System

Abstract

A system is provided.SOLUTION: A system, comprising: means for collecting user reviews from an online evaluation platform; means for pre-processing text of the collected user reviews; means for extracting features of the pre-processed user reviews; means for detecting fraudulent reviews using a machine learning algorithm; and means for excluding the detected fraudulent reviews and displaying only reviews based on true user experience.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] Online review platforms provide extremely useful information to consumers and store owners, but they face problems such as the inclusion of fraudulent reviews, fake reviews, automated posting by bots, and unnatural harmful rumors, which can reduce their reliability. These fraudulent reviews make it difficult for consumers to make fair purchasing decisions. The present invention aims to automatically detect and remove these fraudulent reviews, thereby providing information based on true user experiences and enabling consumers to make accurate decisions. [Means for solving the problem]

[0005] The system of the present invention comprises the following means.

[0006] (1) It is a means of collecting user reviews from online rating platforms. Specifically, it uses an API to obtain review text, rating scores, reviewer information, posting date and time, etc., and stores them in a database.

[0007] (2) A method for preprocessing the collected user review text, which includes removing HTML tags and special characters from the text, tokenizing it, and removing stop words.

[0008] (3) A means of extracting features from preprocessed user reviews. Specifically, it performs sentiment analysis, keyword extraction, and time series analysis to analyze review patterns and trends.

[0009] (4) A method for detecting fraudulent reviews using machine learning algorithms, which identify fraudulent reviews based on a large number of posts from the same IP address, text similarity, and reviewer activity history.

[0010] (5) It is a means to filter out detected fraudulent reviews and display only reviews based on genuine user experiences. It provides filtered review information to users' devices and serves as a reliable source of information.

[0011] This allows users to make accurate purchasing decisions based on highly reliable review information.

[0012] An "online rating platform" is a system that allows users to post ratings and reviews of products and services online, and other users to view that information.

[0013] A "user review" is a comment in text form written by a consumer who has used a product or service, describing their experience and opinions.

[0014] "API" stands for Application Program Interface, an interface that allows different software programs to exchange information.

[0015] A "database" refers to a collection of structured data that allows for efficient management of large amounts of data and quick search and retrieval of required information.

[0016] "Preprocessing" is the process of preparing data for analysis and parsing, such as removing unnecessary elements from text and tokenizing it.

[0017] "Tokenization" is the process of dividing text into its smallest units, such as words or phrases, to make the text easier to analyze.

[0018] "Stop words" are words that have little meaning in text analysis and are unlikely to affect the analysis results. Examples include "no", "wa", and "desu".

[0019] "Sentiment analysis" is a technology that automatically determines emotions such as positive, negative, and neutral from text data.

[0020] "Keyword extraction" is the process of identifying important words and phrases from text data.

[0021] "Time series analysis" is a method for analyzing changes and trends in data over time.

[0022] A "machine learning algorithm" is a computational method for learning patterns based on data and making predictions and classifications.

[0023] "Fraudulent reviews" refer to false ratings that are not based on true user experiences or reviews that have been intentionally manipulated.

[0024] "Detection" means automatically finding items of interest based on specific criteria or algorithms.

[0025] "Filtering" is the process of sorting through data based on specific criteria, extracting what is needed and removing what is not. [Brief explanation of the drawings]

[0026] [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

[0027] 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.

[0028] First, the terms used in the following description will be explained.

[0029] 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).

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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."

[0034] [First embodiment]

[0035] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0036] 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.

[0037] 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).

[0038] 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.

[0039] 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.

[0040] 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.

[0041] 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.

[0042] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0043] 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.

[0044] 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.

[0045] 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.

[0046] 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."

[0047] The present invention provides a system for detecting and filtering out fraudulent reviews based on user reviews collected from an online rating platform, thereby providing a genuine user experience. Specific embodiments of this system are described below.

[0048] System configuration

[0049] 1. Data Collection Methods

[0050] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.) by using an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in a database.

[0051] 2. Pretreatment Methods

[0052] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0053] Remove HTML tags and special characters

[0054] Tokenizing text (splitting it into words and phrases)

[0055] Removal of stop words (e.g., "no", "wa", "desu", etc.)

[0056] 3. Feature Extraction Method

[0057] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0058] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0059] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0060] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0061] 4. How to detect fraudulent reviews

[0062] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0063] Mass posting from the same IP address or concentrated posting at the same time

[0064] Text similarity (check for copy-paste)

[0065] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0066] 5. Review filtering and display methods

[0067] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed on the user's device.

[0068] Program implementation example

[0069] As a concrete example, consider a system that collects restaurant reviews and filters out fraudulent reviews.

[0070] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0071] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0072] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0073] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0074] 5. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0075] The system aims to provide users with reliable information based on genuine reviews, helping consumers make accurate purchasing decisions. It also contributes to improving the credibility of store owners by providing fair reviews.

[0076] The processing flow will be explained below.

[0077] Step 1:

[0078] The server uses the API to collect user reviews from online rating platforms. Specifically, it acquires the text, rating score, reviewer information, and posting date and time of reviews posted within a specified period and stores them in an internal database.

[0079] Step 2:

[0080] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it performs tokenization, dividing the text into words and phrases. Finally, it removes meaningless stop words (e.g., "no," "wa," and "desu").

[0081] Step 3:

[0082] The server extracts specific features from the pre-processed reviews, performs sentiment analysis on the reviews to determine whether they are positive, negative, or neutral, and extracts frequently occurring keywords and phrases to analyze review trends.

[0083] Step 4:

[0084] The server applies machine learning algorithms to detect fraudulent reviews. It checks for large numbers of reviews from the same IP address and detects abnormal patterns, such as a large number of reviews posted in a short period of time or from a new account. It also analyzes the similarity of the review text between reviews.

[0085] Step 5:

[0086] The server filters out reviews that are determined to be fraudulent and only filters out genuine reviews. The filtered reviews are then deleted from the database, leaving only reviews based on genuine user experiences in the dataset.

[0087] Step 6:

[0088] The server sends the filtered, truthful reviews to the user's device, which displays these reviews to the user and makes them available as reference information.

[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] Currently, online rating platforms collect a large number of user reviews, but these often contain fraudulent reviews. These fraudulent reviews prevent consumers from obtaining accurate information and lead to a decline in trust. Particularly problematic are mass postings from the same IP address, copy-pasted text, and mass postings from new accounts in a short period of time.

[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 collecting evaluation information from an online evaluation platform, means for preprocessing the collected text of the evaluation information, means for extracting features of the preprocessed evaluation information, means for analyzing sentiment using natural language processing technology, means for detecting fraudulent information using a machine learning algorithm, and means for filtering out the detected fraudulent information and displaying only evaluation information based on true user experiences, thereby increasing the reliability of the evaluation information and enabling users to make decisions based on accurate information.

[0094] "Online Rating Platform" means an online interface that collects and provides user reviews and rating information.

[0095] "Rating information" is data such as reviews provided by users, rating scores, reviewer information, and posting dates and times.

[0096] "Preprocessing the text" refers to the process of removing unnecessary markup tags and special symbols from the evaluation information, splitting the text, and removing stop words.

[0097] "Feature extraction" means extracting important keywords and patterns from preprocessed text and preparing the data for analysis.

[0098] "Natural language processing technology" is a technology for analyzing text data and understanding emotions and intentions.

[0099] "Analyzing emotions" refers to the process of determining whether the content of text data is positive, negative, or neutral.

[0100] A "machine learning algorithm" is an algorithm that learns patterns from data and uses them to make classifications and predictions.

[0101] "Detecting fraudulent information" refers to the process of identifying fraudulent reviews or abnormal patterns from rating information.

[0102] A "true user experience" is an experience based on the evaluations and impressions of actual users, without any fraudulent manipulation or manipulation whatsoever.

[0103] The "means for displaying" is a method for displaying the filtered and accurate evaluation information on the user's terminal.

[0104] The present invention relates to a system that detects and removes fraudulent information based on evaluation information collected from an online evaluation platform, and provides a genuine user experience. The system consists of the following components:

[0105] 1. Data Collection Methods

[0106] The server uses an API to collect review information from an online review platform. The collected data includes review text, review score, reviewer information, and posting date and time. As a specific example, the server retrieves review information for "Restaurant A" from the API and stores it in a database. To do this, the server sends a request to a specified API endpoint and stores the JSON data returned as a response in the database.

[0107] 2. Data preprocessing methods

[0108] The server preprocesses the collected text data of the evaluation information. Data preprocessing includes removing HTML tags and special symbols, tokenizing the text, and removing stop words. The server uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. Specifically, the evaluation information "It was a great meal. Five stars!" is tokenized as "Great meal 5 stars."

[0109] 3. Feature Extraction Method

[0110] The server extracts important features from the preprocessed text data. It analyzes word frequency and appearance patterns to extract specific keywords. It also analyzes rating scores and posting frequency as time-series data. The server extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and uses the pandas library to analyze time-series data based on the time axis.

[0111] 4. Sentiment analysis tool

[0112] The server performs sentiment analysis of the review information using natural language processing technology. Specifically, it uses machine learning models (e.g., BERT, LSTM) to determine whether the review content is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive.

[0113] 5. How to detect fraudulent reviews

[0114] The server uses machine learning algorithms to detect fraudulent reviews. It analyzes review patterns and reviewer behavior to identify reviews with fraudulent patterns. Specifically, it uses algorithms such as random forests and support vector machines (SVMs) and references database logs to detect suspicious activity.

[0115] 6. Filtering and Display Methods

[0116] The server filters out fraudulent reviews and only displays rating information based on genuine user experiences. To provide the filtered reviews to the user, the server sends the information to the user's device in JSON or HTML format. The device receives this information and displays accurate rating information to the user.

[0117] This system allows users to make accurate decisions based on reliable information, and also contributes to improving the reliability of store owners by providing fair evaluations.

[0118] Examples of prompt statements

[0119] "Design a system that collects user reviews from online rating platforms and detects and filters out fraudulent reviews. Include the following processing steps: data collection, data preprocessing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display. Please also provide examples of specific APIs or libraries."

[0120] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0121] Step 1: Collect data

[0122] The server collects rating information using the API of the online rating platform. As input, it receives the JSON data returned as a response from the API endpoint. This includes the review text, rating score, reviewer information, and posting date and time. The server stores this data in a database. Specifically, it sends an API request, parses the response, extracts the necessary information, and stores it in the database.

[0123] Step 2: Preprocessing the data

[0124] The server preprocesses the collected text data of rating information. As input, it receives the collected text data of rating information. The server removes HTML tags and special symbols, tokenizes the text (divides it into words and phrases), and removes stop words. Specifically, it uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. As output, it obtains clear preprocessed text data.

[0125] Step 3: Feature extraction

[0126] The server extracts specific features from preprocessed text data. The input is preprocessed text data. The server analyzes word frequency and appearance patterns to identify prominent keywords and phrases. It also analyzes rating scores and posting frequency as time-series data. Specifically, it extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and performs data analysis based on the time axis using the pandas library. The output is the extracted feature data.

[0127] Step 4: Sentiment analysis

[0128] The server performs sentiment analysis on preprocessed text data using natural language processing techniques. The input is the preprocessed text data. The server uses a machine learning model (e.g., BERT, LSTM) to determine whether the text is positive, negative, or neutral. Specifically, the server applies the sentiment analysis model and assigns a sentiment label to each review. The output is text data with the sentiment labels.

[0129] Step 5: Detect fraudulent reviews

[0130] The server uses machine learning algorithms to detect fraudulent reviews. The inputs are text data with sentiment labels and reviewer behavioral history. The server detects mass posting from the same IP address, copy-pasted text, and mass posting from new accounts in a short period of time. Specifically, it uses random forests and support vector machines (SVMs) to identify reviews with fraudulent patterns. The output is data determined to be fraudulent reviews.

[0131] Step 6: Filter and view reviews

[0132] The server filters out fraudulent reviews and only reviews based on genuine user experiences. The inputs are data determined to be fraudulent reviews and accurate rating information. The server runs the filtering algorithm and sends the filtered reviews to the user's device in JSON or HTML format. Specifically, the server filters out fraudulent reviews, generates filtered reviews, and sends them to the user's device. The output is accurate rating information that is displayed on the user's device.

[0133] (Application example 1)

[0134] 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."

[0135] On online rating platforms, user reviews are an important source of information for users to evaluate products and services. However, the presence of fraudulent reviews can make accurate evaluations difficult, potentially leading to incorrect decisions. Furthermore, the lack of review statistics based on sentiment analysis makes it difficult to understand the content of reviews. Therefore, it is necessary to obtain reliable reviews, eliminate fraudulent reviews, and analyze and display the sentiment of reviews, thereby enabling users to make appropriate decisions.

[0136] 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.

[0137] In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for excluding the detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for displaying the reviews after excluding the fraudulent reviews and their sentiment analysis on the user's terminal, thereby enabling users to make appropriate decisions based on reliable reviews and the results of the sentiment analysis.

[0138] An "online review platform" is a system that allows users to post reviews of products and services via the Internet and for other users to view those reviews.

[0139] "User reviews" are written opinions and evaluations of products and services posted by users.

[0140] The "means of collection" refers to the method by which the server obtains user reviews from online rating platforms, such as by using APIs or web scraping.

[0141] "Preprocessing" refers to methods for removing unnecessary information from the collected user review text and converting it into a format that is easy to analyze. Specifically, this includes removing HTML tags and special characters, tokenizing, and removing stop words.

[0142] "Means for extracting features" refers to methods for deriving features from preprocessed user reviews. Specifically, methods include sentiment analysis, keyword extraction, and time series analysis.

[0143] A "machine learning algorithm" is a model in which a computer learns patterns from data and makes predictions or classifications based on those patterns.

[0144] "Fraudulent reviews" are false reviews posted with the intent to manipulate ratings.

[0145] The "detection method" is a method of identifying fraudulent reviews using a machine learning algorithm.

[0146] The "display method" refers to a method for displaying only reviews that meet certain conditions on the user's device. This method targets highly reliable reviews that exclude fraudulent reviews.

[0147] "Sentiment analysis" is a technology that extracts emotions and opinions from text data and determines their nature (positive, negative, neutral).

[0148] "Keyword extraction" is a technique for identifying important words and phrases from user reviews.

[0149] "Time series analysis" is a technique for analyzing how data changes over time.

[0150] A "terminal" is an electronic device that a user uses to view and operate information. Examples include smartphones, tablets, and personal computers.

[0151] The present invention relates to a system that analyzes user reviews collected from an online rating platform, eliminates fraudulent reviews, and presents more reliable reviews, particularly for performing sentiment analysis to assist users in making decisions. An embodiment of this system will now be described in detail.

[0152] First, the server collects user reviews from online rating platforms (e.g., shopping sites or review sites) using APIs or web scraping technology to obtain data such as review text, rating scores, reviewer information, and posting date and time, and stores them in a database. Next, the server preprocesses the collected user review text data, which includes steps such as removing HTML tags and special characters, tokenizing, and removing stop words.

[0153] The server extracts specific features from the preprocessed reviews. This involves sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis determines whether a review is positive, negative, or neutral, while keyword extraction identifies important words and phrases that frequently appear. Time series analysis analyzes rating scores and posting frequency over time. After feature extraction, the server uses machine learning algorithms to detect fraudulent reviews. This detection takes into account large numbers of posts from the same IP address, concentrations in time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.).

[0154] Detected fraudulent reviews are filtered out, and the server filters out only reviews based on genuine user experiences. The filtered reviews and their sentiment analysis results are then displayed on the user's device. Users can access this reliable review information using devices such as smartphones and tablets, allowing them to make appropriate decisions based on accurate information.

[0155] The hardware and software used include a server infrastructure capable of high-performance data collection and processing, specifically software libraries such as Python, scikit-learn, BeautifulSoup, requests, and nltk. Reviews are displayed on the user's device using a browser or a dedicated app.

[0156] As a concrete example, consider the case where a user wants to check reviews for "product ID 12345" on "mail order site A." The user launches the smartphone app and enters "product ID 12345." An example of a prompt sentence in this case is as follows:

[0157] "Collect reviews for product ID 12345, preprocess them to a clean form, perform sentiment analysis, and finally detect and filter out fraudulent reviews."

[0158] Based on these prompts, the server can perform the necessary processing and provide users with reliable reviews and their sentiment analysis results, allowing them to select products with confidence.

[0159] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0160] Step 1:

[0161] The server collects user reviews from an online rating platform using an API. As input, it receives data including reviews related to a specified product or service (text, rating score, reviewer information, and posting date and time), and stores the data in a database as output.

[0162] Step 2:

[0163] The server preprocesses the collected user reviews. As input, it uses the text data of reviews stored in the database, removes HTML tags and special characters, tokenizes, and removes stop words. As output, it generates the preprocessed, clean data.

[0164] Step 3:

[0165] The server extracts specific features from the preprocessed reviews. Using the preprocessed text data as input, it performs sentiment analysis, keyword extraction, and time series analysis. As output, it generates the analysis results as a feature vector, which includes the review's emotional state (positive, negative, neutral), key keywords, and the review's posting timestamp.

[0166] Step 4:

[0167] The server uses a machine learning algorithm to detect fraudulent reviews. It uses the feature vectors obtained through feature extraction as input and analyzes patterns such as large numbers of posts from the same IP address, concentration of posts in specific time periods, text similarity, and reviewer activity history. As an output, it lists reviews that are suspected of being fraudulent and generates a detection result.

[0168] Step 5:

[0169] The server filters out the detected fraudulent reviews and filters out only reviews based on genuine user experiences. As input, it uses the fraudulent review detection results and the original review list to filter out fraudulent reviews. As output, it generates a filtered review list that contains only reliable reviews.

[0170] Step 6:

[0171] The server displays the filtered reviews and their sentiment analysis results on the user's device. Using the filtered review list and sentiment analysis results as input, the server displays the reviews on the screen or application of the user's device, allowing the user to check reliable review information and make appropriate decisions.

[0172] Specifically, when a user launches the app on their smartphone and enters "product ID 12345," the server executes each step in sequence based on the prompt, ultimately displaying reliable review information and the results of sentiment analysis on the user's device. When the prompt is entered, "Collect reviews for product ID 12345, preprocess them into a clean format, perform sentiment analysis, and finally detect and filter out fraudulent reviews," the process proceeds according to the instructions.

[0173] 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.

[0174] The present invention is a system that detects and excludes fraudulent reviews based on user reviews collected from an online rating platform, provides a genuine user experience, and combines this with an emotion engine that recognizes user emotions. Specific embodiments of this system are described below.

[0175] System configuration

[0176] 1. Data Collection Methods

[0177] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.). Specifically, it uses an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database.

[0178] 2. Pretreatment Methods

[0179] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0180] Remove HTML tags and special characters

[0181] Tokenizing text (splitting it into words and phrases)

[0182] Removal of stop words (e.g. "no", "wa", "desu", etc.)

[0183] 3. Feature Extraction Method

[0184] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0185] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0186] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0187] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0188] 4. Emotion Engine

[0189] The server recognizes the user's emotions through the emotion engine and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine can grasp the user's emotional state in real time and adjust the way reviews are displayed accordingly.

[0190] 5. How to detect fraudulent reviews

[0191] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0192] Mass posting from the same IP address or concentrated posting at the same time

[0193] Text similarity (check for copy-paste)

[0194] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0195] 6. Review Filtering and Display Methods

[0196] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed.The server then adjusts the filtered reviews based on the user's emotional state using an emotion engine, and finally displays them on the user's device.

[0197] Program implementation example

[0198] As a concrete example, consider a system that collects restaurant reviews, filters out fraudulent reviews, and adjusts the display depending on the user's emotional state.

[0199] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0200] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0201] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0202] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0203] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust the way reviews are displayed accordingly. For example, if the user is in a positive emotional state, positive reviews will be displayed first.

[0204] 6. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0205] The system aims to help users make accurate purchasing decisions based on reliable reviews, helping consumers make accurate decisions. Recognizing users' emotions in real time also enables more personalized information to be provided, improving the browsing experience. Store owners can also benefit from fair reviews, which contributes to increased trust.

[0206] The processing flow will be explained below.

[0207] Step 1:

[0208] The server uses APIs to collect user reviews from online rating platforms. For example, it automatically retrieves review texts, rating scores, reviewer information, and posting dates / times related to specific periods or specific subjects (restaurants, products, etc.), and stores them in an internal database.

[0209] Step 2:

[0210] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it tokenizes the review text and splits it into words and phrases. Finally, it removes stop words (such as "no," "wa," and "desu") and converts it into a format that is easy to analyze.

[0211] Step 3:

[0212] The server extracts specific features from the preprocessed reviews. First, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. Next, it extracts frequently occurring keywords and important phrases from the review text. Finally, it performs time-series analysis to analyze the rating score and posting frequency over time.

[0213] Step 4:

[0214] The server uses an emotion engine to recognize the user's emotions and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine also detects the user's emotional state from facial expressions and voice while viewing reviews, and grasps changes in emotions in real time.

[0215] Step 5:

[0216] The server applies machine learning algorithms to detect fraudulent reviews. It analyzes large numbers of posts from the same IP address, text similarity, and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.) to identify fraudulent reviews. Reviews that are deemed to be non-issues are flagged as genuine reviews.

[0217] Step 6:

[0218] The server adjusts the display method of the filtered true reviews according to the user's emotional state. Depending on the user's emotional state (positive, negative, etc.) recognized by the emotion engine, it decides whether to prioritize positive reviews or balanced reviews.

[0219] Step 7:

[0220] The server then sends the final selected reviews to the user's device, which then displays the received reviews in a format appropriate to the user's emotional state. For example, if the user is in a positive emotional state, more positive reviews will be displayed.

[0221] Step 8:

[0222] Users can view the displayed reviews and make purchasing decisions based on reliable information. When a user posts a new review, that review is also analyzed through the same process.

[0223] Example 2

[0224] 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."

[0225] Online rating platforms are prone to a flood of fraudulent reviews, leading to users making erroneous decisions based on unreliable information. Furthermore, reviews that do not reflect the user's emotional state are not provided, making it difficult to help users make appropriate purchasing decisions.

[0226] 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 user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for analyzing users' emotional tendencies using an emotion engine, means for detecting fraudulent reviews using a machine learning algorithm, and means for excluding detected fraudulent reviews, filtering reviews based on true user experiences, and displaying them according to the user's emotional state. This allows users to make accurate decisions based on reliable reviews and provides information adapted to the user's emotions.

[0227] An "online rating platform" refers to a system on the Internet where users can post ratings and reviews of products and services and other users can view them.

[0228] "User reviews" refers to ratings and comments about products and services posted by users on online rating platforms.

[0229] "Means of collection" refers to the way in which the server automatically obtains user reviews from online rating platforms using technologies such as APIs or web scraping.

[0230] The "means of text preprocessing" refers to a method of removing HTML tags and special characters from the text data of user reviews collected by the server, and performing tokenization and removal of stop words.

[0231] "Means for extracting features" refers to the methods by which the server extracts important information and patterns from pre-processed user reviews, including sentiment analysis, keyword extraction, and time series analysis.

[0232] An "emotion engine" refers to the technology and algorithms used to analyze a user's emotional state from text data, which then classifies emotions as positive, negative, or neutral.

[0233] "Machine learning algorithms" refer to technology that trains models based on past data to classify and predict unknown data. In this case, they are used to detect fraudulent reviews.

[0234] "Fraudulent reviews" are manipulated ratings and false comments that provide biased information that differs from the normal user experience.

[0235] "Filtering means" refers to the method by which the server filters out fraudulent reviews and selects only genuine reviews.

[0236] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[0237] "Means for displaying" refers to the method by which the server selects the filtered reviews and ultimately displays them on the user's device.

[0238] This invention is a system that provides reliable reviews by detecting and filtering out fraudulent reviews based on user reviews collected from online rating platforms. Furthermore, it aims to improve the user experience by combining it with an emotion engine that recognizes the user's emotional state.

[0239] Hardware and software configuration

[0240] The following hardware and software are used to implement this system.

[0241] Server: This is the main component used for data collection, pre-processing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display.

[0242] Database: This is an internal database that stores collected user reviews. It uses a relational database such as MySQL or PostgreSQL.

[0243] API: A means of collecting reviews from online rating platforms, such as a RESTful API.

[0244] Machine learning algorithms are used to detect fraudulent reviews, specifically using machine learning libraries such as Scikit-learn and TensorFlow.

[0245] Sentiment engine: A means of analyzing user sentiment, analyzing review text data using natural language processing technology. Libraries such as NLTK and TextBlob, or APIs dedicated to sentiment analysis (e.g., IBM Watson Natural Language Understanding) are used.

[0246] Data collection

[0247] The server automatically collects user reviews from online rating platforms using APIs. For example, to collect reviews for a specific restaurant, the server uses the endpoint "https: / / api.example.com / reviews?restaurant_id=A".

[0248] Data Preprocessing

[0249] The server preprocesses the collected user review text data, specifically removing HTML tags and special characters, tokenizing, and removing stop words, so that the text data is in a format that is easy to analyze.

[0250] Feature Extraction

[0251] The server extracts specific features from the pre-processed reviews. This step uses a variety of analytical techniques, including sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis involves classifying emotions into "positive," "negative," and "neutral," extracting important keywords, and analyzing review trends over time.

[0252] Emotion Engine

[0253] The server uses an emotion engine to analyze the user's emotional state in real time. The emotion engine determines the user's psychological state based on the content of the reviews and grasps the overall emotional trend. For example, if the user is in a positive emotional state, positive reviews will be displayed preferentially.

[0254] Fraudulent review detection

[0255] The server uses machine learning algorithms to detect fraudulent reviews, which analyze factors such as large numbers of posts from the same IP address, similar text, and reviewer activity history to identify reviews that are likely to be fraudulent.

[0256] Filter and view reviews

[0257] The server filters out fraudulent reviews and only reviews based on genuine user experiences, then adjusts the way reviews are displayed according to the user's emotional state, and finally displays the filtered reviews on the user's device.

[0258] Specific examples

[0259] As a concrete example, the process of collecting, preprocessing, analyzing, detecting fraudulent reviews, and displaying restaurant reviews is shown below.

[0260] 1. The server collects reviews for "Restaurant A" from the API.

[0261] 2. The server removes HTML tags from the collected data and performs tokenization and stop word removal.

[0262] 3. The server performs sentiment analysis, determines the review as positive, and extracts important keywords.

[0263] 4. The server uses machine learning algorithms to detect fraudulent reviews and certify legitimate reviews.

[0264] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust how the review is displayed.

[0265] 6. The server filters the true reviews and displays them on the user's device.

[0266] Prompt Sentence Examples

[0267] By inputting the following prompts into the generative AI model, an explanation of the system and specific examples can be generated.

[0268] Example prompt:

[0269] "Describe a system that detects and filters out fraudulent reviews based on user reviews collected from an online rating platform, and provides a genuine user experience. Also, provide a concrete example of how to utilize an emotion engine that recognizes user emotions."

[0270] In this way, by detailing the specific steps and techniques required to implement the invention, it is possible to accurately understand the invention and provide guidance for others to practice it.

[0271] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0272] Step 1: Data collection methods

[0273] A server automatically collects user reviews from an online rating platform using an API. As input, it uses an API endpoint (e.g., "https: / / api.example.com / reviews?restaurant_id=A"). The server sends the request and stores the response data, such as review text, rating score, reviewer information, and posting date and time, in an internal database. The output is a comprehensive dataset of the collected reviews.

[0274] Step 2: Data preprocessing measures

[0275] The server preprocesses the collected user review text data. Specifically, it removes HTML tags and special characters. For example, Great food " is converted to "Great meal". Next, the text is tokenized and a review such as "Great meal 5 stars" is split into words such as "Great", "Meal", and "5 stars". Furthermore, stop words are removed and "This is a great meal" is converted to "Great meal". The input is the collected text data of reviews, and the output is the preprocessed and clean text data.

[0276] Step 3: Feature extraction methods

[0277] The server extracts specific features from the preprocessed reviews. Specifically, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive. It also extracts important keywords, identifying words such as "amazing," "meal," and "five stars." It then performs time-series analysis to analyze the rating scores and posting frequency over time. The input is the text data of the preprocessed reviews, and the output is the sentiment classification results, extracted keywords, and time-series data.

[0278] Step 4: Emotion Engine

[0279] The server uses an emotion engine to analyze the user's emotional state in real time. Specifically, it understands the user's overall emotional tendency based on the emotion analysis results from the previous step. For example, the text data of a review is input into a natural language processing algorithm to determine the emotional state, such as positive, negative, or neutral. The input is the feature-extracted emotion analysis results, and the output is the user's overall emotional tendency data.

[0280] Step 5: Detecting fraudulent reviews

[0281] The server uses a machine learning algorithm to detect fraudulent reviews. Specific operations include checking for large numbers of posts from the same IP address, concentrated time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.). For example, the machine learning model inputs review metadata to determine whether there is suspicion of fraud. The input is collected and preprocessed review metadata and text data, and the output is the result of the fraudulent review determination.

[0282] Step 6: Filtering and displaying reviews

[0283] The server filters out reviews determined to be fraudulent and filters out only reviews based on genuine user experiences. It also adjusts the way reviews are displayed based on the results of the emotion engine. Specifically, it prioritizes positive or negative reviews depending on the user's emotional state. For example, if a user is in a positive emotional state, positive reviews will be displayed at the top. The input is data on reviews determined to be non-fraudulent and data on the user's emotional state, and the output is the final filtered reviews.

[0284] Through these steps, the system can collect reliable reviews and provide personalized information based on user sentiment.

[0285] (Application example 2)

[0286] 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."

[0287] Online rating platforms face the problem of fraudulent reviews, which can lead to users making purchasing decisions based on unreliable information. Furthermore, they are unable to tailor the display of reviews to reflect the user's emotional state, making it difficult to provide users with information that is appropriate for them.

[0288] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for filtering out detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for recognizing the emotional state of the user and adjusting the display of reviews according to the emotional state. This enables users to make accurate purchasing decisions based on reliable reviews and also enables the provision of personalized information according to the user's emotional state.

[0289] An "online review platform" is a service provided over the Internet that allows users to post and view reviews and ratings of products and services.

[0290] "User reviews" are ratings and comments about products and services posted by users on online rating platforms.

[0291] "Preprocessing" refers to a series of steps performed to convert the collected text data of user reviews into an analyzable format, including removing HTML tags and special characters, tokenizing, and removing stop words.

[0292] "Feature extraction" is a method for extracting meaningful information and patterns from preprocessed user reviews, and includes sentiment analysis, keyword extraction, and time series analysis.

[0293] A "machine learning algorithm" is a method for learning patterns and rules from large amounts of data and performing analysis and classification.

[0294] "Fraudulent reviews" are false reviews that are not based on facts or manipulative reviews posted with a specific intention.

[0295] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[0296] "Adjusting the display of reviews according to emotional state" means optimizing the content and order of reviews to be displayed based on the user's current emotions and mood.

[0297] This invention is a system that detects and excludes fraudulent reviews from users on online rating platforms and displays appropriate reviews according to the user's emotional state. This system mainly consists of a server, a smartphone application, a machine learning algorithm, and an emotion recognition engine.

[0298] System Overview

[0299] The server collects user reviews from online rating platforms, preprocesses the collected reviews, and performs feature extraction. To detect fraudulent reviews, a machine learning algorithm is used. An emotion engine recognizes users' emotional states in real time. Users then receive tailored reviews using a smartphone application.

[0300] Hardware and software used

[0301] The main hardware and software used are servers, smartphones, Python, TextBlob, and Sklearn, which are used for data processing and calculation.

[0302] Data processing and calculation

[0303] 1. Data Collection:

[0304] The server uses APIs to collect user reviews from online rating platforms. The data collected includes review text, rating score, reviewer information, and posting date and time, and is stored in an internal database.

[0305] 2. Pretreatment:

[0306] HTML tags and special characters are removed from the collected review text data, the text is tokenized, and stop words (e.g., "no" and "wa") are removed.

[0307] 3. Feature Extraction:

[0308] After preprocessing, sentiment analysis (determining positive, negative, or neutral), keyword extraction, and time series analysis are performed on the data.

[0309] 4. Fraudulent Review Detection:

[0310] The server uses machine learning algorithms (e.g., Naive Bayes) to detect fraudulent reviews, including criteria such as large numbers of reviews from the same IP address, similar text, and reviews posted by new accounts.

[0311] 5. User emotion recognition and display adjustment:

[0312] The emotion engine detects the user's current emotional state in real time and adjusts how reviews are displayed accordingly: for example, if the user is in a positive emotional state, positive reviews will be displayed first.

[0313] Specific examples

[0314] "As a concrete example of an application to increase the reliability of user reviews on an online shopping site, the server collects product reviews via an API, tokenizes the text data through preprocessing, performs sentiment analysis, and detects and excludes fraudulent reviews using the Naive Bayes algorithm. The system then prioritizes displaying positive or negative reviews depending on the user's emotional state."

[0315] Generative AI model prompt example

[0316] "Write a program in Python to recognize emotions from user-generated reviews on an online shopping site and filter out fraudulent reviews."

[0317] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0318] Step 1:

[0319] The server collects user reviews through the API of the online rating platform. As input, it receives data such as the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database. Specifically, it sends an HTTP request, parses the response data, and stores it.

[0320] Step 2:

[0321] The server preprocesses the collected user review text data. It receives the review text data as input, removes HTML tags and special characters, tokenizes it (divides it into words and phrases), and removes stop words. Specifically, it cleans up the text using regular expressions and natural language processing libraries.

[0322] Step 3:

[0323] The server extracts features from preprocessed user reviews. It receives the preprocessed text data as input and performs sentiment analysis (positive, negative, neutral), keyword extraction, and time series analysis. Specifically, it calculates sentiment scores using the TextBlob library and extracts frequently occurring keywords using Sklearn.

[0324] Step 4:

[0325] The server uses machine learning algorithms to detect fraudulent reviews. It receives collected, preprocessed, and feature-extracted reviews as input and detects fraudulent patterns such as mass posting from the same IP address, text similarity, and posting from new accounts. Specifically, it calculates the probability of fraud using algorithms such as the Naive Bayes algorithm.

[0326] Step 5:

[0327] The server filters out the detected fraudulent reviews and displays only reviews based on genuine user experiences. As input, it receives a list of reviews that have been determined to be fraudulent and generates a list of reliable reviews that excludes them. Specifically, it removes the fraudulent review list from the database and retains the remaining reviews.

[0328] Step 6:

[0329] The server uses an emotion engine to recognize the user's emotional state in real time and adjust the display of reviews according to the emotional state. It receives the user's emotional data and a list of reviews as input, and optimizes the display order and content of the reviews according to the user's emotional state (e.g., positive or negative). Specifically, it analyzes the user's emotional data and runs an algorithm to prioritize reviews that match the user's emotional state.

[0330] Step 7:

[0331] The terminal (smartphone) displays the reliable reviews received from the server to the user. As input, it receives the adjusted review list sent from the server and displays it on the screen. Specifically, it visually presents the reviews through a user interface, allowing the user to easily browse the reviews.

[0332] 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.

[0333] 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.

[0334] 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.

[0335] [Second embodiment]

[0336] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0337] 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.

[0338] 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).

[0339] 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.

[0340] 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.

[0341] 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).

[0342] 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.

[0343] 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.

[0344] 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.

[0345] 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.

[0346] 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.

[0347] 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."

[0348] The present invention provides a system for detecting and filtering out fraudulent reviews based on user reviews collected from an online rating platform, thereby providing a genuine user experience. Specific embodiments of this system are described below.

[0349] System configuration

[0350] 1. Data Collection Methods

[0351] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.) by using an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in a database.

[0352] 2. Pretreatment Methods

[0353] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0354] Remove HTML tags and special characters

[0355] Tokenizing text (splitting it into words and phrases)

[0356] Removal of stop words (e.g., "no", "wa", "desu", etc.)

[0357] 3. Feature Extraction Method

[0358] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0359] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0360] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0361] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0362] 4. How to detect fraudulent reviews

[0363] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0364] Mass posting from the same IP address or concentrated posting at the same time

[0365] Text similarity (check for copy-paste)

[0366] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0367] 5. Review filtering and display methods

[0368] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed on the user's device.

[0369] Program implementation example

[0370] As a concrete example, consider a system that collects restaurant reviews and filters out fraudulent reviews.

[0371] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0372] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0373] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0374] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0375] 5. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0376] The system aims to provide users with reliable information based on genuine reviews, helping consumers make accurate purchasing decisions. It also contributes to improving the credibility of store owners by providing fair reviews.

[0377] The processing flow will be explained below.

[0378] Step 1:

[0379] The server uses the API to collect user reviews from online rating platforms. Specifically, it acquires the text, rating score, reviewer information, and posting date and time of reviews posted within a specified period and stores them in an internal database.

[0380] Step 2:

[0381] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it performs tokenization, dividing the text into words and phrases. Finally, it removes meaningless stop words (e.g., "no," "wa," and "desu").

[0382] Step 3:

[0383] The server extracts specific features from the pre-processed reviews, performs sentiment analysis on the reviews to determine whether they are positive, negative, or neutral, and extracts frequently occurring keywords and phrases to analyze review trends.

[0384] Step 4:

[0385] The server applies machine learning algorithms to detect fraudulent reviews. It checks for large numbers of reviews from the same IP address and detects abnormal patterns, such as a large number of reviews posted in a short period of time or from a new account. It also analyzes the similarity of the review text between reviews.

[0386] Step 5:

[0387] The server filters out reviews that are determined to be fraudulent and only filters out genuine reviews. The filtered reviews are then deleted from the database, leaving only reviews based on genuine user experiences in the dataset.

[0388] Step 6:

[0389] The server sends the filtered, truthful reviews to the user's device, which displays these reviews to the user and makes them available as reference information.

[0390] Example 1

[0391] 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."

[0392] Currently, online rating platforms collect a large number of user reviews, but these often contain fraudulent reviews. These fraudulent reviews prevent consumers from obtaining accurate information and lead to a decline in trust. Particularly problematic are mass postings from the same IP address, copy-pasted text, and mass postings from new accounts in a short period of time.

[0393] 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.

[0394] In this invention, the server includes means for collecting evaluation information from an online evaluation platform, means for preprocessing the collected text of the evaluation information, means for extracting features of the preprocessed evaluation information, means for analyzing sentiment using natural language processing technology, means for detecting fraudulent information using a machine learning algorithm, and means for filtering out the detected fraudulent information and displaying only evaluation information based on true user experiences, thereby increasing the reliability of the evaluation information and enabling users to make decisions based on accurate information.

[0395] "Online Rating Platform" means an online interface that collects and provides user reviews and rating information.

[0396] "Rating information" is data such as reviews provided by users, rating scores, reviewer information, and posting dates and times.

[0397] "Preprocessing the text" refers to the process of removing unnecessary markup tags and special symbols from the evaluation information, splitting the text, and removing stop words.

[0398] "Feature extraction" means extracting important keywords and patterns from preprocessed text and preparing the data for analysis.

[0399] "Natural language processing technology" is a technology for analyzing text data and understanding emotions and intentions.

[0400] "Analyzing emotions" refers to the process of determining whether the content of text data is positive, negative, or neutral.

[0401] A "machine learning algorithm" is an algorithm that learns patterns from data and uses them to make classifications and predictions.

[0402] "Detecting fraudulent information" refers to the process of identifying fraudulent reviews or abnormal patterns from rating information.

[0403] A "true user experience" is an experience based on the evaluations and impressions of actual users, without any fraudulent manipulation or manipulation whatsoever.

[0404] The "means for displaying" is a method for displaying the filtered and accurate evaluation information on the user's terminal.

[0405] The present invention relates to a system that detects and removes fraudulent information based on evaluation information collected from an online evaluation platform, and provides a genuine user experience. The system consists of the following components:

[0406] 1. Data Collection Methods

[0407] The server uses an API to collect review information from an online review platform. The collected data includes review text, review score, reviewer information, and posting date and time. As a specific example, the server retrieves review information for "Restaurant A" from the API and stores it in a database. To do this, the server sends a request to a specified API endpoint and stores the JSON data returned as a response in the database.

[0408] 2. Data preprocessing methods

[0409] The server preprocesses the collected text data of the evaluation information. Data preprocessing includes removing HTML tags and special symbols, tokenizing the text, and removing stop words. The server uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. Specifically, the evaluation information "It was a great meal. Five stars!" is tokenized as "Great meal 5 stars."

[0410] 3. Feature Extraction Method

[0411] The server extracts important features from the preprocessed text data. It analyzes word frequency and appearance patterns to extract specific keywords. It also analyzes rating scores and posting frequency as time-series data. The server extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and uses the pandas library to analyze time-series data based on the time axis.

[0412] 4. Sentiment analysis tool

[0413] The server performs sentiment analysis of the review information using natural language processing technology. Specifically, it uses machine learning models (e.g., BERT, LSTM) to determine whether the review content is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive.

[0414] 5. How to detect fraudulent reviews

[0415] The server uses machine learning algorithms to detect fraudulent reviews. It analyzes review patterns and reviewer behavior to identify reviews with fraudulent patterns. Specifically, it uses algorithms such as random forests and support vector machines (SVMs) and references database logs to detect suspicious activity.

[0416] 6. Filtering and Display Methods

[0417] The server filters out fraudulent reviews and only displays rating information based on genuine user experiences. To provide the filtered reviews to the user, the server sends the information to the user's device in JSON or HTML format. The device receives this information and displays accurate rating information to the user.

[0418] This system allows users to make accurate decisions based on reliable information, and also contributes to improving the reliability of store owners by providing fair evaluations.

[0419] Examples of prompt statements

[0420] "Design a system that collects user reviews from online rating platforms and detects and filters out fraudulent reviews. Include the following processing steps: data collection, data preprocessing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display. Please also provide examples of specific APIs or libraries."

[0421] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0422] Step 1: Collect data

[0423] The server collects rating information using the API of the online rating platform. As input, it receives the JSON data returned as a response from the API endpoint. This includes the review text, rating score, reviewer information, and posting date and time. The server stores this data in a database. Specifically, it sends an API request, parses the response, extracts the necessary information, and stores it in the database.

[0424] Step 2: Preprocessing the data

[0425] The server preprocesses the collected text data of rating information. As input, it receives the collected text data of rating information. The server removes HTML tags and special symbols, tokenizes the text (divides it into words and phrases), and removes stop words. Specifically, it uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. As output, it obtains clear preprocessed text data.

[0426] Step 3: Feature extraction

[0427] The server extracts specific features from preprocessed text data. The input is preprocessed text data. The server analyzes word frequency and appearance patterns to identify prominent keywords and phrases. It also analyzes rating scores and posting frequency as time-series data. Specifically, it extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and performs data analysis based on the time axis using the pandas library. The output is the extracted feature data.

[0428] Step 4: Sentiment analysis

[0429] The server performs sentiment analysis on preprocessed text data using natural language processing techniques. The input is the preprocessed text data. The server uses a machine learning model (e.g., BERT, LSTM) to determine whether the text is positive, negative, or neutral. Specifically, the server applies the sentiment analysis model and assigns a sentiment label to each review. The output is text data with the sentiment labels.

[0430] Step 5: Detect fraudulent reviews

[0431] The server uses machine learning algorithms to detect fraudulent reviews. The inputs are text data with sentiment labels and reviewer behavioral history. The server detects mass posting from the same IP address, copy-pasted text, and mass posting from new accounts in a short period of time. Specifically, it uses random forests and support vector machines (SVMs) to identify reviews with fraudulent patterns. The output is data determined to be fraudulent reviews.

[0432] Step 6: Filter and view reviews

[0433] The server filters out fraudulent reviews and only reviews based on genuine user experiences. The inputs are data determined to be fraudulent reviews and accurate rating information. The server runs the filtering algorithm and sends the filtered reviews to the user's device in JSON or HTML format. Specifically, the server filters out fraudulent reviews, generates filtered reviews, and sends them to the user's device. The output is accurate rating information that is displayed on the user's device.

[0434] (Application example 1)

[0435] 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."

[0436] On online rating platforms, user reviews are an important source of information for users to evaluate products and services. However, the presence of fraudulent reviews can make accurate evaluations difficult, potentially leading to incorrect decisions. Furthermore, the lack of review statistics based on sentiment analysis makes it difficult to understand the content of reviews. Therefore, it is necessary to obtain reliable reviews, eliminate fraudulent reviews, and analyze and display the sentiment of reviews, thereby enabling users to make appropriate decisions.

[0437] 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.

[0438] In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for excluding the detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for displaying the reviews after excluding the fraudulent reviews and their sentiment analysis on the user's terminal, thereby enabling users to make appropriate decisions based on reliable reviews and the results of the sentiment analysis.

[0439] An "online review platform" is a system that allows users to post reviews of products and services via the Internet and for other users to view those reviews.

[0440] "User reviews" are written opinions and evaluations of products and services posted by users.

[0441] The "means of collection" refers to the method by which the server obtains user reviews from online rating platforms, such as by using APIs or web scraping.

[0442] "Preprocessing" refers to methods for removing unnecessary information from the collected user review text and converting it into a format that is easy to analyze. Specifically, this includes removing HTML tags and special characters, tokenizing, and removing stop words.

[0443] "Means for extracting features" refers to methods for deriving features from preprocessed user reviews. Specifically, methods include sentiment analysis, keyword extraction, and time series analysis.

[0444] A "machine learning algorithm" is a model in which a computer learns patterns from data and makes predictions or classifications based on those patterns.

[0445] "Fraudulent reviews" are false reviews posted with the intent to manipulate ratings.

[0446] The "detection method" is a method of identifying fraudulent reviews using a machine learning algorithm.

[0447] The "display method" refers to a method for displaying only reviews that meet certain conditions on the user's device. This method targets highly reliable reviews that exclude fraudulent reviews.

[0448] "Sentiment analysis" is a technology that extracts emotions and opinions from text data and determines their nature (positive, negative, neutral).

[0449] "Keyword extraction" is a technique for identifying important words and phrases from user reviews.

[0450] "Time series analysis" is a technique for analyzing how data changes over time.

[0451] A "terminal" is an electronic device that a user uses to view and operate information. Examples include smartphones, tablets, and personal computers.

[0452] The present invention relates to a system that analyzes user reviews collected from an online rating platform, eliminates fraudulent reviews, and presents more reliable reviews, particularly for performing sentiment analysis to assist users in making decisions. An embodiment of this system will now be described in detail.

[0453] First, the server collects user reviews from online rating platforms (e.g., shopping sites or review sites) using APIs or web scraping technology to obtain data such as review text, rating scores, reviewer information, and posting date and time, and stores them in a database. Next, the server preprocesses the collected user review text data, which includes steps such as removing HTML tags and special characters, tokenizing, and removing stop words.

[0454] The server extracts specific features from the preprocessed reviews. This involves sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis determines whether a review is positive, negative, or neutral, while keyword extraction identifies important words and phrases that frequently appear. Time series analysis analyzes rating scores and posting frequency over time. After feature extraction, the server uses machine learning algorithms to detect fraudulent reviews. This detection takes into account large numbers of posts from the same IP address, concentrations in time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.).

[0455] Detected fraudulent reviews are filtered out, and the server filters out only reviews based on genuine user experiences. The filtered reviews and their sentiment analysis results are then displayed on the user's device. Users can access this reliable review information using devices such as smartphones and tablets, allowing them to make appropriate decisions based on accurate information.

[0456] The hardware and software used include a server infrastructure capable of high-performance data collection and processing, specifically software libraries such as Python, scikit-learn, BeautifulSoup, requests, and nltk. Reviews are displayed on the user's device using a browser or a dedicated app.

[0457] As a concrete example, consider the case where a user wants to check reviews for "product ID 12345" on "mail order site A." The user launches the smartphone app and enters "product ID 12345." An example of a prompt sentence in this case is as follows:

[0458] "Collect reviews for product ID 12345, preprocess them to a clean form, perform sentiment analysis, and finally detect and filter out fraudulent reviews."

[0459] Based on these prompts, the server can perform the necessary processing and provide users with reliable reviews and their sentiment analysis results, allowing them to select products with confidence.

[0460] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0461] Step 1:

[0462] The server collects user reviews from an online rating platform using an API. As input, it receives data including reviews related to a specified product or service (text, rating score, reviewer information, and posting date and time), and stores the data in a database as output.

[0463] Step 2:

[0464] The server preprocesses the collected user reviews. As input, it uses the text data of reviews stored in the database, removes HTML tags and special characters, tokenizes, and removes stop words. As output, it generates the preprocessed, clean data.

[0465] Step 3:

[0466] The server extracts specific features from the preprocessed reviews. Using the preprocessed text data as input, it performs sentiment analysis, keyword extraction, and time series analysis. As output, it generates the analysis results as a feature vector, which includes the review's emotional state (positive, negative, neutral), key keywords, and the review's posting timestamp.

[0467] Step 4:

[0468] The server uses a machine learning algorithm to detect fraudulent reviews. It uses the feature vectors obtained through feature extraction as input and analyzes patterns such as large numbers of posts from the same IP address, concentration of posts in specific time periods, text similarity, and reviewer activity history. As an output, it lists reviews that are suspected of being fraudulent and generates a detection result.

[0469] Step 5:

[0470] The server filters out the detected fraudulent reviews and filters out only reviews based on genuine user experiences. As input, it uses the fraudulent review detection results and the original review list to filter out fraudulent reviews. As output, it generates a filtered review list that contains only reliable reviews.

[0471] Step 6:

[0472] The server displays the filtered reviews and their sentiment analysis results on the user's device. Using the filtered review list and sentiment analysis results as input, the server displays the reviews on the screen or application of the user's device, allowing the user to check reliable review information and make appropriate decisions.

[0473] Specifically, when a user launches the app on their smartphone and enters "product ID 12345," the server executes each step in sequence based on the prompt, ultimately displaying reliable review information and the results of sentiment analysis on the user's device. When the prompt is entered, "Collect reviews for product ID 12345, preprocess them into a clean format, perform sentiment analysis, and finally detect and filter out fraudulent reviews," the process proceeds according to the instructions.

[0474] 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.

[0475] The present invention is a system that detects and excludes fraudulent reviews based on user reviews collected from an online rating platform, provides a genuine user experience, and combines this with an emotion engine that recognizes user emotions. Specific embodiments of this system are described below.

[0476] System configuration

[0477] 1. Data Collection Methods

[0478] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.). Specifically, it uses an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database.

[0479] 2. Pretreatment Methods

[0480] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0481] Remove HTML tags and special characters

[0482] Tokenizing text (splitting it into words and phrases)

[0483] Removal of stop words (e.g. "no", "wa", "desu", etc.)

[0484] 3. Feature Extraction Method

[0485] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0486] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0487] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0488] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0489] 4. Emotion Engine

[0490] The server recognizes the user's emotions through the emotion engine and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine can grasp the user's emotional state in real time and adjust the way reviews are displayed accordingly.

[0491] 5. How to detect fraudulent reviews

[0492] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0493] Mass posting from the same IP address or concentrated posting at the same time

[0494] Text similarity (check for copy-paste)

[0495] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0496] 6. Review Filtering and Display Methods

[0497] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed.The server then adjusts the filtered reviews based on the user's emotional state using an emotion engine, and finally displays them on the user's device.

[0498] Program implementation example

[0499] As a concrete example, consider a system that collects restaurant reviews, filters out fraudulent reviews, and adjusts the display depending on the user's emotional state.

[0500] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0501] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0502] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0503] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0504] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust the way reviews are displayed accordingly. For example, if the user is in a positive emotional state, positive reviews will be displayed first.

[0505] 6. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0506] The system aims to help users make accurate purchasing decisions based on reliable reviews, helping consumers make accurate decisions. Recognizing users' emotions in real time also enables more personalized information to be provided, improving the browsing experience. Store owners can also benefit from fair reviews, which contributes to increased trust.

[0507] The processing flow will be explained below.

[0508] Step 1:

[0509] The server uses APIs to collect user reviews from online rating platforms. For example, it automatically retrieves review texts, rating scores, reviewer information, and posting dates / times related to specific periods or specific subjects (restaurants, products, etc.), and stores them in an internal database.

[0510] Step 2:

[0511] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it tokenizes the review text and splits it into words and phrases. Finally, it removes stop words (such as "no," "wa," and "desu") and converts it into a format that is easy to analyze.

[0512] Step 3:

[0513] The server extracts specific features from the preprocessed reviews. First, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. Next, it extracts frequently occurring keywords and important phrases from the review text. Finally, it performs time-series analysis to analyze the rating score and posting frequency over time.

[0514] Step 4:

[0515] The server uses an emotion engine to recognize the user's emotions and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine also detects the user's emotional state from facial expressions and voice while viewing reviews, and grasps changes in emotions in real time.

[0516] Step 5:

[0517] The server applies machine learning algorithms to detect fraudulent reviews. It analyzes large numbers of posts from the same IP address, text similarity, and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.) to identify fraudulent reviews. Reviews that are deemed to be non-issues are flagged as genuine reviews.

[0518] Step 6:

[0519] The server adjusts the display method of the filtered true reviews according to the user's emotional state. Depending on the user's emotional state (positive, negative, etc.) recognized by the emotion engine, it decides whether to prioritize positive reviews or balanced reviews.

[0520] Step 7:

[0521] The server then sends the final selected reviews to the user's device, which then displays the received reviews in a format appropriate to the user's emotional state. For example, if the user is in a positive emotional state, more positive reviews will be displayed.

[0522] Step 8:

[0523] Users can view the displayed reviews and make purchasing decisions based on reliable information. When a user posts a new review, that review is also analyzed through the same process.

[0524] Example 2

[0525] 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."

[0526] Online rating platforms are prone to a flood of fraudulent reviews, leading to users making erroneous decisions based on unreliable information. Furthermore, reviews that do not reflect the user's emotional state are not provided, making it difficult to help users make appropriate purchasing decisions.

[0527] 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 user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for analyzing users' emotional tendencies using an emotion engine, means for detecting fraudulent reviews using a machine learning algorithm, and means for excluding detected fraudulent reviews, filtering reviews based on true user experiences, and displaying them according to the user's emotional state. This allows users to make accurate decisions based on reliable reviews and provides information adapted to the user's emotions.

[0528] An "online rating platform" refers to a system on the Internet where users can post ratings and reviews of products and services and other users can view them.

[0529] "User reviews" refers to ratings and comments about products and services posted by users on online rating platforms.

[0530] "Means of collection" refers to the way in which the server automatically obtains user reviews from online rating platforms using technologies such as APIs or web scraping.

[0531] The "means of text preprocessing" refers to a method of removing HTML tags and special characters from the text data of user reviews collected by the server, and performing tokenization and removal of stop words.

[0532] "Means for extracting features" refers to the methods by which the server extracts important information and patterns from pre-processed user reviews, including sentiment analysis, keyword extraction, and time series analysis.

[0533] An "emotion engine" refers to the technology and algorithms used to analyze a user's emotional state from text data, which then classifies emotions as positive, negative, or neutral.

[0534] "Machine learning algorithms" refer to technology that trains models based on past data to classify and predict unknown data. In this case, they are used to detect fraudulent reviews.

[0535] "Fraudulent reviews" are manipulated ratings and false comments that provide biased information that differs from the normal user experience.

[0536] "Filtering means" refers to the method by which the server filters out fraudulent reviews and selects only genuine reviews.

[0537] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[0538] "Means for displaying" refers to the method by which the server selects the filtered reviews and ultimately displays them on the user's device.

[0539] This invention is a system that provides reliable reviews by detecting and filtering out fraudulent reviews based on user reviews collected from online rating platforms. Furthermore, it aims to improve the user experience by combining it with an emotion engine that recognizes the user's emotional state.

[0540] Hardware and software configuration

[0541] The following hardware and software are used to implement this system.

[0542] Server: This is the main component used for data collection, pre-processing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display.

[0543] Database: This is an internal database that stores collected user reviews. It uses a relational database such as MySQL or PostgreSQL.

[0544] API: A means of collecting reviews from online rating platforms, such as a RESTful API.

[0545] Machine learning algorithms are used to detect fraudulent reviews, specifically using machine learning libraries such as Scikit-learn and TensorFlow.

[0546] Sentiment engine: A means of analyzing user sentiment, analyzing review text data using natural language processing technology. Libraries such as NLTK and TextBlob, or APIs dedicated to sentiment analysis (e.g., IBM Watson Natural Language Understanding) are used.

[0547] Data collection

[0548] The server automatically collects user reviews from online rating platforms using APIs. For example, to collect reviews for a specific restaurant, the server uses the endpoint "https: / / api.example.com / reviews?restaurant_id=A".

[0549] Data Preprocessing

[0550] The server preprocesses the collected user review text data, specifically removing HTML tags and special characters, tokenizing, and removing stop words, so that the text data is in a format that is easy to analyze.

[0551] Feature Extraction

[0552] The server extracts specific features from the pre-processed reviews. This step uses a variety of analytical techniques, including sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis involves classifying emotions into "positive," "negative," and "neutral," extracting important keywords, and analyzing review trends over time.

[0553] Emotion Engine

[0554] The server uses an emotion engine to analyze the user's emotional state in real time. The emotion engine determines the user's psychological state based on the content of the reviews and grasps the overall emotional trend. For example, if the user is in a positive emotional state, positive reviews will be displayed preferentially.

[0555] Fraudulent review detection

[0556] The server uses machine learning algorithms to detect fraudulent reviews, which analyze factors such as large numbers of posts from the same IP address, similar text, and reviewer activity history to identify reviews that are likely to be fraudulent.

[0557] Filter and view reviews

[0558] The server filters out fraudulent reviews and only reviews based on genuine user experiences, then adjusts the way reviews are displayed according to the user's emotional state, and finally displays the filtered reviews on the user's device.

[0559] Specific examples

[0560] As a concrete example, the process of collecting, preprocessing, analyzing, detecting fraudulent reviews, and displaying restaurant reviews is shown below.

[0561] 1. The server collects reviews for "Restaurant A" from the API.

[0562] 2. The server removes HTML tags from the collected data and performs tokenization and stop word removal.

[0563] 3. The server performs sentiment analysis, determines the review as positive, and extracts important keywords.

[0564] 4. The server uses machine learning algorithms to detect fraudulent reviews and certify legitimate reviews.

[0565] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust how the review is displayed.

[0566] 6. The server filters the true reviews and displays them on the user's device.

[0567] Prompt Sentence Examples

[0568] By inputting the following prompts into the generative AI model, an explanation of the system and specific examples can be generated.

[0569] Example prompt:

[0570] "Describe a system that detects and filters out fraudulent reviews based on user reviews collected from an online rating platform, and provides a genuine user experience. Also, provide a concrete example of how to utilize an emotion engine that recognizes user emotions."

[0571] In this way, by detailing the specific steps and techniques required to implement the invention, it is possible to accurately understand the invention and provide guidance for others to practice it.

[0572] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0573] Step 1: Data collection methods

[0574] A server automatically collects user reviews from an online rating platform using an API. As input, it uses an API endpoint (e.g., "https: / / api.example.com / reviews?restaurant_id=A"). The server sends the request and stores the response data, such as review text, rating score, reviewer information, and posting date and time, in an internal database. The output is a comprehensive dataset of the collected reviews.

[0575] Step 2: Data preprocessing measures

[0576] The server preprocesses the collected user review text data. Specifically, it removes HTML tags and special characters. For example, Great food " is converted to "Great meal". Next, the text is tokenized and a review such as "Great meal 5 stars" is split into words such as "Great", "Meal", and "5 stars". Furthermore, stop words are removed and "This is a great meal" is converted to "Great meal". The input is the collected text data of reviews, and the output is the preprocessed and clean text data.

[0577] Step 3: Feature extraction methods

[0578] The server extracts specific features from the preprocessed reviews. Specifically, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive. It also extracts important keywords, identifying words such as "amazing," "meal," and "five stars." It then performs time-series analysis to analyze the rating scores and posting frequency over time. The input is the text data of the preprocessed reviews, and the output is the sentiment classification results, extracted keywords, and time-series data.

[0579] Step 4: Emotion Engine

[0580] The server uses an emotion engine to analyze the user's emotional state in real time. Specifically, it understands the user's overall emotional tendency based on the emotion analysis results from the previous step. For example, the text data of a review is input into a natural language processing algorithm to determine the emotional state, such as positive, negative, or neutral. The input is the feature-extracted emotion analysis results, and the output is the user's overall emotional tendency data.

[0581] Step 5: Detecting fraudulent reviews

[0582] The server uses a machine learning algorithm to detect fraudulent reviews. Specific operations include checking for large numbers of posts from the same IP address, concentrated time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.). For example, the machine learning model inputs review metadata to determine whether there is suspicion of fraud. The input is collected and preprocessed review metadata and text data, and the output is the result of the fraudulent review determination.

[0583] Step 6: Filtering and displaying reviews

[0584] The server filters out reviews determined to be fraudulent and filters out only reviews based on genuine user experiences. It also adjusts the way reviews are displayed based on the results of the emotion engine. Specifically, it prioritizes positive or negative reviews depending on the user's emotional state. For example, if a user is in a positive emotional state, positive reviews will be displayed at the top. The input is data on reviews determined to be non-fraudulent and data on the user's emotional state, and the output is the final filtered reviews.

[0585] Through these steps, the system can collect reliable reviews and provide personalized information based on user sentiment.

[0586] (Application example 2)

[0587] 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."

[0588] Online rating platforms face the problem of fraudulent reviews, which can lead to users making purchasing decisions based on unreliable information. Furthermore, they are unable to tailor the display of reviews to reflect the user's emotional state, making it difficult to provide users with information that is appropriate for them.

[0589] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for filtering out detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for recognizing the emotional state of the user and adjusting the display of reviews according to the emotional state. This enables users to make accurate purchasing decisions based on reliable reviews and also enables the provision of personalized information according to the user's emotional state.

[0590] An "online review platform" is a service provided over the Internet that allows users to post and view reviews and ratings of products and services.

[0591] "User reviews" are ratings and comments about products and services posted by users on online rating platforms.

[0592] "Preprocessing" refers to a series of steps performed to convert the collected text data of user reviews into an analyzable format, including removing HTML tags and special characters, tokenizing, and removing stop words.

[0593] "Feature extraction" is a method for extracting meaningful information and patterns from preprocessed user reviews, and includes sentiment analysis, keyword extraction, and time series analysis.

[0594] A "machine learning algorithm" is a method for learning patterns and rules from large amounts of data and performing analysis and classification.

[0595] "Fraudulent reviews" are false reviews that are not based on facts or manipulative reviews posted with a specific intention.

[0596] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[0597] "Adjusting the display of reviews according to emotional state" means optimizing the content and order of reviews to be displayed based on the user's current emotions and mood.

[0598] This invention is a system that detects and excludes fraudulent reviews from users on online rating platforms and displays appropriate reviews according to the user's emotional state. This system mainly consists of a server, a smartphone application, a machine learning algorithm, and an emotion recognition engine.

[0599] System Overview

[0600] The server collects user reviews from online rating platforms, preprocesses the collected reviews, and performs feature extraction. To detect fraudulent reviews, a machine learning algorithm is used. An emotion engine recognizes users' emotional states in real time. Users then receive tailored reviews using a smartphone application.

[0601] Hardware and software used

[0602] The main hardware and software used are servers, smartphones, Python, TextBlob, and Sklearn, which are used for data processing and calculation.

[0603] Data processing and calculation

[0604] 1. Data Collection:

[0605] The server uses APIs to collect user reviews from online rating platforms. The data collected includes review text, rating score, reviewer information, and posting date and time, and is stored in an internal database.

[0606] 2. Pretreatment:

[0607] HTML tags and special characters are removed from the collected review text data, the text is tokenized, and stop words (e.g., "no" and "wa") are removed.

[0608] 3. Feature Extraction:

[0609] After preprocessing, sentiment analysis (determining positive, negative, or neutral), keyword extraction, and time series analysis are performed on the data.

[0610] 4. Fraudulent Review Detection:

[0611] The server uses machine learning algorithms (e.g., Naive Bayes) to detect fraudulent reviews, including criteria such as large numbers of reviews from the same IP address, similar text, and reviews posted by new accounts.

[0612] 5. User emotion recognition and display adjustment:

[0613] The emotion engine detects the user's current emotional state in real time and adjusts how reviews are displayed accordingly: for example, if the user is in a positive emotional state, positive reviews will be displayed first.

[0614] Specific examples

[0615] "As a concrete example of an application to increase the reliability of user reviews on an online shopping site, the server collects product reviews via an API, tokenizes the text data through preprocessing, performs sentiment analysis, and detects and excludes fraudulent reviews using the Naive Bayes algorithm. The system then prioritizes displaying positive or negative reviews depending on the user's emotional state."

[0616] Generative AI model prompt example

[0617] "Write a program in Python to recognize emotions from user-generated reviews on an online shopping site and filter out fraudulent reviews."

[0618] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0619] Step 1:

[0620] The server collects user reviews through the API of the online rating platform. As input, it receives data such as the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database. Specifically, it sends an HTTP request, parses the response data, and stores it.

[0621] Step 2:

[0622] The server preprocesses the collected user review text data. It receives the review text data as input, removes HTML tags and special characters, tokenizes it (divides it into words and phrases), and removes stop words. Specifically, it cleans up the text using regular expressions and natural language processing libraries.

[0623] Step 3:

[0624] The server extracts features from preprocessed user reviews. It receives the preprocessed text data as input and performs sentiment analysis (positive, negative, neutral), keyword extraction, and time series analysis. Specifically, it calculates sentiment scores using the TextBlob library and extracts frequently occurring keywords using Sklearn.

[0625] Step 4:

[0626] The server uses machine learning algorithms to detect fraudulent reviews. It receives collected, preprocessed, and feature-extracted reviews as input and detects fraudulent patterns such as mass posting from the same IP address, text similarity, and posting from new accounts. Specifically, it calculates the probability of fraud using algorithms such as the Naive Bayes algorithm.

[0627] Step 5:

[0628] The server filters out the detected fraudulent reviews and displays only reviews based on genuine user experiences. As input, it receives a list of reviews that have been determined to be fraudulent and generates a list of reliable reviews that excludes them. Specifically, it removes the fraudulent review list from the database and retains the remaining reviews.

[0629] Step 6:

[0630] The server uses an emotion engine to recognize the user's emotional state in real time and adjust the display of reviews according to the emotional state. It receives the user's emotional data and a list of reviews as input, and optimizes the display order and content of the reviews according to the user's emotional state (e.g., positive or negative). Specifically, it analyzes the user's emotional data and runs an algorithm to prioritize reviews that match the user's emotional state.

[0631] Step 7:

[0632] The terminal (smartphone) displays the reliable reviews received from the server to the user. As input, it receives the adjusted review list sent from the server and displays it on the screen. Specifically, it visually presents the reviews through a user interface, allowing the user to easily browse the reviews.

[0633] 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.

[0634] 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.

[0635] 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.

[0636] [Third embodiment]

[0637] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0638] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0639] 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).

[0640] 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.

[0641] 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.

[0642] 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).

[0643] 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.

[0644] 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.

[0645] 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.

[0646] 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.

[0647] 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.

[0648] 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."

[0649] The present invention provides a system for detecting and filtering out fraudulent reviews based on user reviews collected from an online rating platform, thereby providing a genuine user experience. Specific embodiments of this system are described below.

[0650] System configuration

[0651] 1. Data Collection Methods

[0652] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.) by using an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in a database.

[0653] 2. Pretreatment Methods

[0654] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0655] Remove HTML tags and special characters

[0656] Tokenizing text (splitting it into words and phrases)

[0657] Removal of stop words (e.g., "no", "wa", "desu", etc.)

[0658] 3. Feature Extraction Method

[0659] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0660] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0661] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0662] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0663] 4. How to detect fraudulent reviews

[0664] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0665] Mass posting from the same IP address or concentrated posting at the same time

[0666] Text similarity (check for copy-paste)

[0667] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0668] 5. Review filtering and display methods

[0669] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed on the user's device.

[0670] Program implementation example

[0671] As a concrete example, consider a system that collects restaurant reviews and filters out fraudulent reviews.

[0672] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0673] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0674] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0675] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0676] 5. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0677] The system aims to provide users with reliable information based on genuine reviews, helping consumers make accurate purchasing decisions. It also contributes to improving the credibility of store owners by providing fair reviews.

[0678] The processing flow will be explained below.

[0679] Step 1:

[0680] The server uses the API to collect user reviews from online rating platforms. Specifically, it acquires the text, rating score, reviewer information, and posting date and time of reviews posted within a specified period and stores them in an internal database.

[0681] Step 2:

[0682] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it performs tokenization, dividing the text into words and phrases. Finally, it removes meaningless stop words (e.g., "no," "wa," and "desu").

[0683] Step 3:

[0684] The server extracts specific features from the pre-processed reviews, performs sentiment analysis on the reviews to determine whether they are positive, negative, or neutral, and extracts frequently occurring keywords and phrases to analyze review trends.

[0685] Step 4:

[0686] The server applies machine learning algorithms to detect fraudulent reviews. It checks for large numbers of reviews from the same IP address and detects abnormal patterns, such as a large number of reviews posted in a short period of time or from a new account. It also analyzes the similarity of the review text between reviews.

[0687] Step 5:

[0688] The server filters out reviews that are determined to be fraudulent and only filters out genuine reviews. The filtered reviews are then deleted from the database, leaving only reviews based on genuine user experiences in the dataset.

[0689] Step 6:

[0690] The server sends the filtered, truthful reviews to the user's device, which displays these reviews to the user and makes them available as reference information.

[0691] Example 1

[0692] 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."

[0693] Currently, online rating platforms collect a large number of user reviews, but these often contain fraudulent reviews. These fraudulent reviews prevent consumers from obtaining accurate information and lead to a decline in trust. Particularly problematic are mass postings from the same IP address, copy-pasted text, and mass postings from new accounts in a short period of time.

[0694] 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.

[0695] In this invention, the server includes means for collecting evaluation information from an online evaluation platform, means for preprocessing the collected text of the evaluation information, means for extracting features of the preprocessed evaluation information, means for analyzing sentiment using natural language processing technology, means for detecting fraudulent information using a machine learning algorithm, and means for filtering out the detected fraudulent information and displaying only evaluation information based on true user experiences, thereby increasing the reliability of the evaluation information and enabling users to make decisions based on accurate information.

[0696] "Online Rating Platform" means an online interface that collects and provides user reviews and rating information.

[0697] "Rating information" is data such as reviews provided by users, rating scores, reviewer information, and posting dates and times.

[0698] "Preprocessing the text" refers to the process of removing unnecessary markup tags and special symbols from the evaluation information, splitting the text, and removing stop words.

[0699] "Feature extraction" means extracting important keywords and patterns from preprocessed text and preparing the data for analysis.

[0700] "Natural language processing technology" is a technology for analyzing text data and understanding emotions and intentions.

[0701] "Analyzing emotions" refers to the process of determining whether the content of text data is positive, negative, or neutral.

[0702] A "machine learning algorithm" is an algorithm that learns patterns from data and uses them to make classifications and predictions.

[0703] "Detecting fraudulent information" refers to the process of identifying fraudulent reviews or abnormal patterns from rating information.

[0704] A "true user experience" is an experience based on the evaluations and impressions of actual users, without any fraudulent manipulation or manipulation whatsoever.

[0705] The "means for displaying" is a method for displaying the filtered and accurate evaluation information on the user's terminal.

[0706] The present invention relates to a system that detects and removes fraudulent information based on evaluation information collected from an online evaluation platform, and provides a genuine user experience. The system consists of the following components:

[0707] 1. Data Collection Methods

[0708] The server uses an API to collect review information from an online review platform. The collected data includes review text, review score, reviewer information, and posting date and time. As a specific example, the server retrieves review information for "Restaurant A" from the API and stores it in a database. To do this, the server sends a request to a specified API endpoint and stores the JSON data returned as a response in the database.

[0709] 2. Data preprocessing methods

[0710] The server preprocesses the collected text data of the evaluation information. Data preprocessing includes removing HTML tags and special symbols, tokenizing the text, and removing stop words. The server uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. Specifically, the evaluation information "It was a great meal. Five stars!" is tokenized as "Great meal 5 stars."

[0711] 3. Feature Extraction Method

[0712] The server extracts important features from the preprocessed text data. It analyzes word frequency and appearance patterns to extract specific keywords. It also analyzes rating scores and posting frequency as time-series data. The server extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and uses the pandas library to analyze time-series data based on the time axis.

[0713] 4. Sentiment analysis tool

[0714] The server performs sentiment analysis of the review information using natural language processing technology. Specifically, it uses machine learning models (e.g., BERT, LSTM) to determine whether the review content is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive.

[0715] 5. How to detect fraudulent reviews

[0716] The server uses machine learning algorithms to detect fraudulent reviews. It analyzes review patterns and reviewer behavior to identify reviews with fraudulent patterns. Specifically, it uses algorithms such as random forests and support vector machines (SVMs) and references database logs to detect suspicious activity.

[0717] 6. Filtering and Display Methods

[0718] The server filters out fraudulent reviews and only displays rating information based on genuine user experiences. To provide the filtered reviews to the user, the server sends the information to the user's device in JSON or HTML format. The device receives this information and displays accurate rating information to the user.

[0719] This system allows users to make accurate decisions based on reliable information, and also contributes to improving the reliability of store owners by providing fair evaluations.

[0720] Examples of prompt statements

[0721] "Design a system that collects user reviews from online rating platforms and detects and filters out fraudulent reviews. Include the following processing steps: data collection, data preprocessing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display. Please also provide examples of specific APIs or libraries."

[0722] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0723] Step 1: Collect data

[0724] The server collects rating information using the API of the online rating platform. As input, it receives the JSON data returned as a response from the API endpoint. This includes the review text, rating score, reviewer information, and posting date and time. The server stores this data in a database. Specifically, it sends an API request, parses the response, extracts the necessary information, and stores it in the database.

[0725] Step 2: Preprocessing the data

[0726] The server preprocesses the collected text data of rating information. As input, it receives the collected text data of rating information. The server removes HTML tags and special symbols, tokenizes the text (divides it into words and phrases), and removes stop words. Specifically, it uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. As output, it obtains clear preprocessed text data.

[0727] Step 3: Feature extraction

[0728] The server extracts specific features from preprocessed text data. The input is preprocessed text data. The server analyzes word frequency and appearance patterns to identify prominent keywords and phrases. It also analyzes rating scores and posting frequency as time-series data. Specifically, it extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and performs data analysis based on the time axis using the pandas library. The output is the extracted feature data.

[0729] Step 4: Sentiment analysis

[0730] The server performs sentiment analysis on preprocessed text data using natural language processing techniques. The input is the preprocessed text data. The server uses a machine learning model (e.g., BERT, LSTM) to determine whether the text is positive, negative, or neutral. Specifically, the server applies the sentiment analysis model and assigns a sentiment label to each review. The output is text data with the sentiment labels.

[0731] Step 5: Detect fraudulent reviews

[0732] The server uses machine learning algorithms to detect fraudulent reviews. The inputs are text data with sentiment labels and reviewer behavioral history. The server detects mass posting from the same IP address, copy-pasted text, and mass posting from new accounts in a short period of time. Specifically, it uses random forests and support vector machines (SVMs) to identify reviews with fraudulent patterns. The output is data determined to be fraudulent reviews.

[0733] Step 6: Filter and view reviews

[0734] The server filters out fraudulent reviews and only reviews based on genuine user experiences. The inputs are data determined to be fraudulent reviews and accurate rating information. The server runs the filtering algorithm and sends the filtered reviews to the user's device in JSON or HTML format. Specifically, the server filters out fraudulent reviews, generates filtered reviews, and sends them to the user's device. The output is accurate rating information that is displayed on the user's device.

[0735] (Application example 1)

[0736] 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."

[0737] On online rating platforms, user reviews are an important source of information for users to evaluate products and services. However, the presence of fraudulent reviews can make accurate evaluations difficult, potentially leading to incorrect decisions. Furthermore, the lack of review statistics based on sentiment analysis makes it difficult to understand the content of reviews. Therefore, it is necessary to obtain reliable reviews, eliminate fraudulent reviews, and analyze and display the sentiment of reviews, thereby enabling users to make appropriate decisions.

[0738] 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.

[0739] In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for excluding the detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for displaying the reviews after excluding the fraudulent reviews and their sentiment analysis on the user's terminal, thereby enabling users to make appropriate decisions based on reliable reviews and the results of the sentiment analysis.

[0740] An "online review platform" is a system that allows users to post reviews of products and services via the Internet and for other users to view those reviews.

[0741] "User reviews" are written opinions and evaluations of products and services posted by users.

[0742] The "means of collection" refers to the method by which the server obtains user reviews from online rating platforms, such as by using APIs or web scraping.

[0743] "Preprocessing" refers to methods for removing unnecessary information from the collected user review text and converting it into a format that is easy to analyze. Specifically, this includes removing HTML tags and special characters, tokenizing, and removing stop words.

[0744] "Means for extracting features" refers to methods for deriving features from preprocessed user reviews. Specifically, methods include sentiment analysis, keyword extraction, and time series analysis.

[0745] A "machine learning algorithm" is a model in which a computer learns patterns from data and makes predictions or classifications based on those patterns.

[0746] "Fraudulent reviews" are false reviews posted with the intent to manipulate ratings.

[0747] The "detection method" is a method of identifying fraudulent reviews using a machine learning algorithm.

[0748] The "display method" refers to a method for displaying only reviews that meet certain conditions on the user's device. This method targets highly reliable reviews that exclude fraudulent reviews.

[0749] "Sentiment analysis" is a technology that extracts emotions and opinions from text data and determines their nature (positive, negative, neutral).

[0750] "Keyword extraction" is a technique for identifying important words and phrases from user reviews.

[0751] "Time series analysis" is a technique for analyzing how data changes over time.

[0752] A "terminal" is an electronic device that a user uses to view and operate information. Examples include smartphones, tablets, and personal computers.

[0753] The present invention relates to a system that analyzes user reviews collected from an online rating platform, eliminates fraudulent reviews, and presents more reliable reviews, particularly for performing sentiment analysis to assist users in making decisions. An embodiment of this system will now be described in detail.

[0754] First, the server collects user reviews from online rating platforms (e.g., shopping sites or review sites) using APIs or web scraping technology to obtain data such as review text, rating scores, reviewer information, and posting date and time, and stores them in a database. Next, the server preprocesses the collected user review text data, which includes steps such as removing HTML tags and special characters, tokenizing, and removing stop words.

[0755] The server extracts specific features from the preprocessed reviews. This involves sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis determines whether a review is positive, negative, or neutral, while keyword extraction identifies important words and phrases that frequently appear. Time series analysis analyzes rating scores and posting frequency over time. After feature extraction, the server uses machine learning algorithms to detect fraudulent reviews. This detection takes into account large numbers of posts from the same IP address, concentrations in time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.).

[0756] Detected fraudulent reviews are filtered out, and the server filters out only reviews based on genuine user experiences. The filtered reviews and their sentiment analysis results are then displayed on the user's device. Users can access this reliable review information using devices such as smartphones and tablets, allowing them to make appropriate decisions based on accurate information.

[0757] The hardware and software used include a server infrastructure capable of high-performance data collection and processing, specifically software libraries such as Python, scikit-learn, BeautifulSoup, requests, and nltk. Reviews are displayed on the user's device using a browser or a dedicated app.

[0758] As a concrete example, consider the case where a user wants to check reviews for "product ID 12345" on "mail order site A." The user launches the smartphone app and enters "product ID 12345." An example of a prompt sentence in this case is as follows:

[0759] "Collect reviews for product ID 12345, preprocess them to a clean form, perform sentiment analysis, and finally detect and filter out fraudulent reviews."

[0760] Based on these prompts, the server can perform the necessary processing and provide users with reliable reviews and their sentiment analysis results, allowing them to select products with confidence.

[0761] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0762] Step 1:

[0763] The server collects user reviews from an online rating platform using an API. As input, it receives data including reviews related to a specified product or service (text, rating score, reviewer information, and posting date and time), and stores the data in a database as output.

[0764] Step 2:

[0765] The server preprocesses the collected user reviews. As input, it uses the text data of reviews stored in the database, removes HTML tags and special characters, tokenizes, and removes stop words. As output, it generates the preprocessed, clean data.

[0766] Step 3:

[0767] The server extracts specific features from the preprocessed reviews. Using the preprocessed text data as input, it performs sentiment analysis, keyword extraction, and time series analysis. As output, it generates the analysis results as a feature vector, which includes the review's emotional state (positive, negative, neutral), key keywords, and the review's posting timestamp.

[0768] Step 4:

[0769] The server uses a machine learning algorithm to detect fraudulent reviews. It uses the feature vectors obtained through feature extraction as input and analyzes patterns such as large numbers of posts from the same IP address, concentration of posts in specific time periods, text similarity, and reviewer activity history. As an output, it lists reviews that are suspected of being fraudulent and generates a detection result.

[0770] Step 5:

[0771] The server filters out the detected fraudulent reviews and filters out only reviews based on genuine user experiences. As input, it uses the fraudulent review detection results and the original review list to filter out fraudulent reviews. As output, it generates a filtered review list that contains only reliable reviews.

[0772] Step 6:

[0773] The server displays the filtered reviews and their sentiment analysis results on the user's device. Using the filtered review list and sentiment analysis results as input, the server displays the reviews on the screen or application of the user's device, allowing the user to check reliable review information and make appropriate decisions.

[0774] Specifically, when a user launches the app on their smartphone and enters "product ID 12345," the server executes each step in sequence based on the prompt, ultimately displaying reliable review information and the results of sentiment analysis on the user's device. When the prompt is entered, "Collect reviews for product ID 12345, preprocess them into a clean format, perform sentiment analysis, and finally detect and filter out fraudulent reviews," the process proceeds according to the instructions.

[0775] 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.

[0776] The present invention is a system that detects and excludes fraudulent reviews based on user reviews collected from an online rating platform, provides a genuine user experience, and combines this with an emotion engine that recognizes user emotions. Specific embodiments of this system are described below.

[0777] System configuration

[0778] 1. Data Collection Methods

[0779] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.). Specifically, it uses an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database.

[0780] 2. Pretreatment Methods

[0781] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0782] Remove HTML tags and special characters

[0783] Tokenizing text (splitting it into words and phrases)

[0784] Removal of stop words (e.g. "no", "wa", "desu", etc.)

[0785] 3. Feature Extraction Method

[0786] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0787] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0788] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0789] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0790] 4. Emotion Engine

[0791] The server recognizes the user's emotions through the emotion engine and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine can grasp the user's emotional state in real time and adjust the way reviews are displayed accordingly.

[0792] 5. How to detect fraudulent reviews

[0793] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0794] Mass posting from the same IP address or concentrated posting at the same time

[0795] Text similarity (check for copy-paste)

[0796] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0797] 6. Review Filtering and Display Methods

[0798] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed.The server then adjusts the filtered reviews based on the user's emotional state using an emotion engine, and finally displays them on the user's device.

[0799] Program implementation example

[0800] As a concrete example, consider a system that collects restaurant reviews, filters out fraudulent reviews, and adjusts the display depending on the user's emotional state.

[0801] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0802] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0803] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0804] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0805] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust the way reviews are displayed accordingly. For example, if the user is in a positive emotional state, positive reviews will be displayed first.

[0806] 6. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0807] The system aims to help users make accurate purchasing decisions based on reliable reviews, helping consumers make accurate decisions. Recognizing users' emotions in real time also enables more personalized information to be provided, improving the browsing experience. Store owners can also benefit from fair reviews, which contributes to increased trust.

[0808] The processing flow will be explained below.

[0809] Step 1:

[0810] The server uses APIs to collect user reviews from online rating platforms. For example, it automatically retrieves review texts, rating scores, reviewer information, and posting dates / times related to specific periods or specific subjects (restaurants, products, etc.), and stores them in an internal database.

[0811] Step 2:

[0812] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it tokenizes the review text and splits it into words and phrases. Finally, it removes stop words (such as "no," "wa," and "desu") and converts it into a format that is easy to analyze.

[0813] Step 3:

[0814] The server extracts specific features from the preprocessed reviews. First, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. Next, it extracts frequently occurring keywords and important phrases from the review text. Finally, it performs time-series analysis to analyze the rating score and posting frequency over time.

[0815] Step 4:

[0816] The server uses an emotion engine to recognize the user's emotions and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine also detects the user's emotional state from facial expressions and voice while viewing reviews, and grasps changes in emotions in real time.

[0817] Step 5:

[0818] The server applies machine learning algorithms to detect fraudulent reviews. It analyzes large numbers of posts from the same IP address, text similarity, and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.) to identify fraudulent reviews. Reviews that are deemed to be non-issues are flagged as genuine reviews.

[0819] Step 6:

[0820] The server adjusts the display method of the filtered true reviews according to the user's emotional state. Depending on the user's emotional state (positive, negative, etc.) recognized by the emotion engine, it decides whether to prioritize positive reviews or balanced reviews.

[0821] Step 7:

[0822] The server then sends the final selected reviews to the user's device, which then displays the received reviews in a format appropriate to the user's emotional state. For example, if the user is in a positive emotional state, more positive reviews will be displayed.

[0823] Step 8:

[0824] Users can view the displayed reviews and make purchasing decisions based on reliable information. When a user posts a new review, that review is also analyzed through the same process.

[0825] Example 2

[0826] 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."

[0827] Online rating platforms are prone to a flood of fraudulent reviews, leading to users making erroneous decisions based on unreliable information. Furthermore, reviews that do not reflect the user's emotional state are not provided, making it difficult to help users make appropriate purchasing decisions.

[0828] 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 user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for analyzing users' emotional tendencies using an emotion engine, means for detecting fraudulent reviews using a machine learning algorithm, and means for excluding detected fraudulent reviews, filtering reviews based on true user experiences, and displaying them according to the user's emotional state. This allows users to make accurate decisions based on reliable reviews and provides information adapted to the user's emotions.

[0829] An "online rating platform" refers to a system on the Internet where users can post ratings and reviews of products and services and other users can view them.

[0830] "User reviews" refers to ratings and comments about products and services posted by users on online rating platforms.

[0831] "Means of collection" refers to the way in which the server automatically obtains user reviews from online rating platforms using technologies such as APIs or web scraping.

[0832] The "means of text preprocessing" refers to a method of removing HTML tags and special characters from the text data of user reviews collected by the server, and performing tokenization and removal of stop words.

[0833] "Means for extracting features" refers to the methods by which the server extracts important information and patterns from pre-processed user reviews, including sentiment analysis, keyword extraction, and time series analysis.

[0834] An "emotion engine" refers to the technology and algorithms used to analyze a user's emotional state from text data, which then classifies emotions as positive, negative, or neutral.

[0835] "Machine learning algorithms" refer to technology that trains models based on past data to classify and predict unknown data. In this case, they are used to detect fraudulent reviews.

[0836] "Fraudulent reviews" are manipulated ratings and false comments that provide biased information that differs from the normal user experience.

[0837] "Filtering means" refers to the method by which the server filters out fraudulent reviews and selects only genuine reviews.

[0838] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[0839] "Means for displaying" refers to the method by which the server selects the filtered reviews and ultimately displays them on the user's device.

[0840] This invention is a system that provides reliable reviews by detecting and filtering out fraudulent reviews based on user reviews collected from online rating platforms. Furthermore, it aims to improve the user experience by combining it with an emotion engine that recognizes the user's emotional state.

[0841] Hardware and software configuration

[0842] The following hardware and software are used to implement this system.

[0843] Server: This is the main component used for data collection, pre-processing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display.

[0844] Database: This is an internal database that stores collected user reviews. It uses a relational database such as MySQL or PostgreSQL.

[0845] API: A means of collecting reviews from online rating platforms, such as a RESTful API.

[0846] Machine learning algorithms are used to detect fraudulent reviews, specifically using machine learning libraries such as Scikit-learn and TensorFlow.

[0847] Sentiment engine: A means of analyzing user sentiment, analyzing review text data using natural language processing technology. Libraries such as NLTK and TextBlob, or APIs dedicated to sentiment analysis (e.g., IBM Watson Natural Language Understanding) are used.

[0848] Data collection

[0849] The server automatically collects user reviews from online rating platforms using APIs. For example, to collect reviews for a specific restaurant, the server uses the endpoint "https: / / api.example.com / reviews?restaurant_id=A".

[0850] Data Preprocessing

[0851] The server preprocesses the collected user review text data, specifically removing HTML tags and special characters, tokenizing, and removing stop words, so that the text data is in a format that is easy to analyze.

[0852] Feature Extraction

[0853] The server extracts specific features from the pre-processed reviews. This step uses a variety of analytical techniques, including sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis involves classifying emotions into "positive," "negative," and "neutral," extracting important keywords, and analyzing review trends over time.

[0854] Emotion Engine

[0855] The server uses an emotion engine to analyze the user's emotional state in real time. The emotion engine determines the user's psychological state based on the content of the reviews and grasps the overall emotional trend. For example, if the user is in a positive emotional state, positive reviews will be displayed preferentially.

[0856] Fraudulent review detection

[0857] The server uses machine learning algorithms to detect fraudulent reviews, which analyze factors such as large numbers of posts from the same IP address, similar text, and reviewer activity history to identify reviews that are likely to be fraudulent.

[0858] Filter and view reviews

[0859] The server filters out fraudulent reviews and only reviews based on genuine user experiences, then adjusts the way reviews are displayed according to the user's emotional state, and finally displays the filtered reviews on the user's device.

[0860] Specific examples

[0861] As a concrete example, the process of collecting, preprocessing, analyzing, detecting fraudulent reviews, and displaying restaurant reviews is shown below.

[0862] 1. The server collects reviews for "Restaurant A" from the API.

[0863] 2. The server removes HTML tags from the collected data and performs tokenization and stop word removal.

[0864] 3. The server performs sentiment analysis, determines the review as positive, and extracts important keywords.

[0865] 4. The server uses machine learning algorithms to detect fraudulent reviews and certify legitimate reviews.

[0866] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust how the review is displayed.

[0867] 6. The server filters the true reviews and displays them on the user's device.

[0868] Prompt Sentence Examples

[0869] By inputting the following prompts into the generative AI model, an explanation of the system and specific examples can be generated.

[0870] Example prompt:

[0871] "Describe a system that detects and filters out fraudulent reviews based on user reviews collected from an online rating platform, and provides a genuine user experience. Also, provide a concrete example of how to utilize an emotion engine that recognizes user emotions."

[0872] In this way, by detailing the specific steps and techniques required to implement the invention, it is possible to accurately understand the invention and provide guidance for others to practice it.

[0873] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0874] Step 1: Data collection methods

[0875] A server automatically collects user reviews from an online rating platform using an API. As input, it uses an API endpoint (e.g., "https: / / api.example.com / reviews?restaurant_id=A"). The server sends the request and stores the response data, such as review text, rating score, reviewer information, and posting date and time, in an internal database. The output is a comprehensive dataset of the collected reviews.

[0876] Step 2: Data preprocessing measures

[0877] The server preprocesses the collected user review text data. Specifically, it removes HTML tags and special characters. For example, Great food " is converted to "Great meal". Next, the text is tokenized and a review such as "Great meal 5 stars" is split into words such as "Great", "Meal", and "5 stars". Furthermore, stop words are removed and "This is a great meal" is converted to "Great meal". The input is the collected text data of reviews, and the output is the preprocessed and clean text data.

[0878] Step 3: Feature extraction methods

[0879] The server extracts specific features from the preprocessed reviews. Specifically, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive. It also extracts important keywords, identifying words such as "amazing," "meal," and "five stars." It then performs time-series analysis to analyze the rating scores and posting frequency over time. The input is the text data of the preprocessed reviews, and the output is the sentiment classification results, extracted keywords, and time-series data.

[0880] Step 4: Emotion Engine

[0881] The server uses an emotion engine to analyze the user's emotional state in real time. Specifically, it understands the user's overall emotional tendency based on the emotion analysis results from the previous step. For example, the text data of a review is input into a natural language processing algorithm to determine the emotional state, such as positive, negative, or neutral. The input is the feature-extracted emotion analysis results, and the output is the user's overall emotional tendency data.

[0882] Step 5: Detecting fraudulent reviews

[0883] The server uses a machine learning algorithm to detect fraudulent reviews. Specific operations include checking for large numbers of posts from the same IP address, concentrated time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.). For example, the machine learning model inputs review metadata to determine whether there is suspicion of fraud. The input is collected and preprocessed review metadata and text data, and the output is the result of the fraudulent review determination.

[0884] Step 6: Filtering and displaying reviews

[0885] The server filters out reviews determined to be fraudulent and filters out only reviews based on genuine user experiences. It also adjusts the way reviews are displayed based on the results of the emotion engine. Specifically, it prioritizes positive or negative reviews depending on the user's emotional state. For example, if a user is in a positive emotional state, positive reviews will be displayed at the top. The input is data on reviews determined to be non-fraudulent and data on the user's emotional state, and the output is the final filtered reviews.

[0886] Through these steps, the system can collect reliable reviews and provide personalized information based on user sentiment.

[0887] (Application example 2)

[0888] 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."

[0889] Online rating platforms face the problem of fraudulent reviews, which can lead to users making purchasing decisions based on unreliable information. Furthermore, they are unable to tailor the display of reviews to reflect the user's emotional state, making it difficult to provide users with information that is appropriate for them.

[0890] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for filtering out detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for recognizing the emotional state of the user and adjusting the display of reviews according to the emotional state. This enables users to make accurate purchasing decisions based on reliable reviews and also enables the provision of personalized information according to the user's emotional state.

[0891] An "online review platform" is a service provided over the Internet that allows users to post and view reviews and ratings of products and services.

[0892] "User reviews" are ratings and comments about products and services posted by users on online rating platforms.

[0893] "Preprocessing" refers to a series of steps performed to convert the collected text data of user reviews into an analyzable format, including removing HTML tags and special characters, tokenizing, and removing stop words.

[0894] "Feature extraction" is a method for extracting meaningful information and patterns from preprocessed user reviews, and includes sentiment analysis, keyword extraction, and time series analysis.

[0895] A "machine learning algorithm" is a method for learning patterns and rules from large amounts of data and performing analysis and classification.

[0896] "Fraudulent reviews" are false reviews that are not based on facts or manipulative reviews posted with a specific intention.

[0897] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[0898] "Adjusting the display of reviews according to emotional state" means optimizing the content and order of reviews to be displayed based on the user's current emotions and mood.

[0899] This invention is a system that detects and excludes fraudulent reviews from users on online rating platforms and displays appropriate reviews according to the user's emotional state. This system mainly consists of a server, a smartphone application, a machine learning algorithm, and an emotion recognition engine.

[0900] System Overview

[0901] The server collects user reviews from online rating platforms, preprocesses the collected reviews, and performs feature extraction. To detect fraudulent reviews, a machine learning algorithm is used. An emotion engine recognizes users' emotional states in real time. Users then receive tailored reviews using a smartphone application.

[0902] Hardware and software used

[0903] The main hardware and software used are servers, smartphones, Python, TextBlob, and Sklearn, which are used for data processing and calculation.

[0904] Data processing and calculation

[0905] 1. Data Collection:

[0906] The server uses APIs to collect user reviews from online rating platforms. The data collected includes review text, rating score, reviewer information, and posting date and time, and is stored in an internal database.

[0907] 2. Pretreatment:

[0908] HTML tags and special characters are removed from the collected review text data, the text is tokenized, and stop words (e.g., "no" and "wa") are removed.

[0909] 3. Feature Extraction:

[0910] After preprocessing, sentiment analysis (determining positive, negative, or neutral), keyword extraction, and time series analysis are performed on the data.

[0911] 4. Fraudulent Review Detection:

[0912] The server uses machine learning algorithms (e.g., Naive Bayes) to detect fraudulent reviews, including criteria such as large numbers of reviews from the same IP address, similar text, and reviews posted by new accounts.

[0913] 5. User emotion recognition and display adjustment:

[0914] The emotion engine detects the user's current emotional state in real time and adjusts how reviews are displayed accordingly: for example, if the user is in a positive emotional state, positive reviews will be displayed first.

[0915] Specific examples

[0916] "As a concrete example of an application to increase the reliability of user reviews on an online shopping site, the server collects product reviews via an API, tokenizes the text data through preprocessing, performs sentiment analysis, and detects and excludes fraudulent reviews using the Naive Bayes algorithm. The system then prioritizes displaying positive or negative reviews depending on the user's emotional state."

[0917] Generative AI model prompt example

[0918] "Write a program in Python to recognize emotions from user-generated reviews on an online shopping site and filter out fraudulent reviews."

[0919] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0920] Step 1:

[0921] The server collects user reviews through the API of the online rating platform. As input, it receives data such as the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database. Specifically, it sends an HTTP request, parses the response data, and stores it.

[0922] Step 2:

[0923] The server preprocesses the collected user review text data. It receives the review text data as input, removes HTML tags and special characters, tokenizes it (divides it into words and phrases), and removes stop words. Specifically, it cleans up the text using regular expressions and natural language processing libraries.

[0924] Step 3:

[0925] The server extracts features from preprocessed user reviews. It receives the preprocessed text data as input and performs sentiment analysis (positive, negative, neutral), keyword extraction, and time series analysis. Specifically, it calculates sentiment scores using the TextBlob library and extracts frequently occurring keywords using Sklearn.

[0926] Step 4:

[0927] The server uses machine learning algorithms to detect fraudulent reviews. It receives collected, preprocessed, and feature-extracted reviews as input and detects fraudulent patterns such as mass posting from the same IP address, text similarity, and posting from new accounts. Specifically, it calculates the probability of fraud using algorithms such as the Naive Bayes algorithm.

[0928] Step 5:

[0929] The server filters out the detected fraudulent reviews and displays only reviews based on genuine user experiences. As input, it receives a list of reviews that have been determined to be fraudulent and generates a list of reliable reviews that excludes them. Specifically, it removes the fraudulent review list from the database and retains the remaining reviews.

[0930] Step 6:

[0931] The server uses an emotion engine to recognize the user's emotional state in real time and adjust the display of reviews according to the emotional state. It receives the user's emotional data and a list of reviews as input, and optimizes the display order and content of the reviews according to the user's emotional state (e.g., positive or negative). Specifically, it analyzes the user's emotional data and runs an algorithm to prioritize reviews that match the user's emotional state.

[0932] Step 7:

[0933] The terminal (smartphone) displays the reliable reviews received from the server to the user. As input, it receives the adjusted review list sent from the server and displays it on the screen. Specifically, it visually presents the reviews through a user interface, allowing the user to easily browse the reviews.

[0934] 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.

[0935] 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.

[0936] 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.

[0937] [Fourth embodiment]

[0938] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[0939] 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.

[0940] 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).

[0941] 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.

[0942] 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.

[0943] 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).

[0944] 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.

[0945] 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.

[0946] 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.

[0947] 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.

[0948] 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.

[0949] 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.

[0950] 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."

[0951] The present invention provides a system for detecting and filtering out fraudulent reviews based on user reviews collected from an online rating platform, thereby providing a genuine user experience. Specific embodiments of this system are described below.

[0952] System configuration

[0953] 1. Data Collection Methods

[0954] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.) by using an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in a database.

[0955] 2. Pretreatment Methods

[0956] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[0957] Remove HTML tags and special characters

[0958] Tokenizing text (splitting it into words and phrases)

[0959] Removal of stop words (e.g., "no", "wa", "desu", etc.)

[0960] 3. Feature Extraction Method

[0961] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[0962] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[0963] Keyword extraction: Identifying important, frequently occurring words and phrases.

[0964] Time series analysis: Analyze evaluation scores and posting frequency over time.

[0965] 4. How to detect fraudulent reviews

[0966] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[0967] Mass posting from the same IP address or concentrated posting at the same time

[0968] Text similarity (check for copy-paste)

[0969] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[0970] 5. Review filtering and display methods

[0971] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed on the user's device.

[0972] Program implementation example

[0973] As a concrete example, consider a system that collects restaurant reviews and filters out fraudulent reviews.

[0974] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[0975] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[0976] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[0977] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[0978] 5. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[0979] The system aims to provide users with reliable information based on genuine reviews, helping consumers make accurate purchasing decisions. It also contributes to improving the credibility of store owners by providing fair reviews.

[0980] The processing flow will be explained below.

[0981] Step 1:

[0982] The server uses the API to collect user reviews from online rating platforms. Specifically, it acquires the text, rating score, reviewer information, and posting date and time of reviews posted within a specified period and stores them in an internal database.

[0983] Step 2:

[0984] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it performs tokenization, dividing the text into words and phrases. Finally, it removes meaningless stop words (e.g., "no," "wa," and "desu").

[0985] Step 3:

[0986] The server extracts specific features from the pre-processed reviews, performs sentiment analysis on the reviews to determine whether they are positive, negative, or neutral, and extracts frequently occurring keywords and phrases to analyze review trends.

[0987] Step 4:

[0988] The server applies machine learning algorithms to detect fraudulent reviews. It checks for large numbers of reviews from the same IP address and detects abnormal patterns, such as a large number of reviews posted in a short period of time or from a new account. It also analyzes the similarity of the review text between reviews.

[0989] Step 5:

[0990] The server filters out reviews that are determined to be fraudulent and only filters out genuine reviews. The filtered reviews are then deleted from the database, leaving only reviews based on genuine user experiences in the dataset.

[0991] Step 6:

[0992] The server sends the filtered, truthful reviews to the user's device, which displays these reviews to the user and makes them available as reference information.

[0993] Example 1

[0994] 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."

[0995] Currently, online rating platforms collect a large number of user reviews, but these often contain fraudulent reviews. These fraudulent reviews prevent consumers from obtaining accurate information and lead to a decline in trust. Particularly problematic are mass postings from the same IP address, copy-pasted text, and mass postings from new accounts in a short period of time.

[0996] 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.

[0997] In this invention, the server includes means for collecting evaluation information from an online evaluation platform, means for preprocessing the collected text of the evaluation information, means for extracting features of the preprocessed evaluation information, means for analyzing sentiment using natural language processing technology, means for detecting fraudulent information using a machine learning algorithm, and means for filtering out the detected fraudulent information and displaying only evaluation information based on true user experiences, thereby increasing the reliability of the evaluation information and enabling users to make decisions based on accurate information.

[0998] "Online Rating Platform" means an online interface that collects and provides user reviews and rating information.

[0999] "Rating information" is data such as reviews provided by users, rating scores, reviewer information, and posting dates and times.

[1000] "Preprocessing the text" refers to the process of removing unnecessary markup tags and special symbols from the evaluation information, splitting the text, and removing stop words.

[1001] "Feature extraction" means extracting important keywords and patterns from preprocessed text and preparing the data for analysis.

[1002] "Natural language processing technology" is a technology for analyzing text data and understanding emotions and intentions.

[1003] "Analyzing emotions" refers to the process of determining whether the content of text data is positive, negative, or neutral.

[1004] A "machine learning algorithm" is an algorithm that learns patterns from data and uses them to make classifications and predictions.

[1005] "Detecting fraudulent information" refers to the process of identifying fraudulent reviews or abnormal patterns from rating information.

[1006] A "true user experience" is an experience based on the evaluations and impressions of actual users, without any fraudulent manipulation or manipulation whatsoever.

[1007] The "means for displaying" is a method for displaying the filtered and accurate evaluation information on the user's terminal.

[1008] The present invention relates to a system that detects and removes fraudulent information based on evaluation information collected from an online evaluation platform, and provides a genuine user experience. The system consists of the following components:

[1009] 1. Data Collection Methods

[1010] The server uses an API to collect review information from an online review platform. The collected data includes review text, review score, reviewer information, and posting date and time. As a specific example, the server retrieves review information for "Restaurant A" from the API and stores it in a database. To do this, the server sends a request to a specified API endpoint and stores the JSON data returned as a response in the database.

[1011] 2. Data preprocessing methods

[1012] The server preprocesses the collected text data of the evaluation information. Data preprocessing includes removing HTML tags and special symbols, tokenizing the text, and removing stop words. The server uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. Specifically, the evaluation information "It was a great meal. Five stars!" is tokenized as "Great meal 5 stars."

[1013] 3. Feature Extraction Method

[1014] The server extracts important features from the preprocessed text data. It analyzes word frequency and appearance patterns to extract specific keywords. It also analyzes rating scores and posting frequency as time-series data. The server extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and uses the pandas library to analyze time-series data based on the time axis.

[1015] 4. Sentiment analysis tool

[1016] The server performs sentiment analysis of the review information using natural language processing technology. Specifically, it uses machine learning models (e.g., BERT, LSTM) to determine whether the review content is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive.

[1017] 5. How to detect fraudulent reviews

[1018] The server uses machine learning algorithms to detect fraudulent reviews. It analyzes review patterns and reviewer behavior to identify reviews with fraudulent patterns. Specifically, it uses algorithms such as random forests and support vector machines (SVMs) and references database logs to detect suspicious activity.

[1019] 6. Filtering and Display Methods

[1020] The server filters out fraudulent reviews and only displays rating information based on genuine user experiences. To provide the filtered reviews to the user, the server sends the information to the user's device in JSON or HTML format. The device receives this information and displays accurate rating information to the user.

[1021] This system allows users to make accurate decisions based on reliable information, and also contributes to improving the reliability of store owners by providing fair evaluations.

[1022] Examples of prompt statements

[1023] "Design a system that collects user reviews from online rating platforms and detects and filters out fraudulent reviews. Include the following processing steps: data collection, data preprocessing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display. Please also provide examples of specific APIs or libraries."

[1024] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1025] Step 1: Collect data

[1026] The server collects rating information using the API of the online rating platform. As input, it receives the JSON data returned as a response from the API endpoint. This includes the review text, rating score, reviewer information, and posting date and time. The server stores this data in a database. Specifically, it sends an API request, parses the response, extracts the necessary information, and stores it in the database.

[1027] Step 2: Preprocessing the data

[1028] The server preprocesses the collected text data of rating information. As input, it receives the collected text data of rating information. The server removes HTML tags and special symbols, tokenizes the text (divides it into words and phrases), and removes stop words. Specifically, it uses regular expressions to remove HTML tags, and a natural language processing library (e.g., NLTK) to perform tokenization and remove stop words. As output, it obtains clear preprocessed text data.

[1029] Step 3: Feature extraction

[1030] The server extracts specific features from preprocessed text data. The input is preprocessed text data. The server analyzes word frequency and appearance patterns to identify prominent keywords and phrases. It also analyzes rating scores and posting frequency as time-series data. Specifically, it extracts important keywords using TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec models, and performs data analysis based on the time axis using the pandas library. The output is the extracted feature data.

[1031] Step 4: Sentiment analysis

[1032] The server performs sentiment analysis on preprocessed text data using natural language processing techniques. The input is the preprocessed text data. The server uses a machine learning model (e.g., BERT, LSTM) to determine whether the text is positive, negative, or neutral. Specifically, the server applies the sentiment analysis model and assigns a sentiment label to each review. The output is text data with the sentiment labels.

[1033] Step 5: Detect fraudulent reviews

[1034] The server uses machine learning algorithms to detect fraudulent reviews. The inputs are text data with sentiment labels and reviewer behavioral history. The server detects mass posting from the same IP address, copy-pasted text, and mass posting from new accounts in a short period of time. Specifically, it uses random forests and support vector machines (SVMs) to identify reviews with fraudulent patterns. The output is data determined to be fraudulent reviews.

[1035] Step 6: Filter and view reviews

[1036] The server filters out fraudulent reviews and only reviews based on genuine user experiences. The inputs are data determined to be fraudulent reviews and accurate rating information. The server runs the filtering algorithm and sends the filtered reviews to the user's device in JSON or HTML format. Specifically, the server filters out fraudulent reviews, generates filtered reviews, and sends them to the user's device. The output is accurate rating information that is displayed on the user's device.

[1037] (Application example 1)

[1038] 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."

[1039] On online rating platforms, user reviews are an important source of information for users to evaluate products and services. However, the presence of fraudulent reviews can make accurate evaluations difficult, potentially leading to incorrect decisions. Furthermore, the lack of review statistics based on sentiment analysis makes it difficult to understand the content of reviews. Therefore, it is necessary to obtain reliable reviews, eliminate fraudulent reviews, and analyze and display the sentiment of reviews, thereby enabling users to make appropriate decisions.

[1040] 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.

[1041] In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for excluding the detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for displaying the reviews after excluding the fraudulent reviews and their sentiment analysis on the user's terminal, thereby enabling users to make appropriate decisions based on reliable reviews and the results of the sentiment analysis.

[1042] An "online review platform" is a system that allows users to post reviews of products and services via the Internet and for other users to view those reviews.

[1043] "User reviews" are written opinions and evaluations of products and services posted by users.

[1044] The "means of collection" refers to the method by which the server obtains user reviews from online rating platforms, such as by using APIs or web scraping.

[1045] "Preprocessing" refers to methods for removing unnecessary information from the collected user review text and converting it into a format that is easy to analyze. Specifically, this includes removing HTML tags and special characters, tokenizing, and removing stop words.

[1046] "Means for extracting features" refers to methods for deriving features from preprocessed user reviews. Specifically, methods include sentiment analysis, keyword extraction, and time series analysis.

[1047] A "machine learning algorithm" is a model in which a computer learns patterns from data and makes predictions or classifications based on those patterns.

[1048] "Fraudulent reviews" are false reviews posted with the intent to manipulate ratings.

[1049] The "detection method" is a method of identifying fraudulent reviews using a machine learning algorithm.

[1050] The "display method" refers to a method for displaying only reviews that meet certain conditions on the user's device. This method targets highly reliable reviews that exclude fraudulent reviews.

[1051] "Sentiment analysis" is a technology that extracts emotions and opinions from text data and determines their nature (positive, negative, neutral).

[1052] "Keyword extraction" is a technique for identifying important words and phrases from user reviews.

[1053] "Time series analysis" is a technique for analyzing how data changes over time.

[1054] A "terminal" is an electronic device that a user uses to view and operate information. Examples include smartphones, tablets, and personal computers.

[1055] The present invention relates to a system that analyzes user reviews collected from an online rating platform, eliminates fraudulent reviews, and presents more reliable reviews, particularly for performing sentiment analysis to assist users in making decisions. An embodiment of this system will now be described in detail.

[1056] First, the server collects user reviews from online rating platforms (e.g., shopping sites or review sites) using APIs or web scraping technology to obtain data such as review text, rating scores, reviewer information, and posting date and time, and stores them in a database. Next, the server preprocesses the collected user review text data, which includes steps such as removing HTML tags and special characters, tokenizing, and removing stop words.

[1057] The server extracts specific features from the preprocessed reviews. This involves sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis determines whether a review is positive, negative, or neutral, while keyword extraction identifies important words and phrases that frequently appear. Time series analysis analyzes rating scores and posting frequency over time. After feature extraction, the server uses machine learning algorithms to detect fraudulent reviews. This detection takes into account large numbers of posts from the same IP address, concentrations in time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.).

[1058] Detected fraudulent reviews are filtered out, and the server filters out only reviews based on genuine user experiences. The filtered reviews and their sentiment analysis results are then displayed on the user's device. Users can access this reliable review information using devices such as smartphones and tablets, allowing them to make appropriate decisions based on accurate information.

[1059] The hardware and software used include a server infrastructure capable of high-performance data collection and processing, specifically software libraries such as Python, scikit-learn, BeautifulSoup, requests, and nltk. Reviews are displayed on the user's device using a browser or a dedicated app.

[1060] As a concrete example, consider the case where a user wants to check reviews for "product ID 12345" on "mail order site A." The user launches the smartphone app and enters "product ID 12345." An example of a prompt sentence in this case is as follows:

[1061] "Collect reviews for product ID 12345, preprocess them to a clean form, perform sentiment analysis, and finally detect and filter out fraudulent reviews."

[1062] Based on these prompts, the server can perform the necessary processing and provide users with reliable reviews and their sentiment analysis results, allowing them to select products with confidence.

[1063] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1064] Step 1:

[1065] The server collects user reviews from an online rating platform using an API. As input, it receives data including reviews related to a specified product or service (text, rating score, reviewer information, and posting date and time), and stores the data in a database as output.

[1066] Step 2:

[1067] The server preprocesses the collected user reviews. As input, it uses the text data of reviews stored in the database, removes HTML tags and special characters, tokenizes, and removes stop words. As output, it generates the preprocessed, clean data.

[1068] Step 3:

[1069] The server extracts specific features from the preprocessed reviews. Using the preprocessed text data as input, it performs sentiment analysis, keyword extraction, and time series analysis. As output, it generates the analysis results as a feature vector, which includes the review's emotional state (positive, negative, neutral), key keywords, and the review's posting timestamp.

[1070] Step 4:

[1071] The server uses a machine learning algorithm to detect fraudulent reviews. It uses the feature vectors obtained through feature extraction as input and analyzes patterns such as large numbers of posts from the same IP address, concentration of posts in specific time periods, text similarity, and reviewer activity history. As an output, it lists reviews that are suspected of being fraudulent and generates a detection result.

[1072] Step 5:

[1073] The server filters out the detected fraudulent reviews and filters out only reviews based on genuine user experiences. As input, it uses the fraudulent review detection results and the original review list to filter out fraudulent reviews. As output, it generates a filtered review list that contains only reliable reviews.

[1074] Step 6:

[1075] The server displays the filtered reviews and their sentiment analysis results on the user's device. Using the filtered review list and sentiment analysis results as input, the server displays the reviews on the screen or application of the user's device, allowing the user to check reliable review information and make appropriate decisions.

[1076] Specifically, when a user launches the app on their smartphone and enters "product ID 12345," the server executes each step in sequence based on the prompt, ultimately displaying reliable review information and the results of sentiment analysis on the user's device. When the prompt is entered, "Collect reviews for product ID 12345, preprocess them into a clean format, perform sentiment analysis, and finally detect and filter out fraudulent reviews," the process proceeds according to the instructions.

[1077] 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.

[1078] The present invention is a system that detects and excludes fraudulent reviews based on user reviews collected from an online rating platform, provides a genuine user experience, and combines this with an emotion engine that recognizes user emotions. Specific embodiments of this system are described below.

[1079] System configuration

[1080] 1. Data Collection Methods

[1081] The server automatically collects user reviews from online rating platforms (e.g., review sites, map services, etc.). Specifically, it uses an API to obtain the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database.

[1082] 2. Pretreatment Methods

[1083] The server pre-processes the collected user review text data. This pre-processing includes the following steps:

[1084] Remove HTML tags and special characters

[1085] Tokenizing text (splitting it into words and phrases)

[1086] Removal of stop words (e.g. "no", "wa", "desu", etc.)

[1087] 3. Feature Extraction Method

[1088] The server extracts specific features from the pre-processed reviews. Feature extraction includes:

[1089] Sentiment analysis: Determine whether a review is positive, negative, or neutral.

[1090] Keyword extraction: Identifying important, frequently occurring words and phrases.

[1091] Time series analysis: Analyze evaluation scores and posting frequency over time.

[1092] 4. Emotion Engine

[1093] The server recognizes the user's emotions through the emotion engine and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine can grasp the user's emotional state in real time and adjust the way reviews are displayed accordingly.

[1094] 5. How to detect fraudulent reviews

[1095] The server uses machine learning algorithms to detect fraudulent reviews, taking into account the following factors:

[1096] Mass posting from the same IP address or concentrated posting at the same time

[1097] Text similarity (check for copy-paste)

[1098] Reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.)

[1099] 6. Review Filtering and Display Methods

[1100] The server filters out reviews that are determined to be fraudulent and only reviews based on genuine user experiences are displayed.The server then adjusts the filtered reviews based on the user's emotional state using an emotion engine, and finally displays them on the user's device.

[1101] Program implementation example

[1102] As a concrete example, consider a system that collects restaurant reviews, filters out fraudulent reviews, and adjusts the display depending on the user's emotional state.

[1103] 1. The server collects reviews for "Restaurant A" through the API. For example, a review such as "Great meal! 5 stars!" is obtained.

[1104] 2. The server strips HTML tags from the collected reviews and tokenizes them, e.g., "Great meal, 5 stars."

[1105] 3. The server performs sentiment analysis and determines the review as positive. It also extracts the frequently occurring keywords "great," "food," and "five stars."

[1106] 4. The server uses a fraudulent review detection algorithm to check if the review is similar to other reviews or if it was posted by a new account. If there are no problems, it is marked as a genuine review.

[1107] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust the way reviews are displayed accordingly. For example, if the user is in a positive emotional state, positive reviews will be displayed first.

[1108] 6. The server displays the authentic review on the user's device. For example, a review such as "It was a great meal. Five stars!" is provided to the user.

[1109] The system aims to help users make accurate purchasing decisions based on reliable reviews, helping consumers make accurate decisions. Recognizing users' emotions in real time also enables more personalized information to be provided, improving the browsing experience. Store owners can also benefit from fair reviews, which contributes to increased trust.

[1110] The processing flow will be explained below.

[1111] Step 1:

[1112] The server uses APIs to collect user reviews from online rating platforms. For example, it automatically retrieves review texts, rating scores, reviewer information, and posting dates / times related to specific periods or specific subjects (restaurants, products, etc.), and stores them in an internal database.

[1113] Step 2:

[1114] The server preprocesses the collected review text data. First, it removes HTML tags and special characters from the text. Next, it tokenizes the review text and splits it into words and phrases. Finally, it removes stop words (such as "no," "wa," and "desu") and converts it into a format that is easy to analyze.

[1115] Step 3:

[1116] The server extracts specific features from the preprocessed reviews. First, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. Next, it extracts frequently occurring keywords and important phrases from the review text. Finally, it performs time-series analysis to analyze the rating score and posting frequency over time.

[1117] Step 4:

[1118] The server uses an emotion engine to recognize the user's emotions and analyzes the user's overall emotional trends based on the emotion analysis results. The emotion engine also detects the user's emotional state from facial expressions and voice while viewing reviews, and grasps changes in emotions in real time.

[1119] Step 5:

[1120] The server applies machine learning algorithms to detect fraudulent reviews. It analyzes large numbers of posts from the same IP address, text similarity, and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.) to identify fraudulent reviews. Reviews that are deemed to be non-issues are flagged as genuine reviews.

[1121] Step 6:

[1122] The server adjusts the display method of the filtered true reviews according to the user's emotional state. Depending on the user's emotional state (positive, negative, etc.) recognized by the emotion engine, it decides whether to prioritize positive reviews or balanced reviews.

[1123] Step 7:

[1124] The server then sends the final selected reviews to the user's device, which then displays the received reviews in a format appropriate to the user's emotional state. For example, if the user is in a positive emotional state, more positive reviews will be displayed.

[1125] Step 8:

[1126] Users can view the displayed reviews and make purchasing decisions based on reliable information. When a user posts a new review, that review is also analyzed through the same process.

[1127] Example 2

[1128] 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."

[1129] Online rating platforms are prone to a flood of fraudulent reviews, leading to users making erroneous decisions based on unreliable information. Furthermore, reviews that do not reflect the user's emotional state are not provided, making it difficult to help users make appropriate purchasing decisions.

[1130] 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 user reviews from an online rating platform, means for preprocessing the text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for analyzing users' emotional tendencies using an emotion engine, means for detecting fraudulent reviews using a machine learning algorithm, and means for excluding detected fraudulent reviews, filtering reviews based on true user experiences, and displaying them according to the user's emotional state. This allows users to make accurate decisions based on reliable reviews and provides information adapted to the user's emotions.

[1131] An "online rating platform" refers to a system on the Internet where users can post ratings and reviews of products and services and other users can view them.

[1132] "User reviews" refers to ratings and comments about products and services posted by users on online rating platforms.

[1133] "Means of collection" refers to the way in which the server automatically obtains user reviews from online rating platforms using technologies such as APIs or web scraping.

[1134] The "means of text preprocessing" refers to a method of removing HTML tags and special characters from the text data of user reviews collected by the server, and performing tokenization and removal of stop words.

[1135] "Means for extracting features" refers to the methods by which the server extracts important information and patterns from pre-processed user reviews, including sentiment analysis, keyword extraction, and time series analysis.

[1136] An "emotion engine" refers to the technology and algorithms used to analyze a user's emotional state from text data, which then classifies emotions as positive, negative, or neutral.

[1137] "Machine learning algorithms" refer to technology that trains models based on past data to classify and predict unknown data. In this case, they are used to detect fraudulent reviews.

[1138] "Fraudulent reviews" are manipulated ratings and false comments that provide biased information that differs from the normal user experience.

[1139] "Filtering means" refers to the method by which the server filters out fraudulent reviews and selects only genuine reviews.

[1140] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[1141] "Means for displaying" refers to the method by which the server selects the filtered reviews and ultimately displays them on the user's device.

[1142] This invention is a system that provides reliable reviews by detecting and filtering out fraudulent reviews based on user reviews collected from online rating platforms. Furthermore, it aims to improve the user experience by combining it with an emotion engine that recognizes the user's emotional state.

[1143] Hardware and software configuration

[1144] The following hardware and software are used to implement this system.

[1145] Server: This is the main component used for data collection, pre-processing, feature extraction, sentiment analysis, fraudulent review detection, and review filtering and display.

[1146] Database: This is an internal database that stores collected user reviews. It uses a relational database such as MySQL or PostgreSQL.

[1147] API: A means of collecting reviews from online rating platforms, such as a RESTful API.

[1148] Machine learning algorithms are used to detect fraudulent reviews, specifically using machine learning libraries such as Scikit-learn and TensorFlow.

[1149] Sentiment engine: A means of analyzing user sentiment, analyzing review text data using natural language processing technology. Libraries such as NLTK and TextBlob, or APIs dedicated to sentiment analysis (e.g., IBM Watson Natural Language Understanding) are used.

[1150] Data collection

[1151] The server automatically collects user reviews from online rating platforms using APIs. For example, to collect reviews for a specific restaurant, the server uses the endpoint "https: / / api.example.com / reviews?restaurant_id=A".

[1152] Data Preprocessing

[1153] The server preprocesses the collected user review text data, specifically removing HTML tags and special characters, tokenizing, and removing stop words, so that the text data is in a format that is easy to analyze.

[1154] Feature Extraction

[1155] The server extracts specific features from the pre-processed reviews. This step uses a variety of analytical techniques, including sentiment analysis, keyword extraction, and time series analysis. Sentiment analysis involves classifying emotions into "positive," "negative," and "neutral," extracting important keywords, and analyzing review trends over time.

[1156] Emotion Engine

[1157] The server uses an emotion engine to analyze the user's emotional state in real time. The emotion engine determines the user's psychological state based on the content of the reviews and grasps the overall emotional trend. For example, if the user is in a positive emotional state, positive reviews will be displayed preferentially.

[1158] Fraudulent review detection

[1159] The server uses machine learning algorithms to detect fraudulent reviews, which analyze factors such as large numbers of posts from the same IP address, similar text, and reviewer activity history to identify reviews that are likely to be fraudulent.

[1160] Filter and view reviews

[1161] The server filters out fraudulent reviews and only reviews based on genuine user experiences, then adjusts the way reviews are displayed according to the user's emotional state, and finally displays the filtered reviews on the user's device.

[1162] Specific examples

[1163] As a concrete example, the process of collecting, preprocessing, analyzing, detecting fraudulent reviews, and displaying restaurant reviews is shown below.

[1164] 1. The server collects reviews for "Restaurant A" from the API.

[1165] 2. The server removes HTML tags from the collected data and performs tokenization and stop word removal.

[1166] 3. The server performs sentiment analysis, determines the review as positive, and extracts important keywords.

[1167] 4. The server uses machine learning algorithms to detect fraudulent reviews and certify legitimate reviews.

[1168] 5. The server uses an emotion engine to recognize the user's emotional state in real time and adjust how the review is displayed.

[1169] 6. The server filters the true reviews and displays them on the user's device.

[1170] Prompt Sentence Examples

[1171] By inputting the following prompts into the generative AI model, an explanation of the system and specific examples can be generated.

[1172] Example prompt:

[1173] "Describe a system that detects and filters out fraudulent reviews based on user reviews collected from an online rating platform, and provides a genuine user experience. Also, provide a concrete example of how to utilize an emotion engine that recognizes user emotions."

[1174] In this way, by detailing the specific steps and techniques required to implement the invention, it is possible to accurately understand the invention and provide guidance for others to practice it.

[1175] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1176] Step 1: Data collection methods

[1177] A server automatically collects user reviews from an online rating platform using an API. As input, it uses an API endpoint (e.g., "https: / / api.example.com / reviews?restaurant_id=A"). The server sends the request and stores the response data, such as review text, rating score, reviewer information, and posting date and time, in an internal database. The output is a comprehensive dataset of the collected reviews.

[1178] Step 2: Data preprocessing measures

[1179] The server preprocesses the collected user review text data. Specifically, it removes HTML tags and special characters. For example, Great food " is converted to "Great meal". Next, the text is tokenized and a review such as "Great meal 5 stars" is split into words such as "Great", "Meal", and "5 stars". Furthermore, stop words are removed and "This is a great meal" is converted to "Great meal". The input is the collected text data of reviews, and the output is the preprocessed and clean text data.

[1180] Step 3: Feature extraction methods

[1181] The server extracts specific features from the preprocessed reviews. Specifically, it performs sentiment analysis to determine whether the review is positive, negative, or neutral. For example, a review that says "The meal was amazing! Five stars!" is classified as positive. It also extracts important keywords, identifying words such as "amazing," "meal," and "five stars." It then performs time-series analysis to analyze the rating scores and posting frequency over time. The input is the text data of the preprocessed reviews, and the output is the sentiment classification results, extracted keywords, and time-series data.

[1182] Step 4: Emotion Engine

[1183] The server uses an emotion engine to analyze the user's emotional state in real time. Specifically, it understands the user's overall emotional tendency based on the emotion analysis results from the previous step. For example, the text data of a review is input into a natural language processing algorithm to determine the emotional state, such as positive, negative, or neutral. The input is the feature-extracted emotion analysis results, and the output is the user's overall emotional tendency data.

[1184] Step 5: Detecting fraudulent reviews

[1185] The server uses a machine learning algorithm to detect fraudulent reviews. Specific operations include checking for large numbers of posts from the same IP address, concentrated time periods, text similarity (checking for copy-paste), and reviewer activity history (new accounts, large numbers of posts in a short period of time, etc.). For example, the machine learning model inputs review metadata to determine whether there is suspicion of fraud. The input is collected and preprocessed review metadata and text data, and the output is the result of the fraudulent review determination.

[1186] Step 6: Filtering and displaying reviews

[1187] The server filters out reviews determined to be fraudulent and filters out only reviews based on genuine user experiences. It also adjusts the way reviews are displayed based on the results of the emotion engine. Specifically, it prioritizes positive or negative reviews depending on the user's emotional state. For example, if a user is in a positive emotional state, positive reviews will be displayed at the top. The input is data on reviews determined to be non-fraudulent and data on the user's emotional state, and the output is the final filtered reviews.

[1188] Through these steps, the system can collect reliable reviews and provide personalized information based on user sentiment.

[1189] (Application example 2)

[1190] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1191] Online rating platforms face the problem of fraudulent reviews, which can lead to users making purchasing decisions based on unreliable information. Furthermore, they are unable to tailor the display of reviews to reflect the user's emotional state, making it difficult to provide users with information that is appropriate for them.

[1192] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting user reviews from an online rating platform, means for preprocessing text of the collected user reviews, means for extracting features of the preprocessed user reviews, means for detecting fraudulent reviews using a machine learning algorithm, means for filtering out detected fraudulent reviews and displaying only reviews based on genuine user experiences, and means for recognizing the emotional state of the user and adjusting the display of reviews according to the emotional state. This enables users to make accurate purchasing decisions based on reliable reviews and also enables the provision of personalized information according to the user's emotional state.

[1193] An "online review platform" is a service provided over the Internet that allows users to post and view reviews and ratings of products and services.

[1194] "User reviews" are ratings and comments about products and services posted by users on online rating platforms.

[1195] "Preprocessing" refers to a series of steps performed to convert the collected text data of user reviews into an analyzable format, including removing HTML tags and special characters, tokenizing, and removing stop words.

[1196] "Feature extraction" is a method for extracting meaningful information and patterns from preprocessed user reviews, and includes sentiment analysis, keyword extraction, and time series analysis.

[1197] A "machine learning algorithm" is a method for learning patterns and rules from large amounts of data and performing analysis and classification.

[1198] "Fraudulent reviews" are false reviews that are not based on facts or manipulative reviews posted with a specific intention.

[1199] "User's emotional state" refers to the psychological state or mood of the user when viewing a review.

[1200] "Adjusting the display of reviews according to emotional state" means optimizing the content and order of reviews to be displayed based on the user's current emotions and mood.

[1201] This invention is a system that detects and excludes fraudulent reviews from users on online rating platforms and displays appropriate reviews according to the user's emotional state. This system mainly consists of a server, a smartphone application, a machine learning algorithm, and an emotion recognition engine.

[1202] System Overview

[1203] The server collects user reviews from online rating platforms, preprocesses the collected reviews, and performs feature extraction. To detect fraudulent reviews, a machine learning algorithm is used. An emotion engine recognizes users' emotional states in real time. Users then receive tailored reviews using a smartphone application.

[1204] Hardware and software used

[1205] The main hardware and software used are servers, smartphones, Python, TextBlob, and Sklearn, which are used for data processing and calculation.

[1206] Data processing and calculation

[1207] 1. Data Collection:

[1208] The server uses APIs to collect user reviews from online rating platforms. The data collected includes review text, rating score, reviewer information, and posting date and time, and is stored in an internal database.

[1209] 2. Pretreatment:

[1210] HTML tags and special characters are removed from the collected review text data, the text is tokenized, and stop words (e.g., "no" and "wa") are removed.

[1211] 3. Feature Extraction:

[1212] After preprocessing, sentiment analysis (determining positive, negative, or neutral), keyword extraction, and time series analysis are performed on the data.

[1213] 4. Fraudulent Review Detection:

[1214] The server uses machine learning algorithms (e.g., Naive Bayes) to detect fraudulent reviews, including criteria such as large numbers of reviews from the same IP address, similar text, and reviews posted by new accounts.

[1215] 5. User emotion recognition and display adjustment:

[1216] The emotion engine detects the user's current emotional state in real time and adjusts how reviews are displayed accordingly: for example, if the user is in a positive emotional state, positive reviews will be displayed first.

[1217] Specific examples

[1218] "As a concrete example of an application to increase the reliability of user reviews on an online shopping site, the server collects product reviews via an API, tokenizes the text data through preprocessing, performs sentiment analysis, and detects and excludes fraudulent reviews using the Naive Bayes algorithm. The system then prioritizes displaying positive or negative reviews depending on the user's emotional state."

[1219] Generative AI model prompt example

[1220] "Write a program in Python to recognize emotions from user-generated reviews on an online shopping site and filter out fraudulent reviews."

[1221] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1222] Step 1:

[1223] The server collects user reviews through the API of the online rating platform. As input, it receives data such as the review text, rating score, reviewer information, and posting date and time, and stores them in an internal database. Specifically, it sends an HTTP request, parses the response data, and stores it.

[1224] Step 2:

[1225] The server preprocesses the collected user review text data. It receives the review text data as input, removes HTML tags and special characters, tokenizes it (divides it into words and phrases), and removes stop words. Specifically, it cleans up the text using regular expressions and natural language processing libraries.

[1226] Step 3:

[1227] The server extracts features from preprocessed user reviews. It receives the preprocessed text data as input and performs sentiment analysis (positive, negative, neutral), keyword extraction, and time series analysis. Specifically, it calculates sentiment scores using the TextBlob library and extracts frequently occurring keywords using Sklearn.

[1228] Step 4:

[1229] The server uses machine learning algorithms to detect fraudulent reviews. It receives collected, preprocessed, and feature-extracted reviews as input and detects fraudulent patterns such as mass posting from the same IP address, text similarity, and posting from new accounts. Specifically, it calculates the probability of fraud using algorithms such as the Naive Bayes algorithm.

[1230] Step 5:

[1231] The server filters out the detected fraudulent reviews and displays only reviews based on genuine user experiences. As input, it receives a list of reviews that have been determined to be fraudulent and generates a list of reliable reviews that excludes them. Specifically, it removes the fraudulent review list from the database and retains the remaining reviews.

[1232] Step 6:

[1233] The server uses an emotion engine to recognize the user's emotional state in real time and adjust the display of reviews according to the emotional state. It receives the user's emotional data and a list of reviews as input, and optimizes the display order and content of the reviews according to the user's emotional state (e.g., positive or negative). Specifically, it analyzes the user's emotional data and runs an algorithm to prioritize reviews that match the user's emotional state.

[1234] Step 7:

[1235] The terminal (smartphone) displays the reliable reviews received from the server to the user. As input, it receives the adjusted review list sent from the server and displays it on the screen. Specifically, it visually presents the reviews through a user interface, allowing the user to easily browse the reviews.

[1236] 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.

[1237] 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.

[1238] 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.

[1239] 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.

[1240] 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.

[1241] 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.

[1242] 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).

[1243] 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.

[1244] 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."

[1245] 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.

[1246] 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).

[1247] 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.

[1248] 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.

[1249] 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.

[1250] 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.

[1251] 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.

[1252] 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.

[1253] 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.

[1254] 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.

[1255] 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.

[1256] 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.

[1257] The following is further disclosed regarding the above embodiment.

[1258] (Claim 1)

[1259] a means for collecting user reviews from an online rating platform;

[1260] A means for preprocessing the collected user review text;

[1261] A means for extracting features of the preprocessed user reviews;

[1262] a means of detecting fraudulent reviews using machine learning algorithms;

[1263] A means to filter out detected fraudulent reviews and display only reviews based on genuine user experiences;

[1264] A system including:

[1265] (Claim 2)

[1266] 2. The system of claim 1, wherein the preprocessing means includes means for removing HTML tags and special characters from the text of the user reviews, tokenizing the text, and removing stop words.

[1267] (Claim 3)

[1268] 2. The system according to claim 1, wherein the feature extraction means includes means for performing sentiment analysis, keyword extraction, and time series analysis of user reviews.

[1269] (Claim 4)

[1270] 2. The system of claim 1, wherein the fraudulent review detection means includes means for detecting fraudulent reviews based on a large number of posts from the same IP address, text similarity, and reviewer activity history.

[1271] (Claim 5)

[1272] 10. The system of claim 1, further comprising means for executing the preprocessing means, the feature extraction means, and the fraudulent review detection means in real time.

[1273] "Example 1"

[1274] (Claim 1)

[1275] a means for collecting evaluation information from an online evaluation platform;

[1276] a means for preprocessing the collected evaluation information text;

[1277] means for extracting features from the preprocessed evaluation information;

[1278] means for analyzing sentiment using natural language processing techniques;

[1279] a means of detecting fraudulent information using machine learning algorithms;

[1280] A means for filtering out detected fraudulent information and displaying only evaluation information based on true user experiences;

[1281] A system including:

[1282] (Claim 2)

[1283] 2. The system of claim 1, wherein the preprocessing means includes means for removing markup tags and special symbols from the text of the rating information, segmenting the text, and removing stop words.

[1284] (Claim 3)

[1285] 2. The system according to claim 1, wherein the feature extraction means includes means for performing sentiment analysis, key word extraction, and time series analysis of the evaluation information.

[1286] "Application Example 1"

[1287] (Claim 1)

[1288] a means for collecting user reviews from an online rating platform;

[1289] A means for preprocessing the collected user review text;

[1290] A means for extracting features of the preprocessed user reviews;

[1291] a means of detecting fraudulent reviews using machine learning algorithms;

[1292] A means to filter out detected fraudulent reviews and display only reviews based on genuine user experiences;

[1293] a means for displaying the reviews and their sentiment analysis after filtering out fraudulent reviews on a user's device;

[1294] A system including:

[1295] (Claim 2)

[1296] 2. The system of claim 1, wherein the preprocessing means includes means for removing HTML tags and special characters from the text of the user reviews, tokenizing the text, and removing stop words.

[1297] (Claim 3)

[1298] 2. The system according to claim 1, wherein the feature extraction means includes means for performing sentiment analysis, keyword extraction, and time series analysis of user reviews.

[1299] "Example 2: Combining Emotion Engines"

[1300] (Claim 1)

[1301] a means for collecting user reviews from an online rating platform;

[1302] A means for preprocessing the collected user review text;

[1303] A means for extracting features of the preprocessed user reviews;

[1304] means for analyzing the emotional tendencies of a user using an emotion engine;

[1305] a means of detecting fraudulent reviews using machine learning algorithms;

[1306] a means for filtering out the detected fraudulent reviews and filtering reviews based on genuine user experiences, and displaying them according to the emotional state of the user;

[1307] A system including:

[1308] (Claim 2)

[1309] 2. The system of claim 1, wherein the preprocessing means includes means for removing HTML tags and special characters from the text of the user reviews, tokenizing the text, and removing stop words.

[1310] (Claim 3)

[1311] 2. The system according to claim 1, wherein the feature extraction means includes means for performing sentiment analysis, keyword extraction, and time series analysis of user reviews.

[1312] "Application example 2 when combining emotion engines"

[1313] (Claim 1)

[1314] a means for collecting user reviews from an online rating platform;

[1315] A means for preprocessing the collected user review text;

[1316] A means for extracting features of the preprocessed user reviews;

[1317] a means of detecting fraudulent reviews using machine learning algorithms;

[1318] A means to filter out detected fraudulent reviews and display only reviews based on genuine user experiences;

[1319] means for recognizing a user's emotional state and adjusting the display of reviews according to the emotional state;

[1320] A system including:

[1321] (Claim 2)

[1322] 2. The system of claim 1, wherein the preprocessing means includes means for removing HTML tags and special characters from the text of the user reviews, tokenizing the text, and removing stop words.

[1323] (Claim 3)

[1324] 2. The system according to claim 1, wherein the feature extraction means includes means for performing sentiment analysis, keyword extraction, and time series analysis of user reviews. [Explanation of symbols]

[1325] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for collecting user reviews from an online rating platform; A means for preprocessing the collected user review text; A means for extracting features of the preprocessed user reviews; a means of detecting fraudulent reviews using machine learning algorithms; A means to filter out detected fraudulent reviews and display only reviews based on genuine user experiences; A system including:

2. 2. The system of claim 1, wherein the preprocessing means includes means for removing HTML tags and special characters from the text of the user reviews, tokenizing the text, and removing stop words.

3. 2. The system according to claim 1, wherein the feature extraction means includes means for performing sentiment analysis, keyword extraction, and time series analysis of user reviews.

4. 2. The system according to claim 1, wherein the fraudulent review detection means includes means for detecting fraudulent reviews based on a large number of posts from the same IP address, text similarity, and reviewer activity history.

5. The system of claim 1 further comprising means for executing the pre-processing means, the feature extraction means, and the fraudulent review detection means in real time.

Images