1: The emotion engine recognizes the user's emotion.

[0984] Step 2: The content, tone, and timing of the virtual influencer's posts are adjusted based on the user's emotions as identified by the emotion engine.

[0985] Step 3: For example, if the virtual influencer recognizes that the user is feeling sad, it will take action that corresponds to the user's emotions, such as posting comforting messages.

[0986] Example 1

[0987] Next, a description will be given of Example 1 of Form 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."

[0988] With conventional crawling systems for social media posts, it was difficult to efficiently collect posts that were rated as interesting and train a generation AI based on them. Furthermore, there was a lack of a mechanism for analyzing the emotions that generated sentences evoke in users and feeding the results back to the generation AI, making it difficult to improve the accuracy of the generation AI. Furthermore, it was not possible to collect and analyze user reactions in real time to use them as training data for the generation AI.

[0989] 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.

[0990] In this invention, the server includes: means for crawling posts rated as interesting from social media; means for additionally training a generation AI based on the crawled posts; means for the additionally trained generation AI to generate sentences that are deemed interesting; means for presenting the generated sentences to users; means for collecting user responses; means including an emotion engine for analyzing the collected user responses; and means for feeding back the analysis results as learning data for the generation AI. This makes it possible to efficiently collect posts rated as interesting and improve the accuracy of the generation AI. Furthermore, by analyzing user emotions in real time and feeding back the results to the generation AI, it is possible to generate sentences that are more suitable for the user.

[0991] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[0992] "Crawling" is the process of automatically collecting information on the Internet and obtaining data based on specific criteria.

[0993] "Generative AI" refers to systems that use artificial intelligence techniques to generate new text or content, particularly in natural language processing.

[0994] "Additive training" is the process of adding new data to an existing machine learning model and retraining it, in order to improve the model's accuracy and performance.

[0995] The term "user" refers to an individual or organization that uses a system or service, and in the present invention particularly refers to a recipient of a generated sentence.

[0996] The "emotion engine" is a system that analyzes data such as the user's facial expressions and tone of voice to recognize the user's emotions.

[0997] "Feedback" is the process of using the system's output or analysis results as input again to improve the system's performance and accuracy.

[0998] A "virtual influencer" is a fictitious character created using artificial intelligence and animation technology who has influence on social media and other online platforms.

[0999] "Platform" refers to a system or environment that serves as the foundation for providing specific services or functions, and on which users can operate.

[1000] This invention is a system that crawls posts rated as interesting on social media, trains a generation AI based on those posts, presents the generated sentences to users, and analyzes user reactions to improve the accuracy of the generation AI. A specific embodiment of this system is described below.

[1001] The server accesses social media platforms and selects posts based on specific keywords, hashtags, and the number of reactions. For example, it crawls posts with the hashtag "funny" or posts with a certain number of "likes." The server then stores these posts in a database.

[1002] As a specific example, the server uses a social networking service's API to search for posts with the hashtag "funny" and crawl posts with more than 100 likes. The server also uses a social networking service's API to search for posts containing a specific keyword (e.g., "funny") and crawl posts with more than 50 comments.

[1003] Example prompt sentence:

[1004] "Using a social media API, create a program that crawls posts with interesting hashtags and retrieves posts with more than 100 likes."

[1005] The server then uses a generative AI model to generate new sentences based on the crawled posts, which are then presented to the user, who can then display the sentences or read them aloud using speech synthesis.

[1006] The user sees and hears the generated sentences. The device's camera and microphone are used to collect the user's reactions. The device then sends the collected data to an emotion engine, which analyzes the user's reactions. The results of this analysis are used as learning data for the generative AI.

[1007] As a concrete example, a generative AI generates the sentence "Today was a very fun day!" and presents it to the user. The device's camera captures the user's smile, and the microphone records the user's laughter. The emotion engine analyzes this data and recognizes that the user is feeling "joy." The results of this analysis are fed back to the generative AI, which improves its ability to generate sentences that make the user feel joyful from the next time onwards.

[1008] Example prompt sentence:

[1009] "Design a system that presents sentences generated by a generative AI to the user, collects the user's reactions using a camera and microphone, and analyzes them with an emotion engine."

[1010] In this way, the server, device, and user work together to efficiently collect posts that are rated as interesting, improving the accuracy of the generation AI. Also, by analyzing user emotions in real time and feeding the results back to the generation AI, it is possible to generate sentences that are more suitable for the user.

[1011] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1012] Step 1:

[1013] The server accesses the API of the social media platform and searches for posts based on specific keywords, hashtags, and the number of reactions. As input, it uses the API endpoint of the social media platform and search criteria (e.g., "funny" hashtag, 100 or more likes). The server receives the post data retrieved from the API and filters it. As output, it obtains a list of posts that meet the criteria. Specifically, the server sends a request to the API endpoint of the social media platform and retrieves posts that match the criteria.

[1014] Step 2:

[1015] The server stores the filtered posts in a database. As input, it uses the list of posts obtained in step 1. The server executes a SQL query to insert the post text and metadata into the database. As output, it gets the post data stored in the database. Specifically, the server stores each post's text, username, number of likes, etc. in the database.

[1016] Step 3:

[1017] The server generates new sentences using a generative AI model based on posts stored in a database. It uses the stored post data as input. The server sends a prompt to the generative AI model to generate a new sentence. The generated sentence is obtained as output. Specifically, the server sends a prompt to a generative AI model such as GPT-3 saying, "Generate a new sentence based on an interesting post."

[1018] Step 4:

[1019] The terminal presents the generated sentence to the user. It uses the sentence generated in step 3 as input. The terminal displays the generated sentence on the screen or reads it to the user using speech synthesis. As output, it obtains the sentence presented to the user. Specifically, the terminal displays the generated sentence on the display and reads it using a speech synthesis engine.

[1020] Step 5:

[1021] The device collects the user's reactions using a camera and a microphone. The inputs are the user's facial expressions and tone of voice. The device captures the user's facial expressions with the camera and records the user's voice with the microphone. The output is the collected reaction data. Specifically, the device captures the user's smile with the camera and records the user's laughter with the microphone.

[1022] Step 6:

[1023] The device sends the collected reaction data to the emotion engine and analyzes the user's emotions. The reaction data collected in step 5 is used as input. The device inputs the data into the emotion engine and recognizes the user's emotions. The output is the analysis result. Specifically, the device inputs image data and voice data into the emotion engine and recognizes the user's emotions (e.g., joy, surprise).

[1024] Step 7:

[1025] The server feeds back the emotion engine's analysis results as learning data for the generative AI. The analysis results obtained in step 6 are used as input. The server feeds back the analysis results to the generative AI model and reflects them in subsequent sentence generation. The output is an improvement in the accuracy of the generative AI model. Specifically, the server inputs the emotion engine's analysis results into the generative AI model and uses them as learning data for the model.

[1026] (Application example 1)

[1027] Next, a description will be given of Application Example 1 of Form 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."

[1028] In conventional SNS, it was difficult to efficiently collect posts that were rated as interesting and generate content that would attract users' interest based on them. In addition, there was a lack of a mechanism to analyze how the generated content affected users' emotions and reflect the results in the next content generation. This made it difficult to continuously attract users' interest and lacked a feedback loop to improve the quality of the content.

[1029] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for presenting the generated sentences to the user and analyzing the user's emotional response, and means for using the analysis results as training data for the generation AI. This makes it possible to efficiently generate content that attracts the user's interest and improve the quality of the content based on the user's emotional response.

[1030] "SNS" is an abbreviation for social networking service, an online platform for users to interact with each other and share information over the Internet.

[1031] "Crawling" is the process of automatically gathering information on a website, a method of gathering posts based on specific keywords or hashtags.

[1032] "Generative AI" is a system that uses artificial intelligence technology to generate new text and content, and has the ability to generate sentences based on specific training data.

[1033] "Additive training" is the process of retraining an existing generative AI model by adding new data to it in order to improve the model's performance.

[1034] An "emotional response" is an emotional reaction a user has to the content presented to them, and is expressed through facial expressions, tone of voice, etc.

[1035] "Analysis" is the process of examining collected data in detail to find specific patterns and trends.

[1036] "Training data" is a dataset that generative AI uses to generate new content and improve the model's performance.

[1037] "Content" is a general term for information and entertainment provided to users, and includes text, images, videos, etc.

[1038] The system for implementing this invention includes a series of processes: crawling posts rated as interesting on social media, conducting additional learning by a generation AI based on those posts, presenting the generated sentences to the user, and analyzing the user's emotional response. Specific embodiments are described below.

[1039] System Configuration

[1040] The server uses web scraping technology to crawl posts that are rated as interesting from social media. Specifically, it uses the requests library and the web scraping library to collect posts that contain a specific hashtag (e.g., "interesting"). The collected posts are filtered based on the number of "likes."

[1041] Next, the server trains the generative AI based on the crawled posts. Here, we use the GPT-3 model from the transformers library. The generative AI generates new content using the crawled posts as input.

[1042] The generated content is presented on the user's device (smartphone or PC). To analyze the user's emotional response, the device's camera is used to capture facial expressions and analyze them using the cv2 (OpenCV) library. The analysis results are processed by an emotion recognition model and sent to the server as training data for the generation AI.

[1043] Hardware and software used

[1044] Hardware: Server, user device (smartphone, PC), camera

[1045] Software: requests library, web scraping library, transformers library, cv2 (OpenCV) library, emotion recognition model

[1046] Specific examples

[1047] For example, the server might input the following prompt to the generator AI:

[1048] Example prompt sentence:

[1049] Generate new content based on this post: "Something really interesting happened today! Interesting and fun"

[1050] The AI ​​then generates new content based on this prompt and presents it to the user. While the user is viewing the content, a camera captures the user's facial expressions, which are then analyzed by an emotion recognition model. The analysis results are sent to the server and reflected in the next content generation.

[1051] In this way, it is possible to efficiently generate content that attracts the user's interest and improve the quality of the content based on the user's emotional response.

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

[1053] Step 1:

[1054] The server crawls posts rated as interesting from social media. Specifically, it uses the requests library to access web pages containing a specific hashtag (e.g., "interesting") and then uses a web scraping library to analyze the posts. The input is the URL of the social media site, and the output is a list of posts rated as interesting.

[1055] Step 2:

[1056] The server further trains the generative AI based on the crawled posts. It uses the GPT-3 model from the transformers library and inputs the crawled posts as prompts. The input is the list of crawled posts, and the output is the generated new content.

[1057] Step 3:

[1058] The server sends the generated content to the user's terminal. The user's terminal displays the generated content. The input is the generated content, and the output is the content displayed on the user's terminal.

[1059] Step 4:

[1060] The user's device uses a camera to capture the user's facial expressions. The captured images are processed using the cv2 (OpenCV) library to analyze the facial expressions. The input is the image captured by the camera, and the output is the analyzed emotion data.

[1061] Step 5:

[1062] The user's device transmits the analyzed emotion data to the server. The input is the analyzed emotion data, and the output is the emotion data transmitted to the server.

[1063] Step 6:

[1064] The server uses the received emotion data as learning data for the generative AI, which is then reflected in the next content generation. The input is emotion data, and the output is an updated generative AI model.

[1065] Example 2

[1066] Next, a description will be given of Example 2 of Form 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."

[1067] Conventional generative AI systems have difficulty generating sentences based on user emotions, and are unable to provide appropriate sentences that reflect the user's emotions. Furthermore, there was a lack of effective ways to train generative AI using posts from social media, resulting in low-quality generated sentences. This led to a decrease in user satisfaction and limited the utility of generative AI.

[1068] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1069] In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for the generation AI to generate sentences based on the user's emotions recognized by the emotion engine, and means for the emotion engine to recognize the user's emotions. This makes it possible to generate appropriate sentences according to the user's emotions and improve user satisfaction.

[1070] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[1071] "Crawling" is the process of automatically visiting web pages on the Internet and gathering information.

[1072] "Generative AI" is a system that uses artificial intelligence technology to automatically generate content such as text and images.

[1073] "Retraining" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's performance.

[1074] An "emotion engine" is a system that recognizes emotions from a user's text, facial expressions, etc. and analyzes those emotions.

[1075] "User" refers to an individual or organization that uses the system and receives the output of the generative AI or emotion engine.

[1076] MODE FOR CARRYING OUT THE INVENTION

[1077] This invention is a system that crawls posts that are rated as interesting on social media and uses those posts for additional learning by the generation AI. Furthermore, it has a function that the emotion engine recognizes the user's emotions and the generation AI generates sentences based on those emotions.

[1078] System program generation

[1079] The server generates a program to crawl posts from social media sites, and the program uses a web scraping tool to collect posts from a specified list of URLs.

[1080] Program processing

[1081] The server saves the crawled posts as text data and formats them as training data for the generative AI. Natural language processing libraries such as NLTK and spaCy are used for text preprocessing. The server then uses the formatted text data for additional training of the generative AI model. This process uses machine learning frameworks such as TensorFlow and PyTorch.

[1082] The device uses software such as OpenCV and Emotion API to recognize emotions from user input and facial expressions. The emotion engine recognizes the user's emotions and feeds that information back to the server. The server inputs the emotion data received from the emotion engine into a generative AI that generates a sentence appropriate to that emotion.

[1083] Specific examples

[1084] As a concrete example, the following prompt sentence can be input to a generative AI model:

[1085] Example prompt sentence:

[1086] 1. "Generate funny jokes based on recent posts."

[1087] 2. "Generate messages that are appropriate for the user's delight."

[1088] Example 1:

[1089] If a user enters the prompt "Generate a funny joke based on recent posts," the server will format the crawled posts as learning data, and the generation AI will generate jokes like the following:

[1090] "Why did computers go to sea? The answer is, they went in search of a network!"

[1091] Example 2:

[1092] When a user inputs the prompt "Please generate a message appropriate for when the user is happy," the device analyzes the user's input text and facial expressions, and the emotion engine recognizes the user's happiness. The server inputs the emotion data received from the emotion engine into the generation AI, which then generates a message like the one below.

[1093] "That's wonderful news! We can feel your joy!"

[1094] In this way, the server and the terminal work together to provide a system that generates appropriate sentences based on crawled posts and user sentiment.

[1095] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1096] Program processing flow

[1097] Step 1: Crawl the posts

[1098] The server crawls posts that are rated as interesting by social media.

[1099] Input: List of social media URLs

[1100] What happens: The server uses a web scraping tool to collect posts from a specified list of URLs.

[1101] Output: Crawled post data

[1102] Step 2: Generate training data

[1103] The server saves the crawled posts as text data and formats them as learning data for the generation AI.

[1104] Input: Crawled post data

[1105] What it does: The server uses natural language processing libraries such as NLTK and spaCy to remove unnecessary HTML tags and special characters from the posted data and clean up the text data.

[1106] Output: Formatted text data

[1107] Step 3: Further training of the generative AI

[1108] The server uses the formatted text data to perform additional training on the generative AI model.

[1109] Input: Formatted text data

[1110] How it works: The server uses machine learning frameworks such as TensorFlow and PyTorch to input new text data into existing generative AI models and update the model's parameters.

[1111] Output: A trained generative AI model

[1112] Step 4: Emotion Recognition with the Emotion Engine

[1113] The device recognizes emotions from the user's input and facial expressions.

[1114] Input: User-entered text and camera footage

[1115] How it works: The device uses software such as OpenCV and Emotion API to analyze the user's input text and camera footage, and the emotion engine identifies the user's emotions. For example, text analysis calculates the frequency of positive words, and video analysis analyzes facial expressions.

[1116] Output: Recognized user emotion data

[1117] Step 5: Sentiment-based sentence generation

[1118] The server feeds back the user's emotions recognized by the emotion engine to the generation AI, which then generates a sentence appropriate to that emotion.

[1119] Input: Recognized user emotion data

[1120] Specific operation: The server inputs the emotion data received from the emotion engine into the generation AI, which then generates a sentence appropriate to that emotion. For example, if the server recognizes that the user is happy, the generation AI will generate a sentence such as, "That's wonderful! I can feel your joy!"

[1121] Output: Sentence generated based on sentiment

[1122] Step 6: Providing generated statements

[1123] The server provides the generated sentences to the user.

[1124] Input: Sentiment-based generated sentence

[1125] Specific operation: The server sends the generated sentence to the user's terminal, allowing the user to view the sentence.

[1126] Output: The generated text provided to the user

[1127] In this way, the server and the terminal work together to create a system that generates appropriate sentences based on crawled posts and user sentiment and provides them to the user.

[1128] (Application example 2)

[1129] Next, a description will be given of Application Example 2 of Form 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."

[1130] Conventional generative AI systems have difficulty generating content based on user emotions, and have been unable to provide content that is appropriate to the user's emotions. In addition, there is a lack of means to train the generative AI based on posts from social media, and the generation of sentences that are rated as interesting has not been sufficiently achieved.

[1131] 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.

[1132] In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for an emotion engine to recognize a user's emotion, and means for the generation AI to generate sentences based on the recognized emotion, thereby making it possible to generate and provide content that is appropriate for the user's emotion.

[1133] "SNS" is an abbreviation for Social Networking Service, an online platform for users to interact with other users over the Internet.

[1134] "Crawling" is a technique for automatically crawling web pages on the Internet and collecting information.

[1135] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new data and content based on given data.

[1136] "Additive training" is the process of adding new data to an existing learning model and retraining it.

[1137] An "emotion engine" is a system that analyzes emotions from a user's text or voice and recognizes those emotions.

[1138] "User emotion" refers to the emotional state, such as joy, sadness, or anger, that a user feels in response to a particular situation or piece of information.

[1139] "Content" means the information and entertainment provided on the Website or Application.

[1140] A "virtual influencer" is a fictitious person created using artificial intelligence and computer graphics who has influence on social media and other online platforms.

[1141] "Platform" refers to the underlying system or environment for providing specific services or applications.

[1142] The embodiments of the present invention will be described in detail below.

[1143] First, the server crawls posts that are rated as interesting from social media. The crawled posts are used for additional learning of the generative AI. The generative AI learns new data based on the crawled posts and improves its ability to generate sentences that are considered interesting.

[1144] Next, the emotion engine recognizes the user's emotion. The emotion engine analyzes emotions from the text or voice input by the user and recognizes their emotional state. For example, if the user inputs "I'm very happy today," the emotion engine recognizes "joy."

[1145] Based on the recognized emotions, the generative AI generates sentences that match the user's emotions. For example, if the user is feeling "joy," the generative AI will generate positive content that amplifies the user's joy. Specifically, it will generate sentences such as, "The weather is very nice today. A new cafe has opened."

[1146] The following hardware and software are used to realize this system. The hardware requires a server and a user device. The software requires a crawler to crawl social media, a generative AI model (e.g., GPT-2), and an emotion engine (e.g., TextBlob).

[1147] For example, if a user types, "I'm very happy today," the emotion engine will recognize "joy," and the generative AI will generate a sentence such as, "Content appropriate for when the user is feeling happy: The weather is very nice today. A new cafe has opened."

[1148] Example prompt sentence:

[1149] Content suitable for delighted users:

[1150] By inputting this prompt into the generative AI, content that matches the user's emotions can be generated.

[1151] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1152] Step 1:

[1153] The server crawls posts from social media that are rated as interesting. Specifically, it uses a crawler to collect posts from social media and selects posts that have received high ratings, such as "likes" and "shares." The input is the social media post data, and the output is a list of posts that are rated as interesting.

[1154] Step 2:

[1155] The server trains the generative AI based on the crawled posts. Specifically, the crawled posts are input into a generative AI model (e.g., GPT-2) and the model is retrained. The input is a list of crawled posts, and the output is the trained generative AI model.

[1156] Step 3:

[1157] The user terminal receives user input. Specifically, the user expresses their emotions through text or voice. The input is the user's text or voice data, and the output is passed directly to the emotion engine.

[1158] Step 4:

[1159] The server uses an emotion engine to recognize the user's emotions. Specifically, it uses an emotion analysis tool such as TextBlob to analyze emotions from the user's input. The input is the user's text or voice data, and the output is the recognized emotion (e.g., joy, sadness).

[1160] Step 5:

[1161] The server uses a generation AI to generate a sentence based on the recognized emotion. Specifically, the recognized emotion is input to the generation AI as a prompt sentence, which generates a sentence appropriate to the emotion. The input is the recognized emotion and the prompt sentence, and the output is the generated sentence.

[1162] Step 6:

[1163] The user terminal displays the generated sentence to the user. Specifically, the generated sentence is displayed on a screen so that the user can view it. The input is the generated sentence, and the output is the content displayed to the user.

[1164] Example 3

[1165] Next, a description will be given of Example 3 of Form Example 3. 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."

[1166] In traditional social media, creating posts that will go viral requires a great deal of time and effort. Virtual influencers are also required to consistently provide fresh and interesting content. However, these tasks are extremely difficult, and there is a lack of efficient methods for doing so. Furthermore, providing appropriate content that reflects users' emotions is also a difficult challenge.

[1167] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[1168] In this invention, the server includes a means for crawling posts rated as interesting from social media, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are considered interesting, a means for a user to input a prompt sentence, a means for inputting the prompt sentence to a generation AI model, a means for the generation AI model to generate sentences, a means for publishing the generated sentences as posts by a virtual influencer, and a means for a user to confirm the generated sentences. This allows users to efficiently create posts that will "go viral," and allows virtual influencers to always provide fresh and interesting content. It also makes it possible to provide appropriate content based on user emotions.

[1169] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[1170] "Crawling" is the process of automatically visiting web pages on the Internet and gathering information.

[1171] "Generative AI" is a system that uses artificial intelligence technology to generate new data and content.

[1172] "Additive learning" is the process of adding new data to an existing learning model and re-training it.

[1173] A "prompt" is text that contains instructions or questions that are input to a generative AI model.

[1174] A "generative AI model" is a specific implementation of generative AI, an algorithm that generates new sentences based on input prompt sentences.

[1175] A "virtual influencer" is an artificially created character who acts as an influential presence on social media.

[1176] "User" means an individual or organization that uses the system.

[1177] A "post" is content such as text, images, and videos that are published on social media.

[1178] A "platform" is a system or environment that serves as the foundation for providing specific services or functions.

[1179] MODE FOR CARRYING OUT THE INVENTION

[1180] This invention is a system that crawls posts on social media that are rated as interesting, and then further trains a generative AI model to generate interesting sentences simply by the user entering a prompt sentence, which are then published as posts by virtual influencers.

[1181] Hardware and software used

[1182] Hardware:

[1183] Server equipped with a high-performance GPU

[1184] software:

[1185] Generative AI models (e.g., GPT-3)

[1186] Emotion engine (e.g. IBM Watson Tone Analyzer)

[1187] Program processing explanation

[1188] The server crawls posts that are rated as interesting from social media. These crawled posts are used for further training of the generative AI model. After further training, the generative AI model generates new sentences based on prompt sentences entered by the user.

[1189] The user accesses the system's interface using a device (e.g., a smartphone or PC) and inputs a prompt sentence, such as "What interesting things happened today?" The server inputs this prompt sentence into the generative AI model, which then generates a new sentence.

[1190] The generated sentences are published by the server on a social networking platform as posts by the virtual influencer, which then become available for followers and other users to view. Users can then check the generated sentences using their devices.

[1191] Specific examples

[1192] Enter the prompt statement:

[1193] A user opens an app on their smartphone and types, "What interesting things happened today?"

[1194] Check the generated statements:

[1195] A user sees a post in the app that says, "Something interesting happened today! That's fun."

[1196] In this way, the server uses the generative AI model to generate interesting sentences in response to user prompts, and supports the creation of posts by virtual influencers and SNS users. The flow of the identification process in Example 3 will be described with reference to FIG. 21.

[1197] Program processing flow

[1198] Step 1:

[1199] The server crawls posts that are rated as interesting by social media.

[1200] Input: SNS URL or API endpoint

[1201] Data processing: Extract the posted content in text format and analyze the evaluation data.

[1202] Output: A list of posts that were rated interesting

[1203] Step 2:

[1204] The server further trains the generative AI model based on the crawled posts.

[1205] Input: A list of posts that were rated as interesting

[1206] Data processing: Feed the post content into the generative AI model and update the model parameters

[1207] Output: A trained generative AI model

[1208] Step 3:

[1209] The user uses the terminal to enter a prompt sentence.

[1210] Input: A prompt entered by the user (e.g., "What interesting things happened today?")

[1211] Data processing: Send the prompt to the server in text format

[1212] Output: Prompt text sent to the server

[1213] Step 4:

[1214] The server inputs the received prompt sentence into the generative AI model.

[1215] Input: The prompt text sent by the user

[1216] Data processing: Input the prompt sentence into the generative AI model to start the sentence generation process.

[1217] Output: The prompt input to the generative AI model

[1218] Step 5:

[1219] The generative AI model generates new sentences based on the prompt sentence.

[1220] Input: The prompt given to the generative AI model

[1221] Data processing: Analyze prompt sentences and generate new sentences based on trained data

[1222] Output: Generated sentence (e.g. "Something interesting happened today! It's interesting.")

[1223] Step 6:

[1224] The server publishes the generated sentences as posts of the virtual influencer.

[1225] Input: Generated statement

[1226] Data processing: Post the generated text using the API of the social media platform

[1227] Output: Virtual influencer posts published on social media

[1228] Step 7:

[1229] The user uses the terminal to review the generated sentences.

[1230] Input: Virtual influencer posts on social media platforms

[1231] Data processing: Displaying the posted content on the device interface

[1232] Output: Generated statements displayed on the terminal

[1233] In this way, the server uses a generative AI model to generate interesting sentences in response to user prompts, helping virtual influencers and social media users create posts.

[1234] (Application example 3)

[1235] Next, a description will be given of Application Example 3 of Form Example 3. 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."

[1236] Conventional virtual influencers and social media posting platforms have had difficulty providing appropriate content that reflects users' emotions. Furthermore, they lacked support for users to create posts that would create buzz, leading to a need for a method to generate effective posts.

[1237] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[1238] In this invention, the server includes means for crawling posts rated as interesting from social media, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for recognizing a user's emotions using an emotion recognition engine, means for adjusting the sentences generated by the generation AI based on the recognized emotions, means for publishing the adjusted sentences as posts by a virtual influencer, and means for providing a platform for users to create their own posts using the generation AI. This makes it possible to provide appropriate content according to user emotions and generate effective "buzzworthy" posts.

[1239] "SNS" is an abbreviation for social networking service, a platform for users to share information and interact online.

[1240] "Crawling" is a technique for automatically collecting information on the Internet, and is a means of obtaining data from specific websites and social media sites.

[1241] "Generative AI" is a type of artificial intelligence technology, an algorithm that generates new text and content based on learned data.

[1242] "Additive training" is the process of adding new data to an existing learning model and retraining it, and is a means of improving the model's performance and accuracy.

[1243] An "emotion recognition engine" is a technology that analyzes emotions from a user's text or voice and recognizes a specific emotional state.

[1244] A "virtual influencer" is a non-existent digital character who acts as an influential presence on social media and other online platforms.

[1245] A "platform" is a system or environment that serves as the foundation for providing specific services or applications.

[1246] "User" refers to an individual or organization that uses the system or service, and in this case refers to a user of the SNS.

[1247] A "post" is content such as text, images, or videos that a user makes public on a social networking site or other online platform.

[1248] "Adjustment" is the process of modifying generated content or system behavior in response to specific conditions or circumstances.

[1249] A system for implementing this invention includes means for crawling posts from social media that are rated as interesting, means for allowing a generation AI to perform additional learning based on the crawled posts, means for the generation AI to generate sentences that are deemed interesting after additional learning, means for recognizing a user's emotions using an emotion recognition engine, means for adjusting the sentences generated by the generation AI based on the recognized emotions, means for publishing the adjusted sentences as posts by a virtual influencer, and means for providing a platform for users to create their own posts using the generation AI.

[1250] System configuration

[1251] This system uses the following hardware and software:

[1252] Hardware: Servers, smartphones

[1253] Software: OpenAI API, EmotionEngine (emotion recognition engine)

[1254] Program processing explanation

[1255] The server crawls posts that are rated as interesting on social media. These crawled posts are used as additional training data for the generation AI. The generation AI then generates new, interesting sentences based on the additional training data.

[1256] When a user accesses the system via their smartphone, the emotion recognition engine analyzes the user's input text and recognizes the user's emotions. Based on the recognized emotions, the sentences generated by the generation AI are adjusted. For example, if the user is feeling sad, the generation AI will generate comforting sentences.

[1257] The adjusted statements are published as posts by virtual influencers, and users can use the platform to create their own posts using generative AI. The platform makes it easy for users to create posts that are intended to go viral.

[1258] Specific examples

[1259] For example, if a user types, "I've been feeling a bit down lately," the emotion recognition engine will recognize the user's emotion as "sadness." Based on this, the generative AI will generate a sentence such as, "I found this great product today! Cheer up!" and publish it as a post by the virtual influencer.

[1260] Prompt Sentence Examples

[1261] If the virtual influencer recognizes that the user is feeling sad, they should generate a comforting post, such as, "I found this great product today! Cheer up!"

[1262] In this way, it becomes possible to provide appropriate content that reflects the user's emotions and to generate effective "buzz" posts.

[1263] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1264] Step 1:

[1265] The server crawls posts rated as interesting from social media. It uses the social media URL or API endpoint as input and obtains the post data rated as interesting as output. Specifically, it uses the social media API to collect posts based on specific hashtags or keywords, and then selects posts with high ratings such as "likes" and "shares."

[1266] Step 2:

[1267] The server further trains the generative AI based on the crawled posts. It uses the crawled post data as input and obtains a trained generative AI model as output. Specifically, it preprocesses the collected post data and supplies it to the generative AI as text data. The generative AI learns new patterns and trends based on this data.

[1268] Step 3:

[1269] The server generates sentences that the trained generative AI finds interesting. It uses the trained generative AI model as input and obtains the generated interesting sentences as output. Specifically, it inputs a prompt sentence to the generative AI and generates new text based on it.

[1270] Step 4:

[1271] The device recognizes the user's emotions using an emotion recognition engine. It uses the user's text input as input and obtains recognized emotion data as output. Specifically, the user provides the input text to the emotion recognition engine, which then analyzes the emotion from the text.

[1272] Step 5:

[1273] The server adjusts the sentences generated by the AI ​​based on the recognized emotions. It uses the recognized emotion data and the generated sentence as input and obtains the adjusted sentence as output. Specifically, it changes the tone and content of the generated sentence according to the emotion data. For example, if the user is feeling sad, it changes the generated sentence to be comforting.

[1274] Step 6:

[1275] The server publishes the adjusted sentence as a post of the virtual influencer. Using the adjusted sentence as input, the server obtains a post published on the SNS as output. Specifically, the adjusted sentence is posted to the SNS from the virtual influencer's account.

[1276] Step 7:

[1277] Users use the platform to create their own posts using generative AI. The platform uses the user's prompt as input and gets the generated post as output. Specifically, the user enters a prompt on the platform, and the generative AI generates a post based on it.

[1278] 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.

[1279] 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> ) 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.

[1280] Another example of generative AI is Gemini (internet search engine). <url: https: gemini.google.com ?hl="ja">) are listed.

[1281] 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.

[1282] [Third embodiment]

[1283] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1284] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[1285] 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).

[1286] 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.

[1287] 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.

[1288] 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).

[1289] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1290] 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.

[1291] 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.

[1292] 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.

[1293] 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.

[1294] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[1295] "Example 1"

[1296] One embodiment of the present invention is a system that includes a means for crawling posts from social media that are rated as interesting. This system selects posts based on specific keywords, hashtags, the number of reactions, etc., and crawls their text. For example, it is conceivable to crawl posts with the hashtag "interesting" or posts with a certain number of "likes."

[1297] "Example 2"

[1298] Next, it is equipped with a means for additional training of the generation AI based on the crawled posts. This AI uses the text of the crawled posts as training data to learn how to generate sentences that will be evaluated as interesting. Specifically, the text of the crawled posts is used as training data for additional training of the existing learning model.

[1299] "Example 3"

[1300] Furthermore, the generative AI, which has undergone additional learning, is equipped with a means to generate sentences that it deems interesting. This generative AI generates new sentences based on the results of its learning. For example, it can generate sentences such as "Something interesting happened today! That's funny" as posts from virtual influencers.

[1301] "Example 4"

[1302] It also provides a means for users to act as virtual influencers using generative AI, and a platform for social media users to create posts that are intended to go viral. Specifically, it publishes interesting sentences created by generative AI as posts by virtual influencers, and provides a platform for social media users to create their own posts using generative AI.

[1303] The processing flow of each embodiment will be described below.

[1304] "Example 1"

[1305] Step 1: The system selects posts from social media based on specific keywords, hashtags, number of reactions, etc. For example, it selects posts with the hashtag "funny" or posts with a certain number of "likes."

[1306] Step 2: The text of the selected posts is crawled. The crawled data is used as training data for the generation AI.

[1307] "Example 2"

[1308] Step 1: Additional training is performed on the generation AI based on the crawled posts. Specifically, additional training is performed on the existing learning model using the text of the crawled posts as training data.

[1309] Step 2: After additional learning, the generation AI generates new sentences based on the results of its learning.

[1310] Step 1: The interesting sentences created by the generative AI are published as posts by virtual influencers.

[1311] Step 2: Provide a platform for social media users to create their own posts using generative AI.

[1312] Example 1

[1313] Next, a description will be given of Example 1 of Form 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."

[1314] There is a demand for a system that can efficiently collect posts that are rated as interesting on social media and automatically generate more attractive content by training a generation AI based on that data. However, with conventional methods, the collection and selection of posts is often done manually, which takes time and effort. In addition, it is difficult to select appropriate posts to use as training data for the generation AI, which results in an issue of inconsistent quality of the generated content.

[1315] 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.

[1316] In this invention, the server includes means for inputting conditions such as specific keywords, hashtags, and the number of reactions, means for accessing the API of the SNS and searching for posts based on the conditions, means for selecting posts that match the conditions from the search results, means for saving the text of the selected posts in a database, means for providing the saved data to the user, means for additional training of the generation AI based on the crawled posts, and means for the additionally trained generation AI to generate sentences that are considered interesting. This makes it possible to efficiently collect posts that are rated as interesting on the SNS and use them as training data for the generation AI, thereby automatically generating more attractive content.

[1317] "Means for entering specific keywords, hashtags, and reaction counts" is a function that provides an interface for users to enter keywords, hashtags, and reaction counts that they specify as search conditions.

[1318] "A means of accessing an SNS's API and searching for posts based on conditions" refers to a function that uses the application programming interface (API) provided by the SNS to search for posts based on conditions specified by the user.

[1319] "Means for selecting posts that match conditions from search results" refers to a function that filters and selects posts that match conditions specified by the user from the search results obtained from the SNS API.

[1320] The "means for saving the text of the selected posts to a database" is a function for storing the text of the filtered posts in a database.

[1321] The "means for providing saved data to users" is a function that provides an interface for displaying posted information saved in the database to users.

[1322] "Means for additional training of the generative AI based on crawled posts" is a function for additional training of the generative AI model using collected post data.

[1323] "Means for a generative AI that has undergone additional training to generate sentences that are deemed interesting" is a function that enables a generative AI model that has undergone additional training to automatically generate sentences that are evaluated as interesting.

[1324] MODE FOR CARRYING OUT THE INVENTION

[1325] This invention is a system that efficiently collects posts that are rated as interesting from social media, and automatically generates more attractive content by training a generation AI based on that data. Specific embodiments of this system are described below.

[1326] 1. Program Generation

[1327] The system's programs are written using programming languages ​​such as Python and JavaScript, and include code to access social networking site APIs and collect posts based on specific criteria.

[1328] 2. Program Processing

[1329] The server performs its processing using the following hardware and software:

[1330] Hardware: The server uses a computer system with a powerful processor and sufficient memory.

[1331] Software: The server uses libraries to access the SNS's API and a database management system (e.g., MySQL).

[1332] The server receives specific keywords, hashtags, and reaction count conditions entered by the user through their device and sends a request to the social networking service's API. For example, if a user specifies the hashtag "funny" and the conditions "more than 100 likes," the server searches for posts based on these conditions.

[1333] The server analyzes the search results returned by the API and selects posts that match the criteria. The selected post text is saved in a database, which stores information such as the post ID, text, hashtags, and number of reactions.

[1334] The server then further trains the generative AI based on the crawled posts. The generative AI model uses an advanced natural language processing model such as GPT-3. After further training, the generative AI automatically generates sentences that it deems interesting.

[1335] 3. Examples of concrete examples and prompts

[1336] As a concrete example, consider the case where you want to use a social media API to crawl posts with the hashtag "interesting." Below is an example of a prompt for that purpose.

[1337] Examples:

[1338] The server uses the social networking site's API to search for posts with the hashtag "funny."

[1339] The server selects posts from the search results that have 100 or more likes.

[1340] The server stores the selected post text in a database.

[1341] Example prompt sentence:

[1342] Create a program that uses a social media API to search for posts with the hashtag "interesting," select posts with more than 100 likes, and save the text of those posts to a database.

[1343] In this way, the server can efficiently crawl posts that are rated as interesting from social media, store them in a database, and use them as learning data for the generative AI, which will enable the AI ​​to automatically generate more attractive content.

[1344] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1345] Program processing flow

[1346] Step 1:

[1347] Users enter conditions such as specific keywords, hashtags, and number of reactions.

[1348] The user inputs search criteria through the device interface, for example, specifying the hashtag "interesting" and the criteria "more than 100 likes."

[1349] Input: Keyword (e.g. "interesting"), reaction count condition (e.g. "100 or more likes")

[1350] Output: The search criteria is sent to the server.

[1351] Step 2:

[1352] The server accesses the SNS API and searches for posts based on the criteria.

[1353] The server accesses the SNS API based on the conditions entered by the user, sends an API request, and searches for posts with the hashtag "interesting."

[1354] Input: Search criteria (e.g. "interesting", reaction count criteria "100 or more likes")

[1355] Data processing: API request generation

[1356] Output: Search results returned from the SNS API

[1357] Step 3:

[1358] The server selects posts that match the search criteria from the search results.

[1359] The server receives the search results returned by the API and selects posts with more than 100 likes. The server extracts information such as the ID, text, and number of reactions from the selected posts.

[1360] Input: Search results from SNS API

[1361] Data processing: Filtering posts that match criteria

[1362] Output: A list of posts that match the criteria

[1363] Step 4:

[1364] The server saves the selected post text to the database.

[1365] The server stores the selected post text in a database, which stores information such as the post ID, text, hashtags, and number of reactions.

[1366] Input: A list of posts that match the criteria

[1367] Data processing: storing submitted information in a database

[1368] Output: Post information saved in the database

[1369] Step 5:

[1370] The server provides the stored data to the user.

[1371] The server provides the user with information about the posts stored in the database, and the user can view a list of the saved posts through the interface of the terminal.

[1372] Input: Post information stored in the database

[1373] Data processing: Providing posted information to users

[1374] Output: A list of posts that the user can see

[1375] Step 6:

[1376] The generation AI further learns based on the posts crawled by the server

[1377] The server uses the crawled post data to further train the generative AI model, which uses an advanced natural language processing model such as GPT-3.

[1378] Input: Post information stored in the database

[1379] Data processing: Additional training for generative AI models

[1380] Output: A trained generative AI model

[1381] Step 7:

[1382] The server performs additional training and the AI ​​generates sentences that it deems interesting.

[1383] The server uses a trained generative AI model to automatically generate sentences that it deems interesting.

[1384] Input: A trained generative AI model

[1385] Data processing: Sentence generation using generative AI

[1386] Output: Funny sentence generated

[1387] In this way, the server can efficiently crawl posts from social media that are rated as interesting, store them in a database, and use them as learning data for the generation AI, making it possible to automatically generate more attractive content.

[1388] (Application example 1)

[1389] Next, a description will be given of Application Example 1 of Form 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."

[1390] Conventional crawling systems for social media posts have had difficulty efficiently collecting and providing users with "interesting" posts that interest them. Furthermore, they lacked the functionality to allow users to customize their feeds to suit their interests, creating a need for an improved user experience. Furthermore, when using generative AI to generate content, it was difficult to accurately generate the "interesting" content users were looking for.

[1391] 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.

[1392] In this invention, the server includes a means for crawling posts rated as interesting from social media, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that it deems interesting, a means for generating a feed for providing the generated sentences to users, and a setting means for users to customize the feed. This makes it possible to efficiently collect "interesting" posts that interest users and provide them to users. It also allows users to customize their feeds to suit their interests, improving the user experience. Furthermore, when generating content using a generation AI, it becomes possible to accurately generate the "interesting" content that users desire.

[1393] "SNS" is an abbreviation for Social Networking Service, an online platform that allows users to share information and interact with other users over the Internet.

[1394] "Crawling" is the process of automatically collecting information on the Internet, and is a technique for obtaining data based on specific keywords or conditions.

[1395] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new information and content based on given data.

[1396] "Additive training" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's accuracy and performance.

[1397] A "feed" is a stream or list of information provided to a user, in which selected content is displayed sequentially based on specific criteria.

[1398] "Configuration means" refers to the interface and functions that allow the user to customize the system's operation and display content.

[1399] A system for implementing this invention includes a means for crawling posts from social media that are rated as interesting, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are deemed interesting, a means for generating a feed to provide the generated sentences to users, and a setting means for users to customize the feed.

[1400] System Program

[1401] The server uses Python and social media APIs to crawl posts that are rated as interesting by social media. Specifically, it accesses the API using social media API credentials and crawls posts based on specific keywords and hashtags. The crawled posts are then filtered based on rating criteria such as the number of likes and retweets.

[1402] Next, the generation AI is trained based on the crawled posts. For the generation AI, an advanced natural language processing model such as GPT-3 is used. The training is carried out using the crawled post data to improve the accuracy and performance of the generation AI.

[1403] After additional learning, the generative AI generates sentences that it deems interesting. These generated sentences are stored on the server as feeds to be provided to users. The feeds are designed to efficiently provide content that users are interested in, and users can customize them to suit their interests.

[1404] Users can browse feeds using devices such as smartphones and PCs and customize their feeds through the settings means, which include an interface for changing the content displayed on the feed and the evaluation criteria.

[1405] Hardware and software used

[1406] Hardware: Servers, smartphones, PCs

[1407] Software: Python, social media APIs, generative AI models (e.g., GPT-3)

[1408] Specific examples

[1409] For example, if a user sets the hashtags "funny" and "hilarious" and crawls posts with more than 100 likes, the server will collect posts based on these conditions and further train the AI. The AI ​​will then generate sentences that it deems funny and display them in the user's feed.

[1410] Prompt Sentence Examples

[1411] Crawl for posts with the hashtags "funny" and "hilarious" and display posts with over 100 likes in your feed.

[1412] In this way, it is possible to efficiently collect and provide "interesting" posts that interest users. It also allows users to customize their feeds to suit their interests, improving the user experience. Furthermore, when generating content using generative AI, it is possible to accurately generate the "interesting" content that users are looking for.

[1413] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1414] Step 1:

[1415] The server crawls posts from social media sites that are rated as interesting. Specifically, it uses Python and social media APIs to collect posts based on specific keywords and hashtags. Evaluation criteria such as keywords, hashtags, and number of likes are given as input. The server searches for posts based on these conditions and outputs the collected post data.

[1416] Step 2:

[1417] The server performs additional training on the generative AI based on the crawled posts. Specifically, the collected post data is input into a generative AI model (e.g., GPT-3) to perform additional training on the model. The crawled post data is given as input, and the trained generative AI model is obtained as output. Through this process, the server improves the accuracy and performance of the generative AI.

[1418] Step 3:

[1419] The server uses a generative AI that has undergone additional training to generate sentences that it deems interesting. Specifically, it inputs a prompt sentence into the generative AI model that has undergone additional training, and outputs the generated sentence. The prompt sentence is given as input, and the generated sentence is obtained as output. The server uses this generated sentence in the next step.

[1420] Step 4:

[1421] The server generates a feed to provide the generated sentences to the user. Specifically, it converts the generated sentences into a feed format and sends it to the user's device. The generated sentences are given as input, and data in feed format is obtained as output. By providing this feed to the user, the server enables the user to efficiently browse content that interests them.

[1422] Step 5:

[1423] Users browse feeds using their devices and customize them through settings. Specifically, users view feeds using smartphones or computers and operate the settings interface to change the content and evaluation criteria of the feed. The user's customization settings are given as input, and a customized feed is obtained as output. Through this process, users can create a feed tailored to their interests.

[1424] Example 2

[1425] Next, a description will be given of Example 2 of Form 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."

[1426] Conventional generative AI models lack the training data to generate sentences that are rated as interesting, resulting in low-quality generated sentences. In addition, the evaluation methods for generated sentences are inadequate, making it difficult to accurately evaluate the performance of generative AI models.

[1427] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes a means for crawling posts on the Internet, a means for cleansing the crawled posts, a means for additionally training the generative AI model using the cleansed text data, and a means for evaluating the additionally trained generative AI model. This enables the generative AI model to generate high-quality sentences that are evaluated as interesting.

[1428] "Means of crawling internet posts" refers to technologies and methods for automatically collecting post data from websites and social networking sites on the internet.

[1429] "Methods for cleansing crawled posts" are techniques and methods for removing unnecessary information from collected post data and normalizing text data.

[1430] "Means for additional training of generative AI models using cleansed text data" refers to techniques and methods for using preprocessed text data to add new knowledge to existing generative AI models and improve their performance.

[1431] "Means for evaluating a trained generative AI model" refers to techniques or methods for measuring the quality of generated sentences in order to evaluate the performance of a trained generative AI model.

[1432] "Means of acting as a virtual character" refers to the technology or method used to create a virtual character using a generative AI model and have that character act on the Internet.

[1433] "Means for providing a platform for internet users to create posts intended to attract attention" refers to technologies and methods that use generative AI models to provide an environment or service that enables internet users to create posts that attract the attention of many people.

[1434] MODE FOR CARRYING OUT THE INVENTION

[1435] The present invention relates to a system for crawling posts on the Internet and additionally training a generative AI model. Specific embodiments of this system are described below.

[1436] Data collection

[1437] The server uses a Python web scraping library to crawl posts on the Internet, collects post data from designated websites and social media sites, and stores it in text files or a database.

[1438] Examples:

[1439] The server retrieves the HTML using requests.get('https: / / example-blog.com') and parses it using web scraping (html.content, 'html.parser'). The post content is extracted from the parsed data and saved in a text file.

[1440] Data Preprocessing

[1441] The server cleanses the collected text data. Specifically, it removes unnecessary HTML tags, deletes special characters, normalizes the data, etc. The server preprocesses the text data using Python's regular expression library (re) and natural language processing libraries (NLTK, spaCy).

[1442] Examples:

[1443] The server removes HTML tags using re.sub('<[^<]+?>', '', text), tokenizes using nltk.word_tokenize(text), and normalizes the text using spacy.load('en_core_web_sm').

[1444] Data training

[1445] The server uses the cleansed text data to further train the generative AI model. The server uses generative AI models such as OpenAI's GPT-3 and GPT-4, and trains the model using the Transformers library.

[1446] Examples:

[1447] The server loads the model and tokenizer using from transformers import GPT2LMHeadModel, GPT2Tokenizer, and starts additional training with model.train(). Use the cleansed text data as training data.

[1448] Evaluating generative AI models

[1449] The server checks the quality of the generated sentences using test data to evaluate the trained generative AI model. The server performs the evaluation using natural language generation evaluation metrics such as the BLEU score and ROUGE score.

[1450] Examples:

[1451] The server calculates the BLEU score by comparing the generated sentence with a reference sentence using from nltk.translate.bleu_score import sentence_bleu, e.g., sentence_bleu([reference], candidate).

[1452] Prompt Sentence Examples

[1453] Examples of prompts to be input to a generative AI model include:

[1454] "Tell us about your fun day."

[1455] "Please tell me a funny story."

[1456] "Tell me about something fun that happened recently."

[1457] Describe a fun time you had with friends.

[1458] In this way, the server further trains the generation AI based on the crawled posts, building a system that learns how to generate sentences that will be evaluated as interesting.

[1459] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1460] Step 1: Data collection

[1461] The server crawls posts on the Internet. As input, it receives the URL of a specified website or social media site. The server uses a Python web scraping library to collect post data from the specified URL. Specifically, it retrieves HTML using requests.get('https: / / example-blog.com') and parses it using web scraping (html.content, 'html.parser'). As output, the collected post data is saved in a text file or database.

[1462] Step 2: Data Preprocessing

[1463] The server cleanses the collected text data. As input, it receives the post data collected in step 1. The server performs tasks such as removing unnecessary HTML tags, deleting special characters, and normalizing the data. Specifically, it removes HTML tags using re.sub('<[^<]+?>', '', text) and tokenizes the text using nltk.word_tokenize(text). It then normalizes the text using spacy.load('en_core_web_sm'). The cleansed text data is obtained as output.

[1464] Step 3: Train the data

[1465] The server uses the cleansed text data to train the generative AI model. As input, it receives the cleansed text data from step 2. The server uses a generative AI model such as OpenAI's GPT-3 or GPT-4, and trains the model using the Transformers library. Specifically, it loads the model and tokenizer using from transformers import GPT2LMHeadModel, GPT2Tokenizer, and starts the additional training with model.train(). The output is the trained generative AI model.

[1466] Step 4: Evaluating the generative AI model

[1467] The server evaluates the trained generative AI model. As input, it receives the trained generative AI model from step 3 and test data. The server performs evaluation using natural language generation evaluation metrics such as the BLEU score and ROUGE score. Specifically, it compares the generated sentence with the reference sentence using from nltk.translate.bleu_score import sentence_bleu and calculates the BLEU score. The evaluation result is obtained as output.

[1468] Step 5: Generate a prompt statement

[1469] The server generates prompt sentences to be input into the generative AI model. As input, it receives requests from the user and system settings. In concrete terms, the user inputs prompt sentences such as "Tell me about a fun day you had" or "Tell me an interesting story." As output, it obtains prompt sentences to be input into the generative AI model.

[1470] (Application example 2)

[1471] Next, a description will be given of Application Example 2 of Form 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."

[1472] Conventional generative AI models lacked the training data needed to generate sentences that would be rated as interesting, making it difficult to generate the high-quality content users demand. Furthermore, they lacked the ability to instantly generate interesting sentences based on keywords or themes entered by users, making it impossible to meet user needs.

[1473] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for generating sentences that the additionally trained generation AI considers to be interesting, means for generating interesting sentences based on keywords and themes entered by the user, and means for displaying the generated sentences to the user. This makes it possible to instantly generate and provide high-quality content that users desire.

[1474] "SNS" is an abbreviation for social networking service, an online platform for users to share information and interact over the Internet.

[1475] "Crawling" is the technique of automatically collecting web pages and posts on the Internet, used by search engines and for data collection purposes.

[1476] "Generative AI" is a system that uses artificial intelligence technology to automatically generate content such as text and images.

[1477] "Retraining" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's performance.

[1478] "Keywords" are specific words or phrases entered by users that serve as instructions for the generative AI to generate content.

[1479] A "theme" is a specific topic or topic entered by the user, which serves as instructions for the generative AI when generating content.

[1480] "User" refers to an individual or organization that uses the system and generates content using generative AI.

[1481] "Content" refers to information such as text and images generated by generative AI and provided to users.

[1482] "Display" is the act of visually presenting the generated content on a user's device.

[1483] A system for implementing the present invention has the following configuration.

[1484] First, the server has a means for crawling posts that are rated as interesting from social media. Specifically, the server runs a crawling program to automatically collect post data from social media platforms on the Internet. This crawling program uses the Python requests library to obtain post data from specified URLs.

[1485] Next, the server has a means to train the generative AI based on the crawled posts. Specifically, the server trains the generative AI model (e.g., GPT-2) using the crawled post data. For this training, the server uses the transformers library, loads a tokenizer and model, tokenizes the post data, and inputs it into the model.

[1486] Furthermore, the server has a means for the generative AI, which has undergone additional training, to generate sentences that it deems interesting. Based on the keywords and themes entered by the user, the generative AI generates interesting sentences. In this generation process, the prompt entered by the user is tokenized using a tokenizer and input into the generative AI model. The generated sentences are then decoded using the tokenizer and displayed to the user.

[1487] Users access the system using a device such as a smartphone and input keywords or themes. For example, if a user inputs the prompt "funny cat stories," the server uses a generative AI model to generate sentences like the following:

[1488] Example prompt sentence:

[1489] Funny cat stories

[1490] Example of generated statement:

[1491] One day, a cat started playing the piano, and its owner, amazed at how perfectly the cat played Beethoven's "Moonlight Sonata," burst into applause.

[1492] In this way, users can instantly generate and enjoy high-quality content.The system is realized using Python programs, the requests library, the transformers library, and a generative AI model (GPT-2).

[1493] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1494] Step 1:

[1495] The server crawls posts that are rated as interesting from social media. Specifically, the server uses the requests library to retrieve post data from the API of the specified social media platform. The input is the SNS API endpoint, and the output is the retrieved post data.

[1496] Step 2:

[1497] The server further trains the generative AI model based on the crawled post data. Specifically, the server uses the transformers library to load the GPT-2 model and tokenizer and tokenize the post data. Then, it further trains the model using the tokenized data. The input is the crawled post data, and the output is the further trained generative AI model.

[1498] Step 3:

[1499] A user accesses the system using a terminal and inputs keywords and themes. Specifically, the user opens the smartphone application and enters a prompt sentence in a text box. The input is the prompt sentence entered by the user, and the output is a request to the server.

[1500] Step 4:

[1501] The server generates funny sentences based on the prompt sentence entered by the user. Specifically, the server tokenizes the prompt sentence using a tokenizer and inputs it into an additionally trained generative AI model. The model outputs the generated token sequence, which the server decodes and converts into a natural language sentence. The input is the prompt sentence entered by the user, and the output is the generated funny sentence.

[1502] Step 5:

[1503] The server displays the generated sentence on the user's device. Specifically, the server sends the generated sentence to the user's smartphone application, which then displays it on the screen. The input is the generated sentence, and the output is the sentence displayed on the user's device.

[1504] Example 3

[1505] Next, a third embodiment of the third embodiment will be described. 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."

[1506] In conventional social media posting, users had to understand trends and think of appropriate content to create posts that would go viral, which took time and effort. Virtual influencers also had to constantly generate new content, which was a significant burden.

[1507] The identification process by the identification processing unit 290 of the data processing device 12 in Example 3 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for the user to input a prompt sentence, means for the terminal to send the prompt sentence to the server, means for the server to generate sentences using the generation AI model, means for the server to send the generated sentences to the terminal, means for the terminal to display the generated sentences to the user, and means for the user to post the generated sentences to SNS. This allows users to easily create posts that will go viral and efficiently carry out virtual influencer activities.

[1508] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[1509] "Crawling" is the process of automatically gathering information on the web, using specific algorithms to crawl through web pages and obtain data.

[1510] "Generative AI" is a system that uses artificial intelligence technology to generate new text and content, particularly utilizing natural language processing technology.

[1511] "Additive training" is the process of adding new data to an existing AI model and retraining it, in order to improve the model's performance and accuracy.

[1512] A "prompt" is input text that gives specific instructions or questions to the generation AI, and determines the direction of the generated content.

[1513] A "terminal" is a device such as a computer or smartphone used by a user, which communicates with a server to send and receive data.

[1514] A "server" is a computer system that processes and stores data on a network, and provides services in response to requests from client terminals.

[1515] A "virtual influencer" is someone who uses an artificially created character or avatar to act as an influential presence on social media.

[1516] A "platform" is a system that serves as the foundation for providing specific services and applications, and on which users can carry out various activities.

[1517] MODE FOR CARRYING OUT THE INVENTION

[1518] This invention is a system that crawls posts that are rated as interesting on social media and performs additional learning on a generation AI based on that data. A specific embodiment of this system is described below.

[1519] Hardware and software used

[1520] Hardware: High-performance servers (e.g., servers equipped with GPUs), user devices (e.g., smartphones, PCs)

[1521] Software: Generative AI models (e.g., OpenAI GPT-4), social media platforms (e.g., social media APIs)

[1522] System configuration

[1523] 1. Server

[1524] The server has a means to crawl posts that are rated as interesting from social media, thereby automatically collecting trending and popular posts.

[1525] The server has a means to further train the generative AI based on the crawled posts, which improves the generative AI model's ability to generate sentences based on the latest trends.

[1526] The server has a means for the generative AI to generate sentences that it thinks are interesting after additional training. The generative AI model generates new sentences based on prompts from the user.

[1527] 2. Terminal

[1528] The terminal has a means for the user to input a prompt sentence, which can be input using a dedicated application or a web interface.

[1529] The terminal has a means to send prompt text to the server, and the entered prompt text is securely sent to the server using the HTTPS protocol.

[1530] The terminal has a means for displaying the generated sentence to the user, and displays the generated sentence received from the server to the user so that the user can confirm it.

[1531] 3. Users

[1532] The user has a means to post the generated sentence to the SNS. After checking the generated sentence and making corrections as necessary, the user posts it to the SNS.

[1533] Specific examples

[1534] For example, if a user wants to create a viral post on social media, they can input the following prompt into the generative AI model:

[1535] Example prompt sentence:

[1536] "Create a funny post based on a recent trend. For example, cat videos are popular, so I'd like a funny post related to that."

[1537] Based on this prompt, the generative AI model generates the following post:

[1538] "Today's cat was jumping like a ninja! Cats are so funny!"

[1539] The server sends the generated message to the user's device, which displays it to the user. The user can then check the message and post it to the SNS.

[1540] In this way, the present invention allows users to easily create posts that will go viral, and also allows virtual influencers to carry out their activities efficiently. The flow of the identification process in the third embodiment will be described with reference to FIG.

[1541] Step 1:

[1542] The user enters a prompt statement.

[1543] The user enters a prompt using a dedicated application or web interface, for example, "Create an interesting post based on recent trends," and the prompt is sent to the device.

[1544] Step 2:

[1545] The terminal sends the prompt to the server.

[1546] The terminal sends the prompt text entered by the user to the server using the HTTPS protocol. The input data is the prompt text, and the output is a request sent to the server.

[1547] Step 3:

[1548] The server generates sentences using a generative AI model.

[1549] The server inputs the received prompt sentence into the generative AI model to generate an interesting sentence. The generative AI model generates a new sentence based on the results of prior additional learning. The input data is the prompt sentence, and the output is the generated sentence.

[1550] Step 4:

[1551] The server sends the generated statement to the terminal.

[1552] The server sends the generated sentence to the terminal using the HTTPS protocol. The input data is the generated sentence, and the output is the response sent to the terminal.

[1553] Step 5:

[1554] The terminal displays the generated statements to the user.

[1555] The terminal displays the received generated sentence in the text area of ​​the application so that the user can check it. The input data is the generated sentence, and the output is what is displayed to the user.

[1556] Step 6:

[1557] The user posts the generated sentence to a social networking site.

[1558] The user checks the generated sentences displayed, makes corrections as necessary, and then posts them to a social networking site. For example, the user posts using the Twitter API. The input data is the generated sentence, and the output is a post to the social networking site.

[1559] (Application example 3)

[1560] Next, a description will be given of Application Example 3 of Form Example 3. 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."

[1561] In conventional virtual stores, product introductions and promotions tend to be monotonous, making it difficult to attract users' interest. Furthermore, there is a lack of support for social media users to create posts that will create buzz, making effective marketing difficult. To solve these issues, a method is needed to generate interesting text using generative AI and utilize it effectively within the virtual store.

[1562] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[1563] In this invention, the server includes means for crawling posts that are rated as interesting on SNS, means for additionally training a generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for introducing and promoting products in a virtual store using the generation AI, and means for users to view posts by virtual influencers through their smart devices. This enables interesting product introductions and promotions that attract users' interest in the virtual store, and also supports SNS users in creating posts that will go viral.

[1564] "SNS" is an abbreviation for social networking service, an online platform for users to share information and interact over the Internet.

[1565] "Crawling" is a technique for automatically crawling web pages on the Internet and collecting information.

[1566] "Generative AI" is a type of artificial intelligence, a system that has the ability to generate new text and content based on learned data.

[1567] "Additive learning" is the process of adding new data to an existing learning model and re-training it.

[1568] A "virtual store" is a virtual store that exists on the Internet and is a platform where users can browse and purchase products online.

[1569] "Promotion" is a marketing activity to promote the sale of a product or service.

[1570] "Smart devices" is a general term for electronic devices that can connect to the Internet and run various applications. Specifically, it includes smartphones and head-mounted displays.

[1571] A "virtual influencer" is a computer-generated character who acts as an influential presence on social media, in virtual stores, and elsewhere.

[1572] "Posting" refers to the act of a user disclosing information or the content of that information on online platforms such as social media and blogs.

[1573] "Going viral" refers to the phenomenon in which a post spreads rapidly on social media and attracts the attention of many people.

[1574] A system for implementing this invention includes a means for crawling posts from social media that are rated as interesting, a means for additionally training a generation AI based on the crawled posts, a means for the additionally trained generation AI to generate sentences that are considered interesting, a means for using the generation AI to introduce and promote products in a virtual store, and a means for users to view posts from virtual influencers via a smart device.

[1575] The server first crawls posts from social media that are rated as interesting. This is done using technology that automatically crawls web pages on the Internet and collects information. Next, the generative AI is trained based on the crawled posts. Training is the process of adding new data to an existing learning model and retraining it.

[1576] The generative AI generates sentences that it deems entertaining based on the results of additional learning. This generative AI is implemented using OpenAI's API. The generated sentences are used to introduce and promote products in the virtual store. Specifically, virtual influencers use the generative AI to generate entertaining sentences and catchphrases, enhancing the appeal of the products.

[1577] Users can view posts from virtual influencers through smart devices (e.g., smartphones or head-mounted displays), which allows them to access detailed product information and purchase links.

[1578] As a concrete example, let's consider the case of generating a description for a smartwatch. The server inputs the following prompt text to the generation AI:

[1579] Product Name: Smartwatch

[1580] Features: Heart rate monitoring, GPS, music playback

[1581] Generate an interesting description of this product.

[1582] Based on this prompt, the AI ​​generates a fun introduction like this:

[1583] "When you wear this smartwatch, it will handle everything from heart rate to GPS to music! It's like a gadget from the future, and it will make your everyday life just like a sci-fi movie!"

[1584] The testimonials generated in this way are provided to users by the virtual influencers in the virtual store, which makes it possible to attract users' interest and effectively communicate the appeal of the products.

[1585] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1586] Step 1:

[1587] The server crawls posts that are rated as interesting on social media. It uses the URL or API endpoint of the social media site as input and collects data on posts that are rated as interesting as output. Specifically, the server launches a web crawler, patrols the social media posts, and extracts posts that have received high ratings, such as "likes" and "shares."

[1588] Step 2:

[1589] The server further trains the generative AI based on the crawled posts. The post data collected in step 1 is used as input, and the trained generative AI model is obtained as output. Specifically, the server preprocesses the collected post data and feeds it back to the generative AI model for re-training.

[1590] Step 3:

[1591] The server generates sentences that the generation AI, which has undergone additional training, thinks are interesting. As input, it uses a prompt sentence containing the product name and features, and as output, it obtains a generated interesting introduction sentence. Specifically, the server inputs the following prompt sentence into the generation AI:

[1592] Product Name: Smartwatch

[1593] Features: Heart rate monitoring, GPS, music playback

[1594] Generate an interesting description of this product.

[1595] Step 4:

[1596] The server uses the generated testimonial to introduce and promote the product in the virtual store. It uses the testimonial generated in step 3 as input and obtains promotional content to be displayed in the virtual store as output. Specifically, the server formats the generated testimonial as a post by the virtual influencer and displays it on the interface of the virtual store.

[1597] Step 5:

[1598] Users view posts by virtual influencers through their smart devices. As input, they use the URL and application of the virtual store, and as output, they obtain a screen on which they can view posts by virtual influencers. Specifically, users access the virtual store using a smartphone or head-mounted display and view posts containing generated testimonials.

[1599] 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.

[1600] "Example 1"

[1601] In one embodiment of the present invention, a sentence generated by a generation AI is combined with an emotion engine that recognizes the user's emotions. Specifically, the sentence generated by the generation AI is presented to the user, and the emotion engine analyzes the user's reaction (e.g., facial expression, tone of voice, etc.). The analysis results are used as training data for the generation AI, improving its ability to generate sentences that correspond to the user's emotions.

[1602] "Example 2"

[1603] In another embodiment of the present invention, a generation AI generates a sentence based on the user's emotion recognized by an emotion engine. Specifically, the emotion engine feeds back the user's emotion (e.g., joy or sadness) recognized by the emotion engine to the generation AI, which then generates a sentence appropriate to that emotion. For example, if the generation AI recognizes that the user is feeling joy, it generates a sentence that amplifies the user's joy.

[1604] "Example 3"

[1605] In yet another embodiment of the present invention, the activity of the virtual influencer is adjusted based on the user's emotion recognized by the emotion engine. Specifically, the content, tone, timing, etc. of the virtual influencer's posts are adjusted according to the user's emotion recognized by the emotion engine. For example, if the emotion engine recognizes that the user is feeling sad, the virtual influencer will take an action corresponding to the user's emotion, such as posting comforting content.

[1606] The processing flow of each embodiment will be described below.

[1607] "Example 1"

[1608] Step 1: The generative AI generates sentences to present to the user.

[1609] Step 2: The user reads the sentence generated by the AI.

[1610] Step 3: The emotion engine analyzes the user's reactions (e.g., facial expressions, tone of voice, etc.).

[1611] Step 4: Use the analysis results as training data for the generation AI to improve its ability to generate sentences that correspond to the user's emotions.

[1612] "Example 2"

[1613] Step 1: The emotion engine recognizes the user's emotion (e.g., joy, sadness, etc.).

[1614] Step 2: The recognized emotion is fed back to the generative AI.

[1615] Step 3: The AI ​​generates a sentence that matches the emotion. For example, if the AI ​​recognizes that the user is feeling joy, it generates a sentence that amplifies the user's joy.

[1616] "Example 3"

[1617] Step 1: The emotion engine recognizes the user's emotion.

[1618] Step 2: The content, tone, and timing of the virtual influencer's posts are adjusted based on the user's emotions as identified by the emotion engine.

[1619] Step 3: For example, if the virtual influencer recognizes that the user is feeling sad, it will take action that corresponds to the user's emotions, such as posting comforting messages.

[1620] Example 1

[1621] Next, a description will be given of Example 1 of Form 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."

[1622] With conventional crawling systems for social media posts, it was difficult to efficiently collect posts that were rated as interesting and train a generation AI based on them. Furthermore, there was a lack of a mechanism for analyzing the emotions that generated sentences evoke in users and feeding the results back to the generation AI, making it difficult to improve the accuracy of the generation AI. Furthermore, it was not possible to collect and analyze user reactions in real time to use them as training data for the generation AI.

[1623] 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.

[1624] In this invention, the server includes: means for crawling posts rated as interesting from social media; means for additionally training a generation AI based on the crawled posts; means for the additionally trained generation AI to generate sentences that are deemed interesting; means for presenting the generated sentences to users; means for collecting user responses; means including an emotion engine for analyzing the collected user responses; and means for feeding back the analysis results as learning data for the generation AI. This makes it possible to efficiently collect posts rated as interesting and improve the accuracy of the generation AI. Furthermore, by analyzing user emotions in real time and feeding back the results to the generation AI, it is possible to generate sentences that are more suitable for the user.

[1625] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[1626] "Crawling" is the process of automatically collecting information on the Internet and obtaining data based on specific criteria.

[1627] "Generative AI" refers to systems that use artificial intelligence techniques to generate new text or content, particularly in natural language processing.

[1628] "Additive training" is the process of adding new data to an existing machine learning model and retraining it, in order to improve the model's accuracy and performance.

[1629] The term "user" refers to an individual or organization that uses a system or service, and in the present invention particularly refers to a recipient of a generated sentence.

[1630] The "emotion engine" is a system that analyzes data such as the user's facial expressions and tone of voice to recognize the user's emotions.

[1631] "Feedback" is the process of using the system's output or analysis results as input again to improve the system's performance and accuracy.

[1632] A "virtual influencer" is a fictitious character created using artificial intelligence and animation technology who has influence on social media and other online platforms.

[1633] "Platform" refers to a system or environment that serves as the foundation for providing specific services or functions, and on which users can operate.

[1634] This invention is a system that crawls posts rated as interesting on social media, trains a generation AI based on those posts, presents the generated sentences to users, and analyzes user reactions to improve the accuracy of the generation AI. A specific embodiment of this system is described below.

[1635] The server accesses social media platforms and selects posts based on specific keywords, hashtags, and the number of reactions. For example, it crawls posts with the hashtag "funny" or posts with a certain number of "likes." The server then stores these posts in a database.

[1636] As a specific example, the server uses a social networking service's API to search for posts with the hashtag "funny" and crawl posts with more than 100 likes. The server also uses a social networking service's API to search for posts containing a specific keyword (e.g., "funny") and crawl posts with more than 50 comments.

[1637] Example prompt sentence:

[1638] "Using a social media API, create a program that crawls posts with interesting hashtags and retrieves posts with more than 100 likes."

[1639] The server then uses a generative AI model to generate new sentences based on the crawled posts, which are then presented to the user, who can then display the sentences or read them aloud using speech synthesis.

[1640] The user sees and hears the generated sentences. The device's camera and microphone are used to collect the user's reactions. The device then sends the collected data to an emotion engine, which analyzes the user's reactions. The results of this analysis are used as learning data for the generative AI.

[1641] As a concrete example, a generative AI generates the sentence "Today was a very fun day!" and presents it to the user. The device's camera captures the user's smile, and the microphone records the user's laughter. The emotion engine analyzes this data and recognizes that the user is feeling "joy." The results of this analysis are fed back to the generative AI, which improves its ability to generate sentences that make the user feel joyful from the next time onwards.

[1642] Example prompt sentence:

[1643] "Design a system that presents sentences generated by a generative AI to the user, collects the user's reactions using a camera and microphone, and analyzes them with an emotion engine."

[1644] In this way, the server, device, and user work together to efficiently collect posts that are rated as interesting, improving the accuracy of the generation AI. Also, by analyzing user emotions in real time and feeding the results back to the generation AI, it is possible to generate sentences that are more suitable for the user.

[1645] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1646] Step 1:

[1647] The server accesses the API of the social media platform and searches for posts based on specific keywords, hashtags, and the number of reactions. As input, it uses the API endpoint of the social media platform and search criteria (e.g., "funny" hashtag, 100 or more likes). The server receives the post data retrieved from the API and filters it. As output, it obtains a list of posts that meet the criteria. Specifically, the server sends a request to the API endpoint of the social media platform and retrieves posts that match the criteria.

[1648] Step 2:

[1649] The server stores the filtered posts in a database. As input, it uses the list of posts obtained in step 1. The server executes a SQL query to insert the post text and metadata into the database. As output, it gets the post data stored in the database. Specifically, the server stores each post's text, username, number of likes, etc. in the database.

[1650] Step 3:

[1651] The server generates new sentences using a generative AI model based on posts stored in a database. It uses the stored post data as input. The server sends a prompt to the generative AI model to generate a new sentence. The generated sentence is obtained as output. Specifically, the server sends a prompt to a generative AI model such as GPT-3 saying, "Generate a new sentence based on an interesting post."

[1652] Step 4:

[1653] The terminal presents the generated sentence to the user. It uses the sentence generated in step 3 as input. The terminal displays the generated sentence on the screen or reads it to the user using speech synthesis. As output, it obtains the sentence presented to the user. Specifically, the terminal displays the generated sentence on the display and reads it using a speech synthesis engine.

[1654] Step 5:

[1655] The device collects the user's reactions using a camera and a microphone. The inputs are the user's facial expressions and tone of voice. The device captures the user's facial expressions with the camera and records the user's voice with the microphone. The output is the collected reaction data. Specifically, the device captures the user's smile with the camera and records the user's laughter with the microphone.

[1656] Step 6:

[1657] The device sends the collected reaction data to the emotion engine and analyzes the user's emotions. The reaction data collected in step 5 is used as input. The device inputs the data into the emotion engine and recognizes the user's emotions. The output is the analysis result. Specifically, the device inputs image data and voice data into the emotion engine and recognizes the user's emotions (e.g., joy, surprise).

[1658] Step 7:

[1659] The server feeds back the emotion engine's analysis results as learning data for the generative AI. The analysis results obtained in step 6 are used as input. The server feeds back the analysis results to the generative AI model and reflects them in subsequent sentence generation. The output is an improvement in the accuracy of the generative AI model. Specifically, the server inputs the emotion engine's analysis results into the generative AI model and uses them as learning data for the model.

[1660] (Application example 1)

[1661] Next, a description will be given of Application Example 1 of Form 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."

[1662] In conventional SNS, it was difficult to efficiently collect posts that were rated as interesting and generate content that would attract users' interest based on them. In addition, there was a lack of a mechanism to analyze how the generated content affected users' emotions and reflect the results in the next content generation. This made it difficult to continuously attract users' interest and lacked a feedback loop to improve the quality of the content.

[1663] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additional training of the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for presenting the generated sentences to the user and analyzing the user's emotional response, and means for using the analysis results as training data for the generation AI. This makes it possible to efficiently generate content that attracts the user's interest and improve the quality of the content based on the user's emotional response.

[1664] "SNS" is an abbreviation for social networking service, an online platform for users to interact with each other and share information over the Internet.

[1665] "Crawling" is the process of automatically gathering information on a website, a method of gathering posts based on specific keywords or hashtags.

[1666] "Generative AI" is a system that uses artificial intelligence technology to generate new text and content, and has the ability to generate sentences based on specific training data.

[1667] "Additive training" is the process of retraining an existing generative AI model by adding new data to it in order to improve the model's performance.

[1668] An "emotional response" is an emotional reaction a user has to the content presented to them, and is expressed through facial expressions, tone of voice, etc.

[1669] "Analysis" is the process of examining collected data in detail to find specific patterns and trends.

[1670] "Training data" is a dataset that generative AI uses to generate new content and improve the model's performance.

[1671] "Content" is a general term for information and entertainment provided to users, and includes text, images, videos, etc.

[1672] The system for implementing this invention includes a series of processes: crawling posts rated as interesting on social media, conducting additional learning by a generation AI based on those posts, presenting the generated sentences to the user, and analyzing the user's emotional response. Specific embodiments are described below.

[1673] System Configuration

[1674] The server uses web scraping technology to crawl posts that are rated as interesting from social media. Specifically, it uses the requests library and the web scraping library to collect posts that contain a specific hashtag (e.g., "interesting"). The collected posts are filtered based on the number of "likes."

[1675] Next, the server trains the generative AI based on the crawled posts. Here, we use the GPT-3 model from the transformers library. The generative AI generates new content using the crawled posts as input.

[1676] The generated content is presented on the user's device (smartphone or PC). To analyze the user's emotional response, the device's camera is used to capture facial expressions and analyze them using the cv2 (OpenCV) library. The analysis results are processed by an emotion recognition model and sent to the server as training data for the generation AI.

[1677] Hardware and software used

[1678] Hardware: Server, user device (smartphone, PC), camera

[1679] Software: requests library, web scraping library, transformers library, cv2 (OpenCV) library, emotion recognition model

[1680] Specific examples

[1681] For example, the server might input the following prompt to the generator AI:

[1682] Example prompt sentence:

[1683] Generate new content based on this post: "Something really interesting happened today! Interesting and fun"

[1684] The AI ​​then generates new content based on this prompt and presents it to the user. While the user is viewing the content, a camera captures the user's facial expressions, which are then analyzed by an emotion recognition model. The analysis results are sent to the server and reflected in the next content generation.

[1685] In this way, it is possible to efficiently generate content that attracts the user's interest and improve the quality of the content based on the user's emotional response.

[1686] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1687] Step 1:

[1688] The server crawls posts rated as interesting from social media. Specifically, it uses the requests library to access web pages containing a specific hashtag (e.g., "interesting") and then uses a web scraping library to analyze the posts. The input is the URL of the social media site, and the output is a list of posts rated as interesting.

[1689] Step 2:

[1690] The server further trains the generative AI based on the crawled posts. It uses the GPT-3 model from the transformers library and inputs the crawled posts as prompts. The input is the list of crawled posts, and the output is the generated new content.

[1691] Step 3:

[1692] The server sends the generated content to the user's terminal. The user's terminal displays the generated content. The input is the generated content, and the output is the content displayed on the user's terminal.

[1693] Step 4:

[1694] The user's device uses a camera to capture the user's facial expressions. The captured images are processed using the cv2 (OpenCV) library to analyze the facial expressions. The input is the image captured by the camera, and the output is the analyzed emotion data.

[1695] Step 5:

[1696] The user's device transmits the analyzed emotion data to the server. The input is the analyzed emotion data, and the output is the emotion data transmitted to the server.

[1697] Step 6:

[1698] The server uses the received emotion data as learning data for the generative AI, which is then reflected in the next content generation. The input is emotion data, and the output is an updated generative AI model.

[1699] Example 2

[1700] Next, a description will be given of Example 2 of Form 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."

[1701] Conventional generative AI systems have difficulty generating sentences based on user emotions, and are unable to provide appropriate sentences that reflect the user's emotions. Furthermore, there was a lack of effective ways to train generative AI using posts from social media, resulting in low-quality generated sentences. This led to a decrease in user satisfaction and limited the utility of generative AI.

[1702] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1703] In this invention, the server includes means for crawling posts rated as interesting from SNS, means for additionally training the generation AI based on the crawled posts, means for the additionally trained generation AI to generate sentences that are considered interesting, means for the generation AI to generate sentences based on the user's emotions recognized by the emotion engine, and means for the emotion engine to recognize the user's emotions. This makes it possible to generate appropriate sentences according to the user's emotions and improve user satisfaction.

[1704] "SNS" is an abbreviation for Social Networking Service, an online platform for users to share information and interact over the Internet.

[1705] "Crawling" is the process of automatically visiting web pages on the Internet and gathering information.

[1706] "Generative AI" is a system that uses artificial intelligence technology to automatically generate content such as text and images.

[1707] "Retraining" is the process of adding new data to an existing learning model and retraining it, in order to improve the model's performance.

[1708] An "emotion engine" is a system that recognizes emotions from a user's text, facial expressions, etc. and analyzes those emotions.

[1709] "User" refers to an individual or organization that uses the system and receives the output of the generative AI or emotion engine.

[1710] MODE FOR CARRYING OUT THE INVENTION

[1711] This invention is a system that crawls posts that are rated as interesting on social media and uses those posts for additional learning by the generation AI. Furthermore, it has a function that the emotion engine recognizes the user's emotions and the generation AI generates sentences based on those emotions.

[1712] System program generation

[1713] The server generates a program to crawl posts from social media sites, and the program uses a web scraping tool to collect posts from a specified list of URLs.

[1714] Program processing

[1715] The server saves the crawled posts as text data and formats them as training data for the generative AI. Natural language processing libraries such as NLTK and spaCy are used for text preprocessing. The server then uses the formatted text data for additional training of the generative AI model. This process uses machine learning frameworks such as TensorFlow and PyTorch.

[1716] The device uses software such as OpenCV and Emotion API to recognize emotions from user input and facial expressions. The emotion engine recognizes the user's emotions and feeds that information back to the server. The server inputs the emotion data received from the emotion engine into a generative AI that generates a sentence appropriate to that emotion.

[1717] Specific examples

[1718] As a concrete example, the following prompt sentence can be input to a generative AI model:

[1719] Example prompt sentence:

[1720] 1. "Generate funny jokes based on recent posts."

[1721] 2. "Generate messages that are appropriate for the user's delight."

[1722] Example 1:

[1723] If a user enters the prompt "Generate a funny joke based on recent posts," the server will format the crawled posts as learning data, and the generation AI will generate jokes like the following:

[1724] "Why did computers go to sea? The answer is, they went in search of a network!"

[1725] Example 2:

[1726] When a user inputs the prompt "Please generate a message appropriate for when the user is happy," the device analyzes the user's input text and facial expressions, and the emotion engine recognizes the user's happiness. The server inputs the emotion data received from the emotion engine into the generation AI, which then generates a message like the one below.

[1727] "That's wonderful news! We can feel your joy!"

[1728] In this way, the server and the terminal work together to provide a system that generates appropriate sentences based on crawled posts and user sentiment.

[1729] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1730] Program processing flow

[1731] Step 1: Crawl the posts

[1732] The server crawls posts that are rated as interesting by social media.

[1733] Input: List of social media URLs

[1734] What happens: The server uses a web scraping tool to collect posts from a specified list of URLs.

[1735] Output: Crawled post data

[1736] Step 2: Generate training data

[1737] The server saves the crawled posts as text data and formats them as learning data for the generation AI.

[1738] Input: Crawled post data

[1739] What it does: The server uses natural language processing libraries such as NLTK and spaCy to remove unnecessary HTML tags and special characters from the posted data and clean up the text data.

[1740] Output:...

Claims

1. A means for recognizing a user's emotion using an emotion engine; A means for generating a post based on a prompt sentence corresponding to the user's emotion by using a generation AI that has been additionally trained based on posts extracted based on a predetermined condition including at least one of a specific keyword, a specific hashtag, and the number of reactions; A means for outputting the generated posted text; A system including:

2. The output means publishes the generated posted text as a post of a fictitious person. The system of claim 1 .

3. A means for presenting the generated post to the user and analyzing an emotional response of the user; A means for using the analysis result of the user's emotional reaction as learning data for the generation AI. The system of claim 1 .

Images