system

Abstract

Provide a system. A method for collecting word-of-mouth data from multiple sources; A means for preprocessing the collected review data; means for evaluating the pre-processed review data using a generative artificial intelligence model to generate a credibility score; A way to filter bad reviews based on their trustworthiness score; A means of storing filtered and reliable review data; a means for providing stored reliable word-of-mouth data in response to a user's search request; A system including:

Description

[Technical Field]

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

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

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

[0004] In recent years, a large amount of word-of-mouth information has been circulating on the Internet, but much of it contains slanderous or false information, and even stealth marketing. Such malicious word-of-mouth can lead ordinary users to make incorrect decisions when selecting products or using services, potentially resulting in harm. Furthermore, obtaining reliable information requires the time-consuming and tedious scrutiny of many reviews. Therefore, there is a need for a system that can automatically filter out malicious reviews and display only reliable reviews. [Means for solving the problem]

[0005] The present invention provides a system including: means for collecting word-of-mouth data from multiple information sources; means for preprocessing the collected word-of-mouth data; means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score; means for filtering malicious word-of-mouth based on the reliability score; means for saving the filtered reliable word-of-mouth data; and means for providing the saved reliable word-of-mouth data in response to a user's search request. This system automatically filters out malicious word-of-mouth, such as slander and stealth marketing, and enables users to quickly and easily obtain reliable information.

[0006] "Source" refers to any internet source of review data, such as a website, database, social media platform, or online forum.

[0007] "Word-of-mouth data" refers to text information shared by users on the Internet, such as product reviews, service experiences, opinions, and impressions.

[0008] "Preprocessing" refers to a series of processes that remove unnecessary information from collected review data and convert it into an analyzable state, including normalization, tokenization, stop word removal, stemming, etc.

[0009] A "generative artificial intelligence model" is a machine learning model trained to perform natural language processing tasks, specifically a model that has the ability to learn meaning and patterns from large amounts of text data and generate or evaluate new text. Examples include the Transformer model and the BERT model.

[0010] A "trustworthiness score" refers to the numerical expression of reliability or maliciousness of review data evaluated by a generative artificial intelligence model.

[0011] "Filtering" refers to the process of automatically removing malicious review data and selecting only useful review data based on the reliability score generated by a generative artificial intelligence model.

[0012] "Storage" refers to storing filtered, reliable review data in a database or storage system and keeping it accessible for later use.

[0013] A "search request" refers to an inquiry or query made by a user to a system to obtain specific information.

[0014] "Providing" refers to the system outputting and displaying stored, reliable word-of-mouth data in response to a user's search request.

[0015] "System" refers to an integrated platform built according to the present invention that collects, pre-processes, evaluates, filters, stores, searches, and provides review data. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0017] 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.

[0018] First, the terms used in the following description will be explained.

[0019] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0020] 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.

[0021] 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.

[0022] 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.

[0023] 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."

[0024] [First embodiment]

[0025] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0026] 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.

[0027] 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).

[0028] 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.

[0029] 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.

[0030] 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.

[0031] 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.

[0032] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] 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.

[0034] 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.

[0035] 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.

[0036] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0037] The present invention is a system that automatically filters malicious reviews from a vast amount of review information and provides users with only highly reliable reviews. Specific operations and embodiments of the system will be described below.

[0038] System Configuration

[0039] The system consists of the following main components:

[0040] 1. Data collection module (server)

[0041] 2. Text preprocessing module (server)

[0042] 3. Evaluation module (server) based on generative artificial intelligence model

[0043] 4. Filtering module (server)

[0044] 5. Database module (server)

[0045] 6. Search and Display Module (Terminal)

[0046] System Operation

[0047] Data collection module (server)

[0048] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtain review information. If an API is provided, it also collects data using that API. The collected raw data is stored in temporary storage.

[0049] Text preprocessing module (server)

[0050] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[0051] Text normalization (removal of special characters and extra whitespace)

[0052] Tokenization (breaking text into words and phrases)

[0053] Stop word removal (elimination of meaningless words)

[0054] Stemming and lemmatization (conversion of words to their root forms)

[0055] Evaluation module (server) based on generative artificial intelligence model

[0056] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The generative AI model analyzes large amounts of text data and is capable of identifying malicious elements such as slander and stealth marketing. This allows each review to be evaluated for credibility.

[0057] Filtering module (server)

[0058] The server filters reviews based on the reliability score obtained from the generative AI model. Specifically, it removes reviews with low reliability scores and leaves only high-score, reliable reviews. This improves the quality of information provided to users.

[0059] Database module (server)

[0060] The filtered, reliable review data is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[0061] Search and Display Module (Terminal)

[0062] When a user searches for specific review information on their device, the device sends the search request to the server. The server retrieves the relevant reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only reliable reviews about "Restaurant A" are displayed.

[0063] Specific examples

[0064] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These reviews are stored in the database and displayed when users search.

[0065] This system allows users to obtain highly reliable word-of-mouth information, enabling them to select products and services with confidence.

[0066] The processing flow will be explained below.

[0067] Step 1: Data Collection (Server)

[0068] The server uses web scraping technology to collect review data from multiple sources. Specifically, it accesses designated review sites, analyzes the HTML data of the pages, and extracts review information. If an API is provided, it also uses the API to obtain structured data.

[0069] Step 2: Text normalization (server)

[0070] Because the collected raw data contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary spaces, emojis, etc. to ensure the review text is clean.

[0071] Step 3: Tokenization (Server)

[0072] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into individual words that are easier to analyze.

[0073] Step 4: Stopword Removal (Server)

[0074] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki," "wa," and "wo." This makes it easier to extract important information.

[0075] Step 5: Stemming and Lemmatization (Server)

[0076] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[0077] Step 6: Evaluation using a generative AI model (server)

[0078] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[0079] Step 7: Filtering malicious reviews (server)

[0080] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[0081] Step 8: Saving to the Database (Server)

[0082] The server stores the filtered and reliable review data in a database, creating an index to enable efficient searches.

[0083] Step 9: Receiving a review search request (terminal, user)

[0084] A user sends a request to search for reviews of a particular product or service from a device, which then sends the request to the server.

[0085] Step 10: Retrieving from the database (server)

[0086] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[0087] Step 11: Returning and displaying search results (server, terminal)

[0088] The server returns the search results to the device, which then displays the received, highly reliable word-of-mouth data to the user, allowing the user to select products and services with confidence.

[0089] Example 1

[0090] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0091] The vast amount of word-of-mouth information contains unreliable and malicious information, making it difficult for users to quickly obtain reliable information. Furthermore, such inappropriate reviews can have a negative impact on product and service ratings. To solve these problems, a system that can automatically detect and filter unreliable reviews is needed.

[0092] 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.

[0093] In this invention, the server includes means for acquiring text data from a plurality of data sources, means for analyzing and preprocessing the acquired text data, means for evaluating the preprocessed text data using a generative model to generate a reliability score, means for filtering inappropriate text data based on the reliability score, means for saving the filtered highly reliable text data, and means for providing the saved highly reliable text data in response to a user search request, thereby enabling the automatic provision of highly reliable word-of-mouth information.

[0094] "Data source" refers to the place or system from which information is obtained, specifically including websites and APIs.

[0095] "Text data" refers to a collection of information expressed as a string of characters, and includes word-of-mouth, reviews, comments, etc.

[0096] "Preprocessing" refers to the process of converting raw text data into a parsable form, and includes text normalization, tokenization, stop word removal, stemming, and lemmatization.

[0097] "Generative models" refer to artificial intelligence algorithms that generate new data based on large amounts of text data, and specifically include Transformer models.

[0098] A "reliability score" is a numerical evaluation of the reliability of text data, with higher reliability being given a higher score.

[0099] "Filtering" refers to the process of removing inappropriate text data based on confidence scores.

[0100] An "index" refers to a data structure that enables fast searches and contains search keys and their location information.

[0101] A "search request" refers to a request made by a user to locate specific information.

[0102] "Storage" refers to the process of saving data for later use, including storing it in a storage system such as a database.

[0103] The present invention is a system for providing reliable word-of-mouth information. The system is comprised of a number of components, each of which performs a specific function.

[0104] Data collection module (server)

[0105] The server has a module that collects review information. This module acquires text data from multiple data sources, specifically using web scraping technology and APIs to collect data. For example, the server uses the API of "review site Y" to collect review data about "restaurant A." This collected data is temporarily stored in storage.

[0106] Text preprocessing module (server)

[0107] The server has a module that preprocesses the collected review data. This module performs tasks such as removing special characters and extra whitespace, normalizing the text, tokenizing, removing stop words, and performing stemming and lemmatization. For example, a review such as "This restaurant is terrible!" can be normalized to "This restaurant is terrible."

[0108] Evaluation module (server) based on generative artificial intelligence model

[0109] After preprocessing, the server evaluates the review data using a generative artificial intelligence model. This model is a Transformer-based generative model that analyzes the reliability of the text and generates a reliability score. For example, a review that says, "This restaurant is terrible!" will receive a low score, while a review that says, "The food is great. I'd like to go again" will receive a high score.

[0110] Filtering module (server)

[0111] The server has a module that filters the review data based on the generated confidence score. Low-confidence reviews are removed, and only high-confidence reviews remain. For example, a review such as "All the staff are friends!" is removed, while a review such as "The food is great. I'd love to go again." remains.

[0112] Database module (server)

[0113] The filtered, highly reliable review data is stored in a database module on the server. This module creates an index of the stored data to enable quick searches. For example, the highly reliable reviews for "Restaurant A" are stored in the database.

[0114] Search and Display Module (Terminal)

[0115] When a user searches for reviews on their device, the device sends the request to the server. The server retrieves the relevant highly reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only highly reliable reviews will be displayed.

[0116] Specific examples

[0117] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data includes defamatory comments such as "This restaurant is terrible," as well as unnatural positive comments such as "All the staff are friends!" When the text preprocessing module removes this noise and the data is evaluated by a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These saved reviews are then displayed when users search. This system allows users to obtain reliable review information, allowing them to choose products and services with confidence.

[0118] Prompt Sentence Examples

[0119] Example prompts to input to a generative AI model:

[0120] "Data collected: 'This restaurant is awful. The food was cold and the waiter was unfriendly.'"

[0121] "Data collected: 'All the staff are friends! It was amazing.'"

[0122] "Data collected: 'The food is amazing. I'd love to go back.'"

[0123] Generate a credibility score for each review.

[0124] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0125] Step 1:

[0126] Data collection

[0127] The server obtains review information from multiple data sources. Specifically, it uses web scraping technology and APIs to collect review data. Inputs include web page URLs and API endpoints. The server sends HTTP requests for these inputs and receives HTML data or structured data as a response. This data is then stored in temporary storage.

[0128] Specific behavior:

[0129] The server receives the URL of "review site Y" as input.

[0130] The server sends an HTTP request to the specified URL and receives HTML data as a response.

[0131] The server analyzes the HTML data and extracts the necessary word-of-mouth information.

[0132] The extracted word-of-mouth information is stored in temporary storage.

[0133] Step 2:

[0134] Text Preprocessing

[0135] The server preprocesses the collected review data. For example, it removes special characters and extra whitespace, normalizes text, tokenizes, removes stop words, and performs stemming and lemmatization. The input includes the collected raw review data. The server preprocesses this input data and outputs clean text data suitable for analysis.

[0136] Specific behavior:

[0137] The server retrieves the raw word-of-mouth data from the temporary storage.

[0138] Normalizes text by removing special characters and extra whitespace.

[0139] Break down (tokenize) the normalized text into words and phrases.

[0140] Stop words are removed and necessary words are converted to their base form using stemming and lemmatization.

[0141] Prepare preprocessed, clean text data.

[0142] Step 3:

[0143] Generating a confidence score

[0144] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The input includes preprocessed, clean review data. The server passes this to the generative AI model and obtains a credibility score for each review as output.

[0145] Specific behavior:

[0146] The server obtains the preprocessed review data.

[0147] Pass word-of-mouth data as input to a generative AI model.

[0148] The AI ​​model analyzes the data and generates a confidence score.

[0149] Get the generated confidence score.

[0150] Step 4:

[0151] filtering

[0152] The server filters the review data based on the generated confidence scores. The input includes the review data and its confidence scores. The server removes reviews with low confidence scores and selects reviews with high scores as output.

[0153] Specific behavior:

[0154] The server obtains the generated word-of-mouth data with the reliability scores.

[0155] Weed out low-trust reviews based on their trust score.

[0156] Select reliable review data with high scores.

[0157] Step 5:

[0158] Saving to a database

[0159] The filtered, reliable review data is stored in a database. The input includes the filtered review data. The server stores this in a database and creates an index to enable quick searches.

[0160] Specific behavior:

[0161] The server obtains filtered, highly reliable word-of-mouth data.

[0162] These data are stored in a database.

[0163] Create indexes in your database to enable efficient searches.

[0164] Step 6:

[0165] Search and Display

[0166] When a user searches for review information on their device, the device sends the request to the server. The input includes the user's search query. The server retrieves relevant, reliable review data from the database and returns it to the device as output. The user can then view it on their device.

[0167] Specific behavior:

[0168] A user inputs a search query through a terminal.

[0169] The terminal sends this search query to the server.

[0170] The server retrieves relevant and reliable word-of-mouth data from the database.

[0171] The server returns the acquired data to the terminal.

[0172] The user checks the search results on the terminal.

[0173] (Application example 1)

[0174] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0175] Conventional word-of-mouth information has been problematic in terms of its reliability, particularly in the case of electronic payments, where inappropriate information is mixed in with product and service reviews. This puts users at risk of making purchasing decisions based on inappropriate information, making it difficult to select appropriate products and services. Furthermore, providing reliable word-of-mouth information in real time is difficult, making it difficult to support timely purchasing decisions. The purpose of this invention is to solve these problems by efficiently providing reliable word-of-mouth information and ensuring users' peace of mind when making electronic payments.

[0176] 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.

[0177] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for storing the filtered reliable word-of-mouth data, means for integrating the filtered reliable word-of-mouth data with an electronic payment function to enable users to obtain reliable word-of-mouth information before purchasing a product or service, and means for providing the stored reliable word-of-mouth data in response to a user's search request, thereby enabling users to easily obtain reliable word-of-mouth information before making an electronic payment and select products or services with confidence.

[0178] "Sources" are data collection sources such as websites and APIs that provide review data.

[0179] "Review Data" refers to opinions, ratings, or impressions written by users about a particular product or service.

[0180] "Preprocessing" is the process of removing noise from collected word-of-mouth data and converting it into a format suitable for analysis.

[0181] A "generative artificial intelligence model" is a machine learning model for analyzing large amounts of text data, and has the ability to evaluate the reliability of word-of-mouth reviews.

[0182] The "trustworthiness score" is the result of an analysis of word-of-mouth data by a generative artificial intelligence model, and is a numerical representation of the trustworthiness of the word-of-mouth.

[0183] "Filtering" is the process of removing low-reliability reviews based on their reliability scores, leaving only high-reliability reviews.

[0184] "Electronic payment function" is a system that allows users to make instant payments when purchasing goods or services.

[0185] A "search request" is a query or request entered by a user to search for specific reviews.

[0186] "Storage" refers to the act of storing filtered, reliable word-of-mouth data in a database.

[0187] "Providing" means displaying the stored highly reliable word-of-mouth data on the user's terminal in response to a user's search request.

[0188] MODE FOR CARRYING OUT THE INVENTION

[0189] This invention is a system that collects word-of-mouth data from multiple sources and provides users with only reliable word-of-mouth information. It can also be integrated with electronic payment functions, allowing users to select products and services with confidence based on reliable word-of-mouth information.

[0190] System Configuration

[0191] The system consists of the following main tools:

[0192] 1. Data Collection Methods

[0193] 2. Text Preprocessing Methods

[0194] 3. Evaluation Methods Using Generative AI Models

[0195] 4. Filtering Methods

[0196] 5. Data storage means

[0197] 6. Electronic payment function integration means

[0198] 7. Search and Display Methods

[0199] Hardware and software used

[0200] Hardware:

[0201] server

[0202] User device (smartphone)

[0203] software:

[0204] Python

[0205] SQLite

[0206] BeautifulSoup

[0207] Scikit-learn

[0208] Gensim

[0209] Explanation of data processing and data calculation

[0210] Data collection methods:

[0211] The server collects review data from multiple sources using web scraping technology and APIs, and the collected raw data is stored in temporary storage.

[0212] Text preprocessing methods:

[0213] The server pre-processes the collected review data and converts it into a format suitable for analysis, including normalization, tokenization, stop-word removal, stemming, and lemmatization.

[0214] Evaluation method using generative artificial intelligence model:

[0215] The server inputs the pre-processed review data into a generative AI model that uses a Transformer-based model to generate a credibility score for each review.

[0216] Filtering methods:

[0217] The server filters reviews based on the reliability scores obtained from the generative AI model, eliminating reviews with low scores and leaving only reviews with high reliability scores.

[0218] Data storage method:

[0219] The filtered, reliable review data is stored in an SQLite database, which is indexed to enable fast searches.

[0220] Electronic payment integration methods:

[0221] When a user searches for reviews before purchasing a product or service, the server retrieves filtered reviews from the database and displays them on the user's terminal. Based on the reviews, the user can make an electronic payment.

[0222] Search and display methods:

[0223] When a user makes a search request, the server retrieves the stored, reliable word-of-mouth data from the database and returns it to the user's device. The user can then check the reliable word-of-mouth information and make a purchase decision with confidence.

[0224] Specific examples

[0225] For example, if a user tries to search for reviews of "Restaurant A," the server collects review data about "Restaurant A" from multiple sources. The collected review data undergoes text preprocessing and is evaluated by a generative artificial intelligence model. After low-reliability reviews are filtered out, the highly reliable review data is stored in a database. When a user searches for "Restaurant A," the filtered, highly reliable review information is displayed on the user's device. After viewing this review information, the user can safely make reservations and payments for "Restaurant A" using the electronic payment function.

[0226] Prompt Sentence Examples

[0227] "How can I search for reviews of Restaurant A using a filtered app that supports electronic payments?"

[0228] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0229] Step 1:

[0230] The server collects review data from multiple sources, including web scraping techniques and APIs, and stores the collected raw data in temporary storage.

[0231] Input: URLs or API endpoints of multiple review sites

[0232] Output: Raw reviews

[0233] Step 2:

[0234] The server preprocesses the collected review data, normalizing and tokenizing the text, removing stop words, stemming, and lemmatizing it to convert it into a format suitable for analysis.

[0235] Input: Raw review data

[0236] Output: Preprocessed review data

[0237] Step 3:

[0238] The server evaluates the preprocessed review data using a generative AI model (Transformer-based) to generate a credibility score. Specifically, the model analyzes the text, identifies slander and stealth marketing, and evaluates its credibility.

[0239] Input: Preprocessed review data

[0240] Output: Review data with credibility scores

[0241] Step 4:

[0242] The server filters out bad reviews based on the generated reliability score, removing reviews with low reliability scores and selecting only reliable reviews with high scores.

[0243] Input: Review data with reliability scores

[0244] Output: Filtered and reliable review data

[0245] Step 5:

[0246] The server stores the filtered, reliable review data in an SQLite database, which is indexed as it is stored, allowing for fast searches of the data.

[0247] Input: Filtered, reliable review data

[0248] Output: Review data stored in a database

[0249] Step 6:

[0250] When a user sends a search request for a specific product or service from a terminal, the server retrieves relevant, reliable word-of-mouth data from the database and returns it to the user terminal.

[0251] Input: User search request

[0252] Output: Reliable review data displayed on the user's device

[0253] Step 7:

[0254] Users select products and services based on the displayed reviews and make purchases using the electronic payment function. The server provides the reviews and electronic payment function in an integrated manner, allowing users to make payments with peace of mind.

[0255] Input: User purchase selection

[0256] Output: Completed electronic payment and purchase details

[0257] 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.

[0258] The present invention combines a system that automatically filters malicious reviews from a vast amount of review information and provides users with only reliable reviews, with an emotion engine that recognizes user emotions. Specific operations and embodiments of this system are described below.

[0259] System Configuration

[0260] The system consists of the following main components:

[0261] 1. Data collection module (server)

[0262] 2. Text preprocessing module (server)

[0263] 3. Evaluation module (server) based on generative artificial intelligence model

[0264] 4. Filtering module (server)

[0265] 5. Emotion engine module (server)

[0266] 6. Database Module (Server)

[0267] 7. Search and Display Module (Terminal)

[0268] System Operation

[0269] Data collection module (server)

[0270] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtains review information. If an API is provided, it also uses that API to obtain structured data. The collected raw data is stored in temporary storage.

[0271] Text preprocessing module (server)

[0272] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[0273] Text normalization (removal of special characters and extra whitespace)

[0274] Tokenization (breaking text into words and phrases)

[0275] Stop word removal (elimination of meaningless words)

[0276] Stemming and lemmatization (conversion of words to their root forms)

[0277] Evaluation module (server) based on generative artificial intelligence model

[0278] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[0279] Filtering module (server)

[0280] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[0281] Emotion engine module (server)

[0282] The emotion engine recognizes users' emotions and individually adjusts the ratings of reviews they view. Specifically, it analyzes users' behavior and review browsing history to determine their current emotional state. Based on this information, it adjusts the credibility score of reviews and provides the most appropriate information for the user.

[0283] Database module (server)

[0284] The reliable review data adjusted by the filtering and sentiment engine is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[0285] Search and Display Module (Terminal)

[0286] When a user searches for specific review information on their device, the device sends the search request to the server. The server then searches the database for relevant, reliable review data and returns the results, adjusted by the emotion engine, to the device. The user can then view these results on their device. For example, if a user searches for reviews of "Restaurant A," only reviews about "Restaurant A" that are highly reliable and match the user's emotional state will be displayed.

[0287] Specific examples

[0288] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing the noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only factual reviews such as "The food is great. I'd like to go again."

[0289] Furthermore, the emotion engine uses the user's past behavioral history to prioritize positive reviews for users who want to avoid negative reviews about restaurants. In this way, reviews are displayed that are in line with the user's emotions. These reviews are stored in the database and are displayed when the user searches.

[0290] This system allows users to obtain highly reliable word-of-mouth information and quickly and easily acquire information that is suited to their own emotional state.

[0291] The processing flow will be explained below.

[0292] Step 1: Data Collection (Server)

[0293] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from review sites and social media platforms and extract review information. If an API is provided, it is also possible to obtain review data using that API. The collected raw data is stored in temporary storage.

[0294] Step 2: Text normalization (server)

[0295] Because the collected raw data contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary spaces, emojis, etc. to ensure the review text is clean.

[0296] Step 3: Tokenization (Server)

[0297] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into units that are easier to analyze.

[0298] Step 4: Stopword Removal (Server)

[0299] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki" (target), "ga" (ga), and "o" (to). This makes it easier to extract important information.

[0300] Step 5: Stemming and Lemmatization (Server)

[0301] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[0302] Step 6: Evaluation using a generative AI model (server)

[0303] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[0304] Step 7: Filtering malicious reviews (server)

[0305] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[0306] Step 8: Recognizing user emotions with the emotion engine (server)

[0307] The server utilizes an emotion engine to recognize the user's current emotional state based on their activity and past browsing history, including their click history, browsing time, and search keywords.

[0308] Step 9: Moderating reviews with the sentiment engine (server)

[0309] Based on the user's emotional state recognized by the emotion engine, the server optimizes reliable reviews for each individual user. For example, if the emotion engine determines that the user prefers positive content, it will prioritize positive reviews.

[0310] Step 10: Saving to the Database (Server)

[0311] The server stores the filtered and adjusted reliable review data in a database, where it creates an index to enable efficient searches.

[0312] Step 11: Receiving a review search request (terminal, user)

[0313] The user sends a request to search for specific reviews from the terminal, which then sends the request to the server.

[0314] Step 12: Retrieving from the Database (Server)

[0315] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[0316] Step 13: Returning and displaying search results (server, terminal)

[0317] The server returns the search results to the device, which then displays the received, highly reliable review data to the user. Because the results have been adjusted by the emotion engine, reviews that best fit the user's emotional state are displayed.

[0318] Example 2

[0319] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0320] In today's information society, unreliable word-of-mouth information is overflowing, making it difficult for consumers to obtain accurate and reliable information. Furthermore, the lack of information provided in response to consumers' emotional state often leads to low user satisfaction. In this situation, there is an urgent need to provide a means to accurately filter malicious word-of-mouth and provide word-of-mouth information that takes into account the user's emotional state.

[0321] 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 means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model and generating a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for analyzing user emotions and adjusting the reliability score, means for storing reliable word-of-mouth data through filtering and emotion analysis, and means for providing the stored reliable word-of-mouth data in response to a user's search request. This makes it possible to quickly and accurately provide reliable word-of-mouth information tailored to the user's emotional state.

[0322] "Sources" are data providers such as online word-of-mouth sites, review sites, and social media platforms.

[0323] "Word-of-mouth data" is text data such as opinions, impressions, and evaluations written by users about products and services.

[0324] "Preprocessing" refers to processes such as normalizing text, tokenizing, removing stop words, stemming, and lemmatizing to make it easier to analyze review data.

[0325] A "generative artificial intelligence model" refers to a large-scale Transformer-based language model used in natural language processing tasks, which analyzes the context of reviews and generates a reliability score.

[0326] The "trust score" is an index that indicates the trustworthiness of a review, calculated by a generative artificial intelligence model by analyzing the context of the review.

[0327] "Filtering" is the process of selecting and removing malicious reviews based on their reliability scores.

[0328] "Emotion analysis" is a process of estimating a user's current emotional state based on their past behavioral patterns and browsing history.

[0329] "Data storage" means storing reliable word-of-mouth data selected through filtering and sentiment analysis in a database.

[0330] A "search request" is a query sent by a user from a terminal in an attempt to search for specific word-of-mouth information.

[0331] "Providing means" refers to the process in which the server returns highly reliable word-of-mouth data that has been filtered and sentiment analyzed to the user.

[0332] "Transformer-based" is an artificial intelligence model architecture built on self-attention mechanisms and deep learning techniques.

[0333] An "index" is a data structure used to speed up database searches and is created for a particular column of data.

[0334] This system preprocesses review data collected from multiple sources, evaluates and filters it using a generative artificial intelligence model, and provides users with highly reliable data. Furthermore, it aims to increase user satisfaction by analyzing user sentiment and adjusting the display of review data.

[0335] System Configuration

[0336] The system consists of the following main components:

[0337] 1. Data collection module (server)

[0338] 2. Text preprocessing module (server)

[0339] 3. Evaluation module (server) based on generative artificial intelligence model

[0340] 4. Filtering module (server)

[0341] 5. Emotion engine module (server)

[0342] 6. Database Module (Server)

[0343] 7. Search and Display Module (Terminal)

[0344] System Operation

[0345] Data collection module (server)

[0346] The server uses BeautifulSoup to retrieve HTML data from multiple review sites, analyze it, and collect review information. It also retrieves structured data via APIs from services that provide them. The collected data is saved in temporary storage. A specific example is a process that creates a list of URLs for "review site A" and extracts reviews from each page.

[0347] Text preprocessing module (server)

[0348] The server performs preprocessing on the collected review data, including normalization, tokenization, stop word removal, stemming, and lemmatization, using Python's NLTK library and SpaCy. For example, it removes noise from sentences like "This restaurant is great" and formats them into a form that is easier to analyze.

[0349] Evaluation module (server) based on generative artificial intelligence model

[0350] The server inputs the preprocessed text data into a generative artificial intelligence model such as GPT-4 (registered trademark). The model analyzes the context of the review and generates a credibility score. An example prompt is, "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[0351] Filtering module (server)

[0352] The server filters the review data based on the generated reliability score, removing reviews with low reliability scores and allowing only the remaining high-score reviews to proceed to the next step, thereby eliminating malicious reviews.

[0353] Emotion engine module (server)

[0354] The emotion engine analyzes a user's past behavior and browsing history to estimate their current emotional state. Based on this, it readjusts the reliability score and provides the most appropriate review information to the user. For example, it prioritizes positive reviews for a user who has avoided negative reviews in the past.

[0355] Database module (server)

[0356] The reliable review data, adjusted by the filtering and sentiment engine, is stored in a database using MySQL (registered trademark) or PostgreSQL, and indexes are created to enable efficient searches.

[0357] Search and Display Module (Terminal)

[0358] When a user enters a search query on their device, the device sends that information to the server. The server searches the database for relevant review data and returns the results to the device. The user can then check the results on their device. For example, if a user searches for reviews of "Restaurant B," only reliable reviews will be displayed.

[0359] This system allows users to quickly and accurately obtain reliable word-of-mouth information, and also provides information tailored to their emotional state.

[0360] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0361] Step 1:

[0362] Data collection

[0363] Input: The server receives a list of URLs for multiple review sites.

[0364] Specific operation: The server uses a web scraping tool such as BeautifulSoup or Selenium to access each review site based on the URL list and obtain HTML data.

[0365] Data processing: Specific tags and attributes are analyzed from the acquired HTML data to extract word-of-mouth information.

[0366] Output: The extracted reviews (raw data) are saved in temporary storage in JSON format.

[0367] Step 2:

[0368] Text Preprocessing

[0369] Input: The server reads the collected raw data (word-of-mouth information) from the storage.

[0370] Specific operation: The server uses Python's NLTK library and SpaCy to perform text normalization, tokenization, stop word removal, stemming, and lemmatization.

[0371] Data processing: Remove special characters, split words, eliminate meaningless words, and convert words to their root forms.

[0372] Output: Preprocessed text data is obtained and put into a usable format for the next step.

[0373] Step 3:

[0374] Evaluation using generative artificial intelligence models

[0375] Input: The server reads the preprocessed text data.

[0376] How it works: The server inputs a prompt into a generative artificial intelligence model (e.g., GPT-4) and begins the process of assessing the trustworthiness of each review.

[0377] Example prompt: "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[0378] Data computation: The AI ​​model analyzes the context and generates a reliability score.

[0379] Output: Data containing the trustworthiness score and rating reason for each review.

[0380] Step 4:

[0381] filtering

[0382] Input: The server reads the data with reliability scores output from the generative artificial intelligence model.

[0383] Specific operation: The server filters the review data based on the reliability score. It sets a threshold and removes reviews with a score below that threshold.

[0384] Data processing: Remove reviews with low scores and keep only reviews with high scores.

[0385] Output: Filtered and reliable word-of-mouth data.

[0386] Step 5:

[0387] Emotion Engine

[0388] Input: The server loads filtered, reliable word-of-mouth data and user behavior history data.

[0389] How it works: The server uses machine learning models to estimate the user's emotional state by analyzing past click history, time spent on the site, search history, etc.

[0390] Data calculation: Estimate the sentiment and recalculate / adjust the reliability score of the review data.

[0391] Output: Word-of-mouth data adjusted according to the user's emotional state.

[0392] Step 6:

[0393] Database storage

[0394] Input: The server loads reliable review data that has been tuned by the sentiment engine.

[0395] Specific operation: The server inserts data into a database such as MySQL or PostgreSQL.

[0396] Data processing: Creating indexes to enable efficient searches when stored.

[0397] Output: Reliable indexed reviews stored in a database.

[0398] Step 7:

[0399] Search and Display

[0400] Input: The terminal receives a search query from a user.

[0401] Specific operation: The terminal sends a search query to the server, which queries the database and retrieves the relevant review data.

[0402] Data processing: Organize search results and format them in a way that is easy for users to view.

[0403] Output: The terminal displays the reliable word-of-mouth results received from the server to the user.

[0404] (Application example 2)

[0405] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0406] Current review filtering systems have difficulty accurately filtering out malicious and defamatory content, resulting in a large amount of information that users find unreliable. Furthermore, they lack the ability to provide appropriate reviews that reflect the user's individual emotional state and interests, making it difficult for users to quickly and accurately obtain the information they need. This creates a demand for reliable review display, especially on online shopping sites, but current systems make it difficult to achieve this.

[0407] 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.

[0408] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious reviews based on the reliability score, means for saving the filtered reliable word-of-mouth data, means for recognizing the emotional state of a user and adjusting the reliability score according to the emotional state, and means for providing the saved reliable word-of-mouth data in response to a user's search request. This enables users to quickly and accurately obtain reliable word-of-mouth information, and further realizes the provision of information optimized for the user's emotional state.

[0409] "Sources" are data providers, including websites and applications on the Internet, from which word-of-mouth data can be obtained.

[0410] "Word-of-mouth data" is text data in which users write their evaluations and opinions about a certain product or service.

[0411] "Preprocessing" refers to a series of processes that convert review data into a form that is easier to analyze, including text normalization and tokenization.

[0412] A "generative artificial intelligence model" is a type of AI technology that learns useful patterns from massive amounts of data and evaluates and predicts new data.

[0413] The "trustworthiness score" is a number calculated as a result of evaluating whether each review contains defamatory or false information.

[0414] "Filtering" is the process of removing bad reviews based on their credibility score.

[0415] "Emotional state" refers to the user's current psychological and emotional state, and is based on their past behavioral history and current operating situation.

[0416] "Data collection means" refers to a program or device for automatically collecting word-of-mouth data.

[0417] "Data preprocessing means" refers to a program or device that formats collected data into a format that is easy to analyze.

[0418] An "evaluation means" is a program or device that analyzes preprocessed data using a generative artificial intelligence model and generates a reliability score.

[0419] A "filtering means" is a program or device for filtering out malicious reviews based on reliability scores.

[0420] "Storage means" refers to a program or device for storing filtered, highly reliable word-of-mouth data.

[0421] An "emotion recognition means" is a program or device for recognizing a user's emotional state and adjusting the reliability score.

[0422] The "providing means" refers to a program or device for displaying or providing highly reliable word-of-mouth data that has been saved in response to a user's search request.

[0423] A "natural language processing engine" is a program or device that provides technology that enables computers to understand, interpret, and generate human language.

[0424] The present invention aims to provide a specific information provision system that allows users to efficiently and accurately obtain reliable word-of-mouth information. This system collects word-of-mouth data from multiple sources, evaluates it using a generative artificial intelligence model, and further adjusts the reliability score based on the user's emotional state to provide reliable word-of-mouth information.

[0425] System Configuration

[0426] The system consists of the following components:

[0427] 1. Data collection method (server)

[0428] 2. Data preprocessing means (server)

[0429] 3. Evaluation Method (Server)

[0430] 4. Filtering Method (Server)

[0431] 5. Storage means (server)

[0432] 6. Emotion Recognition Method (Server)

[0433] 7. Means of provision (terminal)

[0434] Data collection method (server)

[0435] The server collects review information from multiple review data sources using web scraping technology and APIs, and converts the collected data from unstructured data into structured data.

[0436] Data preprocessing means (server)

[0437] The server performs preprocessing on the collected review data, specifically normalizing the text, tokenizing it, removing stop words, etc. This process converts the text data into a format that is easier to analyze.

[0438] Evaluation method (server)

[0439] The server then inputs the pre-processed data into a generative AI model, a Transformer-based AI model, which evaluates the reliability of the data and generates a reliability score for each review.

[0440] Filtering method (server)

[0441] Based on the generated credibility score, the server filters the reviews: reviews with low credibility scores are rejected and only high credibility reviews are passed on to the next processing step.

[0442] Storage method (server)

[0443] The filtered review data is stored in a database, where it is indexed for fast search.

[0444] Emotion recognition means (server)

[0445] The server uses a natural language processing engine to recognize the user's emotional state, determining it based on the user's past browsing history and current operating context, and adjusting the reliability score accordingly.

[0446] Means of provision (terminal)

[0447] When a user searches for specific reviews using their device, the device sends a search request to the server, which searches the stored reliable reviews data and returns results adjusted based on the user's emotional state to the device, where the user can view the results on their device.

[0448] Specific examples

[0449] For example, a server collects review data for electronic products from a shopping website. This data may contain defamatory content or unnatural positive comments. The server preprocesses the data and evaluates it using a generative artificial intelligence model. Inappropriate reviews receive a low reliability score and are filtered out. An emotion recognition tool then reevaluates the reviews based on the user's context, providing tailored review information.

[0450] Prompt Sentence Examples

[0451] "Create a system to automatically extract reliable reviews and provide optimal information based on user sentiment. The specific steps are as follows: collection, preprocessing, rating, filtering, and sentiment adjustment. Please provide a detailed explanation of the process performed at each step, and include any necessary modules or libraries."

[0452] The above system configuration and processing enable users to obtain highly reliable word-of-mouth information efficiently and accurately.

[0453] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0454] Step 1: Data collection

[0455] The server collects review data from multiple sources. To do so, it uses web scraping technology and APIs. Specifically, the server accesses specified URLs or API endpoints to obtain HTML data or structured data. The obtained data is converted from unstructured data to structured data and temporarily stored.

[0456] Input: Source URL or API endpoint

[0457] Output: Collected review data (text format)

[0458] Step 2: Data Preprocessing

[0459] The server preprocesses the collected review data. Specifically, it removes special characters and unnecessary spaces from the data, and uses natural language processing techniques to tokenize and remove stop words. The text data is then converted into a format that is easy to analyze.

[0460] Input: Collected review data

[0461] Output: Preprocessed review data

[0462] Step 3: Evaluation using a generative artificial intelligence model

[0463] The server inputs the preprocessed review data into a generative AI model. A Transformer-based AI model is used here. The model evaluates the reliability of the text data and generates a reliability score for each review. The reliability score evaluates the likelihood that the review contains defamation or stealth marketing.

[0464] Input: Preprocessed review data

[0465] Output: Trustworthiness score for each review

[0466] Step 4: Filtering

[0467] The server filters the reviews based on the generated credibility score. Reviews with low credibility scores are deemed to be defamatory or stealth marketing and are removed. Only reviews with high credibility scores proceed to the next processing step.

[0468] Input: Confidence score

[0469] Output: Filtered and reliable review data

[0470] Step 5: Save Data

[0471] The server stores the filtered, reliable review data in a database, where an index is created to enable fast searches, allowing for quick retrieval of relevant information later.

[0472] Input: Filtered, reliable review data

[0473] Output: Review data stored in the database

[0474] Step 6: Emotion Recognition

[0475] The server recognizes the user's emotional state by analyzing the user's past browsing history and current operating status using a natural language processing engine to determine the user's current emotional state, and then readjusts the reliability score based on this.

[0476] Input: User browsing history, operation status

[0477] Output: Adjusted confidence score

[0478] Step 7: Provide data

[0479] When a user searches for specific review information on their device, the device sends the search request to the server, which then searches the database for relevant, reliable review data and returns results tailored based on the user's emotional state.

[0480] Input: User search request

[0481] Output: Reliable reviews (adjusted based on users' emotional state)

[0482] The above are the specific processing steps from collection to provision of word-of-mouth data.

[0483] 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.

[0484] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0485] 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.

[0486] [Second embodiment]

[0487] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0488] 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.

[0489] 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).

[0490] 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.

[0491] 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.

[0492] 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).

[0493] 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.

[0494] 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.

[0495] 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.

[0496] 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.

[0497] 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.

[0498] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0499] The present invention is a system that automatically filters malicious reviews from a vast amount of review information and provides users with only highly reliable reviews. Specific operations and embodiments of the system will be described below.

[0500] System Configuration

[0501] The system consists of the following main components:

[0502] 1. Data collection module (server)

[0503] 2. Text preprocessing module (server)

[0504] 3. Evaluation module (server) based on generative artificial intelligence model

[0505] 4. Filtering module (server)

[0506] 5. Database module (server)

[0507] 6. Search and Display Module (Terminal)

[0508] System Operation

[0509] Data collection module (server)

[0510] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtain review information. If an API is provided, it also collects data using that API. The collected raw data is stored in temporary storage.

[0511] Text preprocessing module (server)

[0512] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[0513] Text normalization (removal of special characters and extra whitespace)

[0514] Tokenization (breaking text into words and phrases)

[0515] Stop word removal (elimination of meaningless words)

[0516] Stemming and lemmatization (conversion of words to their root forms)

[0517] Evaluation module (server) based on generative artificial intelligence model

[0518] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The generative AI model analyzes large amounts of text data and is capable of identifying malicious elements such as slander and stealth marketing. This allows each review to be evaluated for credibility.

[0519] Filtering module (server)

[0520] The server filters reviews based on the reliability score obtained from the generative AI model. Specifically, it removes reviews with low reliability scores and leaves only high-score, reliable reviews. This improves the quality of information provided to users.

[0521] Database module (server)

[0522] The filtered, reliable review data is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[0523] Search and Display Module (Terminal)

[0524] When a user searches for specific review information on their device, the device sends the search request to the server. The server retrieves the relevant reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only reliable reviews about "Restaurant A" are displayed.

[0525] Specific examples

[0526] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These reviews are stored in the database and displayed when users search.

[0527] This system allows users to obtain highly reliable word-of-mouth information, enabling them to select products and services with confidence.

[0528] The processing flow will be explained below.

[0529] Step 1: Data Collection (Server)

[0530] The server uses web scraping technology to collect review data from multiple sources. Specifically, it accesses designated review sites, analyzes the HTML data of the pages, and extracts review information. If an API is provided, it also uses the API to obtain structured data.

[0531] Step 2: Text normalization (server)

[0532] Because the collected raw data contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary spaces, emojis, etc. to ensure the review text is clean.

[0533] Step 3: Tokenization (Server)

[0534] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into individual words that are easier to analyze.

[0535] Step 4: Stopword Removal (Server)

[0536] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki," "wa," and "wo." This makes it easier to extract important information.

[0537] Step 5: Stemming and Lemmatization (Server)

[0538] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[0539] Step 6: Evaluation using a generative AI model (server)

[0540] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[0541] Step 7: Filtering malicious reviews (server)

[0542] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[0543] Step 8: Saving to the Database (Server)

[0544] The server stores the filtered and reliable review data in a database, creating an index to enable efficient searches.

[0545] Step 9: Receiving a review search request (terminal, user)

[0546] A user sends a request to search for reviews of a particular product or service from a device, which then sends the request to the server.

[0547] Step 10: Retrieving from the database (server)

[0548] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[0549] Step 11: Returning and displaying search results (server, terminal)

[0550] The server returns the search results to the device, which then displays the received, highly reliable word-of-mouth data to the user, allowing the user to select products and services with confidence.

[0551] Example 1

[0552] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0553] The vast amount of word-of-mouth information contains unreliable and malicious information, making it difficult for users to quickly obtain reliable information. Furthermore, such inappropriate reviews can have a negative impact on product and service ratings. To solve these problems, a system that can automatically detect and filter unreliable reviews is needed.

[0554] 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.

[0555] In this invention, the server includes means for acquiring text data from a plurality of data sources, means for analyzing and preprocessing the acquired text data, means for evaluating the preprocessed text data using a generative model to generate a reliability score, means for filtering inappropriate text data based on the reliability score, means for saving the filtered highly reliable text data, and means for providing the saved highly reliable text data in response to a user search request, thereby enabling the automatic provision of highly reliable word-of-mouth information.

[0556] "Data source" refers to the place or system from which information is obtained, specifically including websites and APIs.

[0557] "Text data" refers to a collection of information expressed as a string of characters, and includes word-of-mouth, reviews, comments, etc.

[0558] "Preprocessing" refers to the process of converting raw text data into a parsable form, and includes text normalization, tokenization, stop word removal, stemming, and lemmatization.

[0559] "Generative models" refer to artificial intelligence algorithms that generate new data based on large amounts of text data, and specifically include Transformer models.

[0560] A "reliability score" is a numerical evaluation of the reliability of text data, with higher reliability being given a higher score.

[0561] "Filtering" refers to the process of removing inappropriate text data based on confidence scores.

[0562] An "index" refers to a data structure that enables fast searches and contains search keys and their location information.

[0563] A "search request" refers to a request made by a user to locate specific information.

[0564] "Storage" refers to the process of saving data for later use, including storing it in a storage system such as a database.

[0565] The present invention is a system for providing reliable word-of-mouth information. The system is comprised of a number of components, each of which performs a specific function.

[0566] Data collection module (server)

[0567] The server has a module that collects review information. This module acquires text data from multiple data sources, specifically using web scraping technology and APIs to collect data. For example, the server uses the API of "review site Y" to collect review data about "restaurant A." This collected data is temporarily stored in storage.

[0568] Text preprocessing module (server)

[0569] The server has a module that preprocesses the collected review data. This module performs tasks such as removing special characters and extra whitespace, normalizing the text, tokenizing, removing stop words, and performing stemming and lemmatization. For example, a review such as "This restaurant is terrible!" can be normalized to "This restaurant is terrible."

[0570] Evaluation module (server) based on generative artificial intelligence model

[0571] After preprocessing, the server evaluates the review data using a generative artificial intelligence model. This model is a Transformer-based generative model that analyzes the reliability of the text and generates a reliability score. For example, a review that says, "This restaurant is terrible!" will receive a low score, while a review that says, "The food is great. I'd like to go again" will receive a high score.

[0572] Filtering module (server)

[0573] The server has a module that filters the review data based on the generated confidence score. Low-confidence reviews are removed, and only high-confidence reviews remain. For example, a review such as "All the staff are friends!" is removed, while a review such as "The food is great. I'd love to go again." remains.

[0574] Database module (server)

[0575] The filtered, highly reliable review data is stored in a database module on the server. This module creates an index of the stored data to enable quick searches. For example, the highly reliable reviews for "Restaurant A" are stored in the database.

[0576] Search and Display Module (Terminal)

[0577] When a user searches for reviews on their device, the device sends the request to the server. The server retrieves the relevant highly reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only highly reliable reviews will be displayed.

[0578] Specific examples

[0579] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data includes defamatory comments such as "This restaurant is terrible," as well as unnatural positive comments such as "All the staff are friends!" When the text preprocessing module removes this noise and the data is evaluated by a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These saved reviews are then displayed when users search. This system allows users to obtain reliable review information, allowing them to choose products and services with confidence.

[0580] Prompt Sentence Examples

[0581] Example prompts to input to a generative AI model:

[0582] "Data collected: 'This restaurant is awful. The food was cold and the waiter was unfriendly.'"

[0583] "Data collected: 'All the staff are friends! It was amazing.'"

[0584] "Data collected: 'The food is amazing. I'd love to go back.'"

[0585] Generate a credibility score for each review.

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

[0587] Step 1:

[0588] Data collection

[0589] The server obtains review information from multiple data sources. Specifically, it uses web scraping technology and APIs to collect review data. Inputs include web page URLs and API endpoints. The server sends HTTP requests for these inputs and receives HTML data or structured data as a response. This data is then stored in temporary storage.

[0590] Specific behavior:

[0591] The server receives the URL of "review site Y" as input.

[0592] The server sends an HTTP request to the specified URL and receives HTML data as a response.

[0593] The server analyzes the HTML data and extracts the necessary word-of-mouth information.

[0594] The extracted word-of-mouth information is stored in temporary storage.

[0595] Step 2:

[0596] Text Preprocessing

[0597] The server preprocesses the collected review data. For example, it removes special characters and extra whitespace, normalizes text, tokenizes, removes stop words, and performs stemming and lemmatization. The input includes the collected raw review data. The server preprocesses this input data and outputs clean text data suitable for analysis.

[0598] Specific behavior:

[0599] The server retrieves the raw word-of-mouth data from the temporary storage.

[0600] Normalizes text by removing special characters and extra whitespace.

[0601] Break down (tokenize) the normalized text into words and phrases.

[0602] Stop words are removed and necessary words are converted to their base form using stemming and lemmatization.

[0603] Prepare preprocessed, clean text data.

[0604] Step 3:

[0605] Generating a confidence score

[0606] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The input includes preprocessed, clean review data. The server passes this to the generative AI model and obtains a credibility score for each review as output.

[0607] Specific behavior:

[0608] The server obtains the preprocessed review data.

[0609] Pass word-of-mouth data as input to a generative AI model.

[0610] The AI ​​model analyzes the data and generates a confidence score.

[0611] Get the generated confidence score.

[0612] Step 4:

[0613] filtering

[0614] The server filters the review data based on the generated confidence scores. The input includes the review data and its confidence scores. The server removes reviews with low confidence scores and selects reviews with high scores as output.

[0615] Specific behavior:

[0616] The server obtains the generated word-of-mouth data with the reliability scores.

[0617] Weed out low-trust reviews based on their trust score.

[0618] Select reliable review data with high scores.

[0619] Step 5:

[0620] Saving to a database

[0621] The filtered, reliable review data is stored in a database. The input includes the filtered review data. The server stores this in a database and creates an index to enable quick searches.

[0622] Specific behavior:

[0623] The server obtains filtered, highly reliable word-of-mouth data.

[0624] These data are stored in a database.

[0625] Create indexes in your database to enable efficient searches.

[0626] Step 6:

[0627] Search and Display

[0628] When a user searches for review information on their device, the device sends the request to the server. The input includes the user's search query. The server retrieves relevant, reliable review data from the database and returns it to the device as output. The user can then view it on their device.

[0629] Specific behavior:

[0630] A user inputs a search query through a terminal.

[0631] The terminal sends this search query to the server.

[0632] The server retrieves relevant and reliable word-of-mouth data from the database.

[0633] The server returns the acquired data to the terminal.

[0634] The user checks the search results on the terminal.

[0635] (Application example 1)

[0636] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0637] Conventional word-of-mouth information has been problematic in terms of its reliability, particularly in the case of electronic payments, where inappropriate information is mixed in with product and service reviews. This puts users at risk of making purchasing decisions based on inappropriate information, making it difficult to select appropriate products and services. Furthermore, providing reliable word-of-mouth information in real time is difficult, making it difficult to support timely purchasing decisions. The purpose of this invention is to solve these problems by efficiently providing reliable word-of-mouth information and ensuring users' peace of mind when making electronic payments.

[0638] 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.

[0639] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for storing the filtered reliable word-of-mouth data, means for integrating the filtered reliable word-of-mouth data with an electronic payment function to enable users to obtain reliable word-of-mouth information before purchasing a product or service, and means for providing the stored reliable word-of-mouth data in response to a user's search request, thereby enabling users to easily obtain reliable word-of-mouth information before making an electronic payment and select products or services with confidence.

[0640] "Sources" are data collection sources such as websites and APIs that provide review data.

[0641] "Review Data" refers to opinions, ratings, or impressions written by users about a particular product or service.

[0642] "Preprocessing" is the process of removing noise from collected word-of-mouth data and converting it into a format suitable for analysis.

[0643] A "generative artificial intelligence model" is a machine learning model for analyzing large amounts of text data, and has the ability to evaluate the reliability of word-of-mouth reviews.

[0644] The "trustworthiness score" is the result of an analysis of word-of-mouth data by a generative artificial intelligence model, and is a numerical representation of the trustworthiness of the word-of-mouth.

[0645] "Filtering" is the process of removing low-reliability reviews based on their reliability scores, leaving only high-reliability reviews.

[0646] "Electronic payment function" is a system that allows users to make instant payments when purchasing goods or services.

[0647] A "search request" is a query or request entered by a user to search for specific reviews.

[0648] "Storage" refers to the act of storing filtered, reliable word-of-mouth data in a database.

[0649] "Providing" means displaying the stored highly reliable word-of-mouth data on the user's terminal in response to a user's search request.

[0650] MODE FOR CARRYING OUT THE INVENTION

[0651] This invention is a system that collects word-of-mouth data from multiple sources and provides users with only reliable word-of-mouth information. It can also be integrated with electronic payment functions, allowing users to select products and services with confidence based on reliable word-of-mouth information.

[0652] System Configuration

[0653] The system consists of the following main tools:

[0654] 1. Data Collection Methods

[0655] 2. Text Preprocessing Methods

[0656] 3. Evaluation Methods Using Generative AI Models

[0657] 4. Filtering Methods

[0658] 5. Data storage means

[0659] 6. Electronic payment function integration means

[0660] 7. Search and Display Methods

[0661] Hardware and software used

[0662] Hardware:

[0663] server

[0664] User device (smartphone)

[0665] software:

[0666] Python

[0667] SQLite

[0668] BeautifulSoup

[0669] Scikit-learn

[0670] Gensim

[0671] Explanation of data processing and data calculation

[0672] Data collection methods:

[0673] The server collects review data from multiple sources using web scraping technology and APIs, and the collected raw data is stored in temporary storage.

[0674] Text preprocessing methods:

[0675] The server pre-processes the collected review data and converts it into a format suitable for analysis, including normalization, tokenization, stop-word removal, stemming, and lemmatization.

[0676] Evaluation method using generative artificial intelligence model:

[0677] The server inputs the pre-processed review data into a generative AI model that uses a Transformer-based model to generate a credibility score for each review.

[0678] Filtering methods:

[0679] The server filters reviews based on the reliability scores obtained from the generative AI model, eliminating reviews with low scores and leaving only reviews with high reliability scores.

[0680] Data storage method:

[0681] The filtered, reliable review data is stored in an SQLite database, which is indexed to enable fast searches.

[0682] Electronic payment integration methods:

[0683] When a user searches for reviews before purchasing a product or service, the server retrieves filtered reviews from the database and displays them on the user's terminal. Based on the reviews, the user can make an electronic payment.

[0684] Search and display methods:

[0685] When a user makes a search request, the server retrieves the stored, reliable word-of-mouth data from the database and returns it to the user's device. The user can then check the reliable word-of-mouth information and make a purchase decision with confidence.

[0686] Specific examples

[0687] For example, if a user tries to search for reviews of "Restaurant A," the server collects review data about "Restaurant A" from multiple sources. The collected review data undergoes text preprocessing and is evaluated by a generative artificial intelligence model. After low-reliability reviews are filtered out, the highly reliable review data is stored in a database. When a user searches for "Restaurant A," the filtered, highly reliable review information is displayed on the user's device. After viewing this review information, the user can safely make reservations and payments for "Restaurant A" using the electronic payment function.

[0688] Prompt Sentence Examples

[0689] "How can I search for reviews of Restaurant A using a filtered app that supports electronic payments?"

[0690] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0691] Step 1:

[0692] The server collects review data from multiple sources, including web scraping techniques and APIs, and stores the collected raw data in temporary storage.

[0693] Input: URLs or API endpoints of multiple review sites

[0694] Output: Raw reviews

[0695] Step 2:

[0696] The server preprocesses the collected review data, normalizing and tokenizing the text, removing stop words, stemming, and lemmatizing it to convert it into a format suitable for analysis.

[0697] Input: Raw review data

[0698] Output: Preprocessed review data

[0699] Step 3:

[0700] The server evaluates the preprocessed review data using a generative AI model (Transformer-based) to generate a credibility score. Specifically, the model analyzes the text, identifies slander and stealth marketing, and evaluates its credibility.

[0701] Input: Preprocessed review data

[0702] Output: Review data with credibility scores

[0703] Step 4:

[0704] The server filters out bad reviews based on the generated reliability score, removing reviews with low reliability scores and selecting only reliable reviews with high scores.

[0705] Input: Review data with reliability scores

[0706] Output: Filtered and reliable review data

[0707] Step 5:

[0708] The server stores the filtered, reliable review data in an SQLite database, which is indexed as it is stored, allowing for fast searches of the data.

[0709] Input: Filtered, reliable review data

[0710] Output: Review data stored in a database

[0711] Step 6:

[0712] When a user sends a search request for a specific product or service from a terminal, the server retrieves relevant, reliable word-of-mouth data from the database and returns it to the user terminal.

[0713] Input: User search request

[0714] Output: Reliable review data displayed on the user's device

[0715] Step 7:

[0716] Users select products and services based on the displayed reviews and make purchases using the electronic payment function. The server provides the reviews and electronic payment function in an integrated manner, allowing users to make payments with peace of mind.

[0717] Input: User purchase selection

[0718] Output: Completed electronic payment and purchase details

[0719] 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.

[0720] The present invention combines a system that automatically filters malicious reviews from a vast amount of review information and provides users with only reliable reviews, with an emotion engine that recognizes user emotions. Specific operations and embodiments of this system are described below.

[0721] System Configuration

[0722] The system consists of the following main components:

[0723] 1. Data collection module (server)

[0724] 2. Text preprocessing module (server)

[0725] 3. Evaluation module (server) based on generative artificial intelligence model

[0726] 4. Filtering module (server)

[0727] 5. Emotion engine module (server)

[0728] 6. Database Module (Server)

[0729] 7. Search and Display Module (Terminal)

[0730] System Operation

[0731] Data collection module (server)

[0732] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtains review information. If an API is provided, it also uses that API to obtain structured data. The collected raw data is stored in temporary storage.

[0733] Text preprocessing module (server)

[0734] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[0735] Text normalization (removal of special characters and extra whitespace)

[0736] Tokenization (breaking text into words and phrases)

[0737] Stop word removal (elimination of meaningless words)

[0738] Stemming and lemmatization (conversion of words to their root forms)

[0739] Evaluation module (server) based on generative artificial intelligence model

[0740] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[0741] Filtering module (server)

[0742] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[0743] Emotion engine module (server)

[0744] The emotion engine recognizes users' emotions and individually adjusts the ratings of reviews they view. Specifically, it analyzes users' behavior and review browsing history to determine their current emotional state. Based on this information, it adjusts the credibility score of reviews and provides the most appropriate information for the user.

[0745] Database module (server)

[0746] The reliable review data adjusted by the filtering and sentiment engine is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[0747] Search and Display Module (Terminal)

[0748] When a user searches for specific review information on their device, the device sends the search request to the server. The server then searches the database for relevant, reliable review data and returns the results, adjusted by the emotion engine, to the device. The user can then view these results on their device. For example, if a user searches for reviews of "Restaurant A," only reviews about "Restaurant A" that are highly reliable and match the user's emotional state will be displayed.

[0749] Specific examples

[0750] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing the noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only factual reviews such as "The food is great. I'd like to go again."

[0751] Furthermore, the emotion engine uses the user's past behavioral history to prioritize positive reviews for users who want to avoid negative reviews about restaurants. In this way, reviews are displayed that are in line with the user's emotions. These reviews are stored in the database and are displayed when the user searches.

[0752] This system allows users to obtain highly reliable word-of-mouth information and quickly and easily acquire information that is suited to their own emotional state.

[0753] The processing flow will be explained below.

[0754] Step 1: Data Collection (Server)

[0755] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from review sites and social media platforms and extract review information. If an API is provided, it is also possible to obtain review data using that API. The collected raw data is stored in temporary storage.

[0756] Step 2: Text normalization (server)

[0757] Because the collected raw data contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary spaces, emojis, etc. to ensure the review text is clean.

[0758] Step 3: Tokenization (Server)

[0759] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into units that are easier to analyze.

[0760] Step 4: Stopword Removal (Server)

[0761] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki" (target), "ga" (ga), and "o" (to). This makes it easier to extract important information.

[0762] Step 5: Stemming and Lemmatization (Server)

[0763] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[0764] Step 6: Evaluation using a generative AI model (server)

[0765] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[0766] Step 7: Filtering malicious reviews (server)

[0767] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[0768] Step 8: Recognizing user emotions with the emotion engine (server)

[0769] The server utilizes an emotion engine to recognize the user's current emotional state based on their activity and past browsing history, including their click history, browsing time, and search keywords.

[0770] Step 9: Moderating reviews with the sentiment engine (server)

[0771] Based on the user's emotional state recognized by the emotion engine, the server optimizes reliable reviews for each individual user. For example, if the emotion engine determines that the user prefers positive content, it will prioritize positive reviews.

[0772] Step 10: Saving to the Database (Server)

[0773] The server stores the filtered and adjusted reliable review data in a database, where it creates an index to enable efficient searches.

[0774] Step 11: Receiving a review search request (terminal, user)

[0775] The user sends a request to search for specific reviews from the terminal, which then sends the request to the server.

[0776] Step 12: Retrieving from the Database (Server)

[0777] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[0778] Step 13: Returning and displaying search results (server, terminal)

[0779] The server returns the search results to the device, which then displays the received, highly reliable review data to the user. Because the results have been adjusted by the emotion engine, reviews that best fit the user's emotional state are displayed.

[0780] Example 2

[0781] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0782] In today's information society, unreliable word-of-mouth information is overflowing, making it difficult for consumers to obtain accurate and reliable information. Furthermore, the lack of information provided in response to consumers' emotional state often leads to low user satisfaction. In this situation, there is an urgent need to provide a means to accurately filter malicious word-of-mouth and provide word-of-mouth information that takes into account the user's emotional state.

[0783] 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 means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model and generating a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for analyzing user emotions and adjusting the reliability score, means for storing reliable word-of-mouth data through filtering and emotion analysis, and means for providing the stored reliable word-of-mouth data in response to a user's search request. This makes it possible to quickly and accurately provide reliable word-of-mouth information tailored to the user's emotional state.

[0784] "Sources" are data providers such as online word-of-mouth sites, review sites, and social media platforms.

[0785] "Word-of-mouth data" is text data such as opinions, impressions, and evaluations written by users about products and services.

[0786] "Preprocessing" refers to processes such as normalizing text, tokenizing, removing stop words, stemming, and lemmatizing to make it easier to analyze review data.

[0787] A "generative artificial intelligence model" refers to a large-scale Transformer-based language model used in natural language processing tasks, which analyzes the context of reviews and generates a reliability score.

[0788] The "trust score" is an index that indicates the trustworthiness of a review, calculated by a generative artificial intelligence model by analyzing the context of the review.

[0789] "Filtering" is the process of selecting and removing malicious reviews based on their reliability scores.

[0790] "Emotion analysis" is a process of estimating a user's current emotional state based on their past behavioral patterns and browsing history.

[0791] "Data storage" means storing reliable word-of-mouth data selected through filtering and sentiment analysis in a database.

[0792] A "search request" is a query sent by a user from a terminal in an attempt to search for specific word-of-mouth information.

[0793] "Providing means" refers to the process in which the server returns highly reliable word-of-mouth data that has been filtered and sentiment analyzed to the user.

[0794] "Transformer-based" is an artificial intelligence model architecture built on self-attention mechanisms and deep learning techniques.

[0795] An "index" is a data structure used to speed up database searches and is created for a particular column of data.

[0796] This system preprocesses review data collected from multiple sources, evaluates and filters it using a generative artificial intelligence model, and provides users with highly reliable data. Furthermore, it aims to increase user satisfaction by analyzing user sentiment and adjusting the display of review data.

[0797] System Configuration

[0798] The system consists of the following main components:

[0799] 1. Data collection module (server)

[0800] 2. Text preprocessing module (server)

[0801] 3. Evaluation module (server) based on generative artificial intelligence model

[0802] 4. Filtering module (server)

[0803] 5. Emotion engine module (server)

[0804] 6. Database Module (Server)

[0805] 7. Search and Display Module (Terminal)

[0806] System Operation

[0807] Data collection module (server)

[0808] The server uses BeautifulSoup to retrieve HTML data from multiple review sites, analyze it, and collect review information. It also retrieves structured data via APIs from services that provide them. The collected data is saved in temporary storage. A specific example is a process that creates a list of URLs for "review site A" and extracts reviews from each page.

[0809] Text preprocessing module (server)

[0810] The server performs preprocessing on the collected review data, including normalization, tokenization, stop word removal, stemming, and lemmatization, using Python's NLTK library and SpaCy. For example, it removes noise from sentences like "This restaurant is great" and formats them into a form that is easier to analyze.

[0811] Evaluation module (server) based on generative artificial intelligence model

[0812] The server inputs the preprocessed text data into a generative artificial intelligence model such as GPT-4. The model analyzes the context of the review and generates a credibility score. An example prompt is, "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[0813] Filtering module (server)

[0814] The server filters the review data based on the generated reliability score, removing reviews with low reliability scores and allowing only the remaining high-score reviews to proceed to the next step, thereby eliminating malicious reviews.

[0815] Emotion engine module (server)

[0816] The emotion engine analyzes a user's past behavior and browsing history to estimate their current emotional state. Based on this, it readjusts the reliability score and provides the most appropriate review information to the user. For example, it prioritizes positive reviews for a user who has avoided negative reviews in the past.

[0817] Database module (server)

[0818] The reliable review data, adjusted by the filtering and sentiment engine, is stored in a database using MySQL or PostgreSQL, and indexes are created to enable efficient searches.

[0819] Search and Display Module (Terminal)

[0820] When a user enters a search query on their device, the device sends that information to the server. The server searches the database for relevant review data and returns the results to the device. The user can then check the results on their device. For example, if a user searches for reviews of "Restaurant B," only reliable reviews will be displayed.

[0821] This system allows users to quickly and accurately obtain reliable word-of-mouth information, and also provides information tailored to their emotional state.

[0822] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0823] Step 1:

[0824] Data collection

[0825] Input: The server receives a list of URLs for multiple review sites.

[0826] Specific operation: The server uses a web scraping tool such as BeautifulSoup or Selenium to access each review site based on the URL list and obtain HTML data.

[0827] Data processing: Specific tags and attributes are analyzed from the acquired HTML data to extract word-of-mouth information.

[0828] Output: The extracted reviews (raw data) are saved in temporary storage in JSON format.

[0829] Step 2:

[0830] Text Preprocessing

[0831] Input: The server reads the collected raw data (word-of-mouth information) from the storage.

[0832] Specific operation: The server uses Python's NLTK library and SpaCy to perform text normalization, tokenization, stop word removal, stemming, and lemmatization.

[0833] Data processing: Remove special characters, split words, eliminate meaningless words, and convert words to their root forms.

[0834] Output: Preprocessed text data is obtained and put into a usable format for the next step.

[0835] Step 3:

[0836] Evaluation using generative artificial intelligence models

[0837] Input: The server reads the preprocessed text data.

[0838] How it works: The server inputs a prompt into a generative artificial intelligence model (e.g., GPT-4) and begins the process of assessing the trustworthiness of each review.

[0839] Example prompt: "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[0840] Data computation: The AI ​​model analyzes the context and generates a reliability score.

[0841] Output: Data containing the trustworthiness score and rating reason for each review.

[0842] Step 4:

[0843] filtering

[0844] Input: The server reads the data with reliability scores output from the generative artificial intelligence model.

[0845] Specific operation: The server filters the review data based on the reliability score. It sets a threshold and removes reviews with a score below that threshold.

[0846] Data processing: Remove reviews with low scores and keep only reviews with high scores.

[0847] Output: Filtered and reliable word-of-mouth data.

[0848] Step 5:

[0849] Emotion Engine

[0850] Input: The server loads filtered, reliable word-of-mouth data and user behavior history data.

[0851] How it works: The server uses machine learning models to estimate the user's emotional state by analyzing past click history, time spent on the site, search history, etc.

[0852] Data calculation: Estimate the sentiment and recalculate / adjust the reliability score of the review data.

[0853] Output: Word-of-mouth data adjusted according to the user's emotional state.

[0854] Step 6:

[0855] Database storage

[0856] Input: The server loads reliable review data that has been tuned by the sentiment engine.

[0857] Specific operation: The server inserts data into a database such as MySQL or PostgreSQL.

[0858] Data processing: Creating indexes to enable efficient searches when stored.

[0859] Output: Reliable indexed reviews stored in a database.

[0860] Step 7:

[0861] Search and Display

[0862] Input: The terminal receives a search query from a user.

[0863] Specific operation: The terminal sends a search query to the server, which queries the database and retrieves the relevant review data.

[0864] Data processing: Organize search results and format them in a way that is easy for users to view.

[0865] Output: The terminal displays the reliable word-of-mouth results received from the server to the user.

[0866] (Application example 2)

[0867] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0868] Current review filtering systems have difficulty accurately filtering out malicious and defamatory content, resulting in a large amount of information that users find unreliable. Furthermore, they lack the ability to provide appropriate reviews that reflect the user's individual emotional state and interests, making it difficult for users to quickly and accurately obtain the information they need. This creates a demand for reliable review display, especially on online shopping sites, but current systems make it difficult to achieve this.

[0869] 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.

[0870] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious reviews based on the reliability score, means for saving the filtered reliable word-of-mouth data, means for recognizing the emotional state of a user and adjusting the reliability score according to the emotional state, and means for providing the saved reliable word-of-mouth data in response to a user's search request. This enables users to quickly and accurately obtain reliable word-of-mouth information, and further realizes the provision of information optimized for the user's emotional state.

[0871] "Sources" are data providers, including websites and applications on the Internet, from which word-of-mouth data can be obtained.

[0872] "Word-of-mouth data" is text data in which users write their evaluations and opinions about a certain product or service.

[0873] "Preprocessing" refers to a series of processes that convert review data into a form that is easier to analyze, including text normalization and tokenization.

[0874] A "generative artificial intelligence model" is a type of AI technology that learns useful patterns from massive amounts of data and evaluates and predicts new data.

[0875] The "trustworthiness score" is a number calculated as a result of evaluating whether each review contains defamatory or false information.

[0876] "Filtering" is the process of removing bad reviews based on their credibility score.

[0877] "Emotional state" refers to the user's current psychological and emotional state, and is based on their past behavioral history and current operating situation.

[0878] "Data collection means" refers to a program or device for automatically collecting word-of-mouth data.

[0879] "Data preprocessing means" refers to a program or device that formats collected data into a format that is easy to analyze.

[0880] An "evaluation means" is a program or device that analyzes preprocessed data using a generative artificial intelligence model and generates a reliability score.

[0881] A "filtering means" is a program or device for filtering out malicious reviews based on reliability scores.

[0882] "Storage means" refers to a program or device for storing filtered, highly reliable word-of-mouth data.

[0883] An "emotion recognition means" is a program or device for recognizing a user's emotional state and adjusting the reliability score.

[0884] The "providing means" refers to a program or device for displaying or providing highly reliable word-of-mouth data that has been saved in response to a user's search request.

[0885] A "natural language processing engine" is a program or device that provides technology that enables computers to understand, interpret, and generate human language.

[0886] The present invention aims to provide a specific information provision system that allows users to efficiently and accurately obtain reliable word-of-mouth information. This system collects word-of-mouth data from multiple sources, evaluates it using a generative artificial intelligence model, and further adjusts the reliability score based on the user's emotional state to provide reliable word-of-mouth information.

[0887] System Configuration

[0888] The system consists of the following components:

[0889] 1. Data collection method (server)

[0890] 2. Data preprocessing means (server)

[0891] 3. Evaluation Method (Server)

[0892] 4. Filtering Method (Server)

[0893] 5. Storage means (server)

[0894] 6. Emotion Recognition Method (Server)

[0895] 7. Means of provision (terminal)

[0896] Data collection method (server)

[0897] The server collects review information from multiple review data sources using web scraping technology and APIs, and converts the collected data from unstructured data into structured data.

[0898] Data preprocessing means (server)

[0899] The server performs preprocessing on the collected review data, specifically normalizing the text, tokenizing it, removing stop words, etc. This process converts the text data into a format that is easier to analyze.

[0900] Evaluation method (server)

[0901] The server then inputs the pre-processed data into a generative AI model, a Transformer-based AI model, which evaluates the reliability of the data and generates a reliability score for each review.

[0902] Filtering method (server)

[0903] Based on the generated credibility score, the server filters the reviews: reviews with low credibility scores are rejected and only high credibility reviews are passed on to the next processing step.

[0904] Storage method (server)

[0905] The filtered review data is stored in a database, where it is indexed for fast search.

[0906] Emotion recognition means (server)

[0907] The server uses a natural language processing engine to recognize the user's emotional state, determining it based on the user's past browsing history and current operating context, and adjusting the reliability score accordingly.

[0908] Means of provision (terminal)

[0909] When a user searches for specific reviews using their device, the device sends a search request to the server, which searches the stored reliable reviews data and returns results adjusted based on the user's emotional state to the device, where the user can view the results on their device.

[0910] Specific examples

[0911] For example, a server collects review data for electronic products from a shopping website. This data may contain defamatory content or unnatural positive comments. The server preprocesses the data and evaluates it using a generative artificial intelligence model. Inappropriate reviews receive a low reliability score and are filtered out. An emotion recognition tool then reevaluates the reviews based on the user's context, providing tailored review information.

[0912] Prompt Sentence Examples

[0913] "Create a system to automatically extract reliable reviews and provide optimal information based on user sentiment. The specific steps are as follows: collection, preprocessing, rating, filtering, and sentiment adjustment. Please provide a detailed explanation of the process performed at each step, and include any necessary modules or libraries."

[0914] The above system configuration and processing enable users to obtain highly reliable word-of-mouth information efficiently and accurately.

[0915] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0916] Step 1: Data collection

[0917] The server collects review data from multiple sources. To do so, it uses web scraping technology and APIs. Specifically, the server accesses specified URLs or API endpoints to obtain HTML data or structured data. The obtained data is converted from unstructured data to structured data and temporarily stored.

[0918] Input: Source URL or API endpoint

[0919] Output: Collected review data (text format)

[0920] Step 2: Data Preprocessing

[0921] The server preprocesses the collected review data. Specifically, it removes special characters and unnecessary spaces from the data, and uses natural language processing techniques to tokenize and remove stop words. The text data is then converted into a format that is easy to analyze.

[0922] Input: Collected review data

[0923] Output: Preprocessed review data

[0924] Step 3: Evaluation using a generative artificial intelligence model

[0925] The server inputs the preprocessed review data into a generative AI model. A Transformer-based AI model is used here. The model evaluates the reliability of the text data and generates a reliability score for each review. The reliability score evaluates the likelihood that the review contains defamation or stealth marketing.

[0926] Input: Preprocessed review data

[0927] Output: Trustworthiness score for each review

[0928] Step 4: Filtering

[0929] The server filters the reviews based on the generated credibility score. Reviews with low credibility scores are deemed to be defamatory or stealth marketing and are removed. Only reviews with high credibility scores proceed to the next processing step.

[0930] Input: Confidence score

[0931] Output: Filtered and reliable review data

[0932] Step 5: Save Data

[0933] The server stores the filtered, reliable review data in a database, where an index is created to enable fast searches, allowing for quick retrieval of relevant information later.

[0934] Input: Filtered, reliable review data

[0935] Output: Review data stored in the database

[0936] Step 6: Emotion Recognition

[0937] The server recognizes the user's emotional state by analyzing the user's past browsing history and current operating status using a natural language processing engine to determine the user's current emotional state, and then readjusts the reliability score based on this.

[0938] Input: User browsing history, operation status

[0939] Output: Adjusted confidence score

[0940] Step 7: Provide data

[0941] When a user searches for specific review information on their device, the device sends the search request to the server, which then searches the database for relevant, reliable review data and returns results tailored based on the user's emotional state.

[0942] Input: User search request

[0943] Output: Reliable reviews (adjusted based on users' emotional state)

[0944] The above are the specific processing steps from collection to provision of word-of-mouth data.

[0945] 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.

[0946] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0947] 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.

[0948] [Third embodiment]

[0949] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0950] 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.

[0951] 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).

[0952] 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.

[0953] 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.

[0954] 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).

[0955] 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.

[0956] 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.

[0957] 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.

[0958] 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.

[0959] 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.

[0960] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0961] The present invention is a system that automatically filters malicious reviews from a vast amount of review information and provides users with only highly reliable reviews. Specific operations and embodiments of the system will be described below.

[0962] System Configuration

[0963] The system consists of the following main components:

[0964] 1. Data collection module (server)

[0965] 2. Text preprocessing module (server)

[0966] 3. Evaluation module (server) based on generative artificial intelligence model

[0967] 4. Filtering module (server)

[0968] 5. Database module (server)

[0969] 6. Search and Display Module (Terminal)

[0970] System Operation

[0971] Data collection module (server)

[0972] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtain review information. If an API is provided, it also collects data using that API. The collected raw data is stored in temporary storage.

[0973] Text preprocessing module (server)

[0974] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[0975] Text normalization (removal of special characters and extra whitespace)

[0976] Tokenization (breaking text into words and phrases)

[0977] Stop word removal (elimination of meaningless words)

[0978] Stemming and lemmatization (conversion of words to their root forms)

[0979] Evaluation module (server) based on generative artificial intelligence model

[0980] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The generative AI model analyzes large amounts of text data and is capable of identifying malicious elements such as slander and stealth marketing. This allows each review to be evaluated for credibility.

[0981] Filtering module (server)

[0982] The server filters reviews based on the reliability score obtained from the generative AI model. Specifically, it removes reviews with low reliability scores and leaves only high-score, reliable reviews. This improves the quality of information provided to users.

[0983] Database module (server)

[0984] The filtered, reliable review data is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[0985] Search and Display Module (Terminal)

[0986] When a user searches for specific review information on their device, the device sends the search request to the server. The server retrieves the relevant reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only reliable reviews about "Restaurant A" are displayed.

[0987] Specific examples

[0988] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These reviews are stored in the database and displayed when users search.

[0989] This system allows users to obtain highly reliable word-of-mouth information, enabling them to select products and services with confidence.

[0990] The processing flow will be explained below.

[0991] Step 1: Data Collection (Server)

[0992] The server uses web scraping technology to collect review data from multiple sources. Specifically, it accesses designated review sites, analyzes the HTML data of the pages, and extracts review information. If an API is provided, it also uses the API to obtain structured data.

[0993] Step 2: Text normalization (server)

[0994] Because the collected raw data contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary spaces, emojis, etc. to ensure the review text is clean.

[0995] Step 3: Tokenization (Server)

[0996] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into individual words that are easier to analyze.

[0997] Step 4: Stopword Removal (Server)

[0998] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki," "wa," and "wo." This makes it easier to extract important information.

[0999] Step 5: Stemming and Lemmatization (Server)

[1000] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[1001] Step 6: Evaluation using a generative AI model (server)

[1002] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[1003] Step 7: Filtering malicious reviews (server)

[1004] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[1005] Step 8: Saving to the Database (Server)

[1006] The server stores the filtered and reliable review data in a database, creating an index to enable efficient searches.

[1007] Step 9: Receiving a review search request (terminal, user)

[1008] A user sends a request to search for reviews of a particular product or service from a device, which then sends the request to the server.

[1009] Step 10: Retrieving from the database (server)

[1010] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[1011] Step 11: Returning and displaying search results (server, terminal)

[1012] The server returns the search results to the device, which then displays the received, highly reliable word-of-mouth data to the user, allowing the user to select products and services with confidence.

[1013] Example 1

[1014] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1015] The vast amount of word-of-mouth information contains unreliable and malicious information, making it difficult for users to quickly obtain reliable information. Furthermore, such inappropriate reviews can have a negative impact on product and service ratings. To solve these problems, a system that can automatically detect and filter unreliable reviews is needed.

[1016] 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.

[1017] In this invention, the server includes means for acquiring text data from a plurality of data sources, means for analyzing and preprocessing the acquired text data, means for evaluating the preprocessed text data using a generative model to generate a reliability score, means for filtering inappropriate text data based on the reliability score, means for saving the filtered highly reliable text data, and means for providing the saved highly reliable text data in response to a user search request, thereby enabling the automatic provision of highly reliable word-of-mouth information.

[1018] "Data source" refers to the place or system from which information is obtained, specifically including websites and APIs.

[1019] "Text data" refers to a collection of information expressed as a string of characters, and includes word-of-mouth, reviews, comments, etc.

[1020] "Preprocessing" refers to the process of converting raw text data into a parsable form, and includes text normalization, tokenization, stop word removal, stemming, and lemmatization.

[1021] "Generative models" refer to artificial intelligence algorithms that generate new data based on large amounts of text data, and specifically include Transformer models.

[1022] A "reliability score" is a numerical evaluation of the reliability of text data, with higher reliability being given a higher score.

[1023] "Filtering" refers to the process of removing inappropriate text data based on confidence scores.

[1024] An "index" refers to a data structure that enables fast searches and contains search keys and their location information.

[1025] A "search request" refers to a request made by a user to locate specific information.

[1026] "Storage" refers to the process of saving data for later use, including storing it in a storage system such as a database.

[1027] The present invention is a system for providing reliable word-of-mouth information. The system is comprised of a number of components, each of which performs a specific function.

[1028] Data collection module (server)

[1029] The server has a module that collects review information. This module acquires text data from multiple data sources, specifically using web scraping technology and APIs to collect data. For example, the server uses the API of "review site Y" to collect review data about "restaurant A." This collected data is temporarily stored in storage.

[1030] Text preprocessing module (server)

[1031] The server has a module that preprocesses the collected review data. This module performs tasks such as removing special characters and extra whitespace, normalizing the text, tokenizing, removing stop words, and performing stemming and lemmatization. For example, a review such as "This restaurant is terrible!" can be normalized to "This restaurant is terrible."

[1032] Evaluation module (server) based on generative artificial intelligence model

[1033] After preprocessing, the server evaluates the review data using a generative artificial intelligence model. This model is a Transformer-based generative model that analyzes the reliability of the text and generates a reliability score. For example, a review that says, "This restaurant is terrible!" will receive a low score, while a review that says, "The food is great. I'd like to go again" will receive a high score.

[1034] Filtering module (server)

[1035] The server has a module that filters the review data based on the generated confidence score. Low-confidence reviews are removed, and only high-confidence reviews remain. For example, a review such as "All the staff are friends!" is removed, while a review such as "The food is great. I'd love to go again." remains.

[1036] Database module (server)

[1037] The filtered, highly reliable review data is stored in a database module on the server. This module creates an index of the stored data to enable quick searches. For example, the highly reliable reviews for "Restaurant A" are stored in the database.

[1038] Search and Display Module (Terminal)

[1039] When a user searches for reviews on their device, the device sends the request to the server. The server retrieves the relevant highly reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only highly reliable reviews will be displayed.

[1040] Specific examples

[1041] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data includes defamatory comments such as "This restaurant is terrible," as well as unnatural positive comments such as "All the staff are friends!" When the text preprocessing module removes this noise and the data is evaluated by a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These saved reviews are then displayed when users search. This system allows users to obtain reliable review information, allowing them to choose products and services with confidence.

[1042] Prompt Sentence Examples

[1043] Example prompts to input to a generative AI model:

[1044] "Data collected: 'This restaurant is awful. The food was cold and the waiter was unfriendly.'"

[1045] "Data collected: 'All the staff are friends! It was amazing.'"

[1046] "Data collected: 'The food is amazing. I'd love to go back.'"

[1047] Generate a credibility score for each review.

[1048] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1049] Step 1:

[1050] Data collection

[1051] The server obtains review information from multiple data sources. Specifically, it uses web scraping technology and APIs to collect review data. Inputs include web page URLs and API endpoints. The server sends HTTP requests for these inputs and receives HTML data or structured data as a response. This data is then stored in temporary storage.

[1052] Specific behavior:

[1053] The server receives the URL of "review site Y" as input.

[1054] The server sends an HTTP request to the specified URL and receives HTML data as a response.

[1055] The server analyzes the HTML data and extracts the necessary word-of-mouth information.

[1056] The extracted word-of-mouth information is stored in temporary storage.

[1057] Step 2:

[1058] Text Preprocessing

[1059] The server preprocesses the collected review data. For example, it removes special characters and extra whitespace, normalizes text, tokenizes, removes stop words, and performs stemming and lemmatization. The input includes the collected raw review data. The server preprocesses this input data and outputs clean text data suitable for analysis.

[1060] Specific behavior:

[1061] The server retrieves the raw word-of-mouth data from the temporary storage.

[1062] Normalizes text by removing special characters and extra whitespace.

[1063] Break down (tokenize) the normalized text into words and phrases.

[1064] Stop words are removed and necessary words are converted to their base form using stemming and lemmatization.

[1065] Prepare preprocessed, clean text data.

[1066] Step 3:

[1067] Generating a confidence score

[1068] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The input includes preprocessed, clean review data. The server passes this to the generative AI model and obtains a credibility score for each review as output.

[1069] Specific behavior:

[1070] The server obtains the preprocessed review data.

[1071] Pass word-of-mouth data as input to a generative AI model.

[1072] The AI ​​model analyzes the data and generates a confidence score.

[1073] Get the generated confidence score.

[1074] Step 4:

[1075] filtering

[1076] The server filters the review data based on the generated confidence scores. The input includes the review data and its confidence scores. The server removes reviews with low confidence scores and selects reviews with high scores as output.

[1077] Specific behavior:

[1078] The server obtains the generated word-of-mouth data with the reliability scores.

[1079] Weed out low-trust reviews based on their trust score.

[1080] Select reliable review data with high scores.

[1081] Step 5:

[1082] Saving to a database

[1083] The filtered, reliable review data is stored in a database. The input includes the filtered review data. The server stores this in a database and creates an index to enable quick searches.

[1084] Specific behavior:

[1085] The server obtains filtered, highly reliable word-of-mouth data.

[1086] These data are stored in a database.

[1087] Create indexes in your database to enable efficient searches.

[1088] Step 6:

[1089] Search and Display

[1090] When a user searches for review information on their device, the device sends the request to the server. The input includes the user's search query. The server retrieves relevant, reliable review data from the database and returns it to the device as output. The user can then view it on their device.

[1091] Specific behavior:

[1092] A user inputs a search query through a terminal.

[1093] The terminal sends this search query to the server.

[1094] The server retrieves relevant and reliable word-of-mouth data from the database.

[1095] The server returns the acquired data to the terminal.

[1096] The user checks the search results on the terminal.

[1097] (Application example 1)

[1098] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1099] Conventional word-of-mouth information has been problematic in terms of its reliability, particularly in the case of electronic payments, where inappropriate information is mixed in with product and service reviews. This puts users at risk of making purchasing decisions based on inappropriate information, making it difficult to select appropriate products and services. Furthermore, providing reliable word-of-mouth information in real time is difficult, making it difficult to support timely purchasing decisions. The purpose of this invention is to solve these problems by efficiently providing reliable word-of-mouth information and ensuring users' peace of mind when making electronic payments.

[1100] 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.

[1101] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for storing the filtered reliable word-of-mouth data, means for integrating the filtered reliable word-of-mouth data with an electronic payment function to enable users to obtain reliable word-of-mouth information before purchasing a product or service, and means for providing the stored reliable word-of-mouth data in response to a user's search request, thereby enabling users to easily obtain reliable word-of-mouth information before making an electronic payment and select products or services with confidence.

[1102] "Sources" are data collection sources such as websites and APIs that provide review data.

[1103] "Review Data" refers to opinions, ratings, or impressions written by users about a particular product or service.

[1104] "Preprocessing" is the process of removing noise from collected word-of-mouth data and converting it into a format suitable for analysis.

[1105] A "generative artificial intelligence model" is a machine learning model for analyzing large amounts of text data, and has the ability to evaluate the reliability of word-of-mouth reviews.

[1106] The "trustworthiness score" is the result of an analysis of word-of-mouth data by a generative artificial intelligence model, and is a numerical representation of the trustworthiness of the word-of-mouth.

[1107] "Filtering" is the process of removing low-reliability reviews based on their reliability scores, leaving only high-reliability reviews.

[1108] "Electronic payment function" is a system that allows users to make instant payments when purchasing goods or services.

[1109] A "search request" is a query or request entered by a user to search for specific reviews.

[1110] "Storage" refers to the act of storing filtered, reliable word-of-mouth data in a database.

[1111] "Providing" means displaying the stored highly reliable word-of-mouth data on the user's terminal in response to a user's search request.

[1112] MODE FOR CARRYING OUT THE INVENTION

[1113] This invention is a system that collects word-of-mouth data from multiple sources and provides users with only reliable word-of-mouth information. It can also be integrated with electronic payment functions, allowing users to select products and services with confidence based on reliable word-of-mouth information.

[1114] System Configuration

[1115] The system consists of the following main tools:

[1116] 1. Data Collection Methods

[1117] 2. Text Preprocessing Methods

[1118] 3. Evaluation Methods Using Generative AI Models

[1119] 4. Filtering Methods

[1120] 5. Data storage means

[1121] 6. Electronic payment function integration means

[1122] 7. Search and Display Methods

[1123] Hardware and software used

[1124] Hardware:

[1125] server

[1126] User device (smartphone)

[1127] software:

[1128] Python

[1129] SQLite

[1130] BeautifulSoup

[1131] Scikit-learn

[1132] Gensim

[1133] Explanation of data processing and data calculation

[1134] Data collection methods:

[1135] The server collects review data from multiple sources using web scraping technology and APIs, and the collected raw data is stored in temporary storage.

[1136] Text preprocessing methods:

[1137] The server pre-processes the collected review data and converts it into a format suitable for analysis, including normalization, tokenization, stop-word removal, stemming, and lemmatization.

[1138] Evaluation method using generative artificial intelligence model:

[1139] The server inputs the pre-processed review data into a generative AI model that uses a Transformer-based model to generate a credibility score for each review.

[1140] Filtering methods:

[1141] The server filters reviews based on the reliability scores obtained from the generative AI model, eliminating reviews with low scores and leaving only reviews with high reliability scores.

[1142] Data storage method:

[1143] The filtered, reliable review data is stored in an SQLite database, which is indexed to enable fast searches.

[1144] Electronic payment integration methods:

[1145] When a user searches for reviews before purchasing a product or service, the server retrieves filtered reviews from the database and displays them on the user's terminal. Based on the reviews, the user can make an electronic payment.

[1146] Search and display methods:

[1147] When a user makes a search request, the server retrieves the stored, reliable word-of-mouth data from the database and returns it to the user's device. The user can then check the reliable word-of-mouth information and make a purchase decision with confidence.

[1148] Specific examples

[1149] For example, if a user tries to search for reviews of "Restaurant A," the server collects review data about "Restaurant A" from multiple sources. The collected review data undergoes text preprocessing and is evaluated by a generative artificial intelligence model. After low-reliability reviews are filtered out, the highly reliable review data is stored in a database. When a user searches for "Restaurant A," the filtered, highly reliable review information is displayed on the user's device. After viewing this review information, the user can safely make reservations and payments for "Restaurant A" using the electronic payment function.

[1150] Prompt Sentence Examples

[1151] "How can I search for reviews of Restaurant A using a filtered app that supports electronic payments?"

[1152] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1153] Step 1:

[1154] The server collects review data from multiple sources, including web scraping techniques and APIs, and stores the collected raw data in temporary storage.

[1155] Input: URLs or API endpoints of multiple review sites

[1156] Output: Raw reviews

[1157] Step 2:

[1158] The server preprocesses the collected review data, normalizing and tokenizing the text, removing stop words, stemming, and lemmatizing it to convert it into a format suitable for analysis.

[1159] Input: Raw review data

[1160] Output: Preprocessed review data

[1161] Step 3:

[1162] The server evaluates the preprocessed review data using a generative AI model (Transformer-based) to generate a credibility score. Specifically, the model analyzes the text, identifies slander and stealth marketing, and evaluates its credibility.

[1163] Input: Preprocessed review data

[1164] Output: Review data with credibility scores

[1165] Step 4:

[1166] The server filters out bad reviews based on the generated reliability score, removing reviews with low reliability scores and selecting only reliable reviews with high scores.

[1167] Input: Review data with reliability scores

[1168] Output: Filtered and reliable review data

[1169] Step 5:

[1170] The server stores the filtered, reliable review data in an SQLite database, which is indexed as it is stored, allowing for fast searches of the data.

[1171] Input: Filtered, reliable review data

[1172] Output: Review data stored in a database

[1173] Step 6:

[1174] When a user sends a search request for a specific product or service from a terminal, the server retrieves relevant, reliable word-of-mouth data from the database and returns it to the user terminal.

[1175] Input: User search request

[1176] Output: Reliable review data displayed on the user's device

[1177] Step 7:

[1178] Users select products and services based on the displayed reviews and make purchases using the electronic payment function. The server provides the reviews and electronic payment function in an integrated manner, allowing users to make payments with peace of mind.

[1179] Input: User purchase selection

[1180] Output: Completed electronic payment and purchase details

[1181] 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.

[1182] The present invention combines a system that automatically filters malicious reviews from a vast amount of review information and provides users with only reliable reviews, with an emotion engine that recognizes user emotions. Specific operations and embodiments of this system are described below.

[1183] System Configuration

[1184] The system consists of the following main components:

[1185] 1. Data collection module (server)

[1186] 2. Text preprocessing module (server)

[1187] 3. Evaluation module (server) based on generative artificial intelligence model

[1188] 4. Filtering module (server)

[1189] 5. Emotion engine module (server)

[1190] 6. Database Module (Server)

[1191] 7. Search and Display Module (Terminal)

[1192] System Operation

[1193] Data collection module (server)

[1194] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtains review information. If an API is provided, it also uses that API to obtain structured data. The collected raw data is stored in temporary storage.

[1195] Text preprocessing module (server)

[1196] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[1197] Text normalization (removal of special characters and extra whitespace)

[1198] Tokenization (breaking text into words and phrases)

[1199] Stop word removal (elimination of meaningless words)

[1200] Stemming and lemmatization (conversion of words to their root forms)

[1201] Evaluation module (server) based on generative artificial intelligence model

[1202] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[1203] Filtering module (server)

[1204] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[1205] Emotion engine module (server)

[1206] The emotion engine recognizes users' emotions and individually adjusts the ratings of reviews they view. Specifically, it analyzes users' behavior and review browsing history to determine their current emotional state. Based on this information, it adjusts the credibility score of reviews and provides the most appropriate information for the user.

[1207] Database module (server)

[1208] The reliable review data adjusted by the filtering and sentiment engine is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[1209] Search and Display Module (Terminal)

[1210] When a user searches for specific review information on their device, the device sends the search request to the server. The server then searches the database for relevant, reliable review data and returns the results, adjusted by the emotion engine, to the device. The user can then view these results on their device. For example, if a user searches for reviews of "Restaurant A," only reviews about "Restaurant A" that are highly reliable and match the user's emotional state will be displayed.

[1211] Specific examples

[1212] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing the noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only factual reviews such as "The food is great. I'd like to go again."

[1213] Furthermore, the emotion engine uses the user's past behavioral history to prioritize positive reviews for users who want to avoid negative reviews about restaurants. In this way, reviews are displayed that are in line with the user's emotions. These reviews are stored in the database and are displayed when the user searches.

[1214] This system allows users to obtain highly reliable word-of-mouth information and quickly and easily acquire information that is suited to their own emotional state.

[1215] The processing flow will be explained below.

[1216] Step 1: Data Collection (Server)

[1217] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from review sites and social media platforms and extract review information. If an API is provided, it is also possible to obtain review data using that API. The collected raw data is stored in temporary storage.

[1218] Step 2: Text normalization (server)

[1219] Because the collected raw data contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary spaces, emojis, etc. to ensure the review text is clean.

[1220] Step 3: Tokenization (Server)

[1221] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into units that are easier to analyze.

[1222] Step 4: Stopword Removal (Server)

[1223] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki" (target), "ga" (ga), and "o" (to). This makes it easier to extract important information.

[1224] Step 5: Stemming and Lemmatization (Server)

[1225] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[1226] Step 6: Evaluation using a generative AI model (server)

[1227] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[1228] Step 7: Filtering malicious reviews (server)

[1229] Based on the generated reliability score, the server filters the reviews: reviews with low scores are deemed malicious and removed, while only high-scoring, reliable reviews are advanced to the next step.

[1230] Step 8: Recognizing user emotions with the emotion engine (server)

[1231] The server utilizes an emotion engine to recognize the user's current emotional state based on their activity and past browsing history, including their click history, browsing time, and search keywords.

[1232] Step 9: Moderating reviews with the sentiment engine (server)

[1233] Based on the user's emotional state recognized by the emotion engine, the server optimizes reliable reviews for each individual user. For example, if the emotion engine determines that the user prefers positive content, it will prioritize positive reviews.

[1234] Step 10: Saving to the Database (Server)

[1235] The server stores the filtered and adjusted reliable review data in a database, where it creates an index to enable efficient searches.

[1236] Step 11: Receiving a review search request (terminal, user)

[1237] The user sends a request to search for specific reviews from the terminal, which then sends the request to the server.

[1238] Step 12: Retrieving from the Database (Server)

[1239] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[1240] Step 13: Returning and displaying search results (server, terminal)

[1241] The server returns the search results to the device, which then displays the received, highly reliable review data to the user. Because the results have been adjusted by the emotion engine, reviews that best fit the user's emotional state are displayed.

[1242] Example 2

[1243] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1244] In today's information society, unreliable word-of-mouth information is overflowing, making it difficult for consumers to obtain accurate and reliable information. Furthermore, the lack of information provided in response to consumers' emotional state often leads to low user satisfaction. In this situation, there is an urgent need to provide a means to accurately filter malicious word-of-mouth and provide word-of-mouth information that takes into account the user's emotional state.

[1245] 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 means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model and generating a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for analyzing user emotions and adjusting the reliability score, means for storing reliable word-of-mouth data through filtering and emotion analysis, and means for providing the stored reliable word-of-mouth data in response to a user's search request. This makes it possible to quickly and accurately provide reliable word-of-mouth information tailored to the user's emotional state.

[1246] "Sources" are data providers such as online word-of-mouth sites, review sites, and social media platforms.

[1247] "Word-of-mouth data" is text data such as opinions, impressions, and evaluations written by users about products and services.

[1248] "Preprocessing" refers to processes such as normalizing text, tokenizing, removing stop words, stemming, and lemmatizing to make it easier to analyze review data.

[1249] A "generative artificial intelligence model" refers to a large-scale Transformer-based language model used in natural language processing tasks, which analyzes the context of reviews and generates a reliability score.

[1250] The "trust score" is an index that indicates the trustworthiness of a review, calculated by a generative artificial intelligence model by analyzing the context of the review.

[1251] "Filtering" is the process of selecting and removing malicious reviews based on their reliability scores.

[1252] "Emotion analysis" is a process of estimating a user's current emotional state based on their past behavioral patterns and browsing history.

[1253] "Data storage" means storing reliable word-of-mouth data selected through filtering and sentiment analysis in a database.

[1254] A "search request" is a query sent by a user from a terminal in an attempt to search for specific word-of-mouth information.

[1255] "Providing means" refers to the process in which the server returns highly reliable word-of-mouth data that has been filtered and sentiment analyzed to the user.

[1256] "Transformer-based" is an artificial intelligence model architecture built on self-attention mechanisms and deep learning techniques.

[1257] An "index" is a data structure used to speed up database searches and is created for a particular column of data.

[1258] This system preprocesses review data collected from multiple sources, evaluates and filters it using a generative artificial intelligence model, and provides users with highly reliable data. Furthermore, it aims to increase user satisfaction by analyzing user sentiment and adjusting the display of review data.

[1259] System Configuration

[1260] The system consists of the following main components:

[1261] 1. Data collection module (server)

[1262] 2. Text preprocessing module (server)

[1263] 3. Evaluation module (server) based on generative artificial intelligence model

[1264] 4. Filtering module (server)

[1265] 5. Emotion engine module (server)

[1266] 6. Database Module (Server)

[1267] 7. Search and Display Module (Terminal)

[1268] System Operation

[1269] Data collection module (server)

[1270] The server uses BeautifulSoup to retrieve HTML data from multiple review sites, analyze it, and collect review information. It also retrieves structured data via APIs from services that provide them. The collected data is saved in temporary storage. A specific example is a process that creates a list of URLs for "review site A" and extracts reviews from each page.

[1271] Text preprocessing module (server)

[1272] The server performs preprocessing on the collected review data, including normalization, tokenization, stop word removal, stemming, and lemmatization, using Python's NLTK library and SpaCy. For example, it removes noise from sentences like "This restaurant is great" and formats them into a form that is easier to analyze.

[1273] Evaluation module (server) based on generative artificial intelligence model

[1274] The server inputs the preprocessed text data into a generative artificial intelligence model such as GPT-4. The model analyzes the context of the review and generates a credibility score. An example prompt is, "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[1275] Filtering module (server)

[1276] The server filters the review data based on the generated reliability score, removing reviews with low reliability scores and allowing only the remaining high-score reviews to proceed to the next step, thereby eliminating malicious reviews.

[1277] Emotion engine module (server)

[1278] The emotion engine analyzes a user's past behavior and browsing history to estimate their current emotional state. Based on this, it readjusts the reliability score and provides the most appropriate review information to the user. For example, it prioritizes positive reviews for a user who has avoided negative reviews in the past.

[1279] Database module (server)

[1280] The reliable review data, adjusted by the filtering and sentiment engine, is stored in a database using MySQL or PostgreSQL, and indexes are created to enable efficient searches.

[1281] Search and Display Module (Terminal)

[1282] When a user enters a search query on their device, the device sends that information to the server. The server searches the database for relevant review data and returns the results to the device. The user can then check the results on their device. For example, if a user searches for reviews of "Restaurant B," only reliable reviews will be displayed.

[1283] This system allows users to quickly and accurately obtain reliable word-of-mouth information, and also provides information tailored to their emotional state.

[1284] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1285] Step 1:

[1286] Data collection

[1287] Input: The server receives a list of URLs for multiple review sites.

[1288] Specific operation: The server uses a web scraping tool such as BeautifulSoup or Selenium to access each review site based on the URL list and obtain HTML data.

[1289] Data processing: Specific tags and attributes are analyzed from the acquired HTML data to extract word-of-mouth information.

[1290] Output: The extracted reviews (raw data) are saved in temporary storage in JSON format.

[1291] Step 2:

[1292] Text Preprocessing

[1293] Input: The server reads the collected raw data (word-of-mouth information) from the storage.

[1294] Specific operation: The server uses Python's NLTK library and SpaCy to perform text normalization, tokenization, stop word removal, stemming, and lemmatization.

[1295] Data processing: Remove special characters, split words, eliminate meaningless words, and convert words to their root forms.

[1296] Output: Preprocessed text data is obtained and put into a usable format for the next step.

[1297] Step 3:

[1298] Evaluation using generative artificial intelligence models

[1299] Input: The server reads the preprocessed text data.

[1300] How it works: The server inputs a prompt into a generative artificial intelligence model (e.g., GPT-4) and begins the process of assessing the trustworthiness of each review.

[1301] Example prompt: "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[1302] Data computation: The AI ​​model analyzes the context and generates a reliability score.

[1303] Output: Data containing the trustworthiness score and rating reason for each review.

[1304] Step 4:

[1305] filtering

[1306] Input: The server reads the data with reliability scores output from the generative artificial intelligence model.

[1307] Specific operation: The server filters the review data based on the reliability score. It sets a threshold and removes reviews with a score below that threshold.

[1308] Data processing: Remove reviews with low scores and keep only reviews with high scores.

[1309] Output: Filtered and reliable word-of-mouth data.

[1310] Step 5:

[1311] Emotion Engine

[1312] Input: The server loads filtered, reliable word-of-mouth data and user behavior history data.

[1313] How it works: The server uses machine learning models to estimate the user's emotional state by analyzing past click history, time spent on the site, search history, etc.

[1314] Data calculation: Estimate the sentiment and recalculate / adjust the reliability score of the review data.

[1315] Output: Word-of-mouth data adjusted according to the user's emotional state.

[1316] Step 6:

[1317] Database storage

[1318] Input: The server loads reliable review data that has been tuned by the sentiment engine.

[1319] Specific operation: The server inserts data into a database such as MySQL or PostgreSQL.

[1320] Data processing: Creating indexes to enable efficient searches when stored.

[1321] Output: Reliable indexed reviews stored in a database.

[1322] Step 7:

[1323] Search and Display

[1324] Input: The terminal receives a search query from a user.

[1325] Specific operation: The terminal sends a search query to the server, which queries the database and retrieves the relevant review data.

[1326] Data processing: Organize search results and format them in a way that is easy for users to view.

[1327] Output: The terminal displays the reliable word-of-mouth results received from the server to the user.

[1328] (Application example 2)

[1329] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1330] Current review filtering systems have difficulty accurately filtering out malicious and defamatory content, resulting in a large amount of information that users find unreliable. Furthermore, they lack the ability to provide appropriate reviews that reflect the user's individual emotional state and interests, making it difficult for users to quickly and accurately obtain the information they need. This creates a demand for reliable review display, especially on online shopping sites, but current systems make it difficult to achieve this.

[1331] 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.

[1332] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious reviews based on the reliability score, means for saving the filtered reliable word-of-mouth data, means for recognizing the emotional state of a user and adjusting the reliability score according to the emotional state, and means for providing the saved reliable word-of-mouth data in response to a user's search request. This enables users to quickly and accurately obtain reliable word-of-mouth information, and further realizes the provision of information optimized for the user's emotional state.

[1333] "Sources" are data providers, including websites and applications on the Internet, from which word-of-mouth data can be obtained.

[1334] "Word-of-mouth data" is text data in which users write their evaluations and opinions about a certain product or service.

[1335] "Preprocessing" refers to a series of processes that convert review data into a form that is easier to analyze, including text normalization and tokenization.

[1336] A "generative artificial intelligence model" is a type of AI technology that learns useful patterns from massive amounts of data and evaluates and predicts new data.

[1337] The "trustworthiness score" is a number calculated as a result of evaluating whether each review contains defamatory or false information.

[1338] "Filtering" is the process of removing bad reviews based on their credibility score.

[1339] "Emotional state" refers to the user's current psychological and emotional state, and is based on their past behavioral history and current operating situation.

[1340] "Data collection means" refers to a program or device for automatically collecting word-of-mouth data.

[1341] "Data preprocessing means" refers to a program or device that formats collected data into a format that is easy to analyze.

[1342] An "evaluation means" is a program or device that analyzes preprocessed data using a generative artificial intelligence model and generates a reliability score.

[1343] A "filtering means" is a program or device for filtering out malicious reviews based on reliability scores.

[1344] "Storage means" refers to a program or device for storing filtered, highly reliable word-of-mouth data.

[1345] An "emotion recognition means" is a program or device for recognizing a user's emotional state and adjusting the reliability score.

[1346] The "providing means" refers to a program or device for displaying or providing highly reliable word-of-mouth data that has been saved in response to a user's search request.

[1347] A "natural language processing engine" is a program or device that provides technology that enables computers to understand, interpret, and generate human language.

[1348] The present invention aims to provide a specific information provision system that allows users to efficiently and accurately obtain reliable word-of-mouth information. This system collects word-of-mouth data from multiple sources, evaluates it using a generative artificial intelligence model, and further adjusts the reliability score based on the user's emotional state to provide reliable word-of-mouth information.

[1349] System Configuration

[1350] The system consists of the following components:

[1351] 1. Data collection method (server)

[1352] 2. Data preprocessing means (server)

[1353] 3. Evaluation Method (Server)

[1354] 4. Filtering Method (Server)

[1355] 5. Storage means (server)

[1356] 6. Emotion Recognition Method (Server)

[1357] 7. Means of provision (terminal)

[1358] Data collection method (server)

[1359] The server collects review information from multiple review data sources using web scraping technology and APIs, and converts the collected data from unstructured data into structured data.

[1360] Data preprocessing means (server)

[1361] The server performs preprocessing on the collected review data, specifically normalizing the text, tokenizing it, removing stop words, etc. This process converts the text data into a format that is easier to analyze.

[1362] Evaluation method (server)

[1363] The server then inputs the pre-processed data into a generative AI model, a Transformer-based AI model, which evaluates the reliability of the data and generates a reliability score for each review.

[1364] Filtering method (server)

[1365] Based on the generated credibility score, the server filters the reviews: reviews with low credibility scores are rejected and only high credibility reviews are passed on to the next processing step.

[1366] Storage method (server)

[1367] The filtered review data is stored in a database, where it is indexed for fast search.

[1368] Emotion recognition means (server)

[1369] The server uses a natural language processing engine to recognize the user's emotional state, determining it based on the user's past browsing history and current operating context, and adjusting the reliability score accordingly.

[1370] Means of provision (terminal)

[1371] When a user searches for specific reviews using their device, the device sends a search request to the server, which searches the stored reliable reviews data and returns results adjusted based on the user's emotional state to the device, where the user can view the results on their device.

[1372] Specific examples

[1373] For example, a server collects review data for electronic products from a shopping website. This data may contain defamatory content or unnatural positive comments. The server preprocesses the data and evaluates it using a generative artificial intelligence model. Inappropriate reviews receive a low reliability score and are filtered out. An emotion recognition tool then reevaluates the reviews based on the user's context, providing tailored review information.

[1374] Prompt Sentence Examples

[1375] "Create a system to automatically extract reliable reviews and provide optimal information based on user sentiment. The specific steps are as follows: collection, preprocessing, rating, filtering, and sentiment adjustment. Please provide a detailed explanation of the process performed at each step, and include any necessary modules or libraries."

[1376] The above system configuration and processing enable users to obtain highly reliable word-of-mouth information efficiently and accurately.

[1377] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1378] Step 1: Data collection

[1379] The server collects review data from multiple sources. To do so, it uses web scraping technology and APIs. Specifically, the server accesses specified URLs or API endpoints to obtain HTML data or structured data. The obtained data is converted from unstructured data to structured data and temporarily stored.

[1380] Input: Source URL or API endpoint

[1381] Output: Collected review data (text format)

[1382] Step 2: Data Preprocessing

[1383] The server preprocesses the collected review data. Specifically, it removes special characters and unnecessary spaces from the data, and uses natural language processing techniques to tokenize and remove stop words. The text data is then converted into a format that is easy to analyze.

[1384] Input: Collected review data

[1385] Output: Preprocessed review data

[1386] Step 3: Evaluation using a generative artificial intelligence model

[1387] The server inputs the preprocessed review data into a generative AI model. A Transformer-based AI model is used here. The model evaluates the reliability of the text data and generates a reliability score for each review. The reliability score evaluates the likelihood that the review contains defamation or stealth marketing.

[1388] Input: Preprocessed review data

[1389] Output: Trustworthiness score for each review

[1390] Step 4: Filtering

[1391] The server filters the reviews based on the generated credibility score. Reviews with low credibility scores are deemed to be defamatory or stealth marketing and are removed. Only reviews with high credibility scores proceed to the next processing step.

[1392] Input: Confidence score

[1393] Output: Filtered and reliable review data

[1394] Step 5: Save Data

[1395] The server stores the filtered, reliable review data in a database, where an index is created to enable fast searches, allowing for quick retrieval of relevant information later.

[1396] Input: Filtered, reliable review data

[1397] Output: Review data stored in the database

[1398] Step 6: Emotion Recognition

[1399] The server recognizes the user's emotional state by analyzing the user's past browsing history and current operating status using a natural language processing engine to determine the user's current emotional state, and then readjusts the reliability score based on this.

[1400] Input: User browsing history, operation status

[1401] Output: Adjusted confidence score

[1402] Step 7: Provide data

[1403] When a user searches for specific review information on their device, the device sends the search request to the server, which then searches the database for relevant, reliable review data and returns results tailored based on the user's emotional state.

[1404] Input: User search request

[1405] Output: Reliable reviews (adjusted based on users' emotional state)

[1406] The above are the specific processing steps from collection to provision of word-of-mouth data.

[1407] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1408] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1409] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1410] [Fourth embodiment]

[1411] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1412] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1413] 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).

[1414] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1415] 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.

[1416] 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).

[1417] 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.

[1418] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1419] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1420] 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.

[1421] 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.

[1422] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1423] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1424] The present invention is a system that automatically filters malicious reviews from a vast amount of review information and provides users with only highly reliable reviews. Specific operations and embodiments of the system will be described below.

[1425] System Configuration

[1426] The system consists of the following main components:

[1427] 1. Data collection module (server)

[1428] 2. Text preprocessing module (server)

[1429] 3. Evaluation module (server) based on generative artificial intelligence model

[1430] 4. Filtering module (server)

[1431] 5. Database module (server)

[1432] 6. Search and Display Module (Terminal)

[1433] System Operation

[1434] Data collection module (server)

[1435] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtain review information. If an API is provided, it also collects data using that API. The collected raw data is stored in temporary storage.

[1436] Text preprocessing module (server)

[1437] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[1438] Text normalization (removal of special characters and extra whitespace)

[1439] Tokenization (breaking text into words and phrases)

[1440] Stop word removal (elimination of meaningless words)

[1441] Stemming and lemmatization (conversion of words to their root forms)

[1442] Evaluation module (server) based on generative artificial intelligence model

[1443] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The generative AI model analyzes large amounts of text data and is capable of identifying malicious elements such as slander and stealth marketing. This allows each review to be evaluated for credibility.

[1444] Filtering module (server)

[1445] The server filters reviews based on the reliability score obtained from the generative AI model. Specifically, it removes reviews with low reliability scores and leaves only high-score, reliable reviews. This improves the quality of information provided to users.

[1446] Database module (server)

[1447] The filtered, reliable review data is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[1448] Search and Display Module (Terminal)

[1449] When a user searches for specific review information on their device, the device sends the search request to the server. The server retrieves the relevant reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only reliable reviews about "Restaurant A" are displayed.

[1450] Specific examples

[1451] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These reviews are stored in the database and displayed when users search.

[1452] This system allows users to obtain highly reliable word-of-mouth information, enabling them to select products and services with confidence.

[1453] The processing flow will be explained below.

[1454] Step 1: Data Collection (Server)

[1455] The server uses web scraping technology to collect review data from multiple sources. Specifically, it accesses designated review sites, analyzes the HTML data of the pages, and extracts review information. If an API is provided, it also uses the API to obtain structured data.

[1456] Step 2: Text normalization (server)

[1457] Because the raw data collected contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary whitespace, emojis, etc. to ensure the review text is clean.

[1458] Step 3: Tokenization (Server)

[1459] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into individual words that are easier to analyze.

[1460] Step 4: Stopword Removal (Server)

[1461] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki," "wa," and "wo." This makes it easier to extract important information.

[1462] Step 5: Stemming and Lemmatization (Server)

[1463] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[1464] Step 6: Evaluation using a generative AI model (server)

[1465] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[1466] Step 7: Filtering malicious reviews (server)

[1467] Based on the generated reliability score, the server filters the reviews: low-scoring reviews are deemed malicious and removed, while high-scoring, reliable reviews are advanced to the next step.

[1468] Step 8: Saving to the Database (Server)

[1469] The server stores the filtered and reliable review data in a database, creating an index to enable efficient searches.

[1470] Step 9: Receiving a review search request (terminal, user)

[1471] A user sends a request from their device to search for reviews of a particular product or service, and the device sends this request to the server.

[1472] Step 10: Retrieving from the database (server)

[1473] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[1474] Step 11: Returning and displaying search results (server, terminal)

[1475] The server returns the search results to the device, which then displays the received, highly reliable word-of-mouth data to the user, allowing the user to select products and services with confidence.

[1476] Example 1

[1477] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1478] The vast amount of word-of-mouth information contains unreliable and malicious information, making it difficult for users to quickly obtain reliable information. Furthermore, such inappropriate reviews can have a negative impact on product and service ratings. To solve these problems, a system that can automatically detect and filter unreliable reviews is needed.

[1479] 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.

[1480] In this invention, the server includes means for acquiring text data from a plurality of data sources, means for analyzing and preprocessing the acquired text data, means for evaluating the preprocessed text data using a generative model to generate a reliability score, means for filtering inappropriate text data based on the reliability score, means for saving the filtered highly reliable text data, and means for providing the saved highly reliable text data in response to a user search request, thereby enabling the automatic provision of highly reliable word-of-mouth information.

[1481] "Data source" refers to the place or system from which information is obtained, specifically including websites and APIs.

[1482] "Text data" refers to a collection of information expressed as a string of characters, and includes word-of-mouth, reviews, comments, etc.

[1483] "Preprocessing" refers to the process of converting raw text data into a parsable form, and includes text normalization, tokenization, stop word removal, stemming, and lemmatization.

[1484] "Generative models" refer to artificial intelligence algorithms that generate new data based on large amounts of text data, and specifically include Transformer models.

[1485] A "reliability score" is a numerical evaluation of the reliability of text data, with higher reliability being given a higher score.

[1486] "Filtering" refers to the process of removing inappropriate text data based on confidence scores.

[1487] An "index" refers to a data structure that enables fast searches and contains search keys and their location information.

[1488] A "search request" refers to a request made by a user to locate specific information.

[1489] "Storage" refers to the process of saving data for later use, including storing it in a storage system such as a database.

[1490] The present invention is a system for providing reliable word-of-mouth information. The system is comprised of a number of components, each performing a specific function.

[1491] Data collection module (server)

[1492] The server has a module that collects review information. This module acquires text data from multiple data sources, specifically using web scraping technology and APIs to collect data. For example, the server uses the API of "review site Y" to collect review data about "restaurant A." This collected data is temporarily stored in storage.

[1493] Text preprocessing module (server)

[1494] The server has a module that preprocesses the collected review data. This module performs tasks such as removing special characters and extra whitespace, normalizing the text, tokenizing, removing stop words, and performing stemming and lemmatization. For example, a review such as "This restaurant is terrible!" can be normalized to "This restaurant is terrible."

[1495] Evaluation module (server) based on generative artificial intelligence model

[1496] After preprocessing, the server evaluates the review data using a generative artificial intelligence model. This model is a Transformer-based generative model that analyzes the reliability of the text and generates a reliability score. For example, a review that says, "This restaurant is terrible!" will receive a low score, while a review that says, "The food is great. I'd like to go again" will receive a high score.

[1497] Filtering module (server)

[1498] The server has a module that filters the review data based on the generated confidence score. Low-confidence reviews are removed, and only high-confidence reviews remain. For example, a review such as "All the staff are friends!" is removed, while a review such as "The food is great. I'd love to go again." remains.

[1499] Database module (server)

[1500] The filtered, highly reliable review data is stored in a database module on the server. This module creates an index of the stored data to enable quick searches. For example, the highly reliable reviews for "Restaurant A" are stored in the database.

[1501] Search and Display Module (Terminal)

[1502] When a user searches for reviews on their device, the device sends the request to the server. The server retrieves the relevant highly reliable review data from the database and sends it back to the device. The user can check the results on their device. For example, if a user searches for reviews of "Restaurant A," only highly reliable reviews will be displayed.

[1503] Specific examples

[1504] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data includes defamatory comments such as "This restaurant is terrible," as well as unnatural positive comments such as "All the staff are friends!" When the text preprocessing module removes this noise and the data is evaluated by a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only reviews that are true to their true nature, such as "The food is great. I'd like to go again." These saved reviews are then displayed when users search. This system allows users to obtain reliable review information, allowing them to choose products and services with confidence.

[1505] Prompt Sentence Examples

[1506] Example prompts to input to a generative AI model:

[1507] "Data collected: 'This restaurant is awful. The food was cold and the waiter was unfriendly.'"

[1508] "Data collected: 'All the staff are friends! It was amazing.'"

[1509] "Data collected: 'The food is amazing. I'd love to go back.'"

[1510] Generate a credibility score for each review.

[1511] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1512] Step 1:

[1513] Data collection

[1514] The server obtains review information from multiple data sources. Specifically, it uses web scraping technology and APIs to collect review data. Inputs include web page URLs and API endpoints. The server sends HTTP requests for these inputs and receives HTML data or structured data as a response. This data is then stored in temporary storage.

[1515] Specific behavior:

[1516] The server receives the URL of "review site Y" as input.

[1517] The server sends an HTTP request to the specified URL and receives HTML data as a response.

[1518] The server analyzes the HTML data and extracts the necessary word-of-mouth information.

[1519] The extracted word-of-mouth information is stored in temporary storage.

[1520] Step 2:

[1521] Text Preprocessing

[1522] The server preprocesses the collected review data. For example, it removes special characters and extra whitespace, normalizes text, tokenizes, removes stop words, and performs stemming and lemmatization. The input includes the collected raw review data. The server preprocesses this input data and outputs clean text data suitable for analysis.

[1523] Specific behavior:

[1524] The server retrieves the raw word-of-mouth data from the temporary storage.

[1525] Normalizes text by removing special characters and extra whitespace.

[1526] Break down (tokenize) the normalized text into words and phrases.

[1527] Stop words are removed and necessary words are converted to their base form using stemming and lemmatization.

[1528] Prepare preprocessed, clean text data.

[1529] Step 3:

[1530] Generating a confidence score

[1531] The server inputs the preprocessed review data into a generative AI model to generate a credibility score for each review. The input includes preprocessed, clean review data. The server passes this to the generative AI model and obtains a credibility score for each review as output.

[1532] Specific behavior:

[1533] The server obtains the preprocessed review data.

[1534] Pass word-of-mouth data as input to a generative AI model.

[1535] The AI ​​model analyzes the data and generates a confidence score.

[1536] Get the generated confidence score.

[1537] Step 4:

[1538] filtering

[1539] The server filters the review data based on the generated confidence scores. The input includes the review data and its confidence scores. The server removes reviews with low confidence scores and selects reviews with high scores as output.

[1540] Specific behavior:

[1541] The server obtains the generated word-of-mouth data with the reliability scores.

[1542] Weed out low-trust reviews based on their trust score.

[1543] Select reliable review data with high scores.

[1544] Step 5:

[1545] Saving to a database

[1546] The filtered, reliable review data is stored in a database. The input includes the filtered review data. The server stores this in a database and creates an index to enable quick searches.

[1547] Specific behavior:

[1548] The server obtains filtered, highly reliable word-of-mouth data.

[1549] These data are stored in a database.

[1550] Create indexes in your database to enable efficient searches.

[1551] Step 6:

[1552] Search and Display

[1553] When a user searches for review information on their device, the device sends the request to the server. The input includes the user's search query. The server retrieves relevant, reliable review data from the database and returns it to the device as output. The user can then view it on their device.

[1554] Specific behavior:

[1555] A user inputs a search query through a terminal.

[1556] The terminal sends this search query to the server.

[1557] The server retrieves relevant and reliable word-of-mouth data from the database.

[1558] The server returns the acquired data to the terminal.

[1559] The user checks the search results on the terminal.

[1560] (Application example 1)

[1561] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1562] Conventional word-of-mouth information has been problematic in terms of its reliability, particularly in the case of electronic payments, where inappropriate information is mixed in with product and service reviews. This puts users at risk of making purchasing decisions based on inappropriate information, making it difficult to select appropriate products and services. Furthermore, providing reliable word-of-mouth information in real time is difficult, making it difficult to support timely purchasing decisions. The purpose of this invention is to solve these problems by efficiently providing reliable word-of-mouth information and ensuring users' peace of mind when making electronic payments.

[1563] 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.

[1564] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for storing the filtered reliable word-of-mouth data, means for integrating the filtered reliable word-of-mouth data with an electronic payment function to enable users to obtain reliable word-of-mouth information before purchasing a product or service, and means for providing the stored reliable word-of-mouth data in response to a user's search request, thereby enabling users to easily obtain reliable word-of-mouth information before making an electronic payment and select products or services with confidence.

[1565] "Sources" are data collection sources such as websites and APIs that provide review data.

[1566] "Review Data" refers to opinions, ratings, or impressions written by users about a particular product or service.

[1567] "Preprocessing" is the process of removing noise from collected word-of-mouth data and converting it into a format suitable for analysis.

[1568] A "generative artificial intelligence model" is a machine learning model for analyzing large amounts of text data, and has the ability to evaluate the reliability of word-of-mouth reviews.

[1569] The "trustworthiness score" is the result of an analysis of word-of-mouth data by a generative artificial intelligence model, and is a numerical representation of the trustworthiness of the word-of-mouth.

[1570] "Filtering" is the process of removing low-reliability reviews based on their reliability scores, leaving only high-reliability reviews.

[1571] An "electronic payment function" is a system that allows users to make instant payments when purchasing goods or services.

[1572] A "search request" is a query or request entered by a user to search for specific reviews.

[1573] "Storage" refers to the act of storing filtered, reliable word-of-mouth data in a database.

[1574] "Providing" means displaying the stored highly reliable word-of-mouth data on the user's terminal in response to a user's search request.

[1575] MODE FOR CARRYING OUT THE INVENTION

[1576] This invention is a system that collects word-of-mouth data from multiple sources and provides users with only reliable word-of-mouth information. It can also be integrated with electronic payment functions, allowing users to select products and services with confidence based on reliable word-of-mouth information.

[1577] System Configuration

[1578] The system consists of the following main tools:

[1579] 1. Data Collection Methods

[1580] 2. Text Preprocessing Methods

[1581] 3. Evaluation Methods Using Generative AI Models

[1582] 4. Filtering Methods

[1583] 5. Data storage means

[1584] 6. Electronic payment function integration means

[1585] 7. Search and Display Methods

[1586] Hardware and software used

[1587] Hardware:

[1588] server

[1589] User device (smartphone)

[1590] software:

[1591] Python

[1592] SQLite

[1593] BeautifulSoup

[1594] Scikit-learn

[1595] Gensim

[1596] Explanation of data processing and data calculation

[1597] Data collection methods:

[1598] The server collects review data from multiple sources using web scraping technology and APIs, and the collected raw data is stored in temporary storage.

[1599] Text preprocessing methods:

[1600] The server pre-processes the collected review data and converts it into a format suitable for analysis, including normalization, tokenization, stop-word removal, stemming, and lemmatization.

[1601] Evaluation method using generative artificial intelligence model:

[1602] The server inputs the pre-processed review data into a generative AI model to generate a credibility score for each review. The generative AI model uses a Transformer-based model.

[1603] Filtering methods:

[1604] The server filters reviews based on the reliability score obtained from the generative AI model, eliminating reviews with low scores and leaving only reviews with high reliability scores.

[1605] Data storage method:

[1606] The filtered, reliable review data is stored in an SQLite database, which is indexed to enable fast searches.

[1607] Electronic payment integration methods:

[1608] When a user searches for reviews before purchasing a product or service, the server retrieves filtered reviews from the database and displays them on the user's terminal. Based on the reviews, the user can make an electronic payment.

[1609] Search and display methods:

[1610] When a user makes a search request, the server retrieves the stored, reliable word-of-mouth data from the database and returns it to the user's device. The user can then check the reliable word-of-mouth information and make a purchase decision with confidence.

[1611] Specific examples

[1612] For example, if a user tries to search for reviews of "Restaurant A," the server collects review data about "Restaurant A" from multiple sources. The collected review data undergoes text preprocessing and is evaluated by a generative artificial intelligence model. After low-reliability reviews are filtered out, the highly reliable review data is stored in a database. When a user searches for "Restaurant A," the filtered, highly reliable review information is displayed on the user's device. After viewing this review information, the user can safely make reservations and payments for "Restaurant A" using the electronic payment function.

[1613] Prompt Sentence Examples

[1614] "How can I search for reviews of Restaurant A using a filtered app that supports electronic payments?"

[1615] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1616] Step 1:

[1617] The server collects review data from multiple sources, including web scraping techniques and APIs, and stores the collected raw data in temporary storage.

[1618] Input: URLs or API endpoints of multiple review sites

[1619] Output: Raw reviews

[1620] Step 2:

[1621] The server preprocesses the collected review data, normalizing and tokenizing the text, removing stop words, stemming, and lemmatizing it to convert it into a format suitable for analysis.

[1622] Input: Raw review data

[1623] Output: Preprocessed review data

[1624] Step 3:

[1625] The server evaluates the preprocessed review data using a generative AI model (Transformer-based) to generate a credibility score. Specifically, the model analyzes the text, identifies slander and stealth marketing, and evaluates its credibility.

[1626] Input: Preprocessed review data

[1627] Output: Review data with credibility scores

[1628] Step 4:

[1629] The server filters out bad reviews based on the generated reliability score, removing reviews with low reliability scores and selecting only reliable reviews with high scores.

[1630] Input: Review data with reliability scores

[1631] Output: Filtered and reliable review data

[1632] Step 5:

[1633] The server stores the filtered, reliable review data in an SQLite database, which is indexed as it is stored, allowing for fast searches of the data.

[1634] Input: Filtered and reliable review data

[1635] Output: Review data stored in a database

[1636] Step 6:

[1637] When a user sends a search request for a specific product or service from a terminal, the server retrieves relevant, reliable word-of-mouth data from the database and returns it to the user terminal.

[1638] Input: User search request

[1639] Output: Reliable review data displayed on the user's device

[1640] Step 7:

[1641] Users select products and services based on the displayed reviews and make purchases using the electronic payment function. The server provides an integrated service of review information and electronic payment functions, allowing users to make payments with peace of mind.

[1642] Input: User purchase selection

[1643] Output: Completed electronic payment and purchase details

[1644] 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.

[1645] The present invention combines a system that automatically filters malicious reviews from a vast amount of review information and provides users with only reliable reviews, with an emotion engine that recognizes user emotions. Specific operations and embodiments of this system are described below.

[1646] System Configuration

[1647] The system consists of the following main components:

[1648] 1. Data collection module (server)

[1649] 2. Text preprocessing module (server)

[1650] 3. Evaluation module (server) based on generative artificial intelligence model

[1651] 4. Filtering module (server)

[1652] 5. Emotion engine module (server)

[1653] 6. Database module (server)

[1654] 7. Search and Display Module (Terminal)

[1655] System Operation

[1656] Data collection module (server)

[1657] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from the review sites and obtains review information. If an API is provided, it also uses that API to obtain structured data. The collected raw data is stored in temporary storage.

[1658] Text preprocessing module (server)

[1659] The server preprocesses the collected review data and converts it into a format suitable for analysis. Preprocessing includes the following steps:

[1660] Text normalization (removal of special characters and extra whitespace)

[1661] Tokenization (breaking text into words and phrases)

[1662] Stop word removal (elimination of meaningless words)

[1663] Stemming and lemmatization (conversion of words to their root forms)

[1664] Evaluation module (server) based on generative artificial intelligence model

[1665] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[1666] Filtering module (server)

[1667] Based on the generated reliability score, the server filters the reviews: low-scoring reviews are deemed malicious and removed, while high-scoring, reliable reviews are advanced to the next step.

[1668] Emotion engine module (server)

[1669] The emotion engine recognizes users' emotions and individually adjusts the ratings of reviews they view. Specifically, it analyzes users' behavior and review browsing history to determine their current emotional state. Based on this information, it adjusts the credibility score of reviews and provides the most appropriate information for the user.

[1670] Database module (server)

[1671] The reliable review data adjusted by the filtering and sentiment engine is stored in a database. The server manages this database and creates an index to enable efficient searches, allowing users to quickly retrieve relevant information from large amounts of data.

[1672] Search and Display Module (Terminal)

[1673] When a user searches for specific review information on their device, the device sends the search request to the server. The server then searches the database for relevant, reliable review data and returns the results, adjusted by the emotion engine, to the device. The user can then view these results on their device. For example, if a user searches for reviews of "Restaurant A," only reviews about "Restaurant A" that are highly reliable and match the user's emotional state will be displayed.

[1674] Specific examples

[1675] For example, suppose a server collects review data for "Restaurant A" from "Review Site Y." This data contains defamatory comments such as "This restaurant is terrible" and unnatural positive comments such as "All the staff are friends!". After removing the noise with the text preprocessing module and running the data through a generative AI model, these inappropriate reviews receive a low reliability score. The filtering module then removes these malicious reviews, leaving only factual reviews such as "The food is great. I'd like to go again."

[1676] Furthermore, the emotion engine uses the user's past behavioral history to prioritize positive reviews for users who want to avoid negative reviews about restaurants. In this way, reviews are displayed that are in line with the user's emotions. These reviews are stored in the database and are displayed when the user searches.

[1677] This system allows users to obtain highly reliable word-of-mouth information and quickly and easily acquire information that is suited to their own emotional state.

[1678] The processing flow will be explained below.

[1679] Step 1: Data Collection (Server)

[1680] The server collects review data from multiple review sites. Specifically, it uses web scraping technology to analyze HTML data from review sites and social media platforms and extract review information. If an API is provided, it is also possible to obtain review data using that API. The collected raw data is stored in temporary storage.

[1681] Step 2: Text normalization (server)

[1682] Because the raw data collected contains noise and formatting variations, the server normalizes the text, removing special characters, unnecessary whitespace, emojis, etc. to ensure the review text is clean.

[1683] Step 3: Tokenization (Server)

[1684] The server then tokenizes the normalized text, breaking it down into words and phrases. This process splits the text data into units that are easier to analyze.

[1685] Step 4: Stopword Removal (Server)

[1686] From the tokenized text, the server removes common words (stop words) that have little analytical meaning, such as "teki" (target), "ga" (ga), and "o" (to). This makes it easier to extract important information.

[1687] Step 5: Stemming and Lemmatization (Server)

[1688] The server performs stemming and lemmatization on the remaining words, converting them to their root forms. For example, the verbs "hashiru" (run) and "haseda" (run) are unified into "hashiru" (run).

[1689] Step 6: Evaluation using a generative AI model (server)

[1690] After preprocessing, the server inputs the text data into a generative AI model, which analyzes the review's context and generates a credibility score based on the likelihood that each review contains defamatory or stealth marketing material.

[1691] Step 7: Filtering malicious reviews (server)

[1692] Based on the generated reliability score, the server filters the reviews: low-scoring reviews are deemed malicious and removed, while high-scoring, reliable reviews are advanced to the next step.

[1693] Step 8: Recognizing user emotions with the emotion engine (server)

[1694] The server utilizes an emotion engine to recognize the user's current emotional state based on their activity and past browsing history, including their click history, browsing time, and search keywords.

[1695] Step 9: Moderating reviews with the sentiment engine (server)

[1696] Based on the user's emotional state recognized by the emotion engine, the server optimizes reliable reviews for each individual user. For example, if the emotion engine determines that the user prefers positive content, it will prioritize positive reviews.

[1697] Step 10: Saving to the Database (Server)

[1698] The server stores the filtered and adjusted reliable review data in a database, where it creates an index to enable efficient searches.

[1699] Step 11: Receiving a review search request (terminal, user)

[1700] The user sends a request to search for specific reviews from the terminal, which then sends the request to the server.

[1701] Step 12: Retrieving from the Database (Server)

[1702] Based on the received search request, the server searches the database for relevant and reliable review data, using an index to quickly retrieve the data.

[1703] Step 13: Returning and displaying search results (server, terminal)

[1704] The server returns the search results to the device, which then displays the received, highly reliable review data to the user. Because the results have been adjusted by the emotion engine, reviews that best fit the user's emotional state are displayed.

[1705] Example 2

[1706] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1707] In today's information society, unreliable word-of-mouth information is overflowing, making it difficult for consumers to obtain accurate and reliable information. Furthermore, the lack of information provided that reflects the consumer's emotional state often leads to low user satisfaction. In this situation, there is an urgent need to provide a means to accurately filter malicious word-of-mouth and provide word-of-mouth information that takes the user's emotional state into account.

[1708] 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 means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model and generating a reliability score, means for filtering malicious word-of-mouth based on the reliability score, means for analyzing user emotions and adjusting the reliability score, means for storing reliable word-of-mouth data through filtering and emotion analysis, and means for providing the stored reliable word-of-mouth data in response to a user's search request. This makes it possible to quickly and accurately provide reliable word-of-mouth information tailored to the user's emotional state.

[1709] "Sources" are data providers such as online word-of-mouth sites, review sites, and social media platforms.

[1710] "Word-of-mouth data" is text data such as opinions, impressions, and evaluations written by users about products and services.

[1711] "Preprocessing" refers to processes such as normalizing text, tokenizing, removing stop words, stemming, and lemmatizing to make it easier to analyze review data.

[1712] A "generative artificial intelligence model" refers to a large-scale Transformer-based language model used in natural language processing tasks, which analyzes the context of reviews and generates a reliability score.

[1713] The "trust score" is an index that indicates the trustworthiness of a review, calculated by a generative artificial intelligence model by analyzing the context of the review.

[1714] "Filtering" is the process of selecting and removing malicious reviews based on their reliability scores.

[1715] "Emotion analysis" is a process of estimating a user's current emotional state based on their past behavioral patterns and browsing history.

[1716] "Data storage" means storing reliable word-of-mouth data selected through filtering and sentiment analysis in a database.

[1717] A "search request" is a query sent by a user from a terminal in an attempt to search for specific word-of-mouth information.

[1718] "Providing means" refers to the process in which the server returns highly reliable word-of-mouth data that has been filtered and sentiment analyzed to the user.

[1719] "Transformer-based" is an artificial intelligence model architecture built on self-attention mechanisms and deep learning techniques.

[1720] An "index" is a data structure used to speed up database searches and is created for a particular column of data.

[1721] This system preprocesses review data collected from multiple sources, evaluates and filters it using a generative artificial intelligence model, and provides users with highly reliable data. Furthermore, it aims to increase user satisfaction by analyzing user sentiment and adjusting the display of review data.

[1722] System Configuration

[1723] The system consists of the following main components:

[1724] 1. Data collection module (server)

[1725] 2. Text preprocessing module (server)

[1726] 3. Evaluation module (server) based on generative artificial intelligence model

[1727] 4. Filtering module (server)

[1728] 5. Emotion engine module (server)

[1729] 6. Database Module (Server)

[1730] 7. Search and Display Module (Terminal)

[1731] System Operation

[1732] Data collection module (server)

[1733] The server uses BeautifulSoup to retrieve HTML data from multiple review sites, analyze it, and collect review information. It also retrieves structured data via APIs from services that provide them. The collected data is saved in temporary storage. A specific example is a process that creates a list of URLs for "review site A" and extracts reviews from each page.

[1734] Text preprocessing module (server)

[1735] The server performs preprocessing on the collected review data, including normalization, tokenization, stop word removal, stemming, and lemmatization, using Python's NLTK library and SpaCy. For example, it removes noise from sentences like "This restaurant is great" and formats them into a form that is easier to analyze.

[1736] Evaluation module (server) based on generative artificial intelligence model

[1737] The server inputs the preprocessed text data into a generative artificial intelligence model such as GPT-4. The model analyzes the context of the review and generates a credibility score. An example prompt is, "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[1738] Filtering module (server)

[1739] The server filters the review data based on the generated reliability score, removing reviews with low reliability scores and allowing only the remaining high-score reviews to proceed to the next step, thereby eliminating malicious reviews.

[1740] Emotion engine module (server)

[1741] The emotion engine analyzes a user's past behavior and browsing history to estimate their current emotional state. Based on this, it readjusts the reliability score and provides the most appropriate review information to the user. For example, it prioritizes positive reviews for a user who has avoided negative reviews in the past.

[1742] Database module (server)

[1743] The reliable review data, adjusted by the filtering and sentiment engine, is stored in a database using MySQL or PostgreSQL, and indexes are created to enable efficient searches.

[1744] Search and Display Module (Terminal)

[1745] When a user enters a search query on their device, the device sends that information to the server. The server searches the database for relevant review data and returns the results to the device. The user can then check the results on their device. For example, if a user searches for reviews of "Restaurant B," only reliable reviews will be displayed.

[1746] This system allows users to quickly and accurately obtain reliable word-of-mouth information, and also provides information tailored to their emotional state.

[1747] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1748] Step 1:

[1749] Data collection

[1750] Input: The server receives a list of URLs for multiple review sites.

[1751] Specific operation: The server uses a web scraping tool such as BeautifulSoup or Selenium to access each review site based on the URL list and obtain HTML data.

[1752] Data processing: Specific tags and attributes are analyzed from the acquired HTML data to extract word-of-mouth information.

[1753] Output: The extracted reviews (raw data) are saved in temporary storage in JSON format.

[1754] Step 2:

[1755] Text Preprocessing

[1756] Input: The server reads the collected raw data (word-of-mouth information) from the storage.

[1757] Specific operation: The server uses Python's NLTK library and SpaCy to perform text normalization, tokenization, stop word removal, stemming, and lemmatization.

[1758] Data processing: Remove special characters, split words, eliminate meaningless words, and convert words to their root forms.

[1759] Output: Preprocessed text data is obtained and put into a usable format for the next step.

[1760] Step 3:

[1761] Evaluation using generative artificial intelligence models

[1762] Input: The server reads the preprocessed text data.

[1763] How it works: The server inputs a prompt into a generative artificial intelligence model (e.g., GPT-4) and begins the process of assessing the trustworthiness of each review.

[1764] Example prompt: "Please rate the credibility of the following review: 'This restaurant is terrible.'"

[1765] Data computation: The AI ​​model analyzes the context and generates a reliability score.

[1766] Output: Data containing the trustworthiness score and rating reason for each review.

[1767] Step 4:

[1768] filtering

[1769] Input: The server reads the data with reliability scores output from the generative artificial intelligence model.

[1770] Specific operation: The server filters the review data based on the reliability score. It sets a threshold and removes reviews with a score below that threshold.

[1771] Data processing: Remove reviews with low scores and keep only reviews with high scores.

[1772] Output: Filtered and reliable word-of-mouth data.

[1773] Step 5:

[1774] Emotion Engine

[1775] Input: The server loads filtered, reliable word-of-mouth data and user behavior history data.

[1776] How it works: The server uses machine learning models to estimate the user's emotional state by analyzing past click history, time spent on the site, search history, etc.

[1777] Data calculation: Estimate the sentiment and recalculate / adjust the reliability score of the review data.

[1778] Output: Word-of-mouth data adjusted according to the user's emotional state.

[1779] Step 6:

[1780] Database storage

[1781] Input: The server loads reliable review data that has been tuned by the sentiment engine.

[1782] Specific operation: The server inserts data into a database such as MySQL or PostgreSQL.

[1783] Data processing: Creating indexes to enable efficient searches when stored.

[1784] Output: Reliable indexed reviews stored in a database.

[1785] Step 7:

[1786] Search and Display

[1787] Input: The terminal receives a search query from a user.

[1788] Specific operation: The terminal sends a search query to the server, which queries the database and retrieves the relevant review data.

[1789] Data processing: Organize search results and format them in a way that is easy for users to view.

[1790] Output: The terminal displays the reliable word-of-mouth results received from the server to the user.

[1791] (Application example 2)

[1792] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1793] Current review filtering systems have difficulty accurately filtering out malicious and defamatory content, resulting in a large amount of information that users find unreliable. Furthermore, they lack the ability to provide appropriate reviews that reflect the user's individual emotional state and interests, making it difficult for users to quickly and accurately obtain the information they need. This creates a demand for reliable review display, especially on online shopping sites, but current systems make it difficult to achieve this.

[1794] 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.

[1795] In this invention, the server includes means for collecting word-of-mouth data from multiple information sources, means for preprocessing the collected word-of-mouth data, means for evaluating the preprocessed word-of-mouth data using a generative artificial intelligence model to generate a reliability score, means for filtering malicious reviews based on the reliability score, means for saving the filtered reliable word-of-mouth data, means for recognizing the emotional state of a user and adjusting the reliability score according to the emotional state, and means for providing the saved reliable word-of-mouth data in response to a user's search request. This enables users to quickly and accurately obtain reliable word-of-mouth information, and further realizes the provision of information optimized for the user's emotional state.

[1796] "Sources" are data providers, including websites and applications on the Internet, from which word-of-mouth data can be obtained.

[1797] "Word-of-mouth data" is text data in which users write their evaluations and opinions about a certain product or service.

[1798] "Preprocessing" refers to a series of processes that convert review data into a form that is easier to analyze, including text normalization and tokenization.

[1799] A "generative artificial intelligence model" is a type of AI technology that learns useful patterns from massive amounts of data and evaluates and predicts new data.

[1800] The "trustworthiness score" is a number calculated as a result of evaluating whether each review contains defamatory or false information.

[1801] "Filtering" is the process of removing bad reviews based on their credibility score.

[1802] "Emotional state" refers to the user's current psychological and emotional state, and is based on their past behavioral history and current operating situation.

[1803] "Data collection means" refers to a program or device for automatically collecting word-of-mouth data.

[1804] "Data preprocessing means" refers to a program or device that formats collected data into a format that is easy to analyze.

[1805] An "evaluation means" is a program or device that analyzes preprocessed data using a generative artificial intelligence model and generates a reliability score.

[1806] A "filtering means" is a program or device for filtering out malicious reviews based on reliability scores.

[1807] "Storage means" refers to a program or device for storing filtered, highly reliable word-of-mouth data.

[1808] An "emotion recognition means" is a program or device for recognizing a user's emotional state and adjusting the reliability score.

[1809] The "providing means" refers to a program or device for displaying or providing highly reliable word-of-mouth data that has been saved in response to a user's search request.

[1810] A "natural language processing engine" is a program or device that provides technology that enables computers to understand, interpret, and generate human language.

[1811] The present invention aims to provide a specific information provision system that allows users to efficiently and accurately obtain reliable word-of-mouth information. This system collects word-of-mouth data from multiple sources, evaluates it using a generative artificial intelligence model, and further adjusts the reliability score based on the user's emotional state to provide reliable word-of-mouth information.

[1812] System Configuration

[1813] The system consists of the following components:

[1814] 1. Data collection method (server)

[1815] 2. Data preprocessing means (server)

[1816] 3. Evaluation Method (Server)

[1817] 4. Filtering Method (Server)

[1818] 5. Storage means (server)

[1819] 6. Emotion Recognition Method (Server)

[1820] 7. Means of provision (terminal)

[1821] Data collection method (server)

[1822] The server collects review information from multiple review data sources using web scraping technology and APIs, and converts the collected data from unstructured data into structured data.

[1823] Data preprocessing means (server)

[1824] The server performs preprocessing on the collected review data, specifically normalizing the text, tokenizing it, removing stop words, etc. This process converts the text data into a format that is easier to analyze.

[1825] Evaluation method (server)

[1826] The server then inputs the pre-processed data into a generative AI model, a Transformer-based AI model, which evaluates the reliability of the data and generates a reliability score for each review.

[1827] Filtering method (server)

[1828] Based on the generated credibility score, the server filters the reviews: reviews with low credibility scores are rejected and only high credibility reviews are passed on to the next processing step.

[1829] Storage method (server)

[1830] The filtered review data is stored in a database, where it is indexed for fast search.

[1831] Emotion recognition means (server)

[1832] The server uses a natural language processing engine to recognize the user's emotional state, determining it based on the user's past browsing history and current operating context, and adjusting the reliability score accordingly.

[1833] Means of provision (terminal)

[1834] When a user searches for specific reviews using their device, the device sends a search request to the server, which searches the stored reliable reviews data and returns results adjusted based on the user's emotional state to the device, where the user can view the results on their device.

[1835] Specific examples

[1836] For example, a server collects review data for electronic products from a shopping website. This data may contain defamatory content or unnatural positive comments. The server preprocesses the data and evaluates it using a generative artificial intelligence model. Inappropriate reviews receive a low reliability score and are filtered out. An emotion recognition tool then reevaluates the reviews based on the user's context, providing tailored review information.

[1837] Prompt Sentence Examples

[1838] "Create a system to automatically extract reliable reviews and provide optimal information based on user sentiment. The specific steps are as follows: collection, preprocessing, rating, filtering, and sentiment adjustment. Please provide a detailed explanation of the process performed at each step, and include any necessary modules or libraries."

[1839] The above system configuration and processing enable users to obtain highly reliable word-of-mouth information efficiently and accurately.

[1840] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1841] Step 1: Data collection

[1842] The server collects review data from multiple sources. To do so, it uses web scraping technology and APIs. Specifically, the server accesses specified URLs or API endpoints to obtain HTML data or structured data. The obtained data is converted from unstructured data to structured data and temporarily stored.

[1843] Input: Source URL or API endpoint

[1844] Output: Collected review data (text format)

[1845] Step 2: Data Preprocessing

[1846] The server preprocesses the collected review data. Specifically, it removes special characters and unnecessary spaces from the data, and uses natural language processing techniques to tokenize and remove stop words. The text data is then converted into a format that is easy to analyze.

[1847] Input: Collected review data

[1848] Output: Preprocessed review data

[1849] Step 3: Evaluation using a generative artificial intelligence model

[1850] The server inputs the preprocessed review data into a generative AI model. A Transformer-based AI model is used here. The model evaluates the reliability of the text data and generates a reliability score for each review. The reliability score evaluates the likelihood that the review contains defamation or stealth marketing.

[1851] Input: Preprocessed review data

[1852] Output: Trustworthiness score for each review

[1853] Step 4: Filtering

[1854] The server filters the reviews based on the generated credibility score. Reviews with low credibility scores are deemed to be defamatory or stealth marketing and are removed. Only reviews with high credibility scores proceed to the next processing step.

[1855] Input: Confidence score

[1856] Output: Filtered and reliable review data

[1857] Step 5: Save Data

[1858] The server stores the filtered, reliable review data in a database, where an index is created to enable fast searches, allowing for quick retrieval of relevant information later.

[1859] Input: Filtered, reliable review data

[1860] Output: Review data stored in the database

[1861] Step 6: Emotion Recognition

[1862] The server recognizes the user's emotional state by analyzing the user's past browsing history and current operating status using a natural language processing engine to determine the user's current emotional state, and then readjusts the reliability score based on this.

[1863] Input: User browsing history, operation status

[1864] Output: Adjusted confidence score

[1865] Step 7: Provide data

[1866] When a user searches for specific review information on their device, the device sends the search request to the server, which then searches the database for relevant, reliable review data and returns results tailored based on the user's emotional state.

[1867] Input: User search request

[1868] Output: Reliable reviews (adjusted based on users' emotional state)

[1869] The above are the specific processing steps from collection to provision of word-of-mouth data.

[1870] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1871] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1872] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1873] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1874] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1875] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1876] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1877] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1878] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1879] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1880] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1881] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1882] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1883] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1884] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1885] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1886] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1887] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1888] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1889] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1890] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1891] The following is further disclosed regarding the above embodiment.

[1892] (Claim 1)

[1893] A means of collecting review data from multiple sources;

[1894] A means for preprocessing the collected review data;

[1895] means for evaluating the pre-processed review data using a generative artificial intelligence model to generate a credibility score;

[1896] A way to filter bad reviews based on their trustworthiness score;

[1897] A means of storing filtered and reliable review data;

[1898] a means for providing stored reliable word-of-mouth data in response to a user's search request;

[1899] A system including:

[1900] (Claim 2)

[1901] 2. The system of claim 1, wherein a transformer-based model is used as the generative artificial intelligence model.

[1902] (Claim 3)

[1903] 10. The system of claim 1, further comprising means for indexing the filtered review data as it is stored to enable rapid searching.

[1904] "Example 1"

[1905] (Claim 1)

[1906] a means for retrieving text data from multiple data sources;

[1907] means for analyzing and preprocessing the acquired text data;

[1908] means for evaluating the preprocessed text data using a generative model to generate a confidence score;

[1909] means for filtering inappropriate text data based on the confidence score;

[1910] a means for storing the filtered, highly reliable text data;

[1911] means for providing highly reliable text data stored in response to a user's search request;

[1912] A system including:

[1913] (Claim 2)

[1914] 10. The system of claim 1, which uses a transformer-based generative model.

[1915] (Claim 3)

[1916] 10. The system of claim 1, further comprising means for indexing the filtered text data as it is stored to enable rapid information retrieval.

[1917] "Application Example 1"

[1918] (Claim 1)

[1919] A means of collecting review data from multiple sources;

[1920] A means for preprocessing the collected review data;

[1921] means for evaluating the pre-processed review data using a generative artificial intelligence model to generate a credibility score;

[1922] A way to filter bad reviews based on their trustworthiness score;

[1923] A means of storing filtered and reliable review data;

[1924] A means for integrating the filtered and reliable review data with an electronic payment function to enable users to obtain reliable review information before purchasing a product or service;

[1925] a means for providing stored reliable word-of-mouth data in response to a user's search request;

[1926] A system including:

[1927] (Claim 2)

[1928] 2. The system of claim 1, wherein a transformer-based model is used as the generative artificial intelligence model.

[1929] (Claim 3)

[1930] 10. The system of claim 1, further comprising means for indexing the filtered review data as it is stored to enable rapid searching.

[1931] "Example 2: Combining Emotion Engines"

[1932] (Claim 1)

[1933] A means of collecting review data from multiple sources;

[1934] A means for preprocessing the collected review data;

[1935] means for evaluating the pre-processed review data using a generative artificial intelligence model to generate a credibility score;

[1936] A way to filter bad reviews based on their trustworthiness score;

[1937] means for analyzing user sentiment and adjusting the trustworthiness score;

[1938] A means of storing reliable review data through filtering and sentiment analysis;

[1939] a means for providing stored reliable word-of-mouth data in response to a user's search request;

[1940] A system including:

[1941] (Claim 2)

[1942] 2. The system of claim 1, wherein a transformer-based model is used as the generative artificial intelligence model.

[1943] (Claim 3)

[1944] 10. The system of claim 1, further comprising means for indexing the filtered review data as it is stored to enable rapid searching.

[1945] "Application example 2 when combining emotion engines"

[1946] (Claim 1)

[1947] A means of collecting review data from multiple sources;

[1948] A means for preprocessing the collected review data;

[1949] means for evaluating the pre-processed review data using a generative artificial intelligence model to generate a credibility score;

[1950] A way to filter bad reviews based on their trustworthiness score;

[1951] A means of storing filtered and reliable review data;

[1952] means for recognizing the emotional state of a user and adjusting the confidence score in response to the emotional state;

[1953] a means for providing stored reliable word-of-mouth data in response to a user's search request;

[1954] A system including:

[1955] (Claim 2)

[1956] 2. The system of claim 1, wherein a transformer-based model is used as the generative artificial intelligence model.

[1957] (Claim 3)

[1958] 10. The system of claim 1, further comprising means for indexing the filtered review data as it is stored to enable rapid searching.

[1959] (Claim 4)

[1960] 10. The system of claim 1, wherein the system uses a natural language processing engine to recognize the user's emotional state. [Explanation of symbols]

[1961] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting review data from multiple sources; A means for preprocessing the collected review data; means for evaluating the pre-processed review data using a generative artificial intelligence model to generate a credibility score; A way to filter bad reviews based on their trustworthiness score; A means of storing filtered and reliable review data; a means for providing stored reliable word-of-mouth data in response to a user's search request; A system including:

2. The system of claim 1 , wherein the generative artificial intelligence model is a transformer-based model.

3. 10. The system of claim 1, further comprising means for indexing the filtered review data as it is stored to enable rapid searching.

Images