System

Abstract

A system is provided.SOLUTION: A system comprising: means for receiving a long sentence content input by a user; means for analyzing the received long sentence content to obtain a generation probability according to a generative AI model; means for calculating a believability score based on the obtained generation probability; and means for outputting the calculated believability score.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] As generative AI technology becomes more widespread, it is becoming increasingly difficult to distinguish between content that was created by users with great effort and that was automatically generated. In particular, credibility is highly required for important documents such as motivation letters and application documents, and reviewers must expend a great deal of effort to properly assess this. Given this background, there is a need for a tool that can efficiently evaluate the credibility of content that was truly created by users and distinguish it from content automatically generated by generative AI. [Means for solving the problem]

[0005] To solve the above problems, the present invention provides the following means. The present invention provides a system including: means for receiving long-form content entered by a user; means for analyzing the received long-form content and obtaining a generation probability by a generative AI model; means for calculating a credibility score based on the obtained generation probability; and means for outputting the calculated credibility score. This makes it possible to efficiently distinguish between content truly created by a user and content automatically generated by a generative AI, thereby reducing the burden on reviewers. Furthermore, by using an analytical algorithm for evaluating the generation probability, a highly reliable evaluation can be achieved. Furthermore, the calculation of the credibility score is designed so that a higher score is calculated as the generation probability decreases, allowing users to intuitively understand the credibility of posted content.

[0006] A "user" is a person or operator who performs operations or inputs data into a system.

[0007] "Long-form content" refers to text data of a certain length, such as motivation statements or application documents.

[0008] "Means for receiving" refers to a function or module for receiving long-form content sent by a user on a server or other device.

[0009] "Means for analysis" refers to functions or modules that analyze received long-form content using algorithms or programs and extract specific information (in this case, the probability of generation by the generation AI).

[0010] A "generative AI model" is an artificial intelligence model or algorithm used to automatically generate text data.

[0011] "Generation probability" is a probability value that indicates the likelihood that specific text data was generated by a generative AI model.

[0012] The "authenticity score" is a numerical score that indicates whether the received long-form content was truly created by the user, and is calculated based on the probability of creation.

[0013] "Output means" refers to a function or module for displaying the calculated credibility score on the reviewer's or user's device.

[0014] An "analytic algorithm" is a set of computational steps or methods for analyzing data for a specific purpose. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram illustrating a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0016] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0017] First, the terms used in the following description will be explained.

[0018] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0019] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0020] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0021] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0022] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0023] [First embodiment]

[0024] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0025] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0026] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0028] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0031] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0032] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0033] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0034] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0035] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0036] This invention is a system that analyzes long-form content entered by a user, obtains the probability of generation by a generative AI model, and calculates and outputs a credibility score based on that probability. This system makes it possible to efficiently distinguish between content created by a user and content automatically generated by a generative AI.

[0037] Embodiments of the invention

[0038] System Overview

[0039] 1. User Input

[0040] The user inputs long content such as a reason for applying into an application form or submission form. The user's device sends this input to the server.

[0041] 2. Initial Server Setup

[0042] The server receives input from the user and initializes the OpenAI API, which is used to analyze the user's input and evaluate the likelihood that the text was generated by a generative AI.

[0043] 3. Parsing User Input

[0044] The server analyzes the text received from the user and obtains the generation probability from the generative AI model. For this process, it uses the OpenAI API and extracts the generation probability from the response.

[0045] 4. Calculating the credibility score

[0046] The server calculates a credibility score based on the obtained probability of creation, which is calculated using the formula (1 - probability of creation) 100.

[0047] 5. Outputting the results

[0048] The calculated credibility score is returned to the reviewer or user's device, and the reviewer uses this score to determine the credibility of the submitted content.

[0049] System processing flow

[0050] The user enters long-form content and submits it to the server

[0051] For example, a user enters their motivation for applying into an application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[0052] The server receives user input and initializes the OpenAI API.

[0053] The server stores the text data received from the user and sets the OpenAI API key to prepare the API.

[0054] The server parses the user input

[0055] The server calls the check_authenticity function and sends the user's input text to the OpenAI API to obtain the generation probability. This process uses the following parameters: engine (e.g., text-davinci-003), prompt (a sentence containing the text to be evaluated and the evaluation instructions), maximum number of tokens, and temperature parameter.

[0056] The server analyzes the generation probability

[0057] The server receives the response from the OpenAI API, parses it, and extracts the generated probability using regular expressions, converting the extracted probability value into a floating-point number between 0 and 1.

[0058] The server calculates the credibility score

[0059] The server calculates the credibility score using the calculate_score function. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[0060] The server returns the results

[0061] The server returns the calculated credibility score to the reviewer or user's device, and the reviewer uses this score to judge the credibility of the submitted sentence.

[0062] Specific example explanation

[0063] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server analyzes this probability and calculates a credibility score of 34.6. This score is returned to the reviewer or user's device to help determine whether the submitted statement is the user's own.

[0064] In this way, the present invention enables efficient distinction between content created by users and content generated by AI, reducing the burden on reviewers.

[0065] The processing flow will be explained below.

[0066] Step 1:

[0067] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[0068] Step 2:

[0069] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[0070] Step 3:

[0071] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[0072] Step 4:

[0073] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[0074] Step 5:

[0075] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[0076] Engine:text-davinci-003

[0077] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[0078] Maximum number of tokens: 60

[0079] Temperature parameter: 0.5

[0080] Step 6:

[0081] The server sends the created request to the OpenAI API and waits for a response.

[0082] Step 7:

[0083] The server receives the response returned from the OpenAI API, which includes the generation probability.

[0084] Step 8:

[0085] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[0086] Step 9:

[0087] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[0088] Step 10:

[0089] The server calls the calculate_score function to calculate the credibility score based on the obtained generation probability, using the formula (1 - generation probability) 100.

[0090] Step 11:

[0091] The server generates a result containing the calculated credibility score as an HTTP response and sends it back to the reviewer or device.

[0092] Step 12:

[0093] The terminal displays the credibility score received from the server, and the reviewer uses this to judge the credibility of the submitted statement.

[0094] Example 1

[0095] 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."

[0096] It is difficult to efficiently distinguish whether user-created long-form content was created manually or automatically generated by a generative AI model. This makes credibility assessment difficult and increases the burden on reviewers. The present invention aims to solve this problem by providing a system that accurately and efficiently evaluates the credibility of user-created content.

[0097] 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.

[0098] In this invention, the server includes a means for receiving long-form content input by a user, a means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, and a means for calculating a credibility score based on the obtained generation probability. This makes it possible to evaluate the credibility of content input by a user with high accuracy and efficiently determine whether a submitted document was written by the user himself / herself.

[0099] "User" refers to an individual or group of people who input long-form content into the system.

[0100] "Long content" refers to text data entered by a user that has a certain number of characters or more.

[0101] "Means for receiving" refers to the interface or process by which the server obtains the data entered by the user.

[0102] "Means for analyzing" refers to algorithms or programs for evaluating and analyzing received long-form content.

[0103] A "generative AI model" refers to a system or software that uses artificial intelligence technology to generate text.

[0104] "Generation probability" refers to a number that indicates the likelihood that a generative AI model generated a given piece of text.

[0105] The "credibility score" refers to an index that indicates whether the text was created by the user himself or herself, calculated based on the probability of creation.

[0106] "Means for outputting" refers to an interface or method for providing the calculated credibility score to a user or reviewer.

[0107] "Means of initializing the API" refers to the process of preparing the API for use with the API key and settings of the generated AI model.

[0108] "Means for setting prompt sentences" refers to a method for forming input sentences or instructions to make analysis requests to a generative AI model.

[0109] "Means of analyzing the response" refers to the process of interpreting the response data obtained from the generative AI model and extracting the necessary information.

[0110] "Means for extracting probability values" refers to a method or algorithm for extracting generation probabilities from response data.

[0111] "Terminal" refers to an electronic device used by a user, such as a computer, smartphone, or tablet.

[0112] "Means for evaluating" refers to the set of processes or calculations required to determine the authenticity of a received text.

[0113] The present invention is a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. The following describes in detail an embodiment of the present invention.

[0114] System configuration

[0115] User Input

[0116] A user uses their device (computer, smartphone, tablet, etc.) to enter long-form content into an application or submission form, such as a motivation letter or essay, and clicks the "Submit" button. This entered text data is sent by the device to the server via an HTTP POST request.

[0117] Initial Server Configuration

[0118] The server receives the text data sent from the user's device. Then, the server initializes the generative AI model (e.g., OpenAI API) by creating an API instance using the API key obtained from the configuration file.

[0119] Parsing User Input

[0120] The server calls the check_authenticity function to parse the text entered by the user. This function sets a prompt containing the text to be evaluated and evaluation instructions. Here is an example prompt:

[0121] _Example prompt sentence:_

[0122] "Please rate the likelihood that the following text was generated by a generative AI."

[0123] Obtaining generation probability

[0124] The server sends a request to the OpenAI API endpoint containing the prompt and settings (engine: e.g., text-davinci-003, maximum number of tokens, temperature parameters, etc.) to obtain a generation probability indicating the likelihood that the user's input was generated by the generative AI.

[0125] Analysis of generation probability

[0126] The server parses the response received from the API and extracts the generation probability. In this process, it extracts the probability value from the response JSON data and converts it to a floating point number between 0 and 1 using regular expressions if necessary.

[0127] Calculating the credibility score

[0128] The server calculates the credibility score based on the obtained generation probability. Specifically, it uses the following formula:

[0129] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[0130] For example, if the generation probability is 0.654, the credibility score is 34.6.

[0131] Output of results

[0132] The server returns the calculated credibility score to the reviewer or user's device, allowing the reviewer to evaluate the credibility of the submitted content.

[0133] Specific examples

[0134] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the API responds with "Generation probability 65.4%," the server analyzes this probability and calculates an authenticity score of 34.6. By returning this score to the reviewer or user's device, it can determine whether the submitted document was created by the user themselves.

[0135] In this way, the present invention provides a highly accurate and efficient evaluation system for effectively distinguishing between user-generated content and content generated by a generation AI.

[0136] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0137] Explanation of the processing flow

[0138] Step 1:

[0139] A user enters long-form content into an application or submission form on their own device and clicks the "Send" button. The entered text data is sent from the device to the server. The input data includes the long-form content (e.g., reason for applying) and the user ID. The server then receives an HTTP POST request and extracts the text data.

[0140] Step 2:

[0141] The server saves the received text data and initializes the API of the generative AI model. Specifically, the server obtains the API key from the configuration file and configures the API. Once the API key is correctly configured, the generative AI model is ready to use. The API key is included as input data. The output is an initialized API instance.

[0142] Step 3:

[0143] The server calls the check_authenticity function to analyze the user's input text. This function sets a prompt and generates an analysis request. The prompt includes instructions such as "Please rate the likelihood that the following text was generated by a generative AI." The input data includes the text data and the prompt. The output is an analysis request.

[0144] Step 4:

[0145] The server sends the generated analysis request to the generative AI model endpoint. The request includes settings such as the engine (e.g., text-davinci-003), maximum number of tokens, and temperature parameters. The analysis request and API settings are included as input data. This sends an API request to obtain the generation probability.

[0146] Step 5:

[0147] The server receives the response from the generative AI model and parses the response data. It extracts the generation probability from the JSON-formatted response data. In this process, it uses a regular expression to extract the generation probability and converts it into a floating-point number ranging from 0 to 1. The input data includes the response JSON data. The output is the extracted generation probability.

[0148] Step 6:

[0149] The server calculates the credibility score based on the obtained generation probability using the following formula:

[0150] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[0151] For example, if the generation probability is 0.654, the credibility score is 34.6. The input data includes the generation probability. The output is the calculated credibility score.

[0152] Step 7:

[0153] The server returns the calculated credibility score to the reviewer or user's device. It generates an HTTP response and sends information including the credibility score and related metadata to the user's device. The credibility score and user ID are included as input data, allowing the user or reviewer to confirm the credibility of the submitted document.

[0154] (Application example 1)

[0155] 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."

[0156] There is a challenge in determining with high accuracy whether long-form content sent by users was created by generative AI. There is also a need for a means to quickly and efficiently evaluate the authenticity of emails, documents, etc. In addition, users need help making more reliable judgments by specifically visualizing the authenticity evaluation based on generation probability.

[0157] 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.

[0158] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing the input text and using an analysis algorithm and the generation probability to calculate the credibility score, and means for displaying the credibility score on an electronic device. This makes it possible to quickly evaluate the credibility of long-form content received by a user with high accuracy and visualize the credibility score.

[0159] "Long content" refers to text data entered by a user, and mainly includes the contents of emails, documents, and the like.

[0160] A "generative AI model" is a model that can generate new text using an artificial intelligence algorithm trained on large amounts of text data.

[0161] "Generation probability" is a number that indicates the likelihood that a particular piece of text was generated by a generative AI model.

[0162] The "credibility score" is a numerical value for evaluating the reliability of the text entered by the user based on the generation probability.

[0163] An "analysis algorithm" is a calculation method for analyzing input text and calculating its generation probability.

[0164] "Electronic device" refers to a hardware device that can execute a program and display the results, and specifically includes smartphones, tablets, personal computers, etc.

[0165] "Server" means a network-enabled computer system that processes data received from users and calculates and outputs credibility scores.

[0166] The present invention provides a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. Specific embodiments for implementing the present invention are described below.

[0167] System configuration

[0168] This system mainly consists of the following hardware and software:

[0169] Server: A network-enabled computer system that processes data received from users and calculates and outputs credibility scores. The software used is the Python language and the OpenAI API.

[0170] User device: Any device on which a user enters long-form content, including smartphones, tablets, and PCs.

[0171] Electronic Device: A hardware device for displaying the credibility score. This can be a user terminal.

[0172] Processing flow

[0173] 1. User input:

[0174] The user inputs long-form content such as emails and documents through the device. For example, they input a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills."

[0175] 2. Receiving long-form content:

[0176] The long text content entered by the user is sent from the terminal to the server, which receives the text data and stores it in a database.

[0177] 3. Obtaining generation probabilities:

[0178] The server analyzes the stored text data and obtains the probability of generation using a generative AI model (e.g., OpenAI's text-davinci-003). The prompt used is, "Please indicate the probability that the following text was generated by the generative AI: I want to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[0179] 4. Calculating the credibility score:

[0180] The Python regular expression library "re" is used to analyze the generation probability, and a credibility score is calculated based on the obtained generation probability. For example, if the generation probability is 65.4%, the credibility score is "(1 - 0.654) 100 = 34.6". The calculated credibility score is stored in a database in a specific format.

[0181] 5. Output of credibility score:

[0182] Once the calculation is complete, the server sends the authenticity score to the user's device, where it can be displayed to visually confirm the authenticity of the email or document.

[0183] Specific example explanation

[0184] For example, if a user enters the following as their motivation for applying: "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," this text is sent to the server. The server uses the OpenAI API to obtain the probability of generating this text, and if the generation probability is 65.4%, the credibility score will be 34.6. This credibility score is sent to the user's device, and the user can judge the credibility of the document based on the displayed credibility score.

[0185] In this way, the system of the present invention helps users to quickly and accurately assess the credibility of long-form content.

[0186] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0187] Step 1:

[0188] Collect and send user-supplied long-form content

[0189] A user inputs long content such as an email or document through a terminal and clicks the send button. The input text is sent to the server as an HTTP request. The input may include, for example, a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills." The text data is sent to the server in JSON format.

[0190] Step 2:

[0191] The server receives and stores the long-form content.

[0192] The server receives an HTTP request sent from a user terminal. This request contains long text content, and the server stores this text data in a database. Specifically, the server adds the text data as a record in a database table.

[0193] Step 3:

[0194] Construct prompt sentences for analyzing production probability

[0195] The server analyzes the input text data and constructs a prompt to obtain the generation probability. The specific prompt will be in the form of "Please indicate the probability that the following text was generated by the generative AI: [input text]."

[0196] Step 4:

[0197] Obtaining generation probabilities using OpenAI's API

[0198] The server sends the constructed prompt to the OpenAI API and obtains the generation probability. To do this, the server uses the Python requests library to send an API request and receives the generation probability as a response. The data sent to the API includes the engine (e.g., text-davinci-003), prompt, maximum number of tokens, temperature parameters, etc.

[0199] Step 5:

[0200] Extract generation probabilities from responses and calculate credibility scores

[0201] The server parses the response received from the OpenAI API and extracts the generation probability using a regular expression. After obtaining the generation probability, it calculates a credibility score based on it. For example, if the generation probability is 65.4%, the credibility score is (1 - 0.654) 100 = 34.6. This score is then stored in a separate database table.

[0202] Step 6:

[0203] Send and display the credibility score on the user's device

[0204] The server returns the calculated credibility score to the user device as an HTTP response. The user device displays the received credibility score so that the user can confirm it. For example, the device UI might display "Credibility score: 34.6."

[0205] Specific example explanation

[0206] When a user enters the reason for wanting to work for your company as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," the server receives and stores this text. The server then sends the prompt text to the OpenAI API to obtain the generation probability, and calculates a credibility score based on the result. If the calculated credibility score is 34.6, the score is sent back to the user's device so that the user can view it on the screen.

[0207] This process flow allows the user to efficiently evaluate the credibility of long-form content.

[0208] 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.

[0209] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[0210] Embodiments of the invention

[0211] System Overview

[0212] 1. User Input

[0213] The user enters long content such as motivation for applying and application documents into the application form and clicks the submit button. The user's device then sends this input data to the server.

[0214] 2. Initial Server Setup

[0215] The server receives the long-form content sent by the user, sets the OpenAI API key, and prepares the text for analysis.

[0216] 3. Parsing User Input

[0217] The server sends the user's input text to the OpenAI API to obtain a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[0218] 4. Analysis by Emotion Engine

[0219] The server analyzes the emotions in the user's input text using an emotion engine, which uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[0220] 5. Correction of generation probability

[0221] Based on the analysis results of the emotion engine, the server adjusts the generation probability. For example, if the text is highly emotional, the generation probability is reduced to obtain a more credible score.

[0222] 6. Calculating the credibility score

[0223] The server calculates the credibility score based on the adjusted probability of creation, using the formula (1 - probability of creation) 100.

[0224] 7. Outputting the results

[0225] The server then sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[0226] System processing flow

[0227] The user enters long-form content and submits it to the server

[0228] The user enters their motivation for applying into the application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[0229] The server receives user input and initializes the OpenAI API.

[0230] The server receives the text data sent by the user, sets the OpenAI API key, and prepares the API.

[0231] The server parses the user input

[0232] The server uses the check_authenticity function to send the user's input text to the OpenAI API and obtain the generation probability.

[0233] Text analysis with emotion engine

[0234] The server uses an emotion engine to extract emotional components from the text and classify it, for example, classifying the text as "highly emotional," "neutral," or "positive."

[0235] Generation probability adjustment

[0236] The server adjusts the generated probabilities based on the results of the emotion engine, which allows the probabilities to reflect the emotional content of the text, resulting in a more accurate credibility score.

[0237] Calculating the credibility score

[0238] The server calculates the credibility score using the calculate_score function based on the corrected generation probability. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[0239] Output of results

[0240] The server then sends the calculated credibility score to the reviewer or user's device, who then uses this score to determine the credibility of the submitted sentence.

[0241] Specific example explanation

[0242] The user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjusts the generation probability based on the analysis results. The adjusted generation probability and credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer determines whether the submitted statement was written by the user themselves.

[0243] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[0244] The processing flow will be explained below.

[0245] Step 1:

[0246] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[0247] Step 2:

[0248] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[0249] Step 3:

[0250] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[0251] Step 4:

[0252] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[0253] Step 5:

[0254] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[0255] Engine:text-davinci-003

[0256] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[0257] Maximum number of tokens: 60

[0258] Temperature parameter: 0.5

[0259] Step 6:

[0260] The server sends the created request to the OpenAI API and waits for a response.

[0261] Step 7:

[0262] The server receives the response returned from the OpenAI API, which includes the generation probability.

[0263] Step 8:

[0264] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[0265] Step 9:

[0266] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[0267] Step 10:

[0268] The server uses an emotion engine to analyze the text data entered by the user and extract emotional elements from the text. The emotion engine uses natural language processing technology to perform emotion analysis and classify emotions as positive, negative, neutral, etc.

[0269] Step 11:

[0270] The server adjusts the generation probability based on the analysis results of the emotion engine. Specifically, if the emotion analysis results indicate an abnormally high emotional intensity, the server adjusts the generation probability to more accurately evaluate the credibility.

[0271] Step 12:

[0272] The server calculates the credibility score using the calculate_score function based on the adjusted probability of generation, using the formula (1 - probability of generation) 100.

[0273] Step 13:

[0274] The server generates a result containing the calculated credibility score as an HTTP response and sends it to the reviewer or device.

[0275] Step 14:

[0276] The terminal displays the credibility score received from the server, and the reviewer uses this score to judge the credibility of the sentence submitted by the user.

[0277] Example 2

[0278] 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."

[0279] Conventional systems have difficulty accurately assessing the credibility of long content input by users. Furthermore, simply obtaining the generation probability using a generative AI model does not take into account internal information such as emotional factors, making accurate credibility assessment difficult. Furthermore, because the credibility score is calculated without taking emotional factors into account, the system is vulnerable to the evaluation of text whose emotions have been intentionally manipulated.

[0280] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0281] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for analyzing the received long-form content using an emotion engine and classifying emotional elements, means for correcting the generation probability based on the emotion analysis result, means for calculating a credibility score based on the corrected generation probability, and means for outputting the calculated credibility score. This enables a more accurate and objective credibility assessment of the content input by the user by combining the generation probability by the generative AI model and the emotion analysis result.

[0282] "User" refers to an individual or company that uses the system to input and submit long-form content.

[0283] "Long-form content" refers to relatively long textual data such as motivation letters and application documents.

[0284] "Generative AI model" refers to an artificial intelligence algorithm that calculates the probability of generated text.

[0285] "Generation probability" refers to the quantification of the likelihood that a particular piece of text was generated by a generative AI model.

[0286] "Credibility score" refers to a numerical value that indicates the reliability of a text, calculated based on the probability of its generation.

[0287] "Emotion engine" refers to natural language processing technology for analyzing and classifying emotional elements in text.

[0288] "Sentiment analysis" refers to the process of extracting emotional elements from text and classifying them as positive, negative, neutral, etc.

[0289] "Correct" refers to adjusting the generation probability based on the results of sentiment analysis.

[0290] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[0291] The system of the present invention uses the following hardware and software: The hardware depends on the user's terminal and server, and the software uses the OpenAI API and emotion engine (which uses natural language processing technology).

[0292] First, the user enters long text content, such as their motivation for applying and application documents, into the application form and clicks the submit button. The user's device sends this input data to the server. The server receives the long text content sent by the user and simultaneously sets an OpenAI API key and prepares for text analysis.

[0293] The server then sends the user's input text to the OpenAI API to obtain a generation probability, which is a numerical value indicating the likelihood that the text was generated by generative AI. The server then uses an emotion engine to analyze the emotion in the user's input text. The emotion engine uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[0294] Next, the server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, it will reduce the generation probability to obtain a higher credibility score. The server then calculates the credibility score based on the adjusted generation probability. This calculation is performed using the formula (1 - generation probability) 100.

[0295] Finally, the server sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[0296] Specific example explanation

[0297] As a specific example of operation, a user enters the following reason for applying into the application form and submits it.

[0298] "I am looking to grow through my experience with your company. I am interested in the job description and feel that it will allow me to utilize my skills."

[0299] The server receives this text and uses the OpenAI API to obtain the probability of its creation. For example, if the returned probability of creation is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjust the probability of its creation based on the analysis results. The adjusted probability of creation and a credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer can determine whether the submitted sentence was written by the user.

[0300] Example prompt sentence:

[0301] "The user provided the following motivation:

[0302] "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[0303] Find the probability of generating this text, perform sentiment analysis, and calculate a credibility score."

[0304] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[0305] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0306] Step 1:

[0307] The user enters long content such as motivation for applying and application documents into the application form and clicks the "Submit" button. The device sends an HTTP request including the entered text to the server.

[0308] Input: Long-form content entered into the application form

[0309] Output: HTTP request sent to the server

[0310] Step 2:

[0311] The server receives the text data sent by the user as an HTTP request.

[0312] Input: HTTP request from the terminal

[0313] Output: Received text data

[0314] Step 3:

[0315] The server retrieves the OpenAI API key from a configuration file or environment variable and initializes the API.

[0316] Input: API key stored in a config file or environment variable

[0317] Output: Initialized OpenAI API

[0318] Step 4:

[0319] The server calls the check_authenticity function to send the user's input text to the OpenAI API, which formats the text into an API request and sends it to the OpenAI API.

[0320] Input: User-entered text

[0321] Output: API request sent to the OpenAI API

[0322] Step 5:

[0323] The server receives a response from the OpenAI API that includes a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[0324] Input: Response from OpenAI API

[0325] Output: Generation probability

[0326] Step 6:

[0327] The server uses an emotion engine to extract emotional elements from the text and classify them into categories such as "positive," "negative," and "neutral."

[0328] Input: User-entered text

[0329] Output: Emotion analysis results

[0330] Step 7:

[0331] The server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, the generation probability is reduced by a certain percentage.

[0332] Input: Sentiment analysis results and generation probability

[0333] Output: Corrected generation probabilities

[0334] Step 8:

[0335] The server calculates the credibility score based on the adjusted probability of creation, using the formula "(1 - probability of creation) 100".

[0336] Input: Corrected generation probability

[0337] Output: Belief score

[0338] Step 9:

[0339] The server sends the calculated credibility score to the reviewer or user's device, and sends information such as the credibility score, the corrected generation probability, and the sentiment analysis results via an HTTP response.

[0340] Input: Belief score, corrected generation probability, sentiment analysis result

[0341] Output: Results sent to users and reviewers

[0342] Through the above processing steps, the present invention efficiently distinguishes between content created by users and content generated by generation AI, reducing the burden on reviewers and enabling more accurate credibility assessments.

[0343] (Application example 2)

[0344] 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."

[0345] There is a need for efficient and accurate evaluation of the credibility of reviews posted by users on online shopping sites. However, there are many fraudulent reviews generated by AI and reviews that are misleading based on emotions, and appropriate countermeasures against these are lacking. As a result, it is difficult for buyers to select products based on trustworthy reviews. Therefore, an objective of the present invention is to provide a system that increases the credibility of posted reviews, allowing users to refer to reviews with peace of mind.

[0346] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0347] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing emotions in the long-form content, means for correcting the generation probability based on the result of the emotion analysis, and means for providing the credibility score of the long-form content to a reviewer. This makes it possible to increase the credibility of reviews posted by users and provide highly reliable reviews.

[0348] "User" refers to a person or end user of the System.

[0349] "Long-form content" refers to long texts entered by users, such as motivation for applying, application documents, and reviews.

[0350] "Means for receiving" refers to a method or device by which the server receives data sent from the user.

[0351] The term "analyzing means" refers to a method or device for analyzing the content of received long-form content.

[0352] "Generation probability" refers to the probability that a particular piece of text could have been generated by a generative AI.

[0353] "Means for obtaining generation probability" refers to a method or algorithm for calculating the generation probability of a text.

[0354] "Authenticity Score" refers to a numerical value used to assess the veracity of long-form content.

[0355] "Means for calculating" refers to a method or apparatus for calculating a credibility score based on the generation probability.

[0356] "Outputting means" refers to a method or device that displays the calculated credibility score.

[0357] "Means for analyzing emotions" refers to a method or device for analyzing emotions contained within long-form content.

[0358] "Means for correcting generation probability" refers to a method or device for adjusting generation probability based on the results of sentiment analysis.

[0359] "Reviewer" refers to a person or system that checks the credibility score and determines the veracity of posted content.

[0360] "Means for providing" refers to a method or device for providing the credibility score to the reviewer.

[0361] The system embodying this invention analyzes long-form content entered by users and evaluates its credibility. This system corrects the credibility score by combining generation probability obtained by a generative AI model and emotion analysis by an emotion engine.

[0362] System configuration

[0363] Hardware

[0364] server

[0365] User terminal

[0366] Reviewer device

[0367] software

[0368] OpenAI API

[0369] Emotion engine (natural language processing library such as TextBlob)

[0370] Database

[0371] System processing flow

[0372] 1. User input:

[0373] The user accesses the review submission page of the online shopping site and enters a review.

[0374] For example, a user might enter, "This product is great! I'm really happy with my purchase."

[0375] After completing the input, the user clicks the "Submit button."

[0376] 2. Data reception:

[0377] The user terminal sends an HTTP request containing the entered text to the server.

[0378] The server receives this data.

[0379] 3. Obtaining generation probabilities:

[0380] The server initializes the OpenAI API and sends the reviews entered by the user.

[0381] The OpenAI API returns the probability of generating the text. For example, if the generation probability is "0.35", you will get this value.

[0382] 4. Emotion analysis:

[0383] The server uses a sentiment engine (TextBlob library) to analyze the sentiment of the reviews.

[0384] For example, the text "This product is great! I'm really happy with my purchase" would be classified as "positive."

[0385] 5. Generation probability adjustment:

[0386] The server corrects the generation probability based on the results of the emotion analysis.

[0387] If positive emotions are included, the probability of generation is reduced. For example, if the original generation probability was 0.35, the corrected generation probability becomes 0.31.

[0388] 6. Calculating the credibility score:

[0389] The server calculates a credibility score based on the corrected generation probability.

[0390] Specifically, the score is calculated using the formula "(1 - generation probability) 100". If the corrected generation probability is "0.31", the credibility score will be "69".

[0391] 7. Result output:

[0392] The server provides the calculated credibility score to the reviewer.

[0393] Reviewers use this score to judge the credibility of reviews posted by users.

[0394] Specific examples

[0395] Example prompt sentence:

[0396] "Is this text generated by AI? This product is excellent. I use it every day and have had no issues."

[0397] Example response:

[0398] "0.35"

[0399] This will increase the credibility of reviews posted by users and improve the reliability of reviews on online shopping sites. Reviewers will be able to provide information to users based on highly reliable reviews.

[0400] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0401] Step 1:

[0402] Users can write and submit reviews

[0403] A user accesses a review input form on an online shopping site and enters a review. For example, they might enter, "This product is great! I'm really happy with my purchase." After entering the review, the user clicks the "Submit" button to send the review to the server. At this point, the input is the review text entered by the user, and the output is the review text sent to the server.

[0404] Step 2:

[0405] Server receives review text

[0406] The server receives the review text sent from the user's device. It analyzes the data sent as an HTTP request and extracts the review text. At this point, the input is the HTTP request from the user's device, and the output is the review text.

[0407] Step 3:

[0408] Obtaining generation probability

[0409] The server sends the received review text to the OpenAI API and obtains the generation probability. Specifically, it calls the OpenAI API and sends the review text including the prompt "Is this text generated by AI?". The OpenAI API returns a numerical value indicating the probability that the review text was generated by the generation AI. At this point, the input is the review text, and the output is a numerical value indicating the generation probability.

[0410] Step 4:

[0411] Performing sentiment analysis

[0412] The server uses a sentiment analysis library such as TextBlob to analyze the sentiment of the review text. It identifies positive, negative, or neutral sentiment elements in the review text and classifies the sentiment. At this point, the input is the review text, and the output is the sentiment classification result (positive, negative, neutral, etc.).

[0413] Step 5:

[0414] Generation probability adjustment

[0415] The server corrects the obtained generation probability based on the results of emotion analysis. For example, if a positive emotion is included, the generation probability is reduced to increase credibility. This correction is performed by a numerical calculation using the emotion analysis results and generation probability. The input at this point is the generation probability and the emotion classification result, and the output is the corrected generation probability.

[0416] Step 6:

[0417] Calculating the credibility score

[0418] The server calculates the credibility score based on the adjusted generation probability. Specifically, it uses the formula "(1 - generation probability) 100". If the adjusted generation probability is 0.31, the credibility score is 69. At this point, the input is the adjusted generation probability, and the output is the credibility score.

[0419] Step 7:

[0420] Outputting the results and providing them to reviewers

[0421] The server provides the calculated credibility score to the reviewer. The reviewer checks this score and judges the credibility of the review. The credibility score and the review text are displayed on the reviewer's terminal. At this point, the input is the credibility score, and the output is the credibility score and review text displayed on the reviewer's terminal.

[0422] 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.

[0423] 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.

[0424] 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.

[0425] [Second embodiment]

[0426] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0427] 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.

[0428] 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).

[0429] 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.

[0430] 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.

[0431] 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).

[0432] 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.

[0433] 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.

[0434] 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.

[0435] 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.

[0436] 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.

[0437] 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."

[0438] This invention is a system that analyzes long-form content entered by a user, obtains the probability of generation by a generative AI model, and calculates and outputs a credibility score based on that probability. This system makes it possible to efficiently distinguish between content created by a user and content automatically generated by a generative AI.

[0439] Embodiments of the invention

[0440] System Overview

[0441] 1. User Input

[0442] The user inputs long content such as a reason for applying into an application form or submission form. The user's device sends this input to the server.

[0443] 2. Initial Server Setup

[0444] The server receives input from the user and initializes the OpenAI API, which is used to analyze the user's input and evaluate the likelihood that the text was generated by a generative AI.

[0445] 3. Parsing User Input

[0446] The server analyzes the text received from the user and obtains the generation probability from the generative AI model. For this process, it uses the OpenAI API and extracts the generation probability from the response.

[0447] 4. Calculating the credibility score

[0448] The server calculates a credibility score based on the obtained probability of creation, which is calculated using the formula (1 - probability of creation) 100.

[0449] 5. Outputting the results

[0450] The calculated credibility score is returned to the reviewer or user's device, and the reviewer uses this score to determine the credibility of the submitted content.

[0451] System processing flow

[0452] The user enters long-form content and submits it to the server

[0453] For example, a user enters their motivation for applying into an application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[0454] The server receives user input and initializes the OpenAI API.

[0455] The server stores the text data received from the user and sets the OpenAI API key to prepare the API.

[0456] The server parses the user input

[0457] The server calls the check_authenticity function and sends the user's input text to the OpenAI API to obtain the generation probability. This process uses the following parameters: engine (e.g., text-davinci-003), prompt (a sentence containing the text to be evaluated and the evaluation instructions), maximum number of tokens, and temperature parameter.

[0458] The server analyzes the generation probability

[0459] The server receives the response from the OpenAI API, parses it, and extracts the generated probability using regular expressions, converting the extracted probability value into a floating-point number between 0 and 1.

[0460] The server calculates the credibility score

[0461] The server calculates the credibility score using the calculate_score function. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[0462] The server returns the results

[0463] The server returns the calculated credibility score to the reviewer or user's device, and the reviewer uses this score to judge the credibility of the submitted sentence.

[0464] Specific example explanation

[0465] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server analyzes this probability and calculates a credibility score of 34.6. This score is returned to the reviewer or user's device to help determine whether the submitted statement is the user's own.

[0466] In this way, the present invention enables efficient distinction between content created by users and content generated by AI, reducing the burden on reviewers.

[0467] The processing flow will be explained below.

[0468] Step 1:

[0469] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[0470] Step 2:

[0471] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[0472] Step 3:

[0473] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[0474] Step 4:

[0475] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[0476] Step 5:

[0477] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[0478] Engine:text-davinci-003

[0479] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[0480] Maximum number of tokens: 60

[0481] Temperature parameter: 0.5

[0482] Step 6:

[0483] The server sends the created request to the OpenAI API and waits for a response.

[0484] Step 7:

[0485] The server receives the response returned from the OpenAI API, which includes the generation probability.

[0486] Step 8:

[0487] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[0488] Step 9:

[0489] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[0490] Step 10:

[0491] The server calls the calculate_score function to calculate the credibility score based on the obtained generation probability, using the formula (1 - generation probability) 100.

[0492] Step 11:

[0493] The server generates a result containing the calculated credibility score as an HTTP response and sends it back to the reviewer or device.

[0494] Step 12:

[0495] The terminal displays the credibility score received from the server, and the reviewer uses this to judge the credibility of the submitted statement.

[0496] Example 1

[0497] 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."

[0498] It is difficult to efficiently distinguish whether user-created long-form content was created manually or automatically generated by a generative AI model. This makes credibility assessment difficult and increases the burden on reviewers. The present invention aims to solve this problem by providing a system that accurately and efficiently evaluates the credibility of user-created content.

[0499] 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.

[0500] In this invention, the server includes a means for receiving long-form content input by a user, a means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, and a means for calculating a credibility score based on the obtained generation probability. This makes it possible to evaluate the credibility of content input by a user with high accuracy and efficiently determine whether a submitted document was written by the user himself / herself.

[0501] "User" refers to an individual or group of people who input long-form content into the system.

[0502] "Long content" refers to text data entered by a user that has a certain number of characters or more.

[0503] "Means for receiving" refers to the interface or process by which the server obtains the data entered by the user.

[0504] "Means for analyzing" refers to algorithms or programs for evaluating and analyzing received long-form content.

[0505] A "generative AI model" refers to a system or software that uses artificial intelligence technology to generate text.

[0506] "Generation probability" refers to a number that indicates the likelihood that a generative AI model generated a given piece of text.

[0507] The "credibility score" refers to an index that indicates whether the text was created by the user himself or herself, calculated based on the probability of creation.

[0508] "Means for outputting" refers to an interface or method for providing the calculated credibility score to a user or reviewer.

[0509] "Means of initializing the API" refers to the process of preparing the API for use with the API key and settings of the generated AI model.

[0510] "Means for setting prompt sentences" refers to a method for forming input sentences or instructions to make analysis requests to a generative AI model.

[0511] "Means of analyzing the response" refers to the process of interpreting the response data obtained from the generative AI model and extracting the necessary information.

[0512] "Means for extracting probability values" refers to a method or algorithm for extracting generation probabilities from response data.

[0513] "Terminal" refers to an electronic device used by a user, such as a computer, smartphone, or tablet.

[0514] "Means for evaluating" refers to the set of processes or calculations required to determine the authenticity of a received text.

[0515] The present invention is a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. The following describes in detail an embodiment of the present invention.

[0516] System configuration

[0517] User Input

[0518] A user uses their device (computer, smartphone, tablet, etc.) to enter long-form content into an application or submission form, such as a motivation letter or essay, and clicks the "Submit" button. This entered text data is sent by the device to the server via an HTTP POST request.

[0519] Initial Server Configuration

[0520] The server receives the text data sent from the user's device. Then, the server initializes the generative AI model (e.g., OpenAI API) by creating an API instance using the API key obtained from the configuration file.

[0521] Parsing User Input

[0522] The server calls the check_authenticity function to parse the text entered by the user. This function sets a prompt containing the text to be evaluated and evaluation instructions. Here is an example prompt:

[0523] _Example prompt sentence:_

[0524] "Please rate the likelihood that the following text was generated by a generative AI."

[0525] Obtaining generation probability

[0526] The server sends a request to the OpenAI API endpoint containing the prompt and settings (engine: e.g., text-davinci-003, maximum number of tokens, temperature parameters, etc.) to obtain a generation probability indicating the likelihood that the user's input was generated by the generative AI.

[0527] Analysis of generation probability

[0528] The server parses the response received from the API and extracts the generation probability. In this process, it extracts the probability value from the response JSON data and converts it to a floating point number between 0 and 1 using regular expressions if necessary.

[0529] Calculating the credibility score

[0530] The server calculates the credibility score based on the obtained generation probability. Specifically, it uses the following formula:

[0531] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[0532] For example, if the generation probability is 0.654, the credibility score is 34.6.

[0533] Output of results

[0534] The server returns the calculated credibility score to the reviewer or user's device, allowing the reviewer to evaluate the credibility of the submitted content.

[0535] Specific examples

[0536] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the API responds with "Generation probability 65.4%," the server analyzes this probability and calculates an authenticity score of 34.6. By returning this score to the reviewer or user's device, it can determine whether the submitted document was created by the user themselves.

[0537] In this way, the present invention provides a highly accurate and efficient evaluation system for effectively distinguishing between user-generated content and content generated by a generation AI.

[0538] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0539] Explanation of the processing flow

[0540] Step 1:

[0541] A user enters long-form content into an application or submission form on their own device and clicks the "Send" button. The entered text data is sent from the device to the server. The input data includes the long-form content (e.g., reason for applying) and the user ID. The server then receives an HTTP POST request and extracts the text data.

[0542] Step 2:

[0543] The server saves the received text data and initializes the API of the generative AI model. Specifically, the server obtains the API key from the configuration file and configures the API. Once the API key is correctly configured, the generative AI model is ready to use. The API key is included as input data. The output is an initialized API instance.

[0544] Step 3:

[0545] The server calls the check_authenticity function to analyze the user's input text. This function sets a prompt and generates an analysis request. The prompt includes instructions such as "Please rate the likelihood that the following text was generated by a generative AI." The input data includes the text data and the prompt. The output is an analysis request.

[0546] Step 4:

[0547] The server sends the generated analysis request to the generative AI model endpoint. The request includes settings such as the engine (e.g., text-davinci-003), maximum number of tokens, and temperature parameters. The analysis request and API settings are included as input data. This sends an API request to obtain the generation probability.

[0548] Step 5:

[0549] The server receives the response from the generative AI model and parses the response data. It extracts the generation probability from the JSON-formatted response data. In this process, it uses a regular expression to extract the generation probability and converts it into a floating-point number ranging from 0 to 1. The input data includes the response JSON data. The output is the extracted generation probability.

[0550] Step 6:

[0551] The server calculates the credibility score based on the obtained generation probability using the following formula:

[0552] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[0553] For example, if the generation probability is 0.654, the credibility score is 34.6. The input data includes the generation probability. The output is the calculated credibility score.

[0554] Step 7:

[0555] The server returns the calculated credibility score to the reviewer or user's device. It generates an HTTP response and sends information including the credibility score and related metadata to the user's device. The credibility score and user ID are included as input data, allowing the user or reviewer to confirm the credibility of the submitted document.

[0556] (Application example 1)

[0557] 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."

[0558] There is a challenge in determining with high accuracy whether long-form content sent by users was created by generative AI. There is also a need for a means to quickly and efficiently evaluate the authenticity of emails, documents, etc. In addition, users need help making more reliable judgments by specifically visualizing the authenticity evaluation based on generation probability.

[0559] 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.

[0560] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing the input text and using an analysis algorithm and the generation probability to calculate the credibility score, and means for displaying the credibility score on an electronic device. This makes it possible to quickly evaluate the credibility of long-form content received by a user with high accuracy and visualize the credibility score.

[0561] "Long content" refers to text data entered by a user, and mainly includes the contents of emails, documents, and the like.

[0562] A "generative AI model" is a model that can generate new text using an artificial intelligence algorithm trained on large amounts of text data.

[0563] "Generation probability" is a number that indicates the likelihood that a particular piece of text was generated by a generative AI model.

[0564] The "credibility score" is a numerical value for evaluating the reliability of the text entered by the user based on the generation probability.

[0565] An "analysis algorithm" is a calculation method for analyzing input text and calculating its generation probability.

[0566] "Electronic device" refers to a hardware device that can execute a program and display the results, and specifically includes smartphones, tablets, personal computers, etc.

[0567] "Server" means a network-enabled computer system that processes data received from users and calculates and outputs credibility scores.

[0568] The present invention provides a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. Specific embodiments for implementing the present invention are described below.

[0569] System configuration

[0570] This system mainly consists of the following hardware and software:

[0571] Server: A network-enabled computer system that processes data received from users and calculates and outputs credibility scores. The software used is the Python language and the OpenAI API.

[0572] User device: Any device on which a user enters long-form content, including smartphones, tablets, and PCs.

[0573] Electronic Device: A hardware device for displaying the credibility score. This can be a user terminal.

[0574] Processing flow

[0575] 1. User input:

[0576] The user inputs long-form content such as emails and documents through the device. For example, they input a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills."

[0577] 2. Receiving long-form content:

[0578] The long text content entered by the user is sent from the terminal to the server, which receives the text data and stores it in a database.

[0579] 3. Obtaining generation probabilities:

[0580] The server analyzes the stored text data and obtains the probability of generation using a generative AI model (e.g., OpenAI's text-davinci-003). The prompt used is, "Please indicate the probability that the following text was generated by the generative AI: I want to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[0581] 4. Calculating the credibility score:

[0582] The Python regular expression library "re" is used to analyze the generation probability, and a credibility score is calculated based on the obtained generation probability. For example, if the generation probability is 65.4%, the credibility score is "(1 - 0.654) 100 = 34.6". The calculated credibility score is stored in a database in a specific format.

[0583] 5. Output of credibility score:

[0584] Once the calculation is complete, the server sends the authenticity score to the user's device, where it can be displayed to visually confirm the authenticity of the email or document.

[0585] Specific example explanation

[0586] For example, if a user enters the following as their motivation for applying: "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," this text is sent to the server. The server uses the OpenAI API to obtain the probability of generating this text, and if the generation probability is 65.4%, the credibility score will be 34.6. This credibility score is sent to the user's device, and the user can judge the credibility of the document based on the displayed credibility score.

[0587] In this way, the system of the present invention helps users to quickly and accurately assess the credibility of long-form content.

[0588] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0589] Step 1:

[0590] Collect and send user-supplied long-form content

[0591] A user inputs long content such as an email or document through a terminal and clicks the send button. The input text is sent to the server as an HTTP request. The input may include, for example, a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills." The text data is sent to the server in JSON format.

[0592] Step 2:

[0593] The server receives and stores the long-form content.

[0594] The server receives an HTTP request sent from a user terminal. This request contains long text content, and the server stores this text data in a database. Specifically, the server adds the text data as a record in a database table.

[0595] Step 3:

[0596] Construct prompt sentences for analyzing production probability

[0597] The server analyzes the input text data and constructs a prompt to obtain the generation probability. The specific prompt will be in the form of "Please indicate the probability that the following text was generated by the generative AI: [input text]."

[0598] Step 4:

[0599] Obtaining generation probabilities using OpenAI's API

[0600] The server sends the constructed prompt to the OpenAI API and obtains the generation probability. To do this, the server uses the Python requests library to send an API request and receives the generation probability as a response. The data sent to the API includes the engine (e.g., text-davinci-003), prompt, maximum number of tokens, temperature parameters, etc.

[0601] Step 5:

[0602] Extract generation probabilities from responses and calculate credibility scores

[0603] The server parses the response received from the OpenAI API and extracts the generation probability using a regular expression. After obtaining the generation probability, it calculates a credibility score based on it. For example, if the generation probability is 65.4%, the credibility score is (1 - 0.654) 100 = 34.6. This score is then stored in a separate database table.

[0604] Step 6:

[0605] Send and display the credibility score on the user's device

[0606] The server returns the calculated credibility score to the user device as an HTTP response. The user device displays the received credibility score so that the user can confirm it. For example, the device UI might display "Credibility score: 34.6."

[0607] Specific example explanation

[0608] When a user enters the reason for wanting to work for your company as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," the server receives and stores this text. The server then sends the prompt text to the OpenAI API to obtain the generation probability, and calculates a credibility score based on the result. If the calculated credibility score is 34.6, the score is sent back to the user's device so that the user can view it on the screen.

[0609] This process flow allows the user to efficiently evaluate the credibility of long-form content.

[0610] 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.

[0611] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[0612] Embodiments of the invention

[0613] System Overview

[0614] 1. User Input

[0615] The user enters long content such as motivation for applying and application documents into the application form and clicks the submit button. The user's device then sends this input data to the server.

[0616] 2. Initial Server Setup

[0617] The server receives the long-form content sent by the user, sets the OpenAI API key, and prepares the text for analysis.

[0618] 3. Parsing User Input

[0619] The server sends the user's input text to the OpenAI API to obtain a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[0620] 4. Analysis by Emotion Engine

[0621] The server analyzes the emotions in the user's input text using an emotion engine, which uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[0622] 5. Correction of generation probability

[0623] Based on the analysis results of the emotion engine, the server adjusts the generation probability. For example, if the text is highly emotional, the generation probability is reduced to obtain a more credible score.

[0624] 6. Calculating the credibility score

[0625] The server calculates the credibility score based on the adjusted probability of creation, using the formula (1 - probability of creation) 100.

[0626] 7. Outputting the results

[0627] The server then sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[0628] System processing flow

[0629] The user enters long-form content and submits it to the server

[0630] The user enters their motivation for applying into the application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[0631] The server receives user input and initializes the OpenAI API.

[0632] The server receives the text data sent by the user, sets the OpenAI API key, and prepares the API.

[0633] The server parses the user input

[0634] The server uses the check_authenticity function to send the user's input text to the OpenAI API and obtain the generation probability.

[0635] Text analysis with emotion engine

[0636] The server uses an emotion engine to extract emotional components from the text and classify it, for example, classifying the text as "highly emotional," "neutral," or "positive."

[0637] Generation probability adjustment

[0638] The server adjusts the generated probabilities based on the results of the emotion engine, which allows the probabilities to reflect the emotional content of the text, resulting in a more accurate credibility score.

[0639] Calculating the credibility score

[0640] The server calculates the credibility score using the calculate_score function based on the corrected generation probability. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[0641] Output of results

[0642] The server then sends the calculated credibility score to the reviewer or user's device, who then uses this score to determine the credibility of the submitted sentence.

[0643] Specific example explanation

[0644] The user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjusts the generation probability based on the analysis results. The adjusted generation probability and credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer determines whether the submitted statement was written by the user themselves.

[0645] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[0646] The processing flow will be explained below.

[0647] Step 1:

[0648] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[0649] Step 2:

[0650] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[0651] Step 3:

[0652] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[0653] Step 4:

[0654] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[0655] Step 5:

[0656] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[0657] Engine:text-davinci-003

[0658] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[0659] Maximum number of tokens: 60

[0660] Temperature parameter: 0.5

[0661] Step 6:

[0662] The server sends the created request to the OpenAI API and waits for a response.

[0663] Step 7:

[0664] The server receives the response returned from the OpenAI API, which includes the generation probability.

[0665] Step 8:

[0666] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[0667] Step 9:

[0668] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[0669] Step 10:

[0670] The server uses an emotion engine to analyze the text data entered by the user and extract emotional elements from the text. The emotion engine uses natural language processing technology to perform emotion analysis and classify emotions as positive, negative, neutral, etc.

[0671] Step 11:

[0672] The server adjusts the generation probability based on the analysis results of the emotion engine. Specifically, if the emotion analysis results indicate an abnormally high emotional intensity, the server adjusts the generation probability to more accurately evaluate the credibility.

[0673] Step 12:

[0674] The server calculates the credibility score using the calculate_score function based on the adjusted probability of generation, using the formula (1 - probability of generation) 100.

[0675] Step 13:

[0676] The server generates a result containing the calculated credibility score as an HTTP response and sends it to the reviewer or device.

[0677] Step 14:

[0678] The terminal displays the credibility score received from the server, and the reviewer uses this score to judge the credibility of the sentence submitted by the user.

[0679] Example 2

[0680] 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."

[0681] Conventional systems have difficulty accurately assessing the credibility of long content input by users. Furthermore, simply obtaining the generation probability using a generative AI model does not take into account internal information such as emotional factors, making accurate credibility assessment difficult. Furthermore, because the credibility score is calculated without taking emotional factors into account, the system is vulnerable to the evaluation of text whose emotions have been intentionally manipulated.

[0682] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0683] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for analyzing the received long-form content using an emotion engine and classifying emotional elements, means for correcting the generation probability based on the emotion analysis result, means for calculating a credibility score based on the corrected generation probability, and means for outputting the calculated credibility score. This enables a more accurate and objective credibility assessment of the content input by the user by combining the generation probability by the generative AI model and the emotion analysis result.

[0684] "User" refers to an individual or company that uses the system to input and submit long-form content.

[0685] "Long-form content" refers to relatively long textual data such as motivation letters and application documents.

[0686] "Generative AI model" refers to an artificial intelligence algorithm that calculates the probability of generated text.

[0687] "Generation probability" refers to the quantification of the likelihood that a particular piece of text was generated by a generative AI model.

[0688] "Credibility score" refers to a numerical value that indicates the reliability of a text, calculated based on the probability of its generation.

[0689] "Emotion engine" refers to natural language processing technology for analyzing and classifying emotional elements in text.

[0690] "Sentiment analysis" refers to the process of extracting emotional elements from text and classifying them as positive, negative, neutral, etc.

[0691] "Correct" refers to adjusting the generation probability based on the results of sentiment analysis.

[0692] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[0693] The system of the present invention uses the following hardware and software: The hardware depends on the user's terminal and server, and the software uses the OpenAI API and emotion engine (which uses natural language processing technology).

[0694] First, the user enters long text content, such as their motivation for applying and application documents, into the application form and clicks the submit button. The user's device sends this input data to the server. The server receives the long text content sent by the user and simultaneously sets an OpenAI API key and prepares for text analysis.

[0695] The server then sends the user's input text to the OpenAI API to obtain a generation probability, which is a numerical value indicating the likelihood that the text was generated by generative AI. The server then uses an emotion engine to analyze the emotion in the user's input text. The emotion engine uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[0696] Next, the server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, it will reduce the generation probability to obtain a higher credibility score. The server then calculates the credibility score based on the adjusted generation probability. This calculation is performed using the formula (1 - generation probability) 100.

[0697] Finally, the server sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[0698] Specific example explanation

[0699] As a specific example of operation, a user enters the following reason for applying into the application form and submits it.

[0700] "I am looking to grow through my experience with your company. I am interested in the job description and feel that it will allow me to utilize my skills."

[0701] The server receives this text and uses the OpenAI API to obtain the probability of its creation. For example, if the returned probability of creation is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjust the probability of its creation based on the analysis results. The adjusted probability of creation and a credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer can determine whether the submitted sentence was written by the user.

[0702] Example prompt sentence:

[0703] "The user provided the following motivation:

[0704] "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[0705] Find the probability of generating this text, perform sentiment analysis, and calculate a credibility score."

[0706] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[0707] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0708] Step 1:

[0709] The user enters long content such as motivation for applying and application documents into the application form and clicks the "Submit" button. The device sends an HTTP request including the entered text to the server.

[0710] Input: Long-form content entered into the application form

[0711] Output: HTTP request sent to the server

[0712] Step 2:

[0713] The server receives the text data sent by the user as an HTTP request.

[0714] Input: HTTP request from the terminal

[0715] Output: Received text data

[0716] Step 3:

[0717] The server retrieves the OpenAI API key from a configuration file or environment variable and initializes the API.

[0718] Input: API key stored in a config file or environment variable

[0719] Output: Initialized OpenAI API

[0720] Step 4:

[0721] The server calls the check_authenticity function to send the user's input text to the OpenAI API, which formats the text into an API request and sends it to the OpenAI API.

[0722] Input: User-entered text

[0723] Output: API request sent to the OpenAI API

[0724] Step 5:

[0725] The server receives a response from the OpenAI API that includes a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[0726] Input: Response from OpenAI API

[0727] Output: Generation probability

[0728] Step 6:

[0729] The server uses an emotion engine to extract emotional elements from the text and classify them into categories such as "positive," "negative," and "neutral."

[0730] Input: User-entered text

[0731] Output: Emotion analysis results

[0732] Step 7:

[0733] The server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, the generation probability is reduced by a certain percentage.

[0734] Input: Sentiment analysis results and generation probability

[0735] Output: Corrected generation probabilities

[0736] Step 8:

[0737] The server calculates the credibility score based on the adjusted probability of creation, using the formula "(1 - probability of creation) 100".

[0738] Input: Corrected generation probability

[0739] Output: Belief score

[0740] Step 9:

[0741] The server sends the calculated credibility score to the reviewer or user's device, and sends information such as the credibility score, the corrected generation probability, and the sentiment analysis results via an HTTP response.

[0742] Input: Belief score, corrected generation probability, sentiment analysis result

[0743] Output: Results sent to users and reviewers

[0744] Through the above processing steps, the present invention efficiently distinguishes between content created by users and content generated by generation AI, reducing the burden on reviewers and enabling more accurate credibility assessments.

[0745] (Application example 2)

[0746] 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."

[0747] There is a need for efficient and accurate evaluation of the credibility of reviews posted by users on online shopping sites. However, there are many fraudulent reviews generated by AI and reviews that are misleading based on emotions, and appropriate countermeasures against these are lacking. As a result, it is difficult for buyers to select products based on trustworthy reviews. Therefore, an objective of the present invention is to provide a system that increases the credibility of posted reviews, allowing users to refer to reviews with peace of mind.

[0748] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0749] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing emotions in the long-form content, means for correcting the generation probability based on the result of the emotion analysis, and means for providing the credibility score of the long-form content to a reviewer. This makes it possible to increase the credibility of reviews posted by users and provide highly reliable reviews.

[0750] "User" refers to a person or end user of the System.

[0751] "Long-form content" refers to long texts entered by users, such as motivation for applying, application documents, and reviews.

[0752] "Means for receiving" refers to a method or device by which the server receives data sent from the user.

[0753] The term "analyzing means" refers to a method or device for analyzing the content of received long-form content.

[0754] "Generation probability" refers to the probability that a particular piece of text could have been generated by a generative AI.

[0755] "Means for obtaining generation probability" refers to a method or algorithm for calculating the generation probability of a text.

[0756] "Authenticity Score" refers to a numerical value used to assess the veracity of long-form content.

[0757] "Means for calculating" refers to a method or apparatus for calculating a credibility score based on the generation probability.

[0758] "Outputting means" refers to a method or device that displays the calculated credibility score.

[0759] "Means for analyzing emotions" refers to a method or device for analyzing emotions contained within long-form content.

[0760] "Means for correcting generation probability" refers to a method or device for adjusting generation probability based on the results of sentiment analysis.

[0761] "Reviewer" refers to a person or system that checks the credibility score and determines the veracity of posted content.

[0762] "Means for providing" refers to a method or device for providing the credibility score to the reviewer.

[0763] The system embodying this invention analyzes long-form content entered by users and evaluates its credibility. This system corrects the credibility score by combining generation probability obtained by a generative AI model and emotion analysis by an emotion engine.

[0764] System configuration

[0765] Hardware

[0766] server

[0767] User terminal

[0768] Reviewer device

[0769] software

[0770] OpenAI API

[0771] Emotion engine (natural language processing library such as TextBlob)

[0772] Database

[0773] System processing flow

[0774] 1. User input:

[0775] The user accesses the review submission page of the online shopping site and enters a review.

[0776] For example, a user might enter, "This product is great! I'm really happy with my purchase."

[0777] After completing the input, the user clicks the "Submit button."

[0778] 2. Data reception:

[0779] The user terminal sends an HTTP request containing the entered text to the server.

[0780] The server receives this data.

[0781] 3. Obtaining generation probabilities:

[0782] The server initializes the OpenAI API and sends the reviews entered by the user.

[0783] The OpenAI API returns the probability of generating the text. For example, if the generation probability is "0.35", you will get this value.

[0784] 4. Emotion analysis:

[0785] The server uses a sentiment engine (TextBlob library) to analyze the sentiment of the reviews.

[0786] For example, the text "This product is great! I'm really happy with my purchase" would be classified as "positive."

[0787] 5. Generation probability adjustment:

[0788] The server corrects the generation probability based on the results of the emotion analysis.

[0789] If positive emotions are included, the probability of generation is reduced. For example, if the original generation probability was 0.35, the corrected generation probability becomes 0.31.

[0790] 6. Calculating the credibility score:

[0791] The server calculates a credibility score based on the corrected generation probability.

[0792] Specifically, the score is calculated using the formula "(1 - generation probability) 100". If the corrected generation probability is "0.31", the credibility score will be "69".

[0793] 7. Result output:

[0794] The server provides the calculated credibility score to the reviewer.

[0795] Reviewers use this score to judge the credibility of reviews posted by users.

[0796] Specific examples

[0797] Example prompt sentence:

[0798] "Is this text generated by AI? This product is excellent. I use it every day and have had no issues."

[0799] Example response:

[0800] "0.35"

[0801] This will increase the credibility of reviews posted by users and improve the reliability of reviews on online shopping sites. Reviewers will be able to provide information to users based on highly reliable reviews.

[0802] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0803] Step 1:

[0804] Users can write and submit reviews

[0805] A user accesses a review input form on an online shopping site and enters a review. For example, they might enter, "This product is great! I'm really happy with my purchase." After entering the review, the user clicks the "Submit" button to send the review to the server. At this point, the input is the review text entered by the user, and the output is the review text sent to the server.

[0806] Step 2:

[0807] Server receives review text

[0808] The server receives the review text sent from the user's device. It analyzes the data sent as an HTTP request and extracts the review text. At this point, the input is the HTTP request from the user's device, and the output is the review text.

[0809] Step 3:

[0810] Obtaining generation probability

[0811] The server sends the received review text to the OpenAI API and obtains the generation probability. Specifically, it calls the OpenAI API and sends the review text including the prompt "Is this text generated by AI?". The OpenAI API returns a numerical value indicating the probability that the review text was generated by the generation AI. At this point, the input is the review text, and the output is a numerical value indicating the generation probability.

[0812] Step 4:

[0813] Performing sentiment analysis

[0814] The server uses a sentiment analysis library such as TextBlob to analyze the sentiment of the review text. It identifies positive, negative, or neutral sentiment elements in the review text and classifies the sentiment. At this point, the input is the review text, and the output is the sentiment classification result (positive, negative, neutral, etc.).

[0815] Step 5:

[0816] Generation probability adjustment

[0817] The server corrects the obtained generation probability based on the results of emotion analysis. For example, if a positive emotion is included, the generation probability is reduced to increase credibility. This correction is performed by a numerical calculation using the emotion analysis results and generation probability. The input at this point is the generation probability and the emotion classification result, and the output is the corrected generation probability.

[0818] Step 6:

[0819] Calculating the credibility score

[0820] The server calculates the credibility score based on the adjusted generation probability. Specifically, it uses the formula "(1 - generation probability) 100". If the adjusted generation probability is 0.31, the credibility score is 69. At this point, the input is the adjusted generation probability, and the output is the credibility score.

[0821] Step 7:

[0822] Outputting the results and providing them to reviewers

[0823] The server provides the calculated credibility score to the reviewer. The reviewer checks this score and judges the credibility of the review. The credibility score and the review text are displayed on the reviewer's terminal. At this point, the input is the credibility score, and the output is the credibility score and review text displayed on the reviewer's terminal.

[0824] 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.

[0825] 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.

[0826] 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.

[0827] [Third embodiment]

[0828] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0829] 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.

[0830] 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).

[0831] 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.

[0832] 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.

[0833] 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).

[0834] 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.

[0835] 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.

[0836] 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.

[0837] 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.

[0838] 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.

[0839] 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."

[0840] This invention is a system that analyzes long-form content entered by a user, obtains the probability of generation by a generative AI model, and calculates and outputs a credibility score based on that probability. This system makes it possible to efficiently distinguish between content created by a user and content automatically generated by a generative AI.

[0841] Embodiments of the invention

[0842] System Overview

[0843] 1. User Input

[0844] The user inputs long content such as a reason for applying into an application form or submission form. The user's device sends this input to the server.

[0845] 2. Initial Server Setup

[0846] The server receives input from the user and initializes the OpenAI API, which is used to analyze the user's input and evaluate the likelihood that the text was generated by a generative AI.

[0847] 3. Parsing User Input

[0848] The server analyzes the text received from the user and obtains the generation probability from the generative AI model. For this process, it uses the OpenAI API and extracts the generation probability from the response.

[0849] 4. Calculating the credibility score

[0850] The server calculates a credibility score based on the obtained probability of creation, which is calculated using the formula (1 - probability of creation) 100.

[0851] 5. Outputting the results

[0852] The calculated credibility score is returned to the reviewer or user's device, and the reviewer uses this score to determine the credibility of the submitted content.

[0853] System processing flow

[0854] The user enters long-form content and submits it to the server

[0855] For example, a user enters their motivation for applying into an application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[0856] The server receives user input and initializes the OpenAI API.

[0857] The server stores the text data received from the user and sets the OpenAI API key to prepare the API.

[0858] The server parses the user input

[0859] The server calls the check_authenticity function and sends the user's input text to the OpenAI API to obtain the generation probability. This process uses the following parameters: engine (e.g., text-davinci-003), prompt (a sentence containing the text to be evaluated and the evaluation instructions), maximum number of tokens, and temperature parameter.

[0860] The server analyzes the generation probability

[0861] The server receives the response from the OpenAI API, parses it, and extracts the generated probability using regular expressions, converting the extracted probability value into a floating-point number between 0 and 1.

[0862] The server calculates the credibility score

[0863] The server calculates the credibility score using the calculate_score function. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[0864] The server returns the results

[0865] The server returns the calculated credibility score to the reviewer or user's device, and the reviewer uses this score to judge the credibility of the submitted sentence.

[0866] Specific example explanation

[0867] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server analyzes this probability and calculates a credibility score of 34.6. This score is returned to the reviewer or user's device to help determine whether the submitted statement is the user's own.

[0868] In this way, the present invention enables efficient distinction between content created by users and content generated by AI, reducing the burden on reviewers.

[0869] The processing flow will be explained below.

[0870] Step 1:

[0871] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[0872] Step 2:

[0873] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[0874] Step 3:

[0875] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[0876] Step 4:

[0877] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[0878] Step 5:

[0879] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[0880] Engine:text-davinci-003

[0881] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[0882] Maximum number of tokens: 60

[0883] Temperature parameter: 0.5

[0884] Step 6:

[0885] The server sends the created request to the OpenAI API and waits for a response.

[0886] Step 7:

[0887] The server receives the response returned from the OpenAI API, which includes the generation probability.

[0888] Step 8:

[0889] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[0890] Step 9:

[0891] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[0892] Step 10:

[0893] The server calls the calculate_score function to calculate the credibility score based on the obtained generation probability, using the formula (1 - generation probability) 100.

[0894] Step 11:

[0895] The server generates a result containing the calculated credibility score as an HTTP response and sends it back to the reviewer or device.

[0896] Step 12:

[0897] The terminal displays the credibility score received from the server, and the reviewer uses this to judge the credibility of the submitted statement.

[0898] Example 1

[0899] 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."

[0900] It is difficult to efficiently distinguish whether user-created long-form content was created manually or automatically generated by a generative AI model. This makes credibility assessment difficult and increases the burden on reviewers. The present invention aims to solve this problem by providing a system that accurately and efficiently evaluates the credibility of user-created content.

[0901] 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.

[0902] In this invention, the server includes a means for receiving long-form content input by a user, a means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, and a means for calculating a credibility score based on the obtained generation probability. This makes it possible to evaluate the credibility of content input by a user with high accuracy and efficiently determine whether a submitted document was written by the user himself / herself.

[0903] "User" refers to an individual or group of people who input long-form content into the system.

[0904] "Long content" refers to text data entered by a user that has a certain number of characters or more.

[0905] "Means for receiving" refers to the interface or process by which the server obtains the data entered by the user.

[0906] "Means for analyzing" refers to algorithms or programs for evaluating and analyzing received long-form content.

[0907] A "generative AI model" refers to a system or software that uses artificial intelligence technology to generate text.

[0908] "Generation probability" refers to a number that indicates the likelihood that a generative AI model generated a given piece of text.

[0909] The "credibility score" refers to an index that indicates whether the text was created by the user himself or herself, calculated based on the probability of creation.

[0910] "Means for outputting" refers to an interface or method for providing the calculated credibility score to a user or reviewer.

[0911] "Means of initializing the API" refers to the process of preparing the API for use with the API key and settings of the generated AI model.

[0912] "Means for setting prompt sentences" refers to a method for forming input sentences or instructions to make analysis requests to a generative AI model.

[0913] "Means of analyzing the response" refers to the process of interpreting the response data obtained from the generative AI model and extracting the necessary information.

[0914] "Means for extracting probability values" refers to a method or algorithm for extracting generation probabilities from response data.

[0915] "Terminal" refers to an electronic device used by a user, such as a computer, smartphone, or tablet.

[0916] "Means for evaluating" refers to the set of processes or calculations required to determine the authenticity of a received text.

[0917] The present invention is a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. The following describes in detail an embodiment of the present invention.

[0918] System configuration

[0919] User Input

[0920] A user uses their device (computer, smartphone, tablet, etc.) to enter long-form content into an application or submission form, such as a motivation letter or essay, and clicks the "Submit" button. This entered text data is sent by the device to the server via an HTTP POST request.

[0921] Initial Server Configuration

[0922] The server receives the text data sent from the user's device. Then, the server initializes the generative AI model (e.g., OpenAI API) by creating an API instance using the API key obtained from the configuration file.

[0923] Parsing User Input

[0924] The server calls the check_authenticity function to parse the text entered by the user. This function sets a prompt containing the text to be evaluated and evaluation instructions. Here is an example prompt:

[0925] _Example prompt sentence:_

[0926] "Please rate the likelihood that the following text was generated by a generative AI."

[0927] Obtaining generation probability

[0928] The server sends a request to the OpenAI API endpoint containing the prompt and settings (engine: e.g., text-davinci-003, maximum number of tokens, temperature parameters, etc.) to obtain a generation probability indicating the likelihood that the user's input was generated by the generative AI.

[0929] Analysis of generation probability

[0930] The server parses the response received from the API and extracts the generation probability. In this process, it extracts the probability value from the response JSON data and converts it to a floating point number between 0 and 1 using regular expressions if necessary.

[0931] Calculating the credibility score

[0932] The server calculates the credibility score based on the obtained generation probability. Specifically, it uses the following formula:

[0933] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[0934] For example, if the generation probability is 0.654, the credibility score is 34.6.

[0935] Output of results

[0936] The server returns the calculated credibility score to the reviewer or user's device, allowing the reviewer to evaluate the credibility of the submitted content.

[0937] Specific examples

[0938] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the API responds with "Generation probability 65.4%," the server analyzes this probability and calculates an authenticity score of 34.6. By returning this score to the reviewer or user's device, it can determine whether the submitted document was created by the user themselves.

[0939] In this way, the present invention provides a highly accurate and efficient evaluation system for effectively distinguishing between user-generated content and content generated by a generation AI.

[0940] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0941] Explanation of the processing flow

[0942] Step 1:

[0943] A user enters long-form content into an application or submission form on their own device and clicks the "Send" button. The entered text data is sent from the device to the server. The input data includes the long-form content (e.g., reason for applying) and the user ID. The server then receives an HTTP POST request and extracts the text data.

[0944] Step 2:

[0945] The server saves the received text data and initializes the API of the generative AI model. Specifically, the server obtains the API key from the configuration file and configures the API. Once the API key is correctly configured, the generative AI model is ready to use. The API key is included as input data. The output is an initialized API instance.

[0946] Step 3:

[0947] The server calls the check_authenticity function to analyze the user's input text. This function sets a prompt and generates an analysis request. The prompt includes instructions such as "Please rate the likelihood that the following text was generated by a generative AI." The input data includes the text data and the prompt. The output is an analysis request.

[0948] Step 4:

[0949] The server sends the generated analysis request to the generative AI model endpoint. The request includes settings such as the engine (e.g., text-davinci-003), maximum number of tokens, and temperature parameters. The analysis request and API settings are included as input data. This sends an API request to obtain the generation probability.

[0950] Step 5:

[0951] The server receives the response from the generative AI model and parses the response data. It extracts the generation probability from the JSON-formatted response data. In this process, it uses a regular expression to extract the generation probability and converts it into a floating-point number ranging from 0 to 1. The input data includes the response JSON data. The output is the extracted generation probability.

[0952] Step 6:

[0953] The server calculates the credibility score based on the obtained generation probability using the following formula:

[0954] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[0955] For example, if the generation probability is 0.654, the credibility score is 34.6. The input data includes the generation probability. The output is the calculated credibility score.

[0956] Step 7:

[0957] The server returns the calculated credibility score to the reviewer or user's device. It generates an HTTP response and sends information including the credibility score and related metadata to the user's device. The credibility score and user ID are included as input data, allowing the user or reviewer to confirm the credibility of the submitted document.

[0958] (Application example 1)

[0959] 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."

[0960] There is a challenge in determining with high accuracy whether long-form content sent by users was created by generative AI. There is also a need for a means to quickly and efficiently evaluate the authenticity of emails, documents, etc. In addition, users need help making more reliable judgments by specifically visualizing the authenticity evaluation based on generation probability.

[0961] 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.

[0962] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing the input text and using an analysis algorithm and the generation probability to calculate the credibility score, and means for displaying the credibility score on an electronic device. This makes it possible to quickly evaluate the credibility of long-form content received by a user with high accuracy and visualize the credibility score.

[0963] "Long content" refers to text data entered by a user, and mainly includes the contents of emails, documents, and the like.

[0964] A "generative AI model" is a model that can generate new text using an artificial intelligence algorithm trained on large amounts of text data.

[0965] "Generation probability" is a number that indicates the likelihood that a particular piece of text was generated by a generative AI model.

[0966] The "credibility score" is a numerical value for evaluating the reliability of the text entered by the user based on the generation probability.

[0967] An "analysis algorithm" is a calculation method for analyzing input text and calculating its generation probability.

[0968] "Electronic device" refers to a hardware device that can execute a program and display the results, and specifically includes smartphones, tablets, personal computers, etc.

[0969] "Server" means a network-enabled computer system that processes data received from users and calculates and outputs credibility scores.

[0970] The present invention provides a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. Specific embodiments for implementing the present invention are described below.

[0971] System configuration

[0972] This system mainly consists of the following hardware and software:

[0973] Server: A network-enabled computer system that processes data received from users and calculates and outputs credibility scores. The software used is the Python language and the OpenAI API.

[0974] User device: Any device on which a user enters long-form content, including smartphones, tablets, and PCs.

[0975] Electronic Device: A hardware device for displaying the credibility score. This can be a user terminal.

[0976] Processing flow

[0977] 1. User input:

[0978] The user inputs long-form content such as emails and documents through the device. For example, they input a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills."

[0979] 2. Receiving long-form content:

[0980] The long text content entered by the user is sent from the terminal to the server, which receives the text data and stores it in a database.

[0981] 3. Obtaining generation probabilities:

[0982] The server analyzes the stored text data and obtains the probability of generation using a generative AI model (e.g., OpenAI's text-davinci-003). The prompt used is, "Please indicate the probability that the following text was generated by the generative AI: I want to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[0983] 4. Calculating the credibility score:

[0984] The Python regular expression library "re" is used to analyze the generation probability, and a credibility score is calculated based on the obtained generation probability. For example, if the generation probability is 65.4%, the credibility score is "(1 - 0.654) 100 = 34.6". The calculated credibility score is stored in a database in a specific format.

[0985] 5. Output of credibility score:

[0986] Once the calculation is complete, the server sends the authenticity score to the user's device, where it can be displayed to visually confirm the authenticity of the email or document.

[0987] Specific example explanation

[0988] For example, if a user enters the following as their motivation for applying: "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," this text is sent to the server. The server uses the OpenAI API to obtain the probability of generating this text, and if the generation probability is 65.4%, the credibility score will be 34.6. This credibility score is sent to the user's device, and the user can judge the credibility of the document based on the displayed credibility score.

[0989] In this way, the system of the present invention helps users to quickly and accurately assess the credibility of long-form content.

[0990] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0991] Step 1:

[0992] Collect and send user-supplied long-form content

[0993] A user inputs long content such as an email or document through a terminal and clicks the send button. The input text is sent to the server as an HTTP request. The input may include, for example, a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills." The text data is sent to the server in JSON format.

[0994] Step 2:

[0995] The server receives and stores the long-form content.

[0996] The server receives an HTTP request sent from a user terminal. This request contains long text content, and the server stores this text data in a database. Specifically, the server adds the text data as a record in a database table.

[0997] Step 3:

[0998] Construct prompt sentences for analyzing production probability

[0999] The server analyzes the input text data and constructs a prompt to obtain the generation probability. The specific prompt will be in the form of "Please indicate the probability that the following text was generated by the generative AI: [input text]."

[1000] Step 4:

[1001] Obtaining generation probabilities using OpenAI's API

[1002] The server sends the constructed prompt to the OpenAI API and obtains the generation probability. To do this, the server uses the Python requests library to send an API request and receives the generation probability as a response. The data sent to the API includes the engine (e.g., text-davinci-003), prompt, maximum number of tokens, temperature parameters, etc.

[1003] Step 5:

[1004] Extract generation probabilities from responses and calculate credibility scores

[1005] The server parses the response received from the OpenAI API and extracts the generation probability using a regular expression. After obtaining the generation probability, it calculates a credibility score based on it. For example, if the generation probability is 65.4%, the credibility score is (1 - 0.654) 100 = 34.6. This score is then stored in a separate database table.

[1006] Step 6:

[1007] Send and display the credibility score on the user's device

[1008] The server returns the calculated credibility score to the user device as an HTTP response. The user device displays the received credibility score so that the user can confirm it. For example, the device UI might display "Credibility score: 34.6."

[1009] Specific example explanation

[1010] When a user enters the reason for wanting to work for your company as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," the server receives and stores this text. The server then sends the prompt text to the OpenAI API to obtain the generation probability, and calculates a credibility score based on the result. If the calculated credibility score is 34.6, the score is sent back to the user's device so that the user can view it on the screen.

[1011] This process flow allows the user to efficiently evaluate the credibility of long-form content.

[1012] 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.

[1013] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[1014] Embodiments of the invention

[1015] System Overview

[1016] 1. User Input

[1017] The user enters long content such as motivation for applying and application documents into the application form and clicks the submit button. The user's device then sends this input data to the server.

[1018] 2. Initial Server Setup

[1019] The server receives the long-form content sent by the user, sets the OpenAI API key, and prepares the text for analysis.

[1020] 3. Parsing User Input

[1021] The server sends the user's input text to the OpenAI API to obtain a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[1022] 4. Analysis by Emotion Engine

[1023] The server analyzes the emotions in the user's input text using an emotion engine, which uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[1024] 5. Correction of generation probability

[1025] Based on the analysis results of the emotion engine, the server adjusts the generation probability. For example, if the text is highly emotional, the generation probability is reduced to obtain a more credible score.

[1026] 6. Calculating the credibility score

[1027] The server calculates the credibility score based on the adjusted probability of creation, using the formula (1 - probability of creation) 100.

[1028] 7. Outputting the results

[1029] The server then sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[1030] System processing flow

[1031] The user enters long-form content and submits it to the server

[1032] The user enters their motivation for applying into the application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[1033] The server receives user input and initializes the OpenAI API.

[1034] The server receives the text data sent by the user, sets the OpenAI API key, and prepares the API.

[1035] The server parses the user input

[1036] The server uses the check_authenticity function to send the user's input text to the OpenAI API and obtain the generation probability.

[1037] Text analysis with emotion engine

[1038] The server uses an emotion engine to extract emotional components from the text and classify it, for example, classifying the text as "highly emotional," "neutral," or "positive."

[1039] Generation probability adjustment

[1040] The server adjusts the generated probabilities based on the results of the emotion engine, which allows the probabilities to reflect the emotional content of the text, resulting in a more accurate credibility score.

[1041] Calculating the credibility score

[1042] The server calculates the credibility score using the calculate_score function based on the corrected generation probability. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[1043] Output of results

[1044] The server then sends the calculated credibility score to the reviewer or user's device, who then uses this score to determine the credibility of the submitted sentence.

[1045] Specific example explanation

[1046] The user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjusts the generation probability based on the analysis results. The adjusted generation probability and credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer determines whether the submitted statement was written by the user themselves.

[1047] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[1048] The processing flow will be explained below.

[1049] Step 1:

[1050] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[1051] Step 2:

[1052] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[1053] Step 3:

[1054] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[1055] Step 4:

[1056] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[1057] Step 5:

[1058] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[1059] Engine:text-davinci-003

[1060] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[1061] Maximum number of tokens: 60

[1062] Temperature parameter: 0.5

[1063] Step 6:

[1064] The server sends the created request to the OpenAI API and waits for a response.

[1065] Step 7:

[1066] The server receives the response returned from the OpenAI API, which includes the generation probability.

[1067] Step 8:

[1068] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[1069] Step 9:

[1070] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[1071] Step 10:

[1072] The server uses an emotion engine to analyze the text data entered by the user and extract emotional elements from the text. The emotion engine uses natural language processing technology to perform emotion analysis and classify emotions as positive, negative, neutral, etc.

[1073] Step 11:

[1074] The server adjusts the generation probability based on the analysis results of the emotion engine. Specifically, if the emotion analysis results indicate an abnormally high emotional intensity, the server adjusts the generation probability to more accurately evaluate the credibility.

[1075] Step 12:

[1076] The server calculates the credibility score using the calculate_score function based on the adjusted probability of generation, using the formula (1 - probability of generation) 100.

[1077] Step 13:

[1078] The server generates a result containing the calculated credibility score as an HTTP response and sends it to the reviewer or device.

[1079] Step 14:

[1080] The terminal displays the credibility score received from the server, and the reviewer uses this score to judge the credibility of the sentence submitted by the user.

[1081] Example 2

[1082] 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."

[1083] Conventional systems have difficulty accurately assessing the credibility of long content input by users. Furthermore, simply obtaining the generation probability using a generative AI model does not take into account internal information such as emotional factors, making accurate credibility assessment difficult. Furthermore, because the credibility score is calculated without taking emotional factors into account, the system is vulnerable to the evaluation of text whose emotions have been intentionally manipulated.

[1084] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1085] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for analyzing the received long-form content using an emotion engine and classifying emotional elements, means for correcting the generation probability based on the emotion analysis result, means for calculating a credibility score based on the corrected generation probability, and means for outputting the calculated credibility score. This enables a more accurate and objective credibility assessment of the content input by the user by combining the generation probability by the generative AI model and the emotion analysis result.

[1086] "User" refers to an individual or company that uses the system to input and submit long-form content.

[1087] "Long-form content" refers to relatively long textual data such as motivation letters and application documents.

[1088] "Generative AI model" refers to an artificial intelligence algorithm that calculates the probability of generated text.

[1089] "Generation probability" refers to the quantification of the likelihood that a particular piece of text was generated by a generative AI model.

[1090] "Credibility score" refers to a numerical value that indicates the reliability of a text, calculated based on the probability of its generation.

[1091] "Emotion engine" refers to natural language processing technology for analyzing and classifying emotional elements in text.

[1092] "Sentiment analysis" refers to the process of extracting emotional elements from text and classifying them as positive, negative, neutral, etc.

[1093] "Correct" refers to adjusting the generation probability based on the results of sentiment analysis.

[1094] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[1095] The system of the present invention uses the following hardware and software: The hardware depends on the user's terminal and server, and the software uses the OpenAI API and emotion engine (which uses natural language processing technology).

[1096] First, the user enters long text content, such as their motivation for applying and application documents, into the application form and clicks the submit button. The user's device sends this input data to the server. The server receives the long text content sent by the user and simultaneously sets an OpenAI API key and prepares for text analysis.

[1097] The server then sends the user's input text to the OpenAI API to obtain a generation probability, which is a numerical value indicating the likelihood that the text was generated by generative AI. The server then uses an emotion engine to analyze the emotion in the user's input text. The emotion engine uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[1098] Next, the server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, it will reduce the generation probability to obtain a higher credibility score. The server then calculates the credibility score based on the adjusted generation probability. This calculation is performed using the formula (1 - generation probability) 100.

[1099] Finally, the server sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[1100] Specific example explanation

[1101] As a specific example of operation, a user enters the following reason for applying into the application form and submits it.

[1102] "I am looking to grow through my experience with your company. I am interested in the job description and feel that it will allow me to utilize my skills."

[1103] The server receives this text and uses the OpenAI API to obtain the probability of its creation. For example, if the returned probability of creation is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjust the probability of its creation based on the analysis results. The adjusted probability of creation and a credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer can determine whether the submitted sentence was written by the user.

[1104] Example prompt sentence:

[1105] "The user provided the following motivation:

[1106] "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[1107] Find the probability of generating this text, perform sentiment analysis, and calculate a credibility score."

[1108] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[1109] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1110] Step 1:

[1111] The user enters long content such as motivation for applying and application documents into the application form and clicks the "Submit" button. The device sends an HTTP request including the entered text to the server.

[1112] Input: Long-form content entered into the application form

[1113] Output: HTTP request sent to the server

[1114] Step 2:

[1115] The server receives the text data sent by the user as an HTTP request.

[1116] Input: HTTP request from the terminal

[1117] Output: Received text data

[1118] Step 3:

[1119] The server retrieves the OpenAI API key from a configuration file or environment variable and initializes the API.

[1120] Input: API key stored in a config file or environment variable

[1121] Output: Initialized OpenAI API

[1122] Step 4:

[1123] The server calls the check_authenticity function to send the user's input text to the OpenAI API, which formats the text into an API request and sends it to the OpenAI API.

[1124] Input: User-entered text

[1125] Output: API request sent to the OpenAI API

[1126] Step 5:

[1127] The server receives a response from the OpenAI API that includes a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[1128] Input: Response from OpenAI API

[1129] Output: Generation probability

[1130] Step 6:

[1131] The server uses an emotion engine to extract emotional elements from the text and classify them into categories such as "positive," "negative," and "neutral."

[1132] Input: User-entered text

[1133] Output: Emotion analysis results

[1134] Step 7:

[1135] The server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, the generation probability is reduced by a certain percentage.

[1136] Input: Sentiment analysis results and generation probability

[1137] Output: Corrected generation probabilities

[1138] Step 8:

[1139] The server calculates the credibility score based on the adjusted probability of creation, using the formula "(1 - probability of creation) 100".

[1140] Input: Corrected generation probability

[1141] Output: Belief score

[1142] Step 9:

[1143] The server sends the calculated credibility score to the reviewer or user's device, and sends information such as the credibility score, the corrected generation probability, and the sentiment analysis results via an HTTP response.

[1144] Input: Belief score, corrected generation probability, sentiment analysis result

[1145] Output: Results sent to users and reviewers

[1146] Through the above processing steps, the present invention efficiently distinguishes between content created by users and content generated by generation AI, reducing the burden on reviewers and enabling more accurate credibility assessments.

[1147] (Application example 2)

[1148] 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."

[1149] There is a need for efficient and accurate evaluation of the credibility of reviews posted by users on online shopping sites. However, there are many fraudulent reviews generated by AI and reviews that are misleading based on emotions, and appropriate countermeasures against these are lacking. As a result, it is difficult for buyers to select products based on trustworthy reviews. Therefore, an objective of the present invention is to provide a system that increases the credibility of posted reviews, allowing users to refer to reviews with peace of mind.

[1150] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1151] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing emotions in the long-form content, means for correcting the generation probability based on the result of the emotion analysis, and means for providing the credibility score of the long-form content to a reviewer. This makes it possible to increase the credibility of reviews posted by users and provide highly reliable reviews.

[1152] "User" refers to a person or end user of the System.

[1153] "Long-form content" refers to long texts entered by users, such as motivation for applying, application documents, and reviews.

[1154] "Means for receiving" refers to a method or device by which the server receives data sent from the user.

[1155] The term "analyzing means" refers to a method or device for analyzing the content of received long-form content.

[1156] "Generation probability" refers to the probability that a particular piece of text could have been generated by a generative AI.

[1157] "Means for obtaining generation probability" refers to a method or algorithm for calculating the generation probability of a text.

[1158] "Authenticity Score" refers to a numerical value used to assess the veracity of long-form content.

[1159] "Means for calculating" refers to a method or apparatus for calculating a credibility score based on the generation probability.

[1160] "Outputting means" refers to a method or device that displays the calculated credibility score.

[1161] "Means for analyzing emotions" refers to a method or device for analyzing emotions contained within long-form content.

[1162] "Means for correcting generation probability" refers to a method or device for adjusting generation probability based on the results of sentiment analysis.

[1163] "Reviewer" refers to a person or system that checks the credibility score and determines the veracity of posted content.

[1164] "Means for providing" refers to a method or device for providing the credibility score to the reviewer.

[1165] The system embodying this invention analyzes long-form content entered by users and evaluates its credibility. This system corrects the credibility score by combining generation probability obtained by a generative AI model and emotion analysis by an emotion engine.

[1166] System configuration

[1167] Hardware

[1168] server

[1169] User terminal

[1170] Reviewer device

[1171] software

[1172] OpenAI API

[1173] Emotion engine (natural language processing library such as TextBlob)

[1174] Database

[1175] System processing flow

[1176] 1. User input:

[1177] The user accesses the review submission page of the online shopping site and enters a review.

[1178] For example, a user might enter, "This product is great! I'm really happy with my purchase."

[1179] After completing the input, the user clicks the "Submit button."

[1180] 2. Data reception:

[1181] The user terminal sends an HTTP request containing the entered text to the server.

[1182] The server receives this data.

[1183] 3. Obtaining generation probabilities:

[1184] The server initializes the OpenAI API and sends the reviews entered by the user.

[1185] The OpenAI API returns the probability of generating the text. For example, if the generation probability is "0.35", you will get this value.

[1186] 4. Emotion analysis:

[1187] The server uses a sentiment engine (TextBlob library) to analyze the sentiment of the reviews.

[1188] For example, the text "This product is great! I'm really happy with my purchase" would be classified as "positive."

[1189] 5. Generation probability adjustment:

[1190] The server corrects the generation probability based on the results of the emotion analysis.

[1191] If positive emotions are included, the probability of generation is reduced. For example, if the original generation probability was 0.35, the corrected generation probability becomes 0.31.

[1192] 6. Calculating the credibility score:

[1193] The server calculates a credibility score based on the corrected generation probability.

[1194] Specifically, the score is calculated using the formula "(1 - generation probability) 100". If the corrected generation probability is "0.31", the credibility score will be "69".

[1195] 7. Result output:

[1196] The server provides the calculated credibility score to the reviewer.

[1197] Reviewers use this score to judge the credibility of reviews posted by users.

[1198] Specific examples

[1199] Example prompt sentence:

[1200] "Is this text generated by AI? This product is excellent. I use it every day and have had no issues."

[1201] Example response:

[1202] "0.35"

[1203] This will increase the credibility of reviews posted by users and improve the reliability of reviews on online shopping sites. Reviewers will be able to provide information to users based on highly reliable reviews.

[1204] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1205] Step 1:

[1206] Users can write and submit reviews

[1207] A user accesses a review input form on an online shopping site and enters a review. For example, they might enter, "This product is great! I'm really happy with my purchase." After entering the review, the user clicks the "Submit" button to send the review to the server. At this point, the input is the review text entered by the user, and the output is the review text sent to the server.

[1208] Step 2:

[1209] Server receives review text

[1210] The server receives the review text sent from the user's device. It analyzes the data sent as an HTTP request and extracts the review text. At this point, the input is the HTTP request from the user's device, and the output is the review text.

[1211] Step 3:

[1212] Obtaining generation probability

[1213] The server sends the received review text to the OpenAI API and obtains the generation probability. Specifically, it calls the OpenAI API and sends the review text including the prompt "Is this text generated by AI?". The OpenAI API returns a numerical value indicating the probability that the review text was generated by the generation AI. At this point, the input is the review text, and the output is a numerical value indicating the generation probability.

[1214] Step 4:

[1215] Performing sentiment analysis

[1216] The server uses a sentiment analysis library such as TextBlob to analyze the sentiment of the review text. It identifies positive, negative, or neutral sentiment elements in the review text and classifies the sentiment. At this point, the input is the review text, and the output is the sentiment classification result (positive, negative, neutral, etc.).

[1217] Step 5:

[1218] Generation probability adjustment

[1219] The server corrects the obtained generation probability based on the results of emotion analysis. For example, if a positive emotion is included, the generation probability is reduced to increase credibility. This correction is performed by a numerical calculation using the emotion analysis results and generation probability. The input at this point is the generation probability and the emotion classification result, and the output is the corrected generation probability.

[1220] Step 6:

[1221] Calculating the credibility score

[1222] The server calculates the credibility score based on the adjusted generation probability. Specifically, it uses the formula "(1 - generation probability) 100". If the adjusted generation probability is 0.31, the credibility score is 69. At this point, the input is the adjusted generation probability, and the output is the credibility score.

[1223] Step 7:

[1224] Outputting the results and providing them to reviewers

[1225] The server provides the calculated credibility score to the reviewer. The reviewer checks this score and judges the credibility of the review. The credibility score and the review text are displayed on the reviewer's terminal. At this point, the input is the credibility score, and the output is the credibility score and review text displayed on the reviewer's terminal.

[1226] 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.

[1227] 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.

[1228] 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.

[1229] [Fourth embodiment]

[1230] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1231] 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.

[1232] 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).

[1233] 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.

[1234] 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.

[1235] 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).

[1236] 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.

[1237] 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.

[1238] 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.

[1239] 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.

[1240] 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.

[1241] 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.

[1242] 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."

[1243] This invention is a system that analyzes long-form content entered by a user, obtains the probability of generation by a generative AI model, and calculates and outputs a credibility score based on that probability. This system makes it possible to efficiently distinguish between content created by a user and content automatically generated by a generative AI.

[1244] Embodiments of the invention

[1245] System Overview

[1246] 1. User Input

[1247] The user inputs long content such as a reason for applying into an application form or submission form. The user's device sends this input to the server.

[1248] 2. Initial Server Setup

[1249] The server receives input from the user and initializes the OpenAI API, which is used to analyze the user's input and evaluate the likelihood that the text was generated by a generative AI.

[1250] 3. Parsing User Input

[1251] The server analyzes the text received from the user and obtains the generation probability from the generative AI model. For this process, it uses the OpenAI API and extracts the generation probability from the response.

[1252] 4. Calculating the credibility score

[1253] The server calculates a credibility score based on the obtained probability of creation, which is calculated using the formula (1 - probability of creation) 100.

[1254] 5. Outputting the results

[1255] The calculated credibility score is returned to the reviewer or user's device, and the reviewer uses this score to determine the credibility of the submitted content.

[1256] System processing flow

[1257] The user enters long-form content and submits it to the server

[1258] For example, a user enters their motivation for applying into an application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[1259] The server receives user input and initializes the OpenAI API.

[1260] The server stores the text data received from the user and sets the OpenAI API key to prepare the API.

[1261] The server parses the user input

[1262] The server calls the check_authenticity function and sends the user's input text to the OpenAI API to obtain the generation probability. This process uses the following parameters: engine (e.g., text-davinci-003), prompt (a sentence containing the text to be evaluated and the evaluation instructions), maximum number of tokens, and temperature parameter.

[1263] The server analyzes the generation probability

[1264] The server receives the response from the OpenAI API, parses it, and extracts the generated probability using regular expressions, converting the extracted probability value into a floating-point number between 0 and 1.

[1265] The server calculates the credibility score

[1266] The server calculates the credibility score using the calculate_score function. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[1267] The server returns the results

[1268] The server returns the calculated credibility score to the reviewer or user's device, and the reviewer uses this score to judge the credibility of the submitted sentence.

[1269] Specific example explanation

[1270] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server analyzes this probability and calculates a credibility score of 34.6. This score is returned to the reviewer or user's device to help determine whether the submitted statement is the user's own.

[1271] In this way, the present invention enables efficient distinction between content created by users and content generated by AI, reducing the burden on reviewers.

[1272] The processing flow will be explained below.

[1273] Step 1:

[1274] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[1275] Step 2:

[1276] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[1277] Step 3:

[1278] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[1279] Step 4:

[1280] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[1281] Step 5:

[1282] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[1283] Engine:text-davinci-003

[1284] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[1285] Maximum number of tokens: 60

[1286] Temperature parameter: 0.5

[1287] Step 6:

[1288] The server sends the created request to the OpenAI API and waits for a response.

[1289] Step 7:

[1290] The server receives the response returned from the OpenAI API, which includes the generation probability.

[1291] Step 8:

[1292] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[1293] Step 9:

[1294] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[1295] Step 10:

[1296] The server calls the calculate_score function to calculate the credibility score based on the obtained generation probability, using the formula (1 - generation probability) 100.

[1297] Step 11:

[1298] The server generates a result containing the calculated credibility score as an HTTP response and sends it back to the reviewer or device.

[1299] Step 12:

[1300] The terminal displays the credibility score received from the server, and the reviewer uses this to judge the credibility of the submitted statement.

[1301] Example 1

[1302] 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."

[1303] It is difficult to efficiently distinguish whether user-created long-form content was created manually or automatically generated by a generative AI model. This makes credibility assessment difficult and increases the burden on reviewers. The present invention aims to solve this problem by providing a system that accurately and efficiently evaluates the credibility of user-created content.

[1304] 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.

[1305] In this invention, the server includes a means for receiving long-form content input by a user, a means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, and a means for calculating a credibility score based on the obtained generation probability. This makes it possible to evaluate the credibility of content input by a user with high accuracy and efficiently determine whether a submitted document was written by the user himself / herself.

[1306] "User" refers to an individual or group of people who input long-form content into the system.

[1307] "Long content" refers to text data entered by a user that has a certain number of characters or more.

[1308] "Means for receiving" refers to the interface or process by which the server obtains the data entered by the user.

[1309] "Means for analyzing" refers to algorithms or programs for evaluating and analyzing received long-form content.

[1310] A "generative AI model" refers to a system or software that uses artificial intelligence technology to generate text.

[1311] "Generation probability" refers to a number that indicates the likelihood that a generative AI model generated a given piece of text.

[1312] The "credibility score" refers to an index that indicates whether the text was created by the user himself or herself, calculated based on the probability of creation.

[1313] "Means for outputting" refers to an interface or method for providing the calculated credibility score to a user or reviewer.

[1314] "Means of initializing the API" refers to the process of preparing the API for use with the API key and settings of the generated AI model.

[1315] "Means for setting prompt sentences" refers to a method for forming input sentences or instructions to make analysis requests to a generative AI model.

[1316] "Means of analyzing the response" refers to the process of interpreting the response data obtained from the generative AI model and extracting the necessary information.

[1317] "Means for extracting probability values" refers to a method or algorithm for extracting generation probabilities from response data.

[1318] "Terminal" refers to an electronic device used by a user, such as a computer, smartphone, or tablet.

[1319] "Means for evaluating" refers to the set of processes or calculations required to determine the authenticity of a received text.

[1320] The present invention is a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. The following describes in detail an embodiment of the present invention.

[1321] System configuration

[1322] User Input

[1323] A user uses their device (computer, smartphone, tablet, etc.) to enter long-form content into an application or submission form, such as a motivation letter or essay, and clicks the "Submit" button. This entered text data is sent by the device to the server via an HTTP POST request.

[1324] Initial Server Configuration

[1325] The server receives the text data sent from the user's device. Then, the server initializes the generative AI model (e.g., OpenAI API) by creating an API instance using the API key obtained from the configuration file.

[1326] Parsing User Input

[1327] The server calls the check_authenticity function to parse the text entered by the user. This function sets a prompt containing the text to be evaluated and evaluation instructions. Here is an example prompt:

[1328] _Example prompt sentence:_

[1329] "Please rate the likelihood that the following text was generated by a generative AI."

[1330] Obtaining generation probability

[1331] The server sends a request to the OpenAI API endpoint containing the prompt and settings (engine: e.g., text-davinci-003, maximum number of tokens, temperature parameters, etc.) to obtain a generation probability indicating the likelihood that the user's input was generated by the generative AI.

[1332] Analysis of generation probability

[1333] The server parses the response received from the API and extracts the generation probability. In this process, it extracts the probability value from the response JSON data and converts it to a floating point number between 0 and 1 using regular expressions if necessary.

[1334] Calculating the credibility score

[1335] The server calculates the credibility score based on the obtained generation probability. Specifically, it uses the following formula:

[1336] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[1337] For example, if the generation probability is 0.654, the credibility score is 34.6.

[1338] Output of results

[1339] The server returns the calculated credibility score to the reviewer or user's device, allowing the reviewer to evaluate the credibility of the submitted content.

[1340] Specific examples

[1341] A user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the API responds with "Generation probability 65.4%," the server analyzes this probability and calculates an authenticity score of 34.6. By returning this score to the reviewer or user's device, it can determine whether the submitted document was created by the user themselves.

[1342] In this way, the present invention provides a highly accurate and efficient evaluation system for effectively distinguishing between user-generated content and content generated by a generation AI.

[1343] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1344] Explanation of the processing flow

[1345] Step 1:

[1346] A user enters long-form content into an application or submission form on their own device and clicks the "Send" button. The entered text data is sent from the device to the server. The input data includes the long-form content (e.g., reason for applying) and the user ID. The server then receives an HTTP POST request and extracts the text data.

[1347] Step 2:

[1348] The server saves the received text data and initializes the API of the generative AI model. Specifically, the server obtains the API key from the configuration file and configures the API. Once the API key is correctly configured, the generative AI model is ready to use. The API key is included as input data. The output is an initialized API instance.

[1349] Step 3:

[1350] The server calls the check_authenticity function to analyze the user's input text. This function sets a prompt and generates an analysis request. The prompt includes instructions such as "Please rate the likelihood that the following text was generated by a generative AI." The input data includes the text data and the prompt. The output is an analysis request.

[1351] Step 4:

[1352] The server sends the generated analysis request to the generative AI model endpoint. The request includes settings such as the engine (e.g., text-davinci-003), maximum number of tokens, and temperature parameters. The analysis request and API settings are included as input data. This sends an API request to obtain the generation probability.

[1353] Step 5:

[1354] The server receives the response from the generative AI model and parses the response data. It extracts the generation probability from the JSON-formatted response data. In this process, it uses a regular expression to extract the generation probability and converts it into a floating-point number ranging from 0 to 1. The input data includes the response JSON data. The output is the extracted generation probability.

[1355] Step 6:

[1356] The server calculates the credibility score based on the obtained generation probability using the following formula:

[1357] \[ \text{Credibility score} = (1 - \text{Generation probability}) \times 100 \]

[1358] For example, if the generation probability is 0.654, the credibility score is 34.6. The input data includes the generation probability. The output is the calculated credibility score.

[1359] Step 7:

[1360] The server returns the calculated credibility score to the reviewer or user's device. It generates an HTTP response and sends information including the credibility score and related metadata to the user's device. The credibility score and user ID are included as input data, allowing the user or reviewer to confirm the credibility of the submitted document.

[1361] (Application example 1)

[1362] 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."

[1363] There is a challenge in determining with high accuracy whether long-form content sent by users was created by generative AI. There is also a need for a means to quickly and efficiently evaluate the authenticity of emails, documents, etc. In addition, users need help making more reliable judgments by specifically visualizing the authenticity evaluation based on generation probability.

[1364] 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.

[1365] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing the input text and using an analysis algorithm and the generation probability to calculate the credibility score, and means for displaying the credibility score on an electronic device. This makes it possible to quickly evaluate the credibility of long-form content received by a user with high accuracy and visualize the credibility score.

[1366] "Long content" refers to text data entered by a user, and mainly includes the contents of emails, documents, and the like.

[1367] A "generative AI model" is a model that can generate new text using an artificial intelligence algorithm trained on large amounts of text data.

[1368] "Generation probability" is a number that indicates the likelihood that a particular piece of text was generated by a generative AI model.

[1369] The "credibility score" is a numerical value for evaluating the reliability of the text entered by the user based on the generation probability.

[1370] An "analysis algorithm" is a calculation method for analyzing input text and calculating its generation probability.

[1371] "Electronic device" refers to a hardware device that can execute a program and display the results, and specifically includes smartphones, tablets, personal computers, etc.

[1372] "Server" means a network-enabled computer system that processes data received from users and calculates and outputs credibility scores.

[1373] The present invention provides a system that analyzes long-form content input by a user, obtains generation probabilities using a generative AI model, and calculates and outputs credibility scores based on the probabilities. Specific embodiments for implementing the present invention are described below.

[1374] System configuration

[1375] This system mainly consists of the following hardware and software:

[1376] Server: A network-enabled computer system that processes data received from users and calculates and outputs credibility scores. The software used is the Python language and the OpenAI API.

[1377] User device: Any device on which a user enters long-form content, including smartphones, tablets, and PCs.

[1378] Electronic Device: A hardware device for displaying the credibility score. This can be a user terminal.

[1379] Processing flow

[1380] 1. User input:

[1381] The user inputs long-form content such as emails and documents through the device. For example, they input a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills."

[1382] 2. Receiving long-form content:

[1383] The long text content entered by the user is sent from the terminal to the server, which receives the text data and stores it in a database.

[1384] 3. Obtaining generation probabilities:

[1385] The server analyzes the stored text data and obtains the probability of generation using a generative AI model (e.g., OpenAI's text-davinci-003). The prompt used is, "Please indicate the probability that the following text was generated by the generative AI: I want to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[1386] 4. Calculating the credibility score:

[1387] The Python regular expression library "re" is used to analyze the generation probability, and a credibility score is calculated based on the obtained generation probability. For example, if the generation probability is 65.4%, the credibility score is "(1 - 0.654) 100 = 34.6". The calculated credibility score is stored in a database in a specific format.

[1388] 5. Output of credibility score:

[1389] Once the calculation is complete, the server sends the authenticity score to the user's device, where it can be displayed to visually confirm the authenticity of the email or document.

[1390] Specific example explanation

[1391] For example, if a user enters the following as their motivation for applying: "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," this text is sent to the server. The server uses the OpenAI API to obtain the probability of generating this text, and if the generation probability is 65.4%, the credibility score will be 34.6. This credibility score is sent to the user's device, and the user can judge the credibility of the document based on the displayed credibility score.

[1392] In this way, the system of the present invention helps users to quickly and accurately assess the credibility of long-form content.

[1393] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1394] Step 1:

[1395] Collect and send user-supplied long-form content

[1396] A user inputs long content such as an email or document through a terminal and clicks the send button. The input text is sent to the server as an HTTP request. The input may include, for example, a sentence such as, "I would like to grow through my experience at your company. I am interested in the job content and feel that I can utilize my skills." The text data is sent to the server in JSON format.

[1397] Step 2:

[1398] The server receives and stores the long-form content.

[1399] The server receives an HTTP request sent from a user terminal. This request contains long text content, and the server stores this text data in a database. Specifically, the server adds the text data as a record in a database table.

[1400] Step 3:

[1401] Construct prompt sentences for analyzing production probability

[1402] The server analyzes the input text data and constructs a prompt to obtain the generation probability. The specific prompt will be in the form of "Please indicate the probability that the following text was generated by the generative AI: [input text]."

[1403] Step 4:

[1404] Obtaining generation probabilities using OpenAI's API

[1405] The server sends the constructed prompt to the OpenAI API and obtains the generation probability. To do this, the server uses the Python requests library to send an API request and receives the generation probability as a response. The data sent to the API includes the engine (e.g., text-davinci-003), prompt, maximum number of tokens, temperature parameters, etc.

[1406] Step 5:

[1407] Extract generation probabilities from responses and calculate credibility scores

[1408] The server parses the response received from the OpenAI API and extracts the generation probability using a regular expression. After obtaining the generation probability, it calculates a credibility score based on it. For example, if the generation probability is 65.4%, the credibility score is (1 - 0.654) 100 = 34.6. This score is then stored in a separate database table.

[1409] Step 6:

[1410] Send and display the credibility score on the user's device

[1411] The server returns the calculated credibility score to the user device as an HTTP response. The user device displays the received credibility score so that the user can confirm it. For example, the device UI might display "Credibility score: 34.6."

[1412] Specific example explanation

[1413] When a user enters the reason for wanting to work for your company as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills," the server receives and stores this text. The server then sends the prompt text to the OpenAI API to obtain the generation probability, and calculates a credibility score based on the result. If the calculated credibility score is 34.6, the score is sent back to the user's device so that the user can view it on the screen.

[1414] This process flow allows the user to efficiently evaluate the credibility of long-form content.

[1415] 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.

[1416] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[1417] Embodiments of the invention

[1418] System Overview

[1419] 1. User Input

[1420] The user enters long content such as motivation for applying and application documents into the application form and clicks the submit button. The user's device then sends this input data to the server.

[1421] 2. Initial Server Setup

[1422] The server receives the long-form content sent by the user, sets the OpenAI API key, and prepares the text for analysis.

[1423] 3. Parsing User Input

[1424] The server sends the user's input text to the OpenAI API to obtain a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[1425] 4. Analysis by Emotion Engine

[1426] The server analyzes the emotions in the user's input text using an emotion engine, which uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[1427] 5. Correction of generation probability

[1428] Based on the analysis results of the emotion engine, the server adjusts the generation probability. For example, if the text is highly emotional, the generation probability is reduced to obtain a more credible score.

[1429] 6. Calculating the credibility score

[1430] The server calculates the credibility score based on the adjusted probability of creation, using the formula (1 - probability of creation) 100.

[1431] 7. Outputting the results

[1432] The server then sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[1433] System processing flow

[1434] The user enters long-form content and submits it to the server

[1435] The user enters their motivation for applying into the application form and clicks the "Submit" button. The terminal then sends an HTTP request containing the entered text to the server.

[1436] The server receives user input and initializes the OpenAI API.

[1437] The server receives the text data sent by the user, sets the OpenAI API key, and prepares the API.

[1438] The server parses the user input

[1439] The server uses the check_authenticity function to send the user's input text to the OpenAI API and obtain the generation probability.

[1440] Text analysis with emotion engine

[1441] The server uses an emotion engine to extract emotional components from the text and classify it, for example, classifying the text as "highly emotional," "neutral," or "positive."

[1442] Generation probability adjustment

[1443] The server adjusts the generated probabilities based on the results of the emotion engine, which allows the probabilities to reflect the emotional content of the text, resulting in a more accurate credibility score.

[1444] Calculating the credibility score

[1445] The server calculates the credibility score using the calculate_score function based on the corrected generation probability. For example, if the generation probability is 0.654, the credibility score is (1 - 0.654) 100 = 34.6.

[1446] Output of results

[1447] The server then sends the calculated credibility score to the reviewer or user's device, who then uses this score to determine the credibility of the submitted sentence.

[1448] Specific example explanation

[1449] The user enters their motivation for applying into the application form, such as "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills.", and submits it. The server receives this text and uses the OpenAI API to obtain the generation probability. For example, if the returned generation probability is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjusts the generation probability based on the analysis results. The adjusted generation probability and credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer determines whether the submitted statement was written by the user themselves.

[1450] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[1451] The processing flow will be explained below.

[1452] Step 1:

[1453] The user enters long content, such as a statement of purpose, into the application form. The user then clicks the "Submit" button.

[1454] Step 2:

[1455] The terminal generates an HTTP request including the text data entered by the user and sends the request to the server.

[1456] Step 3:

[1457] The server receives the HTTP request sent from the terminal and extracts the text data entered by the user.

[1458] Step 4:

[1459] The server sets the OpenAI API key and prepares to use the API. The API key is set to authenticate with the OpenAI server.

[1460] Step 5:

[1461] The server makes a request to the OpenAI API to evaluate the text data entered by the user. The request includes the following parameters:

[1462] Engine:text-davinci-003

[1463] Prompt: A sentence containing the text to be evaluated and instructions for evaluation (e.g., "Is the following text generated by an AI? Evaluate the probability:")

[1464] Maximum number of tokens: 60

[1465] Temperature parameter: 0.5

[1466] Step 6:

[1467] The server sends the created request to the OpenAI API and waits for a response.

[1468] Step 7:

[1469] The server receives the response returned from the OpenAI API, which includes the generation probability.

[1470] Step 8:

[1471] The server calls the parse_probability function to extract the generation probability from the API response, specifically by parsing probability values ​​like "65.4%" using regular expressions.

[1472] Step 9:

[1473] The server converts the extracted generation probability into a floating point number in the range of 0 to 1 (e.g., 65.4% -> 0.654).

[1474] Step 10:

[1475] The server uses an emotion engine to analyze the text data entered by the user and extract emotional elements from the text. The emotion engine uses natural language processing technology to perform emotion analysis and classify emotions as positive, negative, neutral, etc.

[1476] Step 11:

[1477] The server adjusts the generation probability based on the analysis results of the emotion engine. Specifically, if the emotion analysis results indicate an abnormally high emotional intensity, the server adjusts the generation probability to more accurately evaluate the credibility.

[1478] Step 12:

[1479] The server calculates the credibility score using the calculate_score function based on the adjusted probability of generation, using the formula (1 - probability of generation) 100.

[1480] Step 13:

[1481] The server generates a result containing the calculated credibility score as an HTTP response and sends it to the reviewer or device.

[1482] Step 14:

[1483] The terminal displays the credibility score received from the server, and the reviewer uses this score to judge the credibility of the sentence submitted by the user.

[1484] Example 2

[1485] 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."

[1486] Conventional systems have difficulty accurately assessing the credibility of long content input by users. Furthermore, simply obtaining the generation probability using a generative AI model does not take into account internal information such as emotional factors, making accurate credibility assessment difficult. Furthermore, because the credibility score is calculated without taking emotional factors into account, the system is vulnerable to the evaluation of text whose emotions have been intentionally manipulated.

[1487] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1488] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for analyzing the received long-form content using an emotion engine and classifying emotional elements, means for correcting the generation probability based on the emotion analysis result, means for calculating a credibility score based on the corrected generation probability, and means for outputting the calculated credibility score. This enables a more accurate and objective credibility assessment of the content input by the user by combining the generation probability by the generative AI model and the emotion analysis result.

[1489] "User" refers to an individual or company that uses the system to input and submit long-form content.

[1490] "Long-form content" refers to relatively long textual data such as motivation letters and application documents.

[1491] "Generative AI model" refers to an artificial intelligence algorithm that calculates the probability of generated text.

[1492] "Generation probability" refers to the quantification of the likelihood that a particular piece of text was generated by a generative AI model.

[1493] "Credibility score" refers to a numerical value that indicates the reliability of a text, calculated based on the probability of its generation.

[1494] "Emotion engine" refers to natural language processing technology for analyzing and classifying emotional elements in text.

[1495] "Sentiment analysis" refers to the process of extracting emotional elements from text and classifying them as positive, negative, neutral, etc.

[1496] "Correct" refers to adjusting the generation probability based on the results of sentiment analysis.

[1497] This invention relates to a system that analyzes long-form content entered by a user and evaluates its credibility. In addition to obtaining generation probabilities using a generative AI model, this invention also incorporates an emotion engine that recognizes the user's emotions.

[1498] The system of the present invention uses the following hardware and software: The hardware depends on the user's terminal and server, and the software uses the OpenAI API and emotion engine (which uses natural language processing technology).

[1499] First, the user enters long text content, such as their motivation for applying and application documents, into the application form and clicks the submit button. The user's device sends this input data to the server. The server receives the long text content sent by the user and simultaneously sets an OpenAI API key and prepares for text analysis.

[1500] The server then sends the user's input text to the OpenAI API to obtain a generation probability, which is a numerical value indicating the likelihood that the text was generated by generative AI. The server then uses an emotion engine to analyze the emotion in the user's input text. The emotion engine uses natural language processing technology to extract emotional elements from the text and classify them as positive, negative, neutral, etc.

[1501] Next, the server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, it will reduce the generation probability to obtain a higher credibility score. The server then calculates the credibility score based on the adjusted generation probability. This calculation is performed using the formula (1 - generation probability) 100.

[1502] Finally, the server sends the calculated credibility score to the reviewer or user's device, allowing the reviewer to determine the credibility of the content submitted by the user based on this score.

[1503] Specific example explanation

[1504] As a specific example of operation, a user enters the following reason for applying into the application form and submits it.

[1505] "I am looking to grow through my experience with your company. I am interested in the job description and feel that it will allow me to utilize my skills."

[1506] The server receives this text and uses the OpenAI API to obtain the probability of its creation. For example, if the returned probability of creation is 65.4%, the server uses an emotion engine to analyze the sentiment in the text and adjust the probability of its creation based on the analysis results. The adjusted probability of creation and a credibility score are calculated and sent back to the reviewer or user. Based on this score, the reviewer can determine whether the submitted sentence was written by the user.

[1507] Example prompt sentence:

[1508] "The user provided the following motivation:

[1509] "I hope to grow through my experience at your company. I'm interested in the job content and feel I can utilize my skills."

[1510] Find the probability of generating this text, perform sentiment analysis, and calculate a credibility score."

[1511] In this way, the present invention efficiently distinguishes between user-generated content and AI-generated content, reducing the burden on reviewers and enabling more accurate credibility assessments.

[1512] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1513] Step 1:

[1514] The user enters long content such as motivation for applying and application documents into the application form and clicks the "Submit" button. The device sends an HTTP request including the entered text to the server.

[1515] Input: Long-form content entered into the application form

[1516] Output: HTTP request sent to the server

[1517] Step 2:

[1518] The server receives the text data sent by the user as an HTTP request.

[1519] Input: HTTP request from the terminal

[1520] Output: Received text data

[1521] Step 3:

[1522] The server retrieves the OpenAI API key from a configuration file or environment variable and initializes the API.

[1523] Input: API key stored in a config file or environment variable

[1524] Output: Initialized OpenAI API

[1525] Step 4:

[1526] The server calls the check_authenticity function to send the user's input text to the OpenAI API, which formats the text into an API request and sends it to the OpenAI API.

[1527] Input: User-entered text

[1528] Output: API request sent to the OpenAI API

[1529] Step 5:

[1530] The server receives a response from the OpenAI API that includes a generation probability, which is a number indicating the likelihood that the text was generated by the generative AI.

[1531] Input: Response from OpenAI API

[1532] Output: Generation probability

[1533] Step 6:

[1534] The server uses an emotion engine to extract emotional elements from the text and classify them into categories such as "positive," "negative," and "neutral."

[1535] Input: User-entered text

[1536] Output: Emotion analysis results

[1537] Step 7:

[1538] The server adjusts the generation probability based on the analysis results of the emotion engine. For example, if the text is highly emotional, the generation probability is reduced by a certain percentage.

[1539] Input: Sentiment analysis results and generation probability

[1540] Output: Corrected generation probabilities

[1541] Step 8:

[1542] The server calculates the credibility score based on the adjusted probability of creation, using the formula "(1 - probability of creation) 100".

[1543] Input: Corrected generation probability

[1544] Output: Belief score

[1545] Step 9:

[1546] The server sends the calculated credibility score to the reviewer or user's device, and sends information such as the credibility score, the corrected generation probability, and the sentiment analysis results via an HTTP response.

[1547] Input: Belief score, corrected generation probability, sentiment analysis result

[1548] Output: Results sent to users and reviewers

[1549] Through the above processing steps, the present invention efficiently distinguishes between content created by users and content generated by generation AI, reducing the burden on reviewers and enabling more accurate credibility assessments.

[1550] (Application example 2)

[1551] 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."

[1552] There is a need for efficient and accurate evaluation of the credibility of reviews posted by users on online shopping sites. However, there are many fraudulent reviews generated by AI and reviews that are misleading based on emotions, and appropriate countermeasures against these are lacking. As a result, it is difficult for buyers to select products based on trustworthy reviews. Therefore, an objective of the present invention is to provide a system that increases the credibility of posted reviews, allowing users to refer to reviews with peace of mind.

[1553] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1554] In this invention, the server includes means for receiving long-form content input by a user, means for analyzing the received long-form content and obtaining a generation probability by a generative AI model, means for calculating a credibility score based on the obtained generation probability, means for outputting the calculated credibility score, means for analyzing emotions in the long-form content, means for correcting the generation probability based on the result of the emotion analysis, and means for providing the credibility score of the long-form content to a reviewer. This makes it possible to increase the credibility of reviews posted by users and provide highly reliable reviews.

[1555] "User" refers to a person or end user of the System.

[1556] "Long-form content" refers to long texts entered by users, such as motivation for applying, application documents, and reviews.

[1557] "Means for receiving" refers to a method or device by which the server receives data sent from the user.

[1558] The term "analyzing means" refers to a method or device for analyzing the content of received long-form content.

[1559] "Generation probability" refers to the probability that a particular piece of text could have been generated by a generative AI.

[1560] "Means for obtaining generation probability" refers to a method or algorithm for calculating the generation probability of a text.

[1561] "Authenticity Score" refers to a numerical value used to assess the veracity of long-form content.

[1562] "Means for calculating" refers to a method or apparatus for calculating a credibility score based on the generation probability.

[1563] "Outputting means" refers to a method or device that displays the calculated credibility score.

[1564] "Means for analyzing emotions" refers to a method or device for analyzing emotions contained within long-form content.

[1565] "Means for correcting generation probability" refers to a method or device for adjusting generation probability based on the results of sentiment analysis.

[1566] "Reviewer" refers to a person or system that checks the credibility score and determines the veracity of posted content.

[1567] "Means for providing" refers to a method or device for providing the credibility score to the reviewer.

[1568] The system embodying this invention analyzes long-form content entered by users and evaluates its credibility. This system corrects the credibility score by combining generation probability obtained by a generative AI model and emotion analysis by an emotion engine.

[1569] System configuration

[1570] Hardware

[1571] server

[1572] User terminal

[1573] Reviewer device

[1574] software

[1575] OpenAI API

[1576] Emotion engine (natural language processing library such as TextBlob)

[1577] Database

[1578] System processing flow

[1579] 1. User input:

[1580] The user accesses the review submission page of the online shopping site and enters a review.

[1581] For example, a user might enter, "This product is great! I'm really happy with my purchase."

[1582] After completing the input, the user clicks the "Submit button."

[1583] 2. Data reception:

[1584] The user terminal sends an HTTP request containing the entered text to the server.

[1585] The server receives this data.

[1586] 3. Obtaining generation probabilities:

[1587] The server initializes the OpenAI API and sends the reviews entered by the user.

[1588] The OpenAI API returns the probability of generating the text. For example, if the generation probability is "0.35", you will get this value.

[1589] 4. Emotion analysis:

[1590] The server uses a sentiment engine (TextBlob library) to analyze the sentiment of the reviews.

[1591] For example, the text "This product is great! I'm really happy with my purchase" would be classified as "positive."

[1592] 5. Generation probability adjustment:

[1593] The server corrects the generation probability based on the results of the emotion analysis.

[1594] If positive emotions are included, the probability of generation is reduced. For example, if the original generation probability was 0.35, the corrected generation probability becomes 0.31.

[1595] 6. Calculating the credibility score:

[1596] The server calculates a credibility score based on the corrected generation probability.

[1597] Specifically, the score is calculated using the formula "(1 - generation probability) 100". If the corrected generation probability is "0.31", the credibility score will be "69".

[1598] 7. Result output:

[1599] The server provides the calculated credibility score to the reviewer.

[1600] Reviewers use this score to judge the credibility of reviews posted by users.

[1601] Specific examples

[1602] Example prompt sentence:

[1603] "Is this text generated by AI? This product is excellent. I use it every day and have had no issues."

[1604] Example response:

[1605] "0.35"

[1606] This will increase the credibility of reviews posted by users and improve the reliability of reviews on online shopping sites. Reviewers will be able to provide information to users based on highly reliable reviews.

[1607] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1608] Step 1:

[1609] Users can write and submit reviews

[1610] A user accesses a review input form on an online shopping site and enters a review. For example, they might enter, "This product is great! I'm really happy with my purchase." After entering the review, the user clicks the "Submit" button to send the review to the server. At this point, the input is the review text entered by the user, and the output is the review text sent to the server.

[1611] Step 2:

[1612] Server receives review text

[1613] The server receives the review text sent from the user's device. It analyzes the data sent as an HTTP request and extracts the review text. At this point, the input is the HTTP request from the user's device, and the output is the review text.

[1614] Step 3:

[1615] Obtaining generation probability

[1616] The server sends the received review text to the OpenAI API and obtains the generation probability. Specifically, it calls the OpenAI API and sends the review text including the prompt "Is this text generated by AI?". The OpenAI API returns a numerical value indicating the probability that the review text was generated by the generation AI. At this point, the input is the review text, and the output is a numerical value indicating the generation probability.

[1617] Step 4:

[1618] Performing sentiment analysis

[1619] The server uses a sentiment analysis library such as TextBlob to analyze the sentiment of the review text. It identifies positive, negative, or neutral sentiment elements in the review text and classifies the sentiment. At this point, the input is the review text, and the output is the sentiment classification result (positive, negative, neutral, etc.).

[1620] Step 5:

[1621] Generation probability adjustment

[1622] The server corrects the obtained generation probability based on the results of emotion analysis. For example, if a positive emotion is included, the generation probability is reduced to increase credibility. This correction is performed by a numerical calculation using the emotion analysis results and generation probability. The input at this point is the generation probability and the emotion classification result, and the output is the corrected generation probability.

[1623] Step 6:

[1624] Calculating the credibility score

[1625] The server calculates the credibility score based on the adjusted generation probability. Specifically, it uses the formula "(1 - generation probability) 100". If the adjusted generation probability is 0.31, the credibility score is 69. At this point, the input is the adjusted generation probability, and the output is the credibility score.

[1626] Step 7:

[1627] Outputting the results and providing them to reviewers

[1628] The server provides the calculated credibility score to the reviewer. The reviewer checks this score and judges the credibility of the review. The credibility score and the review text are displayed on the reviewer's terminal. At this point, the input is the credibility score, and the output is the credibility score and review text displayed on the reviewer's terminal.

[1629] 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.

[1630] 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.

[1631] 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.

[1632] 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.

[1633] 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.

[1634] 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.

[1635] 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).

[1636] 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.

[1637] 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."

[1638] 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.

[1639] 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).

[1640] 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.

[1641] 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.

[1642] 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.

[1643] 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.

[1644] 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.

[1645] 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.

[1646] 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.

[1647] 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.

[1648] 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.

[1649] 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.

[1650] The following is further disclosed regarding the above embodiment.

[1651] (Claim 1)

[1652] means for receiving long-form content input by a user;

[1653] A means for analyzing received long-form content and obtaining the generation probability of the generative AI model;

[1654] means for calculating a credibility score based on the obtained generation probabilities;

[1655] The system includes a means for outputting the calculated credibility score.

[1656] (Claim 2)

[1657] 2. The system according to claim 1, wherein said means for obtaining said generation probability utilizes an analytical algorithm for evaluating said generation probability.

[1658] (Claim 3)

[1659] The system of claim 1 , wherein the credibility score is calculated such that a higher score is calculated for a lower generation probability.

[1660] "Example 1"

[1661] (Claim 1)

[1662] means for receiving long-form content input by a user;

[1663] A means for analyzing received long-form content and obtaining the generation probability of the generative AI model;

[1664] means for calculating a credibility score based on the obtained generation probabilities;

[1665] means for outputting the calculated credibility score;

[1666] A means to initialize the API of the generative AI model, set a prompt sentence and create an analysis request;

[1667] A means for analyzing the generated probability response from the generative AI model and extracting the probability value;

[1668] A means for returning the calculated credibility score to the user or reviewer's device

[1669] A system including:

[1670] (Claim 2)

[1671] 2. The system according to claim 1, wherein said means for obtaining said generation probability utilizes an analytical algorithm for evaluating said generation probability.

[1672] (Claim 3)

[1673] The system of claim 1 , wherein the credibility score is calculated such that a higher score is calculated for a lower generation probability.

[1674] "Application Example 1"

[1675] (Claim 1)

[1676] means for receiving long-form content input by a user;

[1677] A means for analyzing received long-form content and obtaining the generation probability of the generative AI model;

[1678] means for calculating a credibility score based on the obtained generation probabilities;

[1679] means for outputting the calculated credibility score;

[1680] means for analyzing the input text and utilizing analysis algorithms and generation probabilities to calculate a credibility score;

[1681] means for displaying the credibility score on an electronic device;

[1682] A system including:

[1683] (Claim 2)

[1684] 2. The system according to claim 1, wherein said means for obtaining said generation probability utilizes an analytical algorithm for evaluating said generation probability.

[1685] (Claim 3)

[1686] The system of claim 1 , wherein the credibility score is calculated such that a higher score is calculated for a lower generation probability.

[1687] "Example 2: Combining Emotion Engines"

[1688] (Claim 1)

[1689] means for receiving long-form content input by a user;

[1690] A means for analyzing received long-form content and obtaining the generation probability of the generative AI model;

[1691] A means for analyzing received long-form content using an emotion engine and classifying emotion elements;

[1692] A means for correcting the generation probability based on the emotion analysis result;

[1693] means for calculating a credibility score based on the corrected generation probability;

[1694] The system includes a means for outputting the calculated credibility score.

[1695] (Claim 2)

[1696] 2. The system according to claim 1, wherein said means for obtaining said generation probability utilizes an analytical algorithm for evaluating said generation probability.

[1697] (Claim 3)

[1698] The system of claim 1 , wherein the credibility score is calculated such that a higher score is calculated for a lower generation probability.

[1699] "Application example 2 when combining emotion engines"

[1700] (Claim 1)

[1701] means for receiving long-form content input by a user;

[1702] A means for analyzing received long-form content and obtaining the generation probability of the generative AI model;

[1703] means for calculating a credibility score based on the obtained generation probabilities;

[1704] means for outputting the calculated credibility score;

[1705] means for analyzing emotions in the long-form content;

[1706] means for correcting the generation probability based on the result of the emotion analysis;

[1707] a means of providing reviewers with a credibility score for long-form content;

[1708] A system including:

[1709] (Claim 2)

[1710] 2. The system according to claim 1, wherein said means for obtaining said generation probability utilizes an analytical algorithm for evaluating said generation probability.

[1711] (Claim 3)

[1712] The system of claim 1 , wherein the credibility score is calculated such that a higher score is calculated for a lower generation probability. [Explanation of symbols]

[1713] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. means for receiving long-form content input by a user; A means for analyzing received long-form content and obtaining the generation probability of the generative AI model; means for calculating a credibility score based on the obtained generation probabilities; The system includes a means for outputting the calculated credibility score.

2. 2. The system of claim 1, wherein the means for obtaining the generation probability utilizes an analytical algorithm for evaluating the generation probability.

3. The system of claim 1 , wherein the credibility score is calculated such that a higher score is calculated for a lower probability of generation.

Images