System

Abstract

A system is provided.SOLUTION: The system includes a means for receiving data submitted by a student, a means for analyzing the data by a natural language processing engine, a means for applying a probability model for calculating a use suspicion score of a generated AI based on an analysis result, and a means for returning the analysis result to a user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] In educational settings, there are concerns that students' careless use of generative AI will lead to a decline in their writing and thinking skills. Despite some institutions banning or restricting its use in principle, there are still many cases in which compositions generated by generative AI are used as is for assignments, papers, etc. However, there is a problem in that it is difficult to determine at a glance whether a composition has been generated by AI. Therefore, there is a need for a system that can efficiently and accurately determine whether a composition has been generated by AI, without the need for humans to do it manually. [Means for solving the problem]

[0005] To solve the above problems, the present invention proposes the following means. Specifically, the system includes a means for receiving data submitted by students, a means for analyzing the data using a natural language processing engine, a means for applying a probabilistic model that calculates a generative AI usage suspicion score based on the analysis results, and a means for returning the analysis results to the user. The system also includes a stylistic analysis means for comparing the writing style with the characteristics of humans and the generative AI, and a means for generating analysis results for each submitted paragraph, providing more detailed analysis results. This allows for efficient monitoring of the use of generative AI in educational settings.

[0006] "Student" refers to a learner who is affiliated with an educational institution and submits assignments, papers, etc.

[0007] "Means for receiving data" refers to the system components that send the text or file submitted by the user to the server and correctly receive the data.

[0008] A "natural language processing engine" is a group of algorithms and models that analyze text data, and refers to a processing device that performs tokenization, grammar checking, stylistic analysis, etc.

[0009] "Generative AI" refers to an artificial intelligence system that automatically generates sentences, and in particular to the technology of generating natural language text using large-scale language models.

[0010] "Analysis results" refers to the results of analyzing data obtained by a natural language processing engine, including the results of stylistic analysis and the application of probabilistic models.

[0011] A "probabilistic model" refers to a statistical or machine learning model used to probabilistically calculate a specific outcome based on given data.

[0012] A "score" is a number calculated based on certain conditions or evaluation criteria, and in this case refers to a numerical representation of the probability that generative AI is suspected to be used.

[0013] "Means for returning to the user" refers to a system component for presenting or transmitting the analysis results and evaluation scores to the user.

[0014] "Style analysis" refers to the process of analyzing a text's vocabulary choice, sentence rhythm, grammatical structure, etc., and comparing the unique features of the style created by humans with the style created by generative AI.

[0015] A "paragraph" is a group of sentences that make up a sentence, and usually refers to a part of text that forms a logical whole. [Brief explanation of the drawings]

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

[0017] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

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

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

[0020] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0021] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0022] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0024] [First embodiment]

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

[0026] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0027] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0028] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0029] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0030] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0031] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

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

[0033] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0034] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0035] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

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

[0037] The following describes an embodiment of the present invention. The present invention is a system that determines whether a student's submission was created by a generative AI, and specifically describes the roles of the server, terminal, and user.

[0038] Server Processing

[0039] The server receives text data sent by users from their devices. This data is text submitted by students, such as assignments and papers. After the text data received by the server is temporarily stored, it is analyzed by a natural language processing engine. The natural language processing engine first tokenizes the text and analyzes the sentence structure. Next, a stylistic analysis is performed based on the analysis results, and the characteristics of the text written by a human and the text generated by the AI ​​are compared.

[0040] After stylistic analysis, the server applies a probability model to calculate the likelihood that the submitted text is the work of a generative AI. This probability model is a machine learning algorithm trained on pre-collected data, and generates a score to distinguish between generative AI and human writing.

[0041] In addition, the server detects frequently used keywords and phrases in the text and incorporates their frequency into the analysis, generating detailed analysis results for each paragraph.

[0042] Finally, the server generates a report based on these analysis results and sends it to the user's device, including a score indicating the likelihood of using AI and highlighting suspicious parts of each paragraph.

[0043] User operations

[0044] Users upload text data for assignments or papers from their own devices to the system. The uploaded data is received by the server, and once the analysis results are generated, users can check them. If users have any doubts about the use of the generated AI, they can ask questions or conduct additional research on the student.

[0045] Specific examples

[0046] Example 1: Parsing a single essay

[0047] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0048] Example 2: Batch processing multiple essays

[0049] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes the essays individually and generates independent analysis results for each essay. The results are compiled into a single report for all essays and sent back to the user's device. The user then checks the list and closely examines essays suspected of using generative AI.

[0050] The system of the present invention is an effective means of efficiently monitoring the use of generative AI in educational settings and maintaining the quality of student learning.

[0051] The processing flow will be explained below.

[0052] Step 1:

[0053] Users upload text data for assignments and papers from their devices. When doing so, they use the system's upload form and specify the file they wish to submit.

[0054] Step 2:

[0055] The server receives the text data sent by the user and temporarily stores it, checking the file format (txt, docx, pdf, etc.) and converting it into a format that can be analyzed.

[0056] Step 3:

[0057] The server inputs the received text data into a natural language processing engine, which first tokenizes the text and divides sentences into words.

[0058] Step 4:

[0059] The server analyzes the text data for grammatical structure, which involves recognizing sentence structure and checking for grammatical errors.

[0060] Step 5:

[0061] The server runs a stylistic analysis, which analyzes the lexical choice, rhythm, and grammatical patterns of a sentence to identify differences between human and generative AI writing styles.

[0062] Step 6:

[0063] The server then applies a probabilistic model to the parsed data, which uses a pre-trained machine learning algorithm to calculate the probability that a given piece of text is the product of a generative AI.

[0064] Step 7:

[0065] The server detects keywords and phrases frequently used by the AI ​​generator in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0066] Step 8:

[0067] The server generates a detailed analysis for each paragraph, including its stylistic characteristics, probability scores, and keyword detection results.

[0068] Step 9:

[0069] The server compiles the overall analysis results and creates a report, which includes a score for the likelihood of using generative AI and analysis results for each paragraph.

[0070] Step 10:

[0071] The server sends the generated report to the user's device, where the user can review the report and take measures to provide learning guidance or prevent the use of the generated AI as needed.

[0072] Example 1

[0073] 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."

[0074] In modern education, there is a need to efficiently identify student submissions created by generative AI and maintain learning quality. However, existing methods face the challenge of accurately distinguishing between human writing and generative AI writing. Analyzing submissions takes a significant amount of time and effort, especially when there are a large number of submissions. To solve this problem, an automated system is needed.

[0075] 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.

[0076] In this invention, the server includes means for receiving student submission data sent from the terminal, means for analyzing the data using a natural language processing engine that tokenizes the data and analyzes the sentence structure, means for analyzing the writing style based on the analysis results and comparing it with the characteristics of the generative AI, means for applying a probabilistic model that calculates a usage suspicion score for the generative AI based on the analysis results, means for detecting keywords and phrases frequently used by the generative AI and incorporating their frequency of occurrence into the analysis, means for generating detailed analysis results for each paragraph and generating a report based on the results, and means for returning the generated report to the user. This makes it possible to automatically and efficiently determine the possibility that submitted text was created by the generative AI.

[0077] A "terminal" is an electronic device that a user uses to upload text data of assignments or papers to the system.

[0078] A "server" is a central processing unit that receives data sent from user terminals, analyzes, stores, and generates reports.

[0079] A "natural language processing engine" is software or algorithms that tokenize text data, analyze sentence structure, and perform stylistic analysis.

[0080] "Tokenization" is the process of dividing text into words and phrases.

[0081] "Sentence structure analysis" is the process of analyzing the grammatical structure of a sentence.

[0082] "Stylistic analysis" is a method of analyzing the writing style and manner of a piece of text to determine whether it was written by a human or by generative AI.

[0083] A "probability model" is a statistical model for calculating the likelihood that a submitted text is the product of a generative AI.

[0084] A "machine learning algorithm" is a program that learns patterns and rules from data and uses that knowledge to analyze new data.

[0085] "Keyword detection" is the process of extracting specific words and phrases frequently used by generative AI from text.

[0086] "Analysis results" are information and data obtained through the application of natural language processing, stylistic analysis, and probabilistic models.

[0087] A "report" is a document that organizes the analysis results and presents them in a format that is easy for users to understand.

[0088] "Suspicious passages" are sections of text that are deemed likely to have been created by generative AI.

[0089] A "student" is a learner who submits assignments or papers.

[0090] "Submitted data" refers to text data such as assignments and papers that students upload to the system via their terminals.

[0091] This invention is a system that determines whether a student's submission was created by generative AI, and is composed of three parties: a server, a terminal, and a user.

[0092] Server Processing

[0093] The server receives the submitted data from the user's device. This can be done using a web framework such as Python's Flask. The received data is temporarily stored (e.g., in the " / tmp / uploads / " directory).

[0094] The server uses a natural language processing engine to analyze the received text data. Specifically, it uses libraries such as NLTK (Natural Language Toolkit) to tokenize the text and analyze its sentence structure. This breaks down the text data into words and phrases and understands their grammatical structure.

[0095] For stylistic analysis, we use a proprietary stylistic analysis algorithm that analyzes tokenized text features (e.g., word frequency, sentence length, syntactic patterns) to extract its style.

[0096] The server then applies a pre-trained probability model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is from a generative AI. This machine learning algorithm is based on patterns and rules learned from large datasets.

[0097] Additionally, the server detects keywords and phrases frequently used by the AI ​​generator (e.g., "In conclusion," "Based on the results") and calculates their frequency. This information is then incorporated into the results of the stylistic analysis.

[0098] Finally, the server generates a detailed analysis of each paragraph, including a report highlighting suspicious passages (e.g., "result_report.pdf"), which is then sent to the user's device. The user can view the report and identify any suspicious uses of the generated AI.

[0099] User operations

[0100] Users upload assignments and papers submitted by students to the system from their own devices. File uploads can be done through a web interface. The uploaded data is received by the server and temporarily stored.

[0101] The user can then review the analysis results generated by the server. The report includes a score for the suspected use of the generated AI and detailed analysis results with suspicious passages highlighted for each paragraph. The user can then use this information to ask questions of the student or conduct further research.

[0102] Specific examples

[0103] Example 1: Parsing a single essay

[0104] Users upload essays submitted by students (e.g., "assignment.txt") from their own devices to the server. The server receives the text data and uses the NLTK library to tokenize the text, analyze sentence structure, analyze stylistics, apply probabilistic models, and detect keywords. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0105] Example 2: Batch processing multiple essays

[0106] At the end of the semester, users upload multiple essays to the server in bulk. The server analyzes each essay individually and generates independent analysis results for each. The results are compiled into a single report for all essays and sent back to the user's device. The user then reviews the list and closely examines any essays suspected of using generative AI.

[0107] Prompt Sentence Examples

[0108] Determine whether the following sentence was written by a human or by a generative AI.

[0109] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0110] Step 1:

[0111] A user uploads text data (e.g., "assignment.txt") of assignments or papers submitted by students from their own terminal to the system. Specifically, they select a file using the web interface and click the "Upload" button. The input of this operation is the text data submitted by the student, and the output is the file being sent to the server.

[0112] Step 2:

[0113] The server receives text data sent from the user's device and saves it in a temporary storage area (e.g., the " / tmp / uploads / " directory). Specific operations include receiving an HTTP request and saving the data using a file stream. The input of this operation is the received text data, and the output is a temporarily saved file.

[0114] Step 3:

[0115] The server starts analyzing the temporarily stored text data using a natural language processing engine. First, it tokenizes the text using the NLTK library (e.g., it breaks it down into words using the word_tokenize function). The input of this operation is the stored text file, and the output is the tokenized text data.

[0116] Step 4:

[0117] The server performs sentence structure analysis based on the tokenized text data. It uses NLTK's pos_tag function to analyze the part of speech of each word and perform syntactic analysis. The input of this operation is the tokenized text data, and the output is text data with part-of-speech tags added.

[0118] Step 5:

[0119] The server performs stylistic analysis. Based on the results of tokenization and sentence structure analysis, it extracts text features (e.g., word frequency, sentence length, syntactic patterns) and applies a proprietary stylistic analysis algorithm. The input for this operation is text data with part-of-speech tags, and the output is stylistic feature data.

[0120] Step 6:

[0121] The server applies a pre-trained probabilistic model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is the work of a generative AI. Specifically, it inputs feature data and runs a machine learning algorithm to calculate a score. The input of this operation is the stylistic feature data, and the output is a generative AI suspicion score.

[0122] Step 7:

[0123] The server detects keywords and phrases frequently used by the AI. Using a predefined keyword list (e.g., "In conclusion," "Based on the results"), it searches for relevant passages in the text and incorporates their frequency of occurrence into the analysis. The input for this operation is the original text data and the keyword list, and the output is keyword frequency data.

[0124] Step 8:

[0125] The server generates detailed analysis results for each paragraph. Specifically, it combines the results of the stylistic analysis, the AI ​​suspicion score, and the keyword detection results to create a detailed report highlighting suspicious parts of each paragraph. The input of this operation is the analysis result data, and the output is the detailed report data.

[0126] Step 9:

[0127] The server then sends the generated report to the user's device, either as an email attachment or as a downloadable report file via a web interface. The input to this operation is the detailed report data, and the output is the report sent to the user.

[0128] Step 10:

[0129] The user opens the generated report on their device and checks the analysis results. The report includes a score for the suspected use of the generated AI and analysis results with suspicious passages highlighted for each paragraph. The user can use this information to ask questions of the student or conduct further research. The input for this operation is the generated report data, and the output is the user's judgment and response.

[0130] (Application example 1)

[0131] 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."

[0132] In online content distribution services, there is a problem in that it is difficult to determine whether posted text has been created by generative AI. This could lead to an increase in fraudulent posts by generative AI, which could result in a decline in content quality. In addition, a system is needed to quickly and accurately detect posts by generative AI.

[0133] 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.

[0134] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a suspicion score for use of a generation AI based on the analysis result, means for returning the analysis result to the user, and means for displaying a score for determining whether the submitted text was created by the generation AI and suspicious parts. This makes it possible to quickly detect fraudulent posts by generation AI in a content distribution service and maintain content quality.

[0135] "Data submitted by students" refers to text data submitted by students in the form of assignments, essays, etc.

[0136] A "natural language processing engine" is software that analyzes text data and processes it as structured information.

[0137] The "means of analysis" refers to the process of tokenizing text data using a natural language processing engine and performing stylistic analysis and keyword detection.

[0138] The "Generative AI Suspicion Score" is a score that indicates the likelihood that the submitted text was created by a generative AI.

[0139] A "probabilistic model" is a machine learning algorithm trained on pre-collected data to distinguish between generative AI and human-generated text.

[0140] "Means for returning the analysis results to the user" refers to the process of sending the analysis results performed on the server to the user's terminal and displaying them.

[0141] "Means for displaying scores and suspicious parts" refers to a function that displays the suspicion score used by the generating AI and highlights suspicious paragraphs and sentences.

[0142] The present invention provides a system for determining whether posted content in a content distribution service was created by a generation AI. An embodiment of the system will be described below.

[0143] The server receives text data from each user. The received text data is written content posted by the user, such as assignments, papers, and blog posts. This text data is temporarily stored, and then analysis begins using a natural language processing engine. The natural language processing engine uses Python's spaCy, which tokenizes the text and analyzes its sentence structure. It also detects writing style and keywords in the text and compares them with the characteristics of the generation AI.

[0144] The server then applies a probabilistic model to calculate the likelihood that the posted text was created by a generative AI. This model uses a machine learning algorithm, Scikit-Learn, and is trained on previously collected data. The model then generates a score to distinguish between generative AI and human-written text.

[0145] The analysis results in a score for the likelihood of the generated AI being used, which is then sent to the user's device. The user can then check the analysis results via a smartphone app. The analysis results include a score for the likelihood of the generated AI being used and detailed analysis results that highlight suspicious areas. Users can then use this information to correct or recheck the content.

[0146] For example, when a user uses this system before posting a blog post, the system checks in real time whether the content of the post was created by generative AI. When the user enters the blog post, the system displays the analysis results, such as a score of "likely AI-generated: 75%," and highlights suspicious paragraphs. Based on this information, the user can revise the post or check its quality.

[0147] An example of a prompt sentence would be:

[0148] Analyze the following text and determine whether it is the work of a generative AI. Provide a likelihood score for the generative AI and any suspicious areas.

[0149] In this way, generative AI in content distribution services detects fraudulent posts and maintains content quality.

[0150] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0151] Step 1:

[0152] The server receives text data submitted by the user. This data is uploaded by the user using a smartphone app. The input data is written data such as articles or essays. The server temporarily stores the received text data.

[0153] Step 2:

[0154] The server passes the received text data to a natural language processing engine and begins analysis. The natural language processing engine uses Python's spaCy to tokenize the text data and analyze its sentence structure. The input is text data, and the output is the tokenized text and its grammatical structure.

[0155] Step 3:

[0156] The server analyzes writing style and keywords based on the results of tokenization and sentence structure analysis. This analysis compares the characteristics of human-written text with those of the generative AI. Specifically, it detects frequently used keywords and phrases. The input is tokenized text, and the output is a list of writing style features and keywords.

[0157] Step 4:

[0158] The server applies a probabilistic model to calculate a suspicion score for the generative AI. This probabilistic model uses a machine learning algorithm using scikit-learn. Based on the analysis results, a score is calculated to determine whether the text submitted by the student was created by the generative AI. The input is the writing style and keyword features, and the output is the suspicion score for the generative AI.

[0159] Step 5:

[0160] The server generates a detailed analysis result report based on the generated AI's suspected use score. This report includes the generated AI's usability score and detailed analysis results with suspicious areas highlighted. The input is the generated AI's suspected use score and analysis results, and the output is the analysis result report.

[0161] Step 6:

[0162] The server sends the analysis result report to the user's device. The user can check the analysis results through a smartphone app. At this stage, the user can correct or recheck the article. The input is the analysis result report, and the output is the analysis results displayed to the user.

[0163] 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.

[0164] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback. Specifically, the roles of the server, terminal, and user are shown.

[0165] Server Processing

[0166] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine, which first performs tokenization and then grammatical structure analysis. A stylistic analysis is then performed, comparing the stylistic features of the human and generative AI versions.

[0167] The server then applies a probabilistic model to the analyzed data to calculate a probability score for the likelihood that the data was generated by the AI, while also detecting keywords and phrases frequently used by the AI ​​and calculating their frequency of occurrence.

[0168] The server then generates detailed analysis results for each paragraph based on these analysis results, and combines these results to generate a report that includes a score indicating the likelihood of using generative AI and the analysis results for each paragraph.

[0169] Emotion engine processing

[0170] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[0171] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[0172] User operations

[0173] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[0174] Specific examples

[0175] Example 1: Parsing a single essay

[0176] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0177] Example 2: Batch processing multiple essays

[0178] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[0179] The system of the present invention is an effective means for efficiently monitoring the use of generative AI in educational settings and providing feedback that takes into account the user's emotions, thereby achieving more appropriate instruction.

[0180] The processing flow will be explained below.

[0181] Step 1:

[0182] To upload text data for assignments or papers, users access the system's upload form from their terminal, select a file in the upload form, and click the submit button.

[0183] Step 2:

[0184] The server receives the text data sent by the user, temporarily stores the received data, and at the same time checks the file format and converts it into a format that can be analyzed as needed (e.g., txt, docx, pdf, etc.).

[0185] Step 3:

[0186] The server starts the natural language processing engine and prepares to analyze the text data. First, it performs tokenization, dividing the text into words.

[0187] Step 4:

[0188] The server then analyzes the tokenized data for grammatical structure, which involves understanding the structure of the sentence and checking for grammatical errors and patterns.

[0189] Step 5:

[0190] The server performs a stylistic analysis, specifically analyzing the lexical choice, rhythm, and grammatical patterns of a sentence, and comparing these characteristics with human and generative AI writing styles.

[0191] Step 6:

[0192] The server applies a probabilistic model to the parsed data, which calculates a probability score for the likelihood that the data was generated by the generative AI, using a pre-trained machine learning algorithm.

[0193] Step 7:

[0194] The server detects keywords and phrases frequently used by the AI ​​in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0195] Step 8:

[0196] The server generates detailed analysis results for each paragraph, including stylistic characteristics, probability scores, and keyword detection results for each paragraph.

[0197] Step 9:

[0198] The server compiles all the analysis results and creates a report, which includes a score for the likelihood of using a generative AI and the analysis results for each paragraph.

[0199] Step 10:

[0200] The server starts the emotion engine and acquires data to recognize the user's emotions, which are determined from the user's facial expressions, voice, input text, etc.

[0201] Step 11:

[0202] The server adjusts the display of the analysis results based on the results of the emotion engine, for example adding a warning message if the score is high or providing reassurance information if the user appears anxious.

[0203] Step 12:

[0204] The server then sends the generated report to the user's device, which includes the analysis results and emotional feedback.

[0205] Step 13:

[0206] The user checks the report received on the device, considers guidance and countermeasures based on the analysis results and feedback, and provides additional questions and guidance to the student as needed.

[0207] The above is an explanation of the processing steps including the division of roles between the user, server, and terminal, and specific operations.

[0208] Example 2

[0209] 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."

[0210] With conventional systems, it was difficult to analyze the writing style of assignments and papers submitted by students and determine whether they were using generative AI. Furthermore, simply returning the analysis results did not allow for appropriate feedback that took into account the user's level of understanding and psychological state. This made it difficult to provide efficient and effective instruction and support, preventing the maximum educational effect.

[0211] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for receiving data submitted by students, a means for analyzing the data using a natural language processing engine, a means for applying a probabilistic model that calculates a usage suspicion score of the generation AI based on the analysis result, a means for recognizing the analysis result and the user's emotions and adjusting the display method, and a means for returning the analysis result to the user. This makes it possible to accurately determine the use of the generation AI and provide appropriate feedback that takes the user's emotions into consideration.

[0212] A "student" is a learner who is enrolled in an educational institution and submits assignments and papers.

[0213] "Submitted data" refers to text data or document files that students upload to the server as assignments or papers.

[0214] "Receiving means" refers to the function or mechanism by which the server receives the submitted data sent from the user's terminal.

[0215] A "natural language processing engine" is a software program that analyzes text data and performs various language processing operations such as tokenization, grammatical structure analysis, and stylistic analysis.

[0216] A "probability model" is a statistical model or machine learning model used by the server to calculate the suspected use score of the generating AI.

[0217] "Emotion recognition means" refers to software or hardware functions that determine emotions from a user's facial expressions, voice, or text.

[0218] The "display method adjustment means" refers to a function that changes the way the analysis results are presented based on the emotion recognition results, and provides optimal feedback to the user.

[0219] "Return means" refers to the function or mechanism for returning the analysis results to the user.

[0220] "Stylistic analysis" is a method of analyzing a text to determine whether the writing style was created by a human or by a generative AI.

[0221] "Paragraph analysis means" refers to a function or mechanism that generates detailed analysis results for each paragraph of submitted data.

[0222] MODE FOR CARRYING OUT THE INVENTION

[0223] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback.

[0224] Server Processing

[0225] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. For example, the data is cleaned and normalized using Python's pandas library or nltk (Natural Language Toolkit).

[0226] The server then runs a natural language processing engine (e.g., spaCy or BERT) to first tokenize the text, then analyze its grammar. It then performs a stylistic analysis, comparing the stylistic features of the human and generative AI versions of the text. Specifically, it evaluates metrics such as the length of each sentence, the specific syntax it contains, and the vocabulary used.

[0227] Next, the server applies a probabilistic model (e.g., Bayesian inference or machine learning model) to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the generative AI. At the same time, it calculates the frequency of keywords and phrases frequently used by the generative AI. This step uses Python's scikit-learn and tf-idf-based analysis.

[0228] The analysis results are integrated and a report is generated that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. This report is generated using a report template and can be output in HTML or PDF format.

[0229] Emotion engine processing

[0230] Before returning the analysis results to the user, the server uses an emotion engine (e.g., IBM Watson (registered trademark) or Microsoft (registered trademark) Azure (registered trademark) emotion analysis service) to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[0231] Based on the results of emotion recognition, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of generative AI, it adds a message urging the user to take appropriate measures. Also, if the user is feeling anxious, it provides advice to ease their anxiety.

[0232] User operations

[0233] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[0234] Specific examples

[0235] Example 1: Parsing a single essay

[0236] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text, cleans and normalizes the data, and analyzes it using a natural language processing engine. The analysis involves tokenization, grammatical structure analysis, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0237] Specific prompt:

[0238] "Analyze the quality of AI-generated essays submitted by students and provide feedback to students along with the results."

[0239] Example 2: Batch processing multiple essays

[0240] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server receives these essays, prepares them for analysis individually, and then analyzes each essay using a natural language processing engine and an emotion engine. It then generates individual analysis results for each essay, compiles them into a single report, and returns them to the user.

[0241] Specific prompt:

[0242] "Please analyze the quality of AI-generated essays submitted at the end of the semester in a batch process and provide feedback to each student."

[0243] This system is an effective means of efficiently monitoring the use of generative AI in educational settings and providing feedback that takes users' emotions into account, thereby achieving more appropriate instruction.

[0244] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0245] Step 1:

[0246] Submission data is uploaded from the user's device. The user uses their own device to send text data of assignments or papers to the server through a specified interface (browser or application). The input is a text file created by the student, and the output is data stored in a temporary storage area on the server.

[0247] Step 2:

[0248] The server receives the submitted data and stores it in a temporary storage area. The server receives the text data uploaded by the user and checks its format. For example, converting a PDF file to TXT format. The input is the file uploaded in step 1, and the output is the cleaned text data.

[0249] Step 3:

[0250] The server cleans and normalizes the data. Specifically, it uses Python's pandas library and nltk (Natural Language Toolkit) to remove unnecessary characters and line breaks, standardize case, and standardize specific characters. The input is the text data saved in step 2, and the output is cleaned data ready for analysis.

[0251] Step 4:

[0252] The server runs a natural language processing engine and performs tokenization. The server uses a natural language processing engine, such as spaCy or BERT, to split the text into words and phrases. The input is the text data cleaned in step 3, and the output is the tokenized data.

[0253] Step 5:

[0254] The server performs grammatical structure analysis. Using the tokenized data, it analyzes the structure of each sentence and clarifies the relationships between subjects, predicates, objects, etc. The input is the data tokenized in step 4, and the output is the result of analyzing the grammatical structure.

[0255] Step 6:

[0256] The server performs a stylistic analysis. Based on the analyzed sentences, it compares the differences in the specific writing styles of humans and the generative AI. The stylistic analysis includes indicators such as the length of each sentence, the specific syntax contained in the sentence, and the vocabulary used. The input is the result of the grammatical structure analysis in step 5, and the output is the result of the stylistic analysis.

[0257] Step 7:

[0258] The server applies a probabilistic model to calculate a generative AI suspicion score. Based on the analysis results, it uses Bayesian inference and machine learning models to calculate a probability score indicating the likelihood that the data was created by a generative AI. The input is the result of the stylistic analysis in step 6, and the output is a generative AI suspicion score.

[0259] Step 8:

[0260] The server calculates the frequency of keywords and phrases. The generation AI detects frequently used keywords and phrases and calculates their frequency. The input is the same data as the probability score in step 7, and the output is the keyword frequency result.

[0261] Step 9:

[0262] The server integrates the analysis results and generates a report. It integrates the detailed analysis results for each paragraph and generates a report that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. A report template is used to generate this report, and it is output in HTML or PDF format. The input is the analysis results from Step 7 and Step 8, and the output is the final report.

[0263] Step 10:

[0264] The server uses an emotion engine to recognize the user's emotions. Before returning the analysis results to the user, it uses IBM Watson or Microsoft Azure's emotion analysis services to determine the user's emotions from their facial expressions, voice, and text. The input is the user's emotion data, and the output is the emotion recognition results.

[0265] Step 11:

[0266] The server adjusts the display method based on the emotion recognition results. Based on the results of the emotion engine, it changes the way the analysis results are displayed and provides optimal feedback to the user. For example, if there is a high suspicion of the use of generative AI, it adds a message offering careful guidance. The input is the emotion recognition result from step 10, and the output is the analysis result with adjusted feedback.

[0267] Step 12:

[0268] The server returns the analysis results to the user. The final report and sentiment-based feedback are sent to the user's device. The user reviews the report and makes necessary corrections or adjustments to the essay. The input is the analysis results with feedback adjusted in step 11, and the output is the final report displayed on the user's device.

[0269] (Application example 2)

[0270] 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."

[0271] Conventional chatbot systems have limitations in their ability to generate automatic responses to user input. In particular, as text generation using generative AI becomes more common, systems that can appropriately recognize a user's emotions and provide feedback based on those emotions are required. However, systems that provide feedback that takes emotions into account are rare, and have not contributed to improving user satisfaction. It has also been difficult to determine whether or not a generative AI is being used and provide a reliable service. Therefore, the present invention aims to solve these problems by providing a new system that combines emotion recognition and generative AI discrimination.

[0272] 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.

[0273] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a score for suspicion of use of generative AI based on the analysis results, means for returning the analysis results to the user, and an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results. This makes it possible to determine the suspicion of use of generative AI while taking the user's emotions into consideration.

[0274] "Means for receiving data submitted by students" refers to a function for receiving text data such as assignments and papers uploaded by users to the system.

[0275] "Means for analyzing using a natural language processing engine" refers to a function that analyzes uploaded text data using techniques such as tokenization, grammatical structure analysis, and stylistic analysis.

[0276] The "means for applying a probabilistic model to calculate a score for suspected use of generative AI" is a function that calculates a score for the likelihood that a text was generated by generative AI based on the analyzed data.

[0277] The "means for returning the analysis results to the user" is a function for presenting the results of the analysis performed by the server to the user.

[0278] "Means including an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results" refers to a function that determines the user's emotions from their facial expressions, voice, input text, etc., and adjusts the feedback method based on the results.

[0279] The present invention is a system that receives data submitted by users, analyzes it, identifies suspected use of generative AI, and recognizes the user's emotions to provide appropriate feedback. This system clarifies the roles of the server, terminal, and user, and explains how they interact. A specific example is provided below.

[0280] Server Processing

[0281] The server receives text data for assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine (NLP engine) that first performs tokenization. It then performs grammatical structure analysis and stylistic analysis, comparing the stylistic features of human and generative AI writing. Examples of specific software used for analysis include Python's nltk and transformers libraries.

[0282] Next, the server applies a probabilistic model to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the AI. At the same time, it detects keywords and phrases frequently used by the AI ​​and calculates their frequency of occurrence. A possible model to use is the BERT model. This allows for a suspicion score for the AI.

[0283] Emotion engine processing

[0284] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results. The TextBlob library can be used for emotion recognition.

[0285] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[0286] User operations

[0287] Users upload text data of assignments and papers from their own devices to the system. The data uploaded by the user is analyzed on the server, and the results are returned along with the results of the emotion engine. This allows users to receive emotion-based feedback along with the analysis results, enabling them to take better guidance and measures.

[0288] Specific examples

[0289] 1. Analyzing a Single Essay

[0290] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0291] 2. Batch processing multiple essays

[0292] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[0293] Prompt Sentence Examples

[0294] User input: "The delivery was late and the product is damaged."

[0295] Tokenization: ['The', 'delivery', 'was', 'late', 'and', 'the', 'product', 'is', 'damaged', '.']

[0296] Style analysis: Detecting stylistic features with generative AI

[0297] Emotion Recognition Result: Extremely Negative (Emotion Polarity: -1)

[0298] Feedback: "I sincerely apologize for the inconvenience. Can you provide more details or images so that we can resolve this issue for you?"

[0299] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0300] Step 1: User uploads data from device

[0301] ---

[0302] Users upload text data for assignments or papers from their own devices to the system. The input is a text file prepared by the user, and the output is that text file sent to the server. Specifically, users send the text data using the file upload function of a browser or a dedicated application.

[0303] Step 2: The server stores the received data and prepares it for analysis.

[0304] ---

[0305] The server receives data uploaded by the user and stores it in a temporary storage area. The input is the text data sent by the user, and the output is text data ready for analysis. Specifically, the server stores the received file in file storage and prepares it to be passed to the analysis module.

[0306] Step 3: Text analysis using a natural language processing engine

[0307] ---

[0308] The server applies a natural language processing engine to the received text data, performing tokenization, grammatical structure analysis, and stylistic analysis. The input is the stored text data, and the output is the analysis results. Specifically, the server performs tokenization and grammatical analysis using the Python nltk and transformers libraries.

[0309] Step 4: Calculate the generative AI usage suspicion score

[0310] ---

[0311] Based on the analysis results, the server applies a probability model to calculate a probability score indicating the suspected use of generative AI. The input is the analyzed text data, and the output is a probability score indicating the suspected use of generative AI. Specifically, it uses the BERT model to calculate the likelihood that the text was created by generative AI.

[0312] Step 5: Analyze user emotions using an emotion recognition engine

[0313] ---

[0314] The server uses an emotion recognition engine to recognize the user's emotions before returning the analysis results to the user. The input is the text message, voice, or image data entered by the user, and the output is the emotion recognition result. Specifically, it analyzes the user's emotions using the TextBlob library.

[0315] Step 6: Integrating analytical results with emotional feedback

[0316] ---

[0317] The server integrates the generated AI suspicion score and the emotion recognition result to generate a feedback report. The input is the generated AI suspicion score and the emotion recognition result, and the output is the integrated feedback report. Specifically, the server adds a warning message based on the emotion recognition result if careful action is required, and generates feedback according to the user's emotions.

[0318] Step 7: Returning the analysis results to the user

[0319] ---

[0320] The server returns the generated feedback report to the user. The input is the integrated feedback report, and the output is the analysis result and feedback received by the user. Specifically, the server sends the analysis result to the user's device for the user to check.

[0321] Through these steps, users can receive analysis results that determine whether or not to use the generative AI, as well as feedback based on their own emotions.

[0322] 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.

[0323] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0324] 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.

[0325] [Second embodiment]

[0326] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0327] 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.

[0328] 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).

[0329] 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.

[0330] 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.

[0331] 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).

[0332] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0333] 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.

[0334] 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.

[0335] 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.

[0336] 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.

[0337] 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."

[0338] The following describes an embodiment of the present invention. The present invention is a system that determines whether a student's submission was created by a generative AI, and specifically describes the roles of the server, terminal, and user.

[0339] Server Processing

[0340] The server receives text data sent by users from their devices. This data is text submitted by students, such as assignments and papers. After the text data received by the server is temporarily stored, it is analyzed by a natural language processing engine. The natural language processing engine first tokenizes the text and analyzes the sentence structure. Next, a stylistic analysis is performed based on the analysis results, and the characteristics of the text written by a human and the text generated by the AI ​​are compared.

[0341] After stylistic analysis, the server applies a probability model to calculate the likelihood that the submitted text is the work of a generative AI. This probability model is a machine learning algorithm trained on pre-collected data, and generates a score to distinguish between generative AI and human writing.

[0342] In addition, the server detects frequently used keywords and phrases in the text and incorporates their frequency into the analysis, generating detailed analysis results for each paragraph.

[0343] Finally, the server generates a report based on these analysis results and sends it to the user's device, including a score indicating the likelihood of using AI and highlighting suspicious parts of each paragraph.

[0344] User operations

[0345] Users upload text data for assignments or papers from their own devices to the system. The uploaded data is received by the server, and once the analysis results are generated, users can check them. If users have any doubts about the use of the generated AI, they can ask questions or conduct additional research on the student.

[0346] Specific examples

[0347] Example 1: Parsing a single essay

[0348] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0349] Example 2: Batch processing multiple essays

[0350] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes the essays individually and generates independent analysis results for each essay. The results are compiled into a single report for all essays and sent back to the user's device. The user then checks the list and closely examines essays suspected of using generative AI.

[0351] The system of the present invention is an effective means of efficiently monitoring the use of generative AI in educational settings and maintaining the quality of student learning.

[0352] The processing flow will be explained below.

[0353] Step 1:

[0354] Users upload text data for assignments and papers from their devices. When doing so, they use the system's upload form and specify the file they wish to submit.

[0355] Step 2:

[0356] The server receives the text data sent by the user and temporarily stores it, checking the file format (txt, docx, pdf, etc.) and converting it into a format that can be analyzed.

[0357] Step 3:

[0358] The server inputs the received text data into a natural language processing engine, which first tokenizes the text and divides sentences into words.

[0359] Step 4:

[0360] The server analyzes the text data for grammatical structure, which involves recognizing sentence structure and checking for grammatical errors.

[0361] Step 5:

[0362] The server runs a stylistic analysis, which analyzes the lexical choice, rhythm, and grammatical patterns of a sentence to identify differences between human and generative AI writing styles.

[0363] Step 6:

[0364] The server then applies a probabilistic model to the parsed data, which uses a pre-trained machine learning algorithm to calculate the probability that a given piece of text is the product of a generative AI.

[0365] Step 7:

[0366] The server detects keywords and phrases frequently used by the AI ​​generator in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0367] Step 8:

[0368] The server generates a detailed analysis for each paragraph, including its stylistic characteristics, probability scores, and keyword detection results.

[0369] Step 9:

[0370] The server compiles the overall analysis results and creates a report, which includes a score for the likelihood of using generative AI and analysis results for each paragraph.

[0371] Step 10:

[0372] The server sends the generated report to the user's device, where the user can review the report and take measures to provide learning guidance or prevent the use of the generated AI as needed.

[0373] Example 1

[0374] 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."

[0375] In modern education, there is a need to efficiently identify student submissions created by generative AI and maintain learning quality. However, existing methods face the challenge of accurately distinguishing between human writing and generative AI writing. Analyzing submissions takes a significant amount of time and effort, especially when there are a large number of submissions. To solve this problem, an automated system is needed.

[0376] 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.

[0377] In this invention, the server includes means for receiving student submission data sent from the terminal, means for analyzing the data using a natural language processing engine that tokenizes the data and analyzes the sentence structure, means for analyzing the writing style based on the analysis results and comparing it with the characteristics of the generative AI, means for applying a probabilistic model that calculates a usage suspicion score for the generative AI based on the analysis results, means for detecting keywords and phrases frequently used by the generative AI and incorporating their frequency of occurrence into the analysis, means for generating detailed analysis results for each paragraph and generating a report based on the results, and means for returning the generated report to the user. This makes it possible to automatically and efficiently determine the possibility that submitted text was created by the generative AI.

[0378] A "terminal" is an electronic device that a user uses to upload text data of assignments or papers to the system.

[0379] A "server" is a central processing unit that receives data sent from user terminals, analyzes, stores, and generates reports.

[0380] A "natural language processing engine" is software or algorithms that tokenize text data, analyze sentence structure, and perform stylistic analysis.

[0381] "Tokenization" is the process of dividing text into words and phrases.

[0382] "Sentence structure analysis" is the process of analyzing the grammatical structure of a sentence.

[0383] "Stylistic analysis" is a method of analyzing the writing style and manner of a piece of text to determine whether it was written by a human or by generative AI.

[0384] A "probability model" is a statistical model for calculating the likelihood that a submitted text is the product of a generative AI.

[0385] A "machine learning algorithm" is a program that learns patterns and rules from data and uses that knowledge to analyze new data.

[0386] "Keyword detection" is the process of extracting specific words and phrases frequently used by generative AI from text.

[0387] "Analysis results" are information and data obtained through the application of natural language processing, stylistic analysis, and probabilistic models.

[0388] A "report" is a document that organizes the analysis results and presents them in a format that is easy for users to understand.

[0389] "Suspicious passages" are sections of text that are deemed likely to have been created by generative AI.

[0390] A "student" is a learner who submits assignments or papers.

[0391] "Submitted data" refers to text data such as assignments and papers that students upload to the system via their terminals.

[0392] This invention is a system that determines whether a student's submission was created by generative AI, and is composed of three parties: a server, a terminal, and a user.

[0393] Server Processing

[0394] The server receives the submitted data from the user's device. This can be done using a web framework such as Python's Flask. The received data is temporarily stored (e.g., in the " / tmp / uploads / " directory).

[0395] The server uses a natural language processing engine to analyze the received text data. Specifically, it uses libraries such as NLTK (Natural Language Toolkit) to tokenize the text and analyze its sentence structure. This breaks down the text data into words and phrases and understands their grammatical structure.

[0396] For stylistic analysis, we use a proprietary stylistic analysis algorithm that analyzes tokenized text features (e.g., word frequency, sentence length, syntactic patterns) to extract its style.

[0397] The server then applies a pre-trained probability model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is from a generative AI. This machine learning algorithm is based on patterns and rules learned from large datasets.

[0398] Additionally, the server detects keywords and phrases frequently used by the AI ​​generator (e.g., "In conclusion," "Based on the results") and calculates their frequency. This information is then incorporated into the results of the stylistic analysis.

[0399] Finally, the server generates a detailed analysis of each paragraph, including a report highlighting suspicious passages (e.g., "result_report.pdf"), which is then sent to the user's device. The user can view the report and identify any suspicious uses of the generated AI.

[0400] User operations

[0401] Users upload assignments and papers submitted by students to the system from their own devices. File uploads can be done through a web interface. The uploaded data is received by the server and temporarily stored.

[0402] The user can then review the analysis results generated by the server. The report includes a score for the suspected use of the generated AI and detailed analysis results with suspicious passages highlighted for each paragraph. The user can then use this information to ask questions of the student or conduct further research.

[0403] Specific examples

[0404] Example 1: Parsing a single essay

[0405] Users upload essays submitted by students (e.g., "assignment.txt") from their own devices to the server. The server receives the text data and uses the NLTK library to tokenize the text, analyze sentence structure, analyze stylistics, apply probabilistic models, and detect keywords. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0406] Example 2: Batch processing multiple essays

[0407] At the end of the semester, users upload multiple essays to the server in bulk. The server analyzes each essay individually and generates independent analysis results for each. The results are compiled into a single report for all essays and sent back to the user's device. The user then reviews the list and closely examines any essays suspected of using generative AI.

[0408] Prompt Sentence Examples

[0409] Determine whether the following sentence was written by a human or by a generative AI.

[0410] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0411] Step 1:

[0412] A user uploads text data (e.g., "assignment.txt") of assignments or papers submitted by students from their own terminal to the system. Specifically, they select a file using the web interface and click the "Upload" button. The input of this operation is the text data submitted by the student, and the output is the file being sent to the server.

[0413] Step 2:

[0414] The server receives text data sent from the user's device and saves it in a temporary storage area (e.g., the " / tmp / uploads / " directory). Specific operations include receiving an HTTP request and saving the data using a file stream. The input of this operation is the received text data, and the output is a temporarily saved file.

[0415] Step 3:

[0416] The server starts analyzing the temporarily stored text data using a natural language processing engine. First, it tokenizes the text using the NLTK library (e.g., it breaks it down into words using the word_tokenize function). The input of this operation is the stored text file, and the output is the tokenized text data.

[0417] Step 4:

[0418] The server performs sentence structure analysis based on the tokenized text data. It uses NLTK's pos_tag function to analyze the part of speech of each word and perform syntactic analysis. The input of this operation is the tokenized text data, and the output is text data with part-of-speech tags added.

[0419] Step 5:

[0420] The server performs stylistic analysis. Based on the results of tokenization and sentence structure analysis, it extracts text features (e.g., word frequency, sentence length, syntactic patterns) and applies a proprietary stylistic analysis algorithm. The input for this operation is text data with part-of-speech tags, and the output is stylistic feature data.

[0421] Step 6:

[0422] The server applies a pre-trained probabilistic model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is the work of a generative AI. Specifically, it inputs feature data and runs a machine learning algorithm to calculate a score. The input of this operation is the stylistic feature data, and the output is a generative AI suspicion score.

[0423] Step 7:

[0424] The server detects keywords and phrases frequently used by the AI. Using a predefined keyword list (e.g., "In conclusion," "Based on the results"), it searches for relevant passages in the text and incorporates their frequency of occurrence into the analysis. The input for this operation is the original text data and the keyword list, and the output is keyword frequency data.

[0425] Step 8:

[0426] The server generates detailed analysis results for each paragraph. Specifically, it combines the results of the stylistic analysis, the AI ​​suspicion score, and the keyword detection results to create a detailed report highlighting suspicious parts of each paragraph. The input of this operation is the analysis result data, and the output is the detailed report data.

[0427] Step 9:

[0428] The server then sends the generated report to the user's device, either as an email attachment or as a downloadable report file via a web interface. The input to this operation is the detailed report data, and the output is the report sent to the user.

[0429] Step 10:

[0430] The user opens the generated report on their device and checks the analysis results. The report includes a score for the suspected use of the generated AI and analysis results with suspicious passages highlighted for each paragraph. The user can use this information to ask questions of the student or conduct further research. The input for this operation is the generated report data, and the output is the user's judgment and response.

[0431] (Application example 1)

[0432] 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."

[0433] In online content distribution services, there is a problem in that it is difficult to determine whether posted text has been created by generative AI. This could lead to an increase in fraudulent posts by generative AI, which could result in a decline in content quality. In addition, a system is needed to quickly and accurately detect posts by generative AI.

[0434] 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.

[0435] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a suspicion score for use of a generation AI based on the analysis result, means for returning the analysis result to the user, and means for displaying a score for determining whether the submitted text was created by the generation AI and suspicious parts. This makes it possible to quickly detect fraudulent posts by generation AI in a content distribution service and maintain content quality.

[0436] "Data submitted by students" refers to text data submitted by students in the form of assignments, essays, etc.

[0437] A "natural language processing engine" is software that analyzes text data and processes it as structured information.

[0438] The "means of analysis" refers to the process of tokenizing text data using a natural language processing engine and performing stylistic analysis and keyword detection.

[0439] The "Generative AI Suspicion Score" is a score that indicates the likelihood that the submitted text was created by a generative AI.

[0440] A "probabilistic model" is a machine learning algorithm trained on pre-collected data to distinguish between generative AI and human-generated text.

[0441] "Means for returning the analysis results to the user" refers to the process of sending the analysis results performed on the server to the user's terminal and displaying them.

[0442] "Means for displaying scores and suspicious parts" refers to a function that displays the suspicion score used by the generating AI and highlights suspicious paragraphs and sentences.

[0443] The present invention provides a system for determining whether posted content in a content distribution service was created by a generation AI. An embodiment of the system will be described below.

[0444] The server receives text data from each user. The received text data is written content posted by the user, such as assignments, papers, and blog posts. This text data is temporarily stored, and then analysis begins using a natural language processing engine. The natural language processing engine uses Python's spaCy, which tokenizes the text and analyzes its sentence structure. It also detects writing style and keywords in the text and compares them with the characteristics of the generation AI.

[0445] The server then applies a probabilistic model to calculate the likelihood that the posted text was created by a generative AI. This model uses a machine learning algorithm, Scikit-Learn, and is trained on previously collected data. The model then generates a score to distinguish between generative AI and human-written text.

[0446] The analysis results in a score for the likelihood of the generated AI being used, which is then sent to the user's device. The user can then check the analysis results via a smartphone app. The analysis results include a score for the likelihood of the generated AI being used and detailed analysis results that highlight suspicious areas. Users can then use this information to correct or recheck the content.

[0447] For example, when a user uses this system before posting a blog post, the system checks in real time whether the content of the post was created by generative AI. When the user enters the blog post, the system displays the analysis results, such as a score of "likely AI-generated: 75%," and highlights suspicious paragraphs. Based on this information, the user can revise the post or check its quality.

[0448] An example of a prompt sentence would be:

[0449] Analyze the following text and determine whether it is the work of a generative AI. Provide a likelihood score for the generative AI and any suspicious areas.

[0450] In this way, generative AI in content distribution services detects fraudulent posts and maintains content quality.

[0451] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0452] Step 1:

[0453] The server receives text data submitted by the user. This data is uploaded by the user using a smartphone app. The input data is written data such as articles or essays. The server temporarily stores the received text data.

[0454] Step 2:

[0455] The server passes the received text data to a natural language processing engine and begins analysis. The natural language processing engine uses Python's spaCy to tokenize the text data and analyze its sentence structure. The input is text data, and the output is the tokenized text and its grammatical structure.

[0456] Step 3:

[0457] The server analyzes writing style and keywords based on the results of tokenization and sentence structure analysis. This analysis compares the characteristics of human-written text with those of the generative AI. Specifically, it detects frequently used keywords and phrases. The input is tokenized text, and the output is a list of writing style features and keywords.

[0458] Step 4:

[0459] The server applies a probabilistic model to calculate a suspicion score for the generative AI. This probabilistic model uses a machine learning algorithm using scikit-learn. Based on the analysis results, a score is calculated to determine whether the text submitted by the student was created by the generative AI. The input is the writing style and keyword features, and the output is the suspicion score for the generative AI.

[0460] Step 5:

[0461] The server generates a detailed analysis result report based on the generated AI's suspected use score. This report includes the generated AI's usability score and detailed analysis results with suspicious areas highlighted. The input is the generated AI's suspected use score and analysis results, and the output is the analysis result report.

[0462] Step 6:

[0463] The server sends the analysis result report to the user's device. The user can check the analysis results through a smartphone app. At this stage, the user can correct or recheck the article. The input is the analysis result report, and the output is the analysis results displayed to the user.

[0464] 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.

[0465] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback. Specifically, the roles of the server, terminal, and user are shown.

[0466] Server Processing

[0467] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine, which first performs tokenization and then grammatical structure analysis. A stylistic analysis is then performed, comparing the stylistic features of the human and generative AI versions.

[0468] The server then applies a probabilistic model to the analyzed data to calculate a probability score for the likelihood that the data was generated by the AI, while also detecting keywords and phrases frequently used by the AI ​​and calculating their frequency of occurrence.

[0469] The server then generates detailed analysis results for each paragraph based on these analysis results, and combines these results to generate a report that includes a score indicating the likelihood of using generative AI and the analysis results for each paragraph.

[0470] Emotion engine processing

[0471] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[0472] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[0473] User operations

[0474] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[0475] Specific examples

[0476] Example 1: Parsing a single essay

[0477] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0478] Example 2: Batch processing multiple essays

[0479] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[0480] The system of the present invention is an effective means for efficiently monitoring the use of generative AI in educational settings and providing feedback that takes into account the user's emotions, thereby achieving more appropriate instruction.

[0481] The processing flow will be explained below.

[0482] Step 1:

[0483] To upload text data for assignments or papers, users access the system's upload form from their terminal, select a file in the upload form, and click the submit button.

[0484] Step 2:

[0485] The server receives the text data sent by the user, temporarily stores the received data, and at the same time checks the file format and converts it into a format that can be analyzed as needed (e.g., txt, docx, pdf, etc.).

[0486] Step 3:

[0487] The server starts the natural language processing engine and prepares to analyze the text data. First, it performs tokenization, dividing the text into words.

[0488] Step 4:

[0489] The server then analyzes the tokenized data for grammatical structure, which involves understanding the structure of the sentence and checking for grammatical errors and patterns.

[0490] Step 5:

[0491] The server performs a stylistic analysis, specifically analyzing the lexical choice, rhythm, and grammatical patterns of a sentence, and comparing these characteristics with human and generative AI writing styles.

[0492] Step 6:

[0493] The server applies a probabilistic model to the parsed data, which calculates a probability score for the likelihood that the data was generated by the generative AI, using a pre-trained machine learning algorithm.

[0494] Step 7:

[0495] The server detects keywords and phrases frequently used by the AI ​​in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0496] Step 8:

[0497] The server generates detailed analysis results for each paragraph, including stylistic characteristics, probability scores, and keyword detection results for each paragraph.

[0498] Step 9:

[0499] The server compiles all the analysis results and creates a report, which includes a score for the likelihood of using a generative AI and the analysis results for each paragraph.

[0500] Step 10:

[0501] The server starts the emotion engine and acquires data to recognize the user's emotions, which are determined from the user's facial expressions, voice, input text, etc.

[0502] Step 11:

[0503] The server adjusts the display of the analysis results based on the results of the emotion engine, for example adding a warning message if the score is high or providing reassurance information if the user appears anxious.

[0504] Step 12:

[0505] The server then sends the generated report to the user's device, which includes the analysis results and emotional feedback.

[0506] Step 13:

[0507] The user checks the report received on the device, considers guidance and countermeasures based on the analysis results and feedback, and provides additional questions and guidance to the student as needed.

[0508] The above is an explanation of the processing steps including the division of roles between the user, server, and terminal, and specific operations.

[0509] Example 2

[0510] 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."

[0511] With conventional systems, it was difficult to analyze the writing style of assignments and papers submitted by students and determine whether they were using generative AI. Furthermore, simply returning the analysis results did not allow for appropriate feedback that took into account the user's level of understanding and psychological state. This made it difficult to provide efficient and effective instruction and support, preventing the maximum educational effect.

[0512] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for receiving data submitted by students, a means for analyzing the data using a natural language processing engine, a means for applying a probabilistic model that calculates a usage suspicion score of the generation AI based on the analysis result, a means for recognizing the analysis result and the user's emotions and adjusting the display method, and a means for returning the analysis result to the user. This makes it possible to accurately determine the use of the generation AI and provide appropriate feedback that takes the user's emotions into consideration.

[0513] A "student" is a learner who is enrolled in an educational institution and submits assignments and papers.

[0514] "Submitted data" refers to text data or document files that students upload to the server as assignments or papers.

[0515] "Receiving means" refers to the function or mechanism by which the server receives the submitted data sent from the user's terminal.

[0516] A "natural language processing engine" is a software program that analyzes text data and performs various language processing operations such as tokenization, grammatical structure analysis, and stylistic analysis.

[0517] A "probability model" is a statistical model or machine learning model used by the server to calculate the suspected use score of the generating AI.

[0518] "Emotion recognition means" refers to software or hardware functions that determine emotions from a user's facial expressions, voice, or text.

[0519] The "display method adjustment means" refers to a function that changes the way the analysis results are presented based on the emotion recognition results, and provides optimal feedback to the user.

[0520] "Return means" refers to the function or mechanism for returning the analysis results to the user.

[0521] "Stylistic analysis" is a method of analyzing a text to determine whether the writing style was created by a human or by a generative AI.

[0522] "Paragraph analysis means" refers to a function or mechanism that generates detailed analysis results for each paragraph of submitted data.

[0523] MODE FOR CARRYING OUT THE INVENTION

[0524] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback.

[0525] Server Processing

[0526] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. For example, the data is cleaned and normalized using Python's pandas library or nltk (Natural Language Toolkit).

[0527] The server then runs a natural language processing engine (e.g., spaCy or BERT) to first tokenize the text, then analyze its grammar. It then performs a stylistic analysis, comparing the stylistic features of the human and generative AI versions of the text. Specifically, it evaluates metrics such as the length of each sentence, the specific syntax it contains, and the vocabulary used.

[0528] Next, the server applies a probabilistic model (e.g., Bayesian inference or machine learning model) to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the generative AI. At the same time, it calculates the frequency of keywords and phrases frequently used by the generative AI. This step uses Python's scikit-learn and tf-idf-based analysis.

[0529] The analysis results are integrated and a report is generated that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. This report is generated using a report template and can be output in HTML or PDF format.

[0530] Emotion engine processing

[0531] Before returning the analysis results to the user, the server uses an emotion engine (such as IBM Watson or Microsoft Azure's emotion analysis service) to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[0532] Based on the results of emotion recognition, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of generative AI, it adds a message urging the user to take appropriate measures. Also, if the user is feeling anxious, it provides advice to ease their anxiety.

[0533] User operations

[0534] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[0535] Specific examples

[0536] Example 1: Parsing a single essay

[0537] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text, cleans and normalizes the data, and analyzes it using a natural language processing engine. The analysis involves tokenization, grammatical structure analysis, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0538] Specific prompt:

[0539] "Analyze the quality of AI-generated essays submitted by students and provide feedback to students along with the results."

[0540] Example 2: Batch processing multiple essays

[0541] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server receives these essays, prepares them for analysis individually, and then analyzes each essay using a natural language processing engine and an emotion engine. It then generates individual analysis results for each essay, compiles them into a single report, and returns them to the user.

[0542] Specific prompt:

[0543] "Please analyze the quality of AI-generated essays submitted at the end of the semester in a batch process and provide feedback to each student."

[0544] This system is an effective means of efficiently monitoring the use of generative AI in educational settings and providing feedback that takes users' emotions into account, thereby achieving more appropriate instruction.

[0545] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0546] Step 1:

[0547] Submission data is uploaded from the user's device. The user uses their own device to send text data of assignments or papers to the server through a specified interface (browser or application). The input is a text file created by the student, and the output is data stored in a temporary storage area on the server.

[0548] Step 2:

[0549] The server receives the submitted data and stores it in a temporary storage area. The server receives the text data uploaded by the user and checks its format. For example, converting a PDF file to TXT format. The input is the file uploaded in step 1, and the output is the cleaned text data.

[0550] Step 3:

[0551] The server cleans and normalizes the data. Specifically, it uses Python's pandas library and nltk (Natural Language Toolkit) to remove unnecessary characters and line breaks, standardize case, and standardize specific characters. The input is the text data saved in step 2, and the output is cleaned data ready for analysis.

[0552] Step 4:

[0553] The server runs a natural language processing engine and performs tokenization. The server uses a natural language processing engine, such as spaCy or BERT, to split the text into words and phrases. The input is the text data cleaned in step 3, and the output is the tokenized data.

[0554] Step 5:

[0555] The server performs grammatical structure analysis. Using the tokenized data, it analyzes the structure of each sentence and clarifies the relationships between subjects, predicates, objects, etc. The input is the data tokenized in step 4, and the output is the result of analyzing the grammatical structure.

[0556] Step 6:

[0557] The server performs a stylistic analysis. Based on the analyzed sentences, it compares the differences in the specific writing styles of humans and the generative AI. The stylistic analysis includes indicators such as the length of each sentence, the specific syntax contained in the sentence, and the vocabulary used. The input is the result of the grammatical structure analysis in step 5, and the output is the result of the stylistic analysis.

[0558] Step 7:

[0559] The server applies a probabilistic model to calculate a generative AI suspicion score. Based on the analysis results, it uses Bayesian inference and machine learning models to calculate a probability score indicating the likelihood that the data was created by a generative AI. The input is the result of the stylistic analysis in step 6, and the output is a generative AI suspicion score.

[0560] Step 8:

[0561] The server calculates the frequency of keywords and phrases. The generation AI detects frequently used keywords and phrases and calculates their frequency. The input is the same data as the probability score in step 7, and the output is the keyword frequency result.

[0562] Step 9:

[0563] The server integrates the analysis results and generates a report. It integrates the detailed analysis results for each paragraph and generates a report that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. A report template is used to generate this report, and it is output in HTML or PDF format. The input is the analysis results from Step 7 and Step 8, and the output is the final report.

[0564] Step 10:

[0565] The server uses an emotion engine to recognize the user's emotions. Before returning the analysis results to the user, it uses IBM Watson or Microsoft Azure's emotion analysis services to determine the user's emotions from their facial expressions, voice, and text. The input is the user's emotion data, and the output is the emotion recognition results.

[0566] Step 11:

[0567] The server adjusts the display method based on the emotion recognition results. Based on the results of the emotion engine, it changes the way the analysis results are displayed and provides optimal feedback to the user. For example, if there is a high suspicion of the use of generative AI, it adds a message offering careful guidance. The input is the emotion recognition result from step 10, and the output is the analysis result with adjusted feedback.

[0568] Step 12:

[0569] The server returns the analysis results to the user. The final report and sentiment-based feedback are sent to the user's device. The user reviews the report and makes necessary corrections or adjustments to the essay. The input is the analysis results with feedback adjusted in step 11, and the output is the final report displayed on the user's device.

[0570] (Application example 2)

[0571] 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."

[0572] Conventional chatbot systems have limitations in their ability to generate automatic responses to user input. In particular, as text generation using generative AI becomes more common, systems that can appropriately recognize a user's emotions and provide feedback based on those emotions are required. However, systems that provide feedback that takes emotions into account are rare, and have not contributed to improving user satisfaction. It has also been difficult to determine whether or not a generative AI is being used and provide a reliable service. Therefore, the present invention aims to solve these problems by providing a new system that combines emotion recognition and generative AI discrimination.

[0573] 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.

[0574] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a score for suspicion of use of generative AI based on the analysis results, means for returning the analysis results to the user, and an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results. This makes it possible to determine the suspicion of use of generative AI while taking the user's emotions into consideration.

[0575] "Means for receiving data submitted by students" refers to a function for receiving text data such as assignments and papers uploaded by users to the system.

[0576] "Means for analyzing using a natural language processing engine" refers to a function that analyzes uploaded text data using techniques such as tokenization, grammatical structure analysis, and stylistic analysis.

[0577] The "means for applying a probabilistic model to calculate a score for suspected use of generative AI" is a function that calculates a score for the likelihood that a text was generated by generative AI based on the analyzed data.

[0578] The "means for returning the analysis results to the user" is a function for presenting the results of the analysis performed by the server to the user.

[0579] "Means including an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results" refers to a function that determines the user's emotions from their facial expressions, voice, input text, etc., and adjusts the feedback method based on the results.

[0580] The present invention is a system that receives data submitted by users, analyzes it, identifies suspected use of generative AI, and recognizes the user's emotions to provide appropriate feedback. This system clarifies the roles of the server, terminal, and user, and explains how they interact. A specific example is provided below.

[0581] Server Processing

[0582] The server receives text data for assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine (NLP engine) that first performs tokenization. It then performs grammatical structure analysis and stylistic analysis, comparing the stylistic features of human and generative AI writing. Examples of specific software used for analysis include Python's nltk and transformers libraries.

[0583] Next, the server applies a probabilistic model to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the AI. At the same time, it detects keywords and phrases frequently used by the AI ​​and calculates their frequency of occurrence. A possible model to use is the BERT model. This allows for a suspicion score for the AI.

[0584] Emotion engine processing

[0585] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results. The TextBlob library can be used for emotion recognition.

[0586] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[0587] User operations

[0588] Users upload text data of assignments and papers from their own devices to the system. The data uploaded by the user is analyzed on the server, and the results are returned along with the results of the emotion engine. This allows users to receive emotion-based feedback along with the analysis results, enabling them to take better guidance and measures.

[0589] Specific examples

[0590] 1. Analyzing a Single Essay

[0591] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0592] 2. Batch processing multiple essays

[0593] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[0594] Prompt Sentence Examples

[0595] User input: "The delivery was late and the product is damaged."

[0596] Tokenization: ['The', 'delivery', 'was', 'late', 'and', 'the', 'product', 'is', 'damaged', '.']

[0597] Style analysis: Detecting stylistic features with generative AI

[0598] Emotion Recognition Result: Extremely Negative (Emotion Polarity: -1)

[0599] Feedback: "I sincerely apologize for the inconvenience. Can you provide more details or images so that we can resolve this issue for you?"

[0600] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0601] Step 1: User uploads data from device

[0602] ---

[0603] Users upload text data for assignments or papers from their own devices to the system. The input is a text file prepared by the user, and the output is that text file sent to the server. Specifically, users send the text data using the file upload function of a browser or a dedicated application.

[0604] Step 2: The server stores the received data and prepares it for analysis.

[0605] ---

[0606] The server receives data uploaded by the user and stores it in a temporary storage area. The input is the text data sent by the user, and the output is text data ready for analysis. Specifically, the server stores the received file in file storage and prepares it to be passed to the analysis module.

[0607] Step 3: Text analysis using a natural language processing engine

[0608] ---

[0609] The server applies a natural language processing engine to the received text data, performing tokenization, grammatical structure analysis, and stylistic analysis. The input is the stored text data, and the output is the analysis results. Specifically, the server performs tokenization and grammatical analysis using the Python nltk and transformers libraries.

[0610] Step 4: Calculate the generative AI usage suspicion score

[0611] ---

[0612] Based on the analysis results, the server applies a probability model to calculate a probability score indicating the suspected use of generative AI. The input is the analyzed text data, and the output is a probability score indicating the suspected use of generative AI. Specifically, it uses the BERT model to calculate the likelihood that the text was created by generative AI.

[0613] Step 5: Analyze user emotions using an emotion recognition engine

[0614] ---

[0615] The server uses an emotion recognition engine to recognize the user's emotions before returning the analysis results to the user. The input is the text message, voice, or image data entered by the user, and the output is the emotion recognition result. Specifically, it analyzes the user's emotions using the TextBlob library.

[0616] Step 6: Integrating analytical results with emotional feedback

[0617] ---

[0618] The server integrates the generated AI suspicion score and the emotion recognition result to generate a feedback report. The input is the generated AI suspicion score and the emotion recognition result, and the output is the integrated feedback report. Specifically, the server adds a warning message based on the emotion recognition result if careful action is required, and generates feedback according to the user's emotions.

[0619] Step 7: Returning the analysis results to the user

[0620] ---

[0621] The server returns the generated feedback report to the user. The input is the integrated feedback report, and the output is the analysis result and feedback received by the user. Specifically, the server sends the analysis result to the user's device for the user to check.

[0622] Through these steps, users can receive analysis results that determine whether or not to use the generative AI, as well as feedback based on their own emotions.

[0623] 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.

[0624] 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.

[0625] 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.

[0626] [Third embodiment]

[0627] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0628] 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.

[0629] 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).

[0630] 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.

[0631] 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.

[0632] 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).

[0633] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0634] 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.

[0635] 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.

[0636] 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.

[0637] 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.

[0638] 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."

[0639] The following describes an embodiment of the present invention. The present invention is a system that determines whether a student's submission was created by a generative AI, and specifically describes the roles of the server, terminal, and user.

[0640] Server Processing

[0641] The server receives text data sent by users from their devices. This data is text submitted by students, such as assignments and papers. After the text data received by the server is temporarily stored, it is analyzed by a natural language processing engine. The natural language processing engine first tokenizes the text and analyzes the sentence structure. Next, a stylistic analysis is performed based on the analysis results, and the characteristics of the text written by a human and the text generated by the AI ​​are compared.

[0642] After stylistic analysis, the server applies a probability model to calculate the likelihood that the submitted text is the work of a generative AI. This probability model is a machine learning algorithm trained on pre-collected data, and generates a score to distinguish between generative AI and human writing.

[0643] In addition, the server detects frequently used keywords and phrases in the text and incorporates their frequency into the analysis, generating detailed analysis results for each paragraph.

[0644] Finally, the server generates a report based on these analysis results and sends it to the user's device, including a score indicating the likelihood of using AI and highlighting suspicious parts of each paragraph.

[0645] User operations

[0646] Users upload text data for assignments or papers from their own devices to the system. The uploaded data is received by the server, and once the analysis results are generated, users can check them. If users have any doubts about the use of the generated AI, they can ask questions or conduct additional research on the student.

[0647] Specific examples

[0648] Example 1: Parsing a single essay

[0649] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0650] Example 2: Batch processing multiple essays

[0651] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes the essays individually and generates independent analysis results for each essay. The results are compiled into a single report for all essays and sent back to the user's device. The user then checks the list and closely examines essays suspected of using generative AI.

[0652] The system of the present invention is an effective means of efficiently monitoring the use of generative AI in educational settings and maintaining the quality of student learning.

[0653] The processing flow will be explained below.

[0654] Step 1:

[0655] Users upload text data for assignments and papers from their devices. When doing so, they use the system's upload form and specify the file they wish to submit.

[0656] Step 2:

[0657] The server receives the text data sent by the user and temporarily stores it, checking the file format (txt, docx, pdf, etc.) and converting it into a format that can be analyzed.

[0658] Step 3:

[0659] The server inputs the received text data into a natural language processing engine, which first tokenizes the text and divides sentences into words.

[0660] Step 4:

[0661] The server analyzes the text data for grammatical structure, which involves recognizing sentence structure and checking for grammatical errors.

[0662] Step 5:

[0663] The server runs a stylistic analysis, which analyzes the lexical choice, rhythm, and grammatical patterns of a sentence to identify differences between human and generative AI writing styles.

[0664] Step 6:

[0665] The server then applies a probabilistic model to the parsed data, which uses a pre-trained machine learning algorithm to calculate the probability that a given piece of text is the product of a generative AI.

[0666] Step 7:

[0667] The server detects keywords and phrases frequently used by the AI ​​generator in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0668] Step 8:

[0669] The server generates a detailed analysis for each paragraph, including its stylistic characteristics, probability scores, and keyword detection results.

[0670] Step 9:

[0671] The server compiles the overall analysis results and creates a report, which includes a score for the likelihood of using generative AI and analysis results for each paragraph.

[0672] Step 10:

[0673] The server sends the generated report to the user's device, where the user can review the report and take measures to provide learning guidance or prevent the use of the generated AI as needed.

[0674] Example 1

[0675] 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."

[0676] In modern education, there is a need to efficiently identify student submissions created by generative AI and maintain learning quality. However, existing methods face the challenge of accurately distinguishing between human writing and generative AI writing. Analyzing submissions takes a significant amount of time and effort, especially when there are a large number of submissions. To solve this problem, an automated system is needed.

[0677] 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.

[0678] In this invention, the server includes means for receiving student submission data sent from the terminal, means for analyzing the data using a natural language processing engine that tokenizes the data and analyzes the sentence structure, means for analyzing the writing style based on the analysis results and comparing it with the characteristics of the generative AI, means for applying a probabilistic model that calculates a usage suspicion score for the generative AI based on the analysis results, means for detecting keywords and phrases frequently used by the generative AI and incorporating their frequency of occurrence into the analysis, means for generating detailed analysis results for each paragraph and generating a report based on the results, and means for returning the generated report to the user. This makes it possible to automatically and efficiently determine the possibility that submitted text was created by the generative AI.

[0679] A "terminal" is an electronic device that a user uses to upload text data of assignments or papers to the system.

[0680] A "server" is a central processing unit that receives data sent from user terminals, analyzes, stores, and generates reports.

[0681] A "natural language processing engine" is software or algorithms that tokenize text data, analyze sentence structure, and perform stylistic analysis.

[0682] "Tokenization" is the process of dividing text into words and phrases.

[0683] "Sentence structure analysis" is the process of analyzing the grammatical structure of a sentence.

[0684] "Stylistic analysis" is a method of analyzing the writing style and manner of a piece of text to determine whether it was written by a human or by generative AI.

[0685] A "probability model" is a statistical model for calculating the likelihood that a submitted text is the product of a generative AI.

[0686] A "machine learning algorithm" is a program that learns patterns and rules from data and uses that knowledge to analyze new data.

[0687] "Keyword detection" is the process of extracting specific words and phrases frequently used by generative AI from text.

[0688] "Analysis results" are information and data obtained through the application of natural language processing, stylistic analysis, and probabilistic models.

[0689] A "report" is a document that organizes the analysis results and presents them in a format that is easy for users to understand.

[0690] "Suspicious passages" are sections of text that are deemed likely to have been created by generative AI.

[0691] A "student" is a learner who submits assignments or papers.

[0692] "Submitted data" refers to text data such as assignments and papers that students upload to the system via their terminals.

[0693] This invention is a system that determines whether a student's submission was created by generative AI, and is composed of three parties: a server, a terminal, and a user.

[0694] Server Processing

[0695] The server receives the submitted data from the user's device. This can be done using a web framework such as Python's Flask. The received data is temporarily stored (e.g., in the " / tmp / uploads / " directory).

[0696] The server uses a natural language processing engine to analyze the received text data. Specifically, it uses libraries such as NLTK (Natural Language Toolkit) to tokenize the text and analyze its sentence structure. This breaks down the text data into words and phrases and understands their grammatical structure.

[0697] For stylistic analysis, we use a proprietary stylistic analysis algorithm that analyzes tokenized text features (e.g., word frequency, sentence length, syntactic patterns) to extract its style.

[0698] The server then applies a pre-trained probability model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is from a generative AI. This machine learning algorithm is based on patterns and rules learned from large datasets.

[0699] Additionally, the server detects keywords and phrases frequently used by the AI ​​generator (e.g., "In conclusion," "Based on the results") and calculates their frequency. This information is then incorporated into the results of the stylistic analysis.

[0700] Finally, the server generates a detailed analysis of each paragraph, including a report highlighting suspicious passages (e.g., "result_report.pdf"), which is then sent to the user's device. The user can view the report and identify any suspicious uses of the generated AI.

[0701] User operations

[0702] Users upload assignments and papers submitted by students to the system from their own devices. File uploads can be done through a web interface. The uploaded data is received by the server and temporarily stored.

[0703] The user can then review the analysis results generated by the server. The report includes a score for the suspected use of the generated AI and detailed analysis results with suspicious passages highlighted for each paragraph. The user can then use this information to ask questions of the student or conduct further research.

[0704] Specific examples

[0705] Example 1: Parsing a single essay

[0706] Users upload essays submitted by students (e.g., "assignment.txt") from their own devices to the server. The server receives the text data and uses the NLTK library to tokenize the text, analyze sentence structure, analyze stylistics, apply probabilistic models, and detect keywords. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0707] Example 2: Batch processing multiple essays

[0708] At the end of the semester, users upload multiple essays to the server in bulk. The server analyzes each essay individually and generates independent analysis results for each. The results are compiled into a single report for all essays and sent back to the user's device. The user then reviews the list and closely examines any essays suspected of using generative AI.

[0709] Prompt Sentence Examples

[0710] Determine whether the following sentence was written by a human or by a generative AI.

[0711] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0712] Step 1:

[0713] A user uploads text data (e.g., "assignment.txt") of assignments or papers submitted by students from their own terminal to the system. Specifically, they select a file using the web interface and click the "Upload" button. The input of this operation is the text data submitted by the student, and the output is the file being sent to the server.

[0714] Step 2:

[0715] The server receives text data sent from the user's device and saves it in a temporary storage area (e.g., the " / tmp / uploads / " directory). Specific operations include receiving an HTTP request and saving the data using a file stream. The input of this operation is the received text data, and the output is a temporarily saved file.

[0716] Step 3:

[0717] The server starts analyzing the temporarily stored text data using a natural language processing engine. First, it tokenizes the text using the NLTK library (e.g., it breaks it down into words using the word_tokenize function). The input of this operation is the stored text file, and the output is the tokenized text data.

[0718] Step 4:

[0719] The server performs sentence structure analysis based on the tokenized text data. It uses NLTK's pos_tag function to analyze the part of speech of each word and perform syntactic analysis. The input of this operation is the tokenized text data, and the output is text data with part-of-speech tags added.

[0720] Step 5:

[0721] The server performs stylistic analysis. Based on the results of tokenization and sentence structure analysis, it extracts text features (e.g., word frequency, sentence length, syntactic patterns) and applies a proprietary stylistic analysis algorithm. The input for this operation is text data with part-of-speech tags, and the output is stylistic feature data.

[0722] Step 6:

[0723] The server applies a pre-trained probabilistic model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is the work of a generative AI. Specifically, it inputs feature data and runs a machine learning algorithm to calculate a score. The input of this operation is the stylistic feature data, and the output is a generative AI suspicion score.

[0724] Step 7:

[0725] The server detects keywords and phrases frequently used by the AI. Using a predefined keyword list (e.g., "In conclusion," "Based on the results"), it searches for relevant passages in the text and incorporates their frequency of occurrence into the analysis. The input for this operation is the original text data and the keyword list, and the output is keyword frequency data.

[0726] Step 8:

[0727] The server generates detailed analysis results for each paragraph. Specifically, it combines the results of the stylistic analysis, the AI ​​suspicion score, and the keyword detection results to create a detailed report highlighting suspicious parts of each paragraph. The input of this operation is the analysis result data, and the output is the detailed report data.

[0728] Step 9:

[0729] The server then sends the generated report to the user's device, either as an email attachment or as a downloadable report file via a web interface. The input to this operation is the detailed report data, and the output is the report sent to the user.

[0730] Step 10:

[0731] The user opens the generated report on their device and checks the analysis results. The report includes a score for the suspected use of the generated AI and analysis results with suspicious passages highlighted for each paragraph. The user can use this information to ask questions of the student or conduct further research. The input for this operation is the generated report data, and the output is the user's judgment and response.

[0732] (Application example 1)

[0733] 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."

[0734] In online content distribution services, there is a problem in that it is difficult to determine whether posted text has been created by generative AI. This could lead to an increase in fraudulent posts by generative AI, which could result in a decline in content quality. In addition, a system is needed to quickly and accurately detect posts by generative AI.

[0735] 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.

[0736] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a suspicion score for use of a generation AI based on the analysis result, means for returning the analysis result to the user, and means for displaying a score for determining whether the submitted text was created by the generation AI and suspicious parts. This makes it possible to quickly detect fraudulent posts by generation AI in a content distribution service and maintain content quality.

[0737] "Data submitted by students" refers to text data submitted by students in the form of assignments, essays, etc.

[0738] A "natural language processing engine" is software that analyzes text data and processes it as structured information.

[0739] The "means of analysis" refers to the process of tokenizing text data using a natural language processing engine and performing stylistic analysis and keyword detection.

[0740] The "Generative AI Suspicion Score" is a score that indicates the likelihood that the submitted text was created by a generative AI.

[0741] A "probabilistic model" is a machine learning algorithm trained on pre-collected data to distinguish between generative AI and human-generated text.

[0742] "Means for returning the analysis results to the user" refers to the process of sending the analysis results performed on the server to the user's terminal and displaying them.

[0743] "Means for displaying scores and suspicious parts" refers to a function that displays the suspicion score used by the generating AI and highlights suspicious paragraphs and sentences.

[0744] The present invention provides a system for determining whether posted content in a content distribution service was created by a generation AI. An embodiment of the system will be described below.

[0745] The server receives text data from each user. The received text data is written content posted by the user, such as assignments, papers, and blog posts. This text data is temporarily stored, and then analysis begins using a natural language processing engine. The natural language processing engine uses Python's spaCy, which tokenizes the text and analyzes its sentence structure. It also detects writing style and keywords in the text and compares them with the characteristics of the generation AI.

[0746] The server then applies a probabilistic model to calculate the likelihood that the posted text was created by a generative AI. This model uses a machine learning algorithm, Scikit-Learn, and is trained on previously collected data. The model then generates a score to distinguish between generative AI and human-written text.

[0747] The analysis results in a score for the likelihood of the generated AI being used, which is then sent to the user's device. The user can then check the analysis results via a smartphone app. The analysis results include a score for the likelihood of the generated AI being used and detailed analysis results that highlight suspicious areas. Users can then use this information to correct or recheck the content.

[0748] For example, when a user uses this system before posting a blog post, the system checks in real time whether the content of the post was created by generative AI. When the user enters the blog post, the system displays the analysis results, such as a score of "likely AI-generated: 75%," and highlights suspicious paragraphs. Based on this information, the user can revise the post or check its quality.

[0749] An example of a prompt sentence would be:

[0750] Analyze the following text and determine whether it is the work of a generative AI. Provide a likelihood score for the generative AI and any suspicious areas.

[0751] In this way, generative AI in content distribution services detects fraudulent posts and maintains content quality.

[0752] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0753] Step 1:

[0754] The server receives text data submitted by the user. This data is uploaded by the user using a smartphone app. The input data is written data such as articles or essays. The server temporarily stores the received text data.

[0755] Step 2:

[0756] The server passes the received text data to a natural language processing engine and begins analysis. The natural language processing engine uses Python's spaCy to tokenize the text data and analyze its sentence structure. The input is text data, and the output is the tokenized text and its grammatical structure.

[0757] Step 3:

[0758] The server analyzes writing style and keywords based on the results of tokenization and sentence structure analysis. This analysis compares the characteristics of human-written text with those of the generative AI. Specifically, it detects frequently used keywords and phrases. The input is tokenized text, and the output is a list of writing style features and keywords.

[0759] Step 4:

[0760] The server applies a probabilistic model to calculate a suspicion score for the generative AI. This probabilistic model uses a machine learning algorithm using scikit-learn. Based on the analysis results, a score is calculated to determine whether the text submitted by the student was created by the generative AI. The input is the writing style and keyword features, and the output is the suspicion score for the generative AI.

[0761] Step 5:

[0762] The server generates a detailed analysis result report based on the generated AI's suspected use score. This report includes the generated AI's usability score and detailed analysis results with suspicious areas highlighted. The input is the generated AI's suspected use score and analysis results, and the output is the analysis result report.

[0763] Step 6:

[0764] The server sends the analysis result report to the user's device. The user can check the analysis results through a smartphone app. At this stage, the user can correct or recheck the article. The input is the analysis result report, and the output is the analysis results displayed to the user.

[0765] 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.

[0766] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback. Specifically, the roles of the server, terminal, and user are shown.

[0767] Server Processing

[0768] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine, which first performs tokenization and then grammatical structure analysis. A stylistic analysis is then performed, comparing the stylistic features of the human and generative AI versions.

[0769] The server then applies a probabilistic model to the analyzed data to calculate a probability score for the likelihood that the data was generated by the AI, while also detecting keywords and phrases frequently used by the AI ​​and calculating their frequency of occurrence.

[0770] The server then generates detailed analysis results for each paragraph based on these analysis results, and combines these results to generate a report that includes a score indicating the likelihood of using generative AI and the analysis results for each paragraph.

[0771] Emotion engine processing

[0772] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[0773] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[0774] User operations

[0775] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[0776] Specific examples

[0777] Example 1: Parsing a single essay

[0778] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0779] Example 2: Batch processing multiple essays

[0780] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[0781] The system of the present invention is an effective means for efficiently monitoring the use of generative AI in educational settings and providing feedback that takes into account the user's emotions, thereby achieving more appropriate instruction.

[0782] The processing flow will be explained below.

[0783] Step 1:

[0784] To upload text data for assignments or papers, users access the system's upload form from their terminal, select a file in the upload form, and click the submit button.

[0785] Step 2:

[0786] The server receives the text data sent by the user, temporarily stores the received data, and at the same time checks the file format and converts it into a format that can be analyzed as needed (e.g., txt, docx, pdf, etc.).

[0787] Step 3:

[0788] The server starts the natural language processing engine and prepares to analyze the text data. First, it performs tokenization, dividing the text into words.

[0789] Step 4:

[0790] The server then analyzes the tokenized data for grammatical structure, which involves understanding the structure of the sentence and checking for grammatical errors and patterns.

[0791] Step 5:

[0792] The server performs a stylistic analysis, specifically analyzing the lexical choice, rhythm, and grammatical patterns of a sentence, and comparing these characteristics with human and generative AI writing styles.

[0793] Step 6:

[0794] The server applies a probabilistic model to the parsed data, which calculates a probability score for the likelihood that the data was generated by the generative AI, using a pre-trained machine learning algorithm.

[0795] Step 7:

[0796] The server detects keywords and phrases frequently used by the AI ​​in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0797] Step 8:

[0798] The server generates detailed analysis results for each paragraph, including stylistic characteristics, probability scores, and keyword detection results for each paragraph.

[0799] Step 9:

[0800] The server compiles all the analysis results and creates a report, which includes a score for the likelihood of using a generative AI and the analysis results for each paragraph.

[0801] Step 10:

[0802] The server starts the emotion engine and acquires data to recognize the user's emotions, which are determined from the user's facial expressions, voice, input text, etc.

[0803] Step 11:

[0804] The server adjusts the display of the analysis results based on the results of the emotion engine, for example adding a warning message if the score is high or providing reassurance information if the user appears anxious.

[0805] Step 12:

[0806] The server then sends the generated report to the user's device, which includes the analysis results and emotional feedback.

[0807] Step 13:

[0808] The user checks the report received on the device, considers guidance and countermeasures based on the analysis results and feedback, and provides additional questions and guidance to the student as needed.

[0809] The above is an explanation of the processing steps including the division of roles between the user, server, and terminal, and specific operations.

[0810] Example 2

[0811] 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."

[0812] With conventional systems, it was difficult to analyze the writing style of assignments and papers submitted by students and determine whether they were using generative AI. Furthermore, simply returning the analysis results did not allow for appropriate feedback that took into account the user's level of understanding and psychological state. This made it difficult to provide efficient and effective instruction and support, preventing the maximum educational effect.

[0813] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for receiving data submitted by students, a means for analyzing the data using a natural language processing engine, a means for applying a probabilistic model that calculates a usage suspicion score of the generation AI based on the analysis result, a means for recognizing the analysis result and the user's emotions and adjusting the display method, and a means for returning the analysis result to the user. This makes it possible to accurately determine the use of the generation AI and provide appropriate feedback that takes the user's emotions into consideration.

[0814] A "student" is a learner who is enrolled in an educational institution and submits assignments and papers.

[0815] "Submitted data" refers to text data or document files that students upload to the server as assignments or papers.

[0816] "Receiving means" refers to the function or mechanism by which the server receives the submitted data sent from the user's terminal.

[0817] A "natural language processing engine" is a software program that analyzes text data and performs various language processing operations such as tokenization, grammatical structure analysis, and stylistic analysis.

[0818] A "probability model" is a statistical model or machine learning model used by the server to calculate the suspected use score of the generating AI.

[0819] "Emotion recognition means" refers to software or hardware functions that determine emotions from a user's facial expressions, voice, or text.

[0820] The "display method adjustment means" refers to a function that changes the way the analysis results are presented based on the emotion recognition results, and provides optimal feedback to the user.

[0821] "Return means" refers to the function or mechanism for returning the analysis results to the user.

[0822] "Stylistic analysis" is a method of analyzing a text to determine whether the writing style was created by a human or by a generative AI.

[0823] "Paragraph analysis means" refers to a function or mechanism that generates detailed analysis results for each paragraph of submitted data.

[0824] MODE FOR CARRYING OUT THE INVENTION

[0825] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback.

[0826] Server Processing

[0827] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. For example, the data is cleaned and normalized using Python's pandas library or nltk (Natural Language Toolkit).

[0828] The server then runs a natural language processing engine (e.g., spaCy or BERT) to first tokenize the text, then analyze its grammar. It then performs a stylistic analysis, comparing the stylistic features of the human and generative AI versions of the text. Specifically, it evaluates metrics such as the length of each sentence, the specific syntax it contains, and the vocabulary used.

[0829] Next, the server applies a probabilistic model (e.g., Bayesian inference or machine learning model) to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the generative AI. At the same time, it calculates the frequency of keywords and phrases frequently used by the generative AI. This step uses Python's scikit-learn and tf-idf-based analysis.

[0830] The analysis results are integrated and a report is generated that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. This report is generated using a report template and can be output in HTML or PDF format.

[0831] Emotion engine processing

[0832] Before returning the analysis results to the user, the server uses an emotion engine (such as IBM Watson or Microsoft Azure's emotion analysis service) to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[0833] Based on the results of emotion recognition, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of generative AI, it adds a message urging the user to take appropriate measures. Also, if the user is feeling anxious, it provides advice to ease their anxiety.

[0834] User operations

[0835] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[0836] Specific examples

[0837] Example 1: Parsing a single essay

[0838] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text, cleans and normalizes the data, and analyzes it using a natural language processing engine. The analysis involves tokenization, grammatical structure analysis, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0839] Specific prompt:

[0840] "Analyze the quality of AI-generated essays submitted by students and provide feedback to students along with the results."

[0841] Example 2: Batch processing multiple essays

[0842] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server receives these essays, prepares them for analysis individually, and then analyzes each essay using a natural language processing engine and an emotion engine. It then generates individual analysis results for each essay, compiles them into a single report, and returns them to the user.

[0843] Specific prompt:

[0844] "Please analyze the quality of AI-generated essays submitted at the end of the semester in a batch process and provide feedback to each student."

[0845] This system is an effective means of efficiently monitoring the use of generative AI in educational settings and providing feedback that takes users' emotions into account, thereby achieving more appropriate instruction.

[0846] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0847] Step 1:

[0848] Submission data is uploaded from the user's device. The user uses their own device to send text data of assignments or papers to the server through a specified interface (browser or application). The input is a text file created by the student, and the output is data stored in a temporary storage area on the server.

[0849] Step 2:

[0850] The server receives the submitted data and stores it in a temporary storage area. The server receives the text data uploaded by the user and checks its format. For example, converting a PDF file to TXT format. The input is the file uploaded in step 1, and the output is the cleaned text data.

[0851] Step 3:

[0852] The server cleans and normalizes the data. Specifically, it uses Python's pandas library and nltk (Natural Language Toolkit) to remove unnecessary characters and line breaks, standardize case, and standardize specific characters. The input is the text data saved in step 2, and the output is cleaned data ready for analysis.

[0853] Step 4:

[0854] The server runs a natural language processing engine and performs tokenization. The server uses a natural language processing engine, such as spaCy or BERT, to split the text into words and phrases. The input is the text data cleaned in step 3, and the output is the tokenized data.

[0855] Step 5:

[0856] The server performs grammatical structure analysis. Using the tokenized data, it analyzes the structure of each sentence and clarifies the relationships between subjects, predicates, objects, etc. The input is the data tokenized in step 4, and the output is the result of analyzing the grammatical structure.

[0857] Step 6:

[0858] The server performs a stylistic analysis. Based on the analyzed sentences, it compares the differences in the specific writing styles of humans and the generative AI. The stylistic analysis includes indicators such as the length of each sentence, the specific syntax contained in the sentence, and the vocabulary used. The input is the result of the grammatical structure analysis in step 5, and the output is the result of the stylistic analysis.

[0859] Step 7:

[0860] The server applies a probabilistic model to calculate a generative AI suspicion score. Based on the analysis results, it uses Bayesian inference and machine learning models to calculate a probability score indicating the likelihood that the data was created by a generative AI. The input is the result of the stylistic analysis in step 6, and the output is a generative AI suspicion score.

[0861] Step 8:

[0862] The server calculates the frequency of keywords and phrases. The generation AI detects frequently used keywords and phrases and calculates their frequency. The input is the same data as the probability score in step 7, and the output is the keyword frequency result.

[0863] Step 9:

[0864] The server integrates the analysis results and generates a report. It integrates the detailed analysis results for each paragraph and generates a report that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. A report template is used to generate this report, and it is output in HTML or PDF format. The input is the analysis results from Step 7 and Step 8, and the output is the final report.

[0865] Step 10:

[0866] The server uses an emotion engine to recognize the user's emotions. Before returning the analysis results to the user, it uses IBM Watson or Microsoft Azure's emotion analysis services to determine the user's emotions from their facial expressions, voice, and text. The input is the user's emotion data, and the output is the emotion recognition results.

[0867] Step 11:

[0868] The server adjusts the display method based on the emotion recognition results. Based on the results of the emotion engine, it changes the way the analysis results are displayed and provides optimal feedback to the user. For example, if there is a high suspicion of the use of generative AI, it adds a message offering careful guidance. The input is the emotion recognition result from step 10, and the output is the analysis result with adjusted feedback.

[0869] Step 12:

[0870] The server returns the analysis results to the user. The final report and sentiment-based feedback are sent to the user's device. The user reviews the report and makes necessary corrections or adjustments to the essay. The input is the analysis results with feedback adjusted in step 11, and the output is the final report displayed on the user's device.

[0871] (Application example 2)

[0872] 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."

[0873] Conventional chatbot systems have limitations in their ability to generate automatic responses to user input. In particular, as text generation using generative AI becomes more common, systems that can appropriately recognize a user's emotions and provide feedback based on those emotions are required. However, systems that provide feedback that takes emotions into account are rare, and have not contributed to improving user satisfaction. It has also been difficult to determine whether or not a generative AI is being used and provide a reliable service. Therefore, the present invention aims to solve these problems by providing a new system that combines emotion recognition and generative AI discrimination.

[0874] 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.

[0875] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a score for suspicion of use of generative AI based on the analysis results, means for returning the analysis results to the user, and an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results. This makes it possible to determine the suspicion of use of generative AI while taking the user's emotions into consideration.

[0876] "Means for receiving data submitted by students" refers to a function for receiving text data such as assignments and papers uploaded by users to the system.

[0877] "Means for analyzing using a natural language processing engine" refers to a function that analyzes uploaded text data using techniques such as tokenization, grammatical structure analysis, and stylistic analysis.

[0878] The "means for applying a probabilistic model to calculate a score for suspected use of generative AI" is a function that calculates a score for the likelihood that a text was generated by generative AI based on the analyzed data.

[0879] The "means for returning the analysis results to the user" is a function for presenting the results of the analysis performed by the server to the user.

[0880] "Means including an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results" refers to a function that determines the user's emotions from their facial expressions, voice, input text, etc., and adjusts the feedback method based on the results.

[0881] The present invention is a system that receives data submitted by users, analyzes it, identifies suspected use of generative AI, and recognizes the user's emotions to provide appropriate feedback. This system clarifies the roles of the server, terminal, and user, and explains how they interact. A specific example is provided below.

[0882] Server Processing

[0883] The server receives text data for assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine (NLP engine) that first performs tokenization. It then performs grammatical structure analysis and stylistic analysis, comparing the stylistic features of human and generative AI writing. Examples of specific software used for analysis include Python's nltk and transformers libraries.

[0884] Next, the server applies a probabilistic model to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the AI. At the same time, it detects keywords and phrases frequently used by the AI ​​and calculates their frequency of occurrence. A possible model to use is the BERT model. This allows for a suspicion score for the AI.

[0885] Emotion engine processing

[0886] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results. The TextBlob library can be used for emotion recognition.

[0887] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[0888] User operations

[0889] Users upload text data of assignments and papers from their own devices to the system. The data uploaded by the user is analyzed on the server, and the results are returned along with the results of the emotion engine. This allows users to receive emotion-based feedback along with the analysis results, enabling them to take better guidance and measures.

[0890] Specific examples

[0891] 1. Analyzing a Single Essay

[0892] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[0893] 2. Batch processing multiple essays

[0894] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[0895] Prompt Sentence Examples

[0896] User input: "The delivery was late and the product is damaged."

[0897] Tokenization: ['The', 'delivery', 'was', 'late', 'and', 'the', 'product', 'is', 'damaged', '.']

[0898] Style analysis: Detecting stylistic features with generative AI

[0899] Emotion Recognition Result: Extremely Negative (Emotion Polarity: -1)

[0900] Feedback: "I sincerely apologize for the inconvenience. Can you provide more details or images so that we can resolve this issue for you?"

[0901] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0902] Step 1: User uploads data from device

[0903] ---

[0904] Users upload text data for assignments or papers from their own devices to the system. The input is a text file prepared by the user, and the output is that text file sent to the server. Specifically, users send the text data using the file upload function of a browser or a dedicated application.

[0905] Step 2: The server stores the received data and prepares it for analysis.

[0906] ---

[0907] The server receives data uploaded by the user and stores it in a temporary storage area. The input is the text data sent by the user, and the output is text data ready for analysis. Specifically, the server stores the received file in file storage and prepares it to be passed to the analysis module.

[0908] Step 3: Text analysis using a natural language processing engine

[0909] ---

[0910] The server applies a natural language processing engine to the received text data, performing tokenization, grammatical structure analysis, and stylistic analysis. The input is the stored text data, and the output is the analysis results. Specifically, the server performs tokenization and grammatical analysis using the Python nltk and transformers libraries.

[0911] Step 4: Calculate the generative AI usage suspicion score

[0912] ---

[0913] Based on the analysis results, the server applies a probability model to calculate a probability score indicating the suspected use of generative AI. The input is the analyzed text data, and the output is a probability score indicating the suspected use of generative AI. Specifically, it uses the BERT model to calculate the likelihood that the text was created by generative AI.

[0914] Step 5: Analyze user emotions using an emotion recognition engine

[0915] ---

[0916] The server uses an emotion recognition engine to recognize the user's emotions before returning the analysis results to the user. The input is the text message, voice, or image data entered by the user, and the output is the emotion recognition result. Specifically, it analyzes the user's emotions using the TextBlob library.

[0917] Step 6: Integrating analytical results with emotional feedback

[0918] ---

[0919] The server integrates the generated AI suspicion score and the emotion recognition result to generate a feedback report. The input is the generated AI suspicion score and the emotion recognition result, and the output is the integrated feedback report. Specifically, the server adds a warning message based on the emotion recognition result if careful action is required, and generates feedback according to the user's emotions.

[0920] Step 7: Returning the analysis results to the user

[0921] ---

[0922] The server returns the generated feedback report to the user. The input is the integrated feedback report, and the output is the analysis result and feedback received by the user. Specifically, the server sends the analysis result to the user's device for the user to check.

[0923] Through these steps, users can receive analysis results that determine whether or not to use the generative AI, as well as feedback based on their own emotions.

[0924] 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.

[0925] 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.

[0926] 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.

[0927] [Fourth embodiment]

[0928] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[0929] 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.

[0930] 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).

[0931] 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.

[0932] 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.

[0933] 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).

[0934] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0935] 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.

[0936] 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.

[0937] 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.

[0938] 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.

[0939] 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.

[0940] 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."

[0941] The following describes an embodiment of the present invention. The present invention is a system that determines whether a student's submission was created by a generative AI, and specifically describes the roles of the server, terminal, and user.

[0942] Server Processing

[0943] The server receives text data sent by users from their devices. This data is text submitted by students, such as assignments and papers. After the text data received by the server is temporarily stored, it is analyzed by a natural language processing engine. The natural language processing engine first tokenizes the text and analyzes the sentence structure. Next, a stylistic analysis is performed based on the analysis results, and the characteristics of the text written by a human and the text generated by the AI ​​are compared.

[0944] After stylistic analysis, the server applies a probability model to calculate the likelihood that the submitted text is the work of a generative AI. This probability model is a machine learning algorithm trained on pre-collected data, and generates a score to distinguish between generative AI and human writing.

[0945] In addition, the server detects frequently used keywords and phrases in the text and incorporates their frequency into the analysis, generating detailed analysis results for each paragraph.

[0946] Finally, the server generates a report based on these analysis results and sends it to the user's device, including a score indicating the likelihood of using AI and highlighting suspicious parts of each paragraph.

[0947] User operations

[0948] Users upload text data for assignments or papers from their own devices to the system. The uploaded data is received by the server, and once the analysis results are generated, users can check them. If users have any doubts about the use of the generated AI, they can ask questions or conduct additional research on the student.

[0949] Specific examples

[0950] Example 1: Parsing a single essay

[0951] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[0952] Example 2: Batch processing multiple essays

[0953] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes the essays individually and generates independent analysis results for each essay. The results are compiled into a single report for all essays and sent back to the user's device. The user then checks the list and closely examines essays suspected of using generative AI.

[0954] The system of the present invention is an effective means of efficiently monitoring the use of generative AI in educational settings and maintaining the quality of student learning.

[0955] The processing flow will be explained below.

[0956] Step 1:

[0957] Users upload text data for assignments and papers from their devices. When doing so, they use the system's upload form and specify the file they wish to submit.

[0958] Step 2:

[0959] The server receives the text data sent by the user and temporarily stores it, checking the file format (txt, docx, pdf, etc.) and converting it into a format that can be analyzed.

[0960] Step 3:

[0961] The server inputs the received text data into a natural language processing engine, which first tokenizes the text and divides sentences into words.

[0962] Step 4:

[0963] The server analyzes the text data for grammatical structure, which involves recognizing sentence structure and checking for grammatical errors.

[0964] Step 5:

[0965] The server runs a stylistic analysis, which analyzes the lexical choice, rhythm, and grammatical patterns of a sentence to identify differences between human and generative AI writing styles.

[0966] Step 6:

[0967] The server then applies a probabilistic model to the parsed data, which uses a pre-trained machine learning algorithm to calculate the probability that a given piece of text is the product of a generative AI.

[0968] Step 7:

[0969] The server detects keywords and phrases frequently used by the AI ​​generator in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[0970] Step 8:

[0971] The server generates a detailed analysis for each paragraph, including its stylistic characteristics, probability scores, and keyword detection results.

[0972] Step 9:

[0973] The server compiles the overall analysis results and creates a report, which includes a score for the likelihood of using generative AI and analysis results for each paragraph.

[0974] Step 10:

[0975] The server sends the generated report to the user's device, where the user can review the report and take measures to provide learning guidance or prevent the use of the generated AI as needed.

[0976] Example 1

[0977] 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."

[0978] In modern education, there is a need to efficiently identify student submissions created by generative AI and maintain learning quality. However, existing methods face the challenge of accurately distinguishing between human writing and generative AI writing. Analyzing submissions takes a significant amount of time and effort, especially when there are a large number of submissions. To solve this problem, an automated system is needed.

[0979] 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.

[0980] In this invention, the server includes means for receiving student submission data sent from the terminal, means for analyzing the data using a natural language processing engine that tokenizes the data and analyzes the sentence structure, means for analyzing the writing style based on the analysis results and comparing it with the characteristics of the generative AI, means for applying a probabilistic model that calculates a usage suspicion score for the generative AI based on the analysis results, means for detecting keywords and phrases frequently used by the generative AI and incorporating their frequency of occurrence into the analysis, means for generating detailed analysis results for each paragraph and generating a report based on the results, and means for returning the generated report to the user. This makes it possible to automatically and efficiently determine the possibility that submitted text was created by the generative AI.

[0981] A "terminal" is an electronic device that a user uses to upload text data of assignments or papers to the system.

[0982] A "server" is a central processing unit that receives data sent from user terminals, analyzes, stores, and generates reports.

[0983] A "natural language processing engine" is software or algorithms that tokenize text data, analyze sentence structure, and perform stylistic analysis.

[0984] "Tokenization" is the process of dividing text into words and phrases.

[0985] "Sentence structure analysis" is the process of analyzing the grammatical structure of a sentence.

[0986] "Stylistic analysis" is a method of analyzing the writing style and manner of a piece of text to determine whether it was written by a human or by generative AI.

[0987] A "probability model" is a statistical model for calculating the likelihood that a submitted text is the product of a generative AI.

[0988] A "machine learning algorithm" is a program that learns patterns and rules from data and uses that knowledge to analyze new data.

[0989] "Keyword detection" is the process of extracting specific words and phrases frequently used by generative AI from text.

[0990] "Analysis results" are information and data obtained through the application of natural language processing, stylistic analysis, and probabilistic models.

[0991] A "report" is a document that organizes the analysis results and presents them in a format that is easy for users to understand.

[0992] "Suspicious passages" are sections of text that are deemed likely to have been created by generative AI.

[0993] A "student" is a learner who submits assignments or papers.

[0994] "Submitted data" refers to text data such as assignments and papers that students upload to the system via their terminals.

[0995] This invention is a system that determines whether a student's submission was created by generative AI, and is composed of three parties: a server, a terminal, and a user.

[0996] Server Processing

[0997] The server receives the submitted data from the user's device. This can be done using a web framework such as Python's Flask. The received data is temporarily stored (e.g., in the " / tmp / uploads / " directory).

[0998] The server uses a natural language processing engine to analyze the received text data. Specifically, it uses libraries such as NLTK (Natural Language Toolkit) to tokenize the text and analyze its sentence structure. This breaks down the text data into words and phrases and understands their grammatical structure.

[0999] For stylistic analysis, we use a proprietary stylistic analysis algorithm that analyzes tokenized text features (e.g., word frequency, sentence length, syntactic patterns) to extract its style.

[1000] The server then applies a pre-trained probability model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is from a generative AI. This machine learning algorithm is based on patterns and rules learned from large datasets.

[1001] Additionally, the server detects keywords and phrases frequently used by the AI ​​generator (e.g., "In conclusion," "Based on the results") and calculates their frequency. This information is then incorporated into the results of the stylistic analysis.

[1002] Finally, the server generates a detailed analysis of each paragraph, including a report highlighting suspicious passages (e.g., "result_report.pdf"), which is then sent to the user's device. The user can view the report and identify any suspicious uses of the generated AI.

[1003] User operations

[1004] Users upload assignments and papers submitted by students to the system from their own devices. File uploads can be done through a web interface. The uploaded data is received by the server and temporarily stored.

[1005] The user can then review the analysis results generated by the server. The report includes a score for the suspected use of the generated AI and detailed analysis results with suspicious passages highlighted for each paragraph. The user can then use this information to ask questions of the student or conduct further research.

[1006] Specific examples

[1007] Example 1: Parsing a single essay

[1008] Users upload essays submitted by students (e.g., "assignment.txt") from their own devices to the server. The server receives the text data and uses the NLTK library to tokenize the text, analyze sentence structure, analyze stylistics, apply probabilistic models, and detect keywords. The final analysis results are returned to the device as a report in a format that the user can intuitively understand.

[1009] Example 2: Batch processing multiple essays

[1010] At the end of the semester, users upload multiple essays to the server in bulk. The server analyzes each essay individually and generates independent analysis results for each. The results are compiled into a single report for all essays and sent back to the user's device. The user then reviews the list and closely examines any essays suspected of using generative AI.

[1011] Prompt Sentence Examples

[1012] Determine whether the following sentence was written by a human or by a generative AI.

[1013] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1014] Step 1:

[1015] A user uploads text data (e.g., "assignment.txt") of assignments or papers submitted by students from their own terminal to the system. Specifically, they select a file using the web interface and click the "Upload" button. The input of this operation is the text data submitted by the student, and the output is the file being sent to the server.

[1016] Step 2:

[1017] The server receives text data sent from the user's device and saves it in a temporary storage area (e.g., the " / tmp / uploads / " directory). Specific operations include receiving an HTTP request and saving the data using a file stream. The input of this operation is the received text data, and the output is a temporarily saved file.

[1018] Step 3:

[1019] The server starts analyzing the temporarily stored text data using a natural language processing engine. First, it tokenizes the text using the NLTK library (e.g., it breaks it down into words using the word_tokenize function). The input of this operation is the stored text file, and the output is the tokenized text data.

[1020] Step 4:

[1021] The server performs sentence structure analysis based on the tokenized text data. It uses NLTK's pos_tag function to analyze the part of speech of each word and perform syntactic analysis. The input of this operation is the tokenized text data, and the output is text data with part-of-speech tags added.

[1022] Step 5:

[1023] The server performs stylistic analysis. Based on the results of tokenization and sentence structure analysis, it extracts text features (e.g., word frequency, sentence length, syntactic patterns) and applies a proprietary stylistic analysis algorithm. The input for this operation is text data with part-of-speech tags, and the output is stylistic feature data.

[1024] Step 6:

[1025] The server applies a pre-trained probabilistic model (e.g., the BERT model using Hugging Face's Transformers library) to calculate the likelihood that the text is the work of a generative AI. Specifically, it inputs feature data and runs a machine learning algorithm to calculate a score. The input of this operation is the stylistic feature data, and the output is a generative AI suspicion score.

[1026] Step 7:

[1027] The server detects keywords and phrases frequently used by the AI. Using a predefined keyword list (e.g., "In conclusion," "Based on the results"), it searches for relevant passages in the text and incorporates their frequency of occurrence into the analysis. The input for this operation is the original text data and the keyword list, and the output is keyword frequency data.

[1028] Step 8:

[1029] The server generates detailed analysis results for each paragraph. Specifically, it combines the results of the stylistic analysis, the AI ​​suspicion score, and the keyword detection results to create a detailed report highlighting suspicious parts of each paragraph. The input of this operation is the analysis result data, and the output is the detailed report data.

[1030] Step 9:

[1031] The server then sends the generated report to the user's device, either as an email attachment or as a downloadable report file via a web interface. The input to this operation is the detailed report data, and the output is the report sent to the user.

[1032] Step 10:

[1033] The user opens the generated report on their device and checks the analysis results. The report includes a score for the suspected use of the generated AI and analysis results with suspicious passages highlighted for each paragraph. The user can use this information to ask questions of the student or conduct further research. The input for this operation is the generated report data, and the output is the user's judgment and response.

[1034] (Application example 1)

[1035] 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."

[1036] In online content distribution services, there is a problem in that it is difficult to determine whether posted text has been created by generative AI. This could lead to an increase in fraudulent posts by generative AI, which could result in a decline in content quality. In addition, a system is needed to quickly and accurately detect posts by generative AI.

[1037] 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.

[1038] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a suspicion score for use of a generation AI based on the analysis result, means for returning the analysis result to the user, and means for displaying a score for determining whether the submitted text was created by the generation AI and suspicious parts. This makes it possible to quickly detect fraudulent posts by generation AI in a content distribution service and maintain content quality.

[1039] "Data submitted by students" refers to text data submitted by students in the form of assignments, essays, etc.

[1040] A "natural language processing engine" is software that analyzes text data and processes it as structured information.

[1041] The "means of analysis" refers to the process of tokenizing text data using a natural language processing engine and performing stylistic analysis and keyword detection.

[1042] The "Generative AI Suspicion Score" is a score that indicates the likelihood that the submitted text was created by a generative AI.

[1043] A "probabilistic model" is a machine learning algorithm trained on pre-collected data to distinguish between generative AI and human-generated text.

[1044] "Means for returning the analysis results to the user" refers to the process of sending the analysis results performed on the server to the user's terminal and displaying them.

[1045] "Means for displaying scores and suspicious parts" refers to a function that displays the suspicion score used by the generating AI and highlights suspicious paragraphs and sentences.

[1046] The present invention provides a system for determining whether posted content in a content distribution service was created by a generation AI. An embodiment of the system will be described below.

[1047] The server receives text data from each user. The received text data is written content posted by the user, such as assignments, papers, and blog posts. This text data is temporarily stored, and then analysis begins using a natural language processing engine. The natural language processing engine uses Python's spaCy, which tokenizes the text and analyzes its sentence structure. It also detects writing style and keywords in the text and compares them with the characteristics of the generation AI.

[1048] The server then applies a probabilistic model to calculate the likelihood that the posted text was created by a generative AI. This model uses a machine learning algorithm, Scikit-Learn, and is trained on previously collected data. The model then generates a score to distinguish between generative AI and human-written text.

[1049] The analysis results in a score for the likelihood of the generated AI being used, which is then sent to the user's device. The user can then check the analysis results via a smartphone app. The analysis results include a score for the likelihood of the generated AI being used and detailed analysis results that highlight suspicious areas. Users can then use this information to correct or recheck the content.

[1050] For example, when a user uses this system before posting a blog post, the system checks in real time whether the content of the post was created by generative AI. When the user enters the blog post, the system displays the analysis results, such as a score of "likely AI-generated: 75%," and highlights suspicious paragraphs. Based on this information, the user can revise the post or check its quality.

[1051] An example of a prompt sentence would be:

[1052] Analyze the following text and determine whether it is the work of a generative AI. Provide a likelihood score for the generative AI and any suspicious areas.

[1053] In this way, generative AI in content distribution services detects fraudulent posts and maintains content quality.

[1054] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1055] Step 1:

[1056] The server receives text data submitted by the user. This data is uploaded by the user using a smartphone app. The input data is written data such as articles or essays. The server temporarily stores the received text data.

[1057] Step 2:

[1058] The server passes the received text data to a natural language processing engine and begins analysis. The natural language processing engine uses Python's spaCy to tokenize the text data and analyze its sentence structure. The input is text data, and the output is the tokenized text and its grammatical structure.

[1059] Step 3:

[1060] The server analyzes writing style and keywords based on the results of tokenization and sentence structure analysis. This analysis compares the characteristics of human-written text with those of the generative AI. Specifically, it detects frequently used keywords and phrases. The input is tokenized text, and the output is a list of writing style features and keywords.

[1061] Step 4:

[1062] The server applies a probabilistic model to calculate a suspicion score for the generative AI. This probabilistic model uses a machine learning algorithm using scikit-learn. Based on the analysis results, a score is calculated to determine whether the text submitted by the student was created by the generative AI. The input is the writing style and keyword features, and the output is the suspicion score for the generative AI.

[1063] Step 5:

[1064] The server generates a detailed analysis result report based on the generated AI's suspected use score. This report includes the generated AI's usability score and detailed analysis results with suspicious areas highlighted. The input is the generated AI's suspected use score and analysis results, and the output is the analysis result report.

[1065] Step 6:

[1066] The server sends the analysis result report to the user's device. The user can check the analysis results through a smartphone app. At this stage, the user can correct or recheck the article. The input is the analysis result report, and the output is the analysis results displayed to the user.

[1067] 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.

[1068] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback. Specifically, the roles of the server, terminal, and user are shown.

[1069] Server Processing

[1070] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine, which first performs tokenization and then grammatical structure analysis. A stylistic analysis is then performed, comparing the stylistic features of the human and generative AI versions.

[1071] The server then applies a probabilistic model to the analyzed data to calculate a probability score for the likelihood that the data was generated by the AI, while also detecting keywords and phrases frequently used by the AI ​​and calculating their frequency of occurrence.

[1072] The server then generates detailed analysis results for each paragraph based on these analysis results, and combines these results to generate a report that includes a score indicating the likelihood of using generative AI and the analysis results for each paragraph.

[1073] Emotion engine processing

[1074] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[1075] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[1076] User operations

[1077] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[1078] Specific examples

[1079] Example 1: Parsing a single essay

[1080] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[1081] Example 2: Batch processing multiple essays

[1082] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[1083] The system of the present invention is an effective means for efficiently monitoring the use of generative AI in educational settings and providing feedback that takes into account the user's emotions, thereby achieving more appropriate instruction.

[1084] The processing flow will be explained below.

[1085] Step 1:

[1086] To upload text data for assignments or papers, users access the system's upload form from their terminal, select a file in the upload form, and click the submit button.

[1087] Step 2:

[1088] The server receives the text data sent by the user, temporarily stores the received data, and at the same time checks the file format and converts it into a format that can be analyzed as needed (e.g., txt, docx, pdf, etc.).

[1089] Step 3:

[1090] The server starts the natural language processing engine and prepares to analyze the text data. First, it performs tokenization, dividing the text into words.

[1091] Step 4:

[1092] The server then analyzes the tokenized data for grammatical structure, which involves understanding the structure of the sentence and checking for grammatical errors and patterns.

[1093] Step 5:

[1094] The server performs a stylistic analysis, specifically analyzing the lexical choice, rhythm, and grammatical patterns of a sentence, and comparing these characteristics with human and generative AI writing styles.

[1095] Step 6:

[1096] The server applies a probabilistic model to the parsed data, which calculates a probability score for the likelihood that the data was generated by the generative AI, using a pre-trained machine learning algorithm.

[1097] Step 7:

[1098] The server detects keywords and phrases frequently used by the AI ​​in the text, using a search algorithm based on specific patterns and calculating their frequency of occurrence.

[1099] Step 8:

[1100] The server generates detailed analysis results for each paragraph, including stylistic characteristics, probability scores, and keyword detection results for each paragraph.

[1101] Step 9:

[1102] The server compiles all the analysis results and creates a report, which includes a score for the likelihood of using a generative AI and the analysis results for each paragraph.

[1103] Step 10:

[1104] The server starts the emotion engine and acquires data to recognize the user's emotions, which are determined from the user's facial expressions, voice, input text, etc.

[1105] Step 11:

[1106] The server adjusts the display of the analysis results based on the results of the emotion engine, for example adding a warning message if the score is high or providing reassurance information if the user appears anxious.

[1107] Step 12:

[1108] The server then sends the generated report to the user's device, which includes the analysis results and emotional feedback.

[1109] Step 13:

[1110] The user checks the report received on the device, considers guidance and countermeasures based on the analysis results and feedback, and provides additional questions and guidance to the student as needed.

[1111] The above is an explanation of the processing steps including the division of roles between the user, server, and terminal, and specific operations.

[1112] Example 2

[1113] 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."

[1114] With conventional systems, it was difficult to analyze the writing style of assignments and papers submitted by students and determine whether they were using generative AI. Furthermore, simply returning the analysis results did not allow for appropriate feedback that took into account the user's level of understanding and psychological state. This made it difficult to provide efficient and effective instruction and support, preventing the maximum educational effect.

[1115] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for receiving data submitted by students, a means for analyzing the data using a natural language processing engine, a means for applying a probabilistic model that calculates a usage suspicion score of the generation AI based on the analysis result, a means for recognizing the analysis result and the user's emotions and adjusting the display method, and a means for returning the analysis result to the user. This makes it possible to accurately determine the use of the generation AI and provide appropriate feedback that takes the user's emotions into consideration.

[1116] A "student" is a learner who is enrolled in an educational institution and submits assignments and papers.

[1117] "Submitted data" refers to text data or document files that students upload to the server as assignments or papers.

[1118] "Receiving means" refers to the function or mechanism by which the server receives the submitted data sent from the user's terminal.

[1119] A "natural language processing engine" is a software program that analyzes text data and performs various language processing operations such as tokenization, grammatical structure analysis, and stylistic analysis.

[1120] A "probability model" is a statistical model or machine learning model used by the server to calculate the suspected use score of the generating AI.

[1121] "Emotion recognition means" refers to software or hardware functions that determine emotions from a user's facial expressions, voice, or text.

[1122] The "display method adjustment means" refers to a function that changes the way the analysis results are presented based on the emotion recognition results, and provides optimal feedback to the user.

[1123] "Return means" refers to the function or mechanism for returning the analysis results to the user.

[1124] "Stylistic analysis" is a method of analyzing a text to determine whether the writing style was created by a human or by a generative AI.

[1125] "Paragraph analysis means" refers to a function or mechanism that generates detailed analysis results for each paragraph of submitted data.

[1126] MODE FOR CARRYING OUT THE INVENTION

[1127] The following describes an embodiment of the present invention. The present invention is a system that uses a generative AI to distinguish between sentences, and by combining it with an emotion engine that recognizes the user's emotions, it provides more appropriate feedback.

[1128] Server Processing

[1129] The server receives the text data of assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. For example, the data is cleaned and normalized using Python's pandas library or nltk (Natural Language Toolkit).

[1130] The server then runs a natural language processing engine (e.g., spaCy or BERT) to first tokenize the text, then analyze its grammar. It then performs a stylistic analysis, comparing the stylistic features of the human and generative AI versions of the text. Specifically, it evaluates metrics such as the length of each sentence, the specific syntax it contains, and the vocabulary used.

[1131] Next, the server applies a probabilistic model (e.g., Bayesian inference or machine learning model) to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the generative AI. At the same time, it calculates the frequency of keywords and phrases frequently used by the generative AI. This step uses Python's scikit-learn and tf-idf-based analysis.

[1132] The analysis results are integrated and a report is generated that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. This report is generated using a report template and can be output in HTML or PDF format.

[1133] Emotion engine processing

[1134] Before returning the analysis results to the user, the server uses an emotion engine (such as IBM Watson or Microsoft Azure's emotion analysis service) to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results.

[1135] Based on the results of emotion recognition, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of generative AI, it adds a message urging the user to take appropriate measures. Also, if the user is feeling anxious, it provides advice to ease their anxiety.

[1136] User operations

[1137] Users upload text data for assignments and papers to the system from their own devices, and receive emotional feedback along with analysis results returned from the server, enabling them to take better guidance and measures.

[1138] Specific examples

[1139] Example 1: Parsing a single essay

[1140] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text, cleans and normalizes the data, and analyzes it using a natural language processing engine. The analysis involves tokenization, grammatical structure analysis, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[1141] Specific prompt:

[1142] "Analyze the quality of AI-generated essays submitted by students and provide feedback to students along with the results."

[1143] Example 2: Batch processing multiple essays

[1144] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server receives these essays, prepares them for analysis individually, and then analyzes each essay using a natural language processing engine and an emotion engine. It then generates individual analysis results for each essay, compiles them into a single report, and returns them to the user.

[1145] Specific prompt:

[1146] "Please analyze the quality of AI-generated essays submitted at the end of the semester in a batch process and provide feedback to each student."

[1147] This system is an effective means of efficiently monitoring the use of generative AI in educational settings and providing feedback that takes users' emotions into account, thereby achieving more appropriate instruction.

[1148] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1149] Step 1:

[1150] Submission data is uploaded from the user's device. The user uses their own device to send text data of assignments or papers to the server through a specified interface (browser or application). The input is a text file created by the student, and the output is data stored in a temporary storage area on the server.

[1151] Step 2:

[1152] The server receives the submitted data and stores it in a temporary storage area. The server receives the text data uploaded by the user and checks its format. For example, converting a PDF file to TXT format. The input is the file uploaded in step 1, and the output is the cleaned text data.

[1153] Step 3:

[1154] The server cleans and normalizes the data. Specifically, it uses Python's pandas library and nltk (Natural Language Toolkit) to remove unnecessary characters and line breaks, standardize case, and standardize specific characters. The input is the text data saved in step 2, and the output is cleaned data ready for analysis.

[1155] Step 4:

[1156] The server runs a natural language processing engine and performs tokenization. The server uses a natural language processing engine, such as spaCy or BERT, to split the text into words and phrases. The input is the text data cleaned in step 3, and the output is the tokenized data.

[1157] Step 5:

[1158] The server performs grammatical structure analysis. Using the tokenized data, it analyzes the structure of each sentence and clarifies the relationships between subjects, predicates, objects, etc. The input is the data tokenized in step 4, and the output is the result of analyzing the grammatical structure.

[1159] Step 6:

[1160] The server performs a stylistic analysis. Based on the analyzed sentences, it compares the differences in the specific writing styles of humans and the generative AI. The stylistic analysis includes indicators such as the length of each sentence, the specific syntax contained in the sentence, and the vocabulary used. The input is the result of the grammatical structure analysis in step 5, and the output is the result of the stylistic analysis.

[1161] Step 7:

[1162] The server applies a probabilistic model to calculate a generative AI suspicion score. Based on the analysis results, it uses Bayesian inference and machine learning models to calculate a probability score indicating the likelihood that the data was created by a generative AI. The input is the result of the stylistic analysis in step 6, and the output is a generative AI suspicion score.

[1163] Step 8:

[1164] The server calculates the frequency of keywords and phrases. The generation AI detects frequently used keywords and phrases and calculates their frequency. The input is the same data as the probability score in step 7, and the output is the keyword frequency result.

[1165] Step 9:

[1166] The server integrates the analysis results and generates a report. It integrates the detailed analysis results for each paragraph and generates a report that includes a score for the likelihood of using generative AI and the analysis results for each paragraph. A report template is used to generate this report, and it is output in HTML or PDF format. The input is the analysis results from Step 7 and Step 8, and the output is the final report.

[1167] Step 10:

[1168] The server uses an emotion engine to recognize the user's emotions. Before returning the analysis results to the user, it uses IBM Watson or Microsoft Azure's emotion analysis services to determine the user's emotions from their facial expressions, voice, and text. The input is the user's emotion data, and the output is the emotion recognition results.

[1169] Step 11:

[1170] The server adjusts the display method based on the emotion recognition results. Based on the results of the emotion engine, it changes the way the analysis results are displayed and provides optimal feedback to the user. For example, if there is a high suspicion of the use of generative AI, it adds a message offering careful guidance. The input is the emotion recognition result from step 10, and the output is the analysis result with adjusted feedback.

[1171] Step 12:

[1172] The server returns the analysis results to the user. The final report and sentiment-based feedback are sent to the user's device. The user reviews the report and makes necessary corrections or adjustments to the essay. The input is the analysis results with feedback adjusted in step 11, and the output is the final report displayed on the user's device.

[1173] (Application example 2)

[1174] 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."

[1175] Conventional chatbot systems have limitations in their ability to generate automatic responses to user input. In particular, as text generation using generative AI becomes more common, systems that can appropriately recognize a user's emotions and provide feedback based on those emotions are required. However, systems that provide feedback that takes emotions into account are rare, and have not contributed to improving user satisfaction. It has also been difficult to determine whether or not a generative AI is being used and provide a reliable service. Therefore, the present invention aims to solve these problems by providing a new system that combines emotion recognition and generative AI discrimination.

[1176] 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.

[1177] In this invention, the server includes means for receiving data submitted by students, means for analyzing the data using a natural language processing engine, means for applying a probabilistic model that calculates a score for suspicion of use of generative AI based on the analysis results, means for returning the analysis results to the user, and an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results. This makes it possible to determine the suspicion of use of generative AI while taking the user's emotions into consideration.

[1178] "Means for receiving data submitted by students" refers to a function for receiving text data such as assignments and papers uploaded by users to the system.

[1179] "Means for analyzing using a natural language processing engine" refers to a function that analyzes uploaded text data using techniques such as tokenization, grammatical structure analysis, and stylistic analysis.

[1180] The "means for applying a probabilistic model to calculate a score for suspected use of generative AI" is a function that calculates a score for the likelihood that a text was generated by generative AI based on the analyzed data.

[1181] The "means for returning the analysis results to the user" is a function for presenting the results of the analysis performed by the server to the user.

[1182] "Means including an emotion recognition engine that recognizes the user's emotions and adjusts the display method of the analysis results" refers to a function that determines the user's emotions from their facial expressions, voice, input text, etc., and adjusts the feedback method based on the results.

[1183] The present invention is a system that receives data submitted by users, analyzes it, identifies suspected use of generative AI, and recognizes the user's emotions to provide appropriate feedback. This system clarifies the roles of the server, terminal, and user, and explains how they interact. A specific example is provided below.

[1184] Server Processing

[1185] The server receives text data for assignments and papers uploaded by users from their devices. The received data is temporarily stored and prepared for analysis. The server then launches a natural language processing engine (NLP engine) that first performs tokenization. It then performs grammatical structure analysis and stylistic analysis, comparing the stylistic features of human and generative AI writing. Examples of specific software used for analysis include Python's nltk and transformers libraries.

[1186] Next, the server applies a probabilistic model to the analyzed data to calculate a probability score indicating the likelihood that the data was generated by the AI. At the same time, it detects keywords and phrases frequently used by the AI ​​and calculates their frequency of occurrence. A possible model to use is the BERT model. This allows for a suspicion score for the AI.

[1187] Emotion engine processing

[1188] Before returning the analysis results to the user, the server uses an emotion engine to recognize the user's emotions. The emotion engine determines emotions from the user's facial expressions, voice, input text, etc., and generates emotion recognition results. The TextBlob library can be used for emotion recognition.

[1189] Based on the results of the emotion engine, the server adjusts the way the analysis results are displayed. For example, if there is a high suspicion of the use of generative AI, it adds a message urging the user to take appropriate measures. Also, if the emotion recognition results make the user feel uneasy, it provides advice to ease their anxiety.

[1190] User operations

[1191] Users upload text data of assignments and papers from their own devices to the system. The data uploaded by the user is analyzed on the server, and the results are returned along with the results of the emotion engine. This allows users to receive emotion-based feedback along with the analysis results, enabling them to take better guidance and measures.

[1192] Specific examples

[1193] 1. Analyzing a Single Essay

[1194] Users upload essays (e.g., txt files) submitted by students from their own devices to the server. The server receives the text and analyzes it using a natural language processing engine. The analysis involves tokenization, stylistic analysis, application of probabilistic models, and keyword detection. The final analysis results, along with the results of the sentiment engine, are returned to the user in a report in an intuitive format.

[1195] 2. Batch processing multiple essays

[1196] At the end of the semester, users upload a large number of essays (multiple files) to the server in one batch. The server analyzes each essay individually, performs emotion recognition for each essay using an emotion engine, and generates individual analysis results. These results are then compiled and returned to the user's device as an overall report. The user then reviews the report, including the emotion information, and closely examines essays suspected of using generative AI.

[1197] Prompt Sentence Examples

[1198] User input: "The delivery was late and the product is damaged."

[1199] Tokenization: ['The', 'delivery', 'was', 'late', 'and', 'the', 'product', 'is', 'damaged', '.']

[1200] Style analysis: Detecting stylistic features with generative AI

[1201] Emotion Recognition Result: Extremely Negative (Emotion Polarity: -1)

[1202] Feedback: "I sincerely apologize for the inconvenience. Can you provide more details or images so that we can resolve this issue for you?"

[1203] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1204] Step 1: User uploads data from device

[1205] ---

[1206] Users upload text data for assignments or papers from their own devices to the system. The input is a text file prepared by the user, and the output is that text file sent to the server. Specifically, users send the text data using the file upload function of a browser or a dedicated application.

[1207] Step 2: The server stores the received data and prepares it for analysis.

[1208] ---

[1209] The server receives data uploaded by the user and stores it in a temporary storage area. The input is the text data sent by the user, and the output is text data ready for analysis. Specifically, the server stores the received file in file storage and prepares it to be passed to the analysis module.

[1210] Step 3: Text analysis using a natural language processing engine

[1211] ---

[1212] The server applies a natural language processing engine to the received text data, performing tokenization, grammatical structure analysis, and stylistic analysis. The input is the stored text data, and the output is the analysis results. Specifically, the server performs tokenization and grammatical analysis using the Python nltk and transformers libraries.

[1213] Step 4: Calculate the generative AI usage suspicion score

[1214] ---

[1215] Based on the analysis results, the server applies a probability model to calculate a probability score indicating the suspected use of generative AI. The input is the analyzed text data, and the output is a probability score indicating the suspected use of generative AI. Specifically, it uses the BERT model to calculate the likelihood that the text was created by generative AI.

[1216] Step 5: Analyze user emotions using an emotion recognition engine

[1217] ---

[1218] The server uses an emotion recognition engine to recognize the user's emotions before returning the analysis results to the user. The input is the text message, voice, or image data entered by the user, and the output is the emotion recognition result. Specifically, it analyzes the user's emotions using the TextBlob library.

[1219] Step 6: Integrating analytical results with emotional feedback

[1220] ---

[1221] The server integrates the generated AI suspicion score and the emotion recognition result to generate a feedback report. The input is the generated AI suspicion score and the emotion recognition result, and the output is the integrated feedback report. Specifically, the server adds a warning message based on the emotion recognition result if careful action is required, and generates feedback according to the user's emotions.

[1222] Step 7: Returning the analysis results to the user

[1223] ---

[1224] The server returns the generated feedback report to the user. The input is the integrated feedback report, and the output is the analysis result and feedback received by the user. Specifically, the server sends the analysis result to the user's device for the user to check.

[1225] Through these steps, users can receive analysis results that determine whether or not to use the generative AI, as well as feedback based on their own emotions.

[1226] 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.

[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 robot 414.

[1229] 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.

[1230] 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.

[1231] 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.

[1232] 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).

[1233] 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.

[1234] 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."

[1235] 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.

[1236] 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).

[1237] 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.

[1238] 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.

[1239] 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.

[1240] 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.

[1241] 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.

[1242] 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.

[1243] 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.

[1244] 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.

[1245] 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.

[1246] 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.

[1247] The following is further disclosed regarding the above embodiment.

[1248] (Claim 1)

[1249] a means for receiving student submitted data;

[1250] A means for analyzing the data using a natural language processing engine;

[1251] A means for applying a probabilistic model to calculate a suspected use score of the generating AI based on the analysis results;

[1252] means for returning the analysis results to the user;

[1253] A system including:

[1254] (Claim 2)

[1255] 10. The system of claim 1, further comprising means for comparing the style of the text with human and generative AI characteristics through stylistic analysis.

[1256] (Claim 3)

[1257] 10. The system of claim 1, further comprising: means for generating an analysis result for each submitted paragraph.

[1258] "Example 1"

[1259] (Claim 1)

[1260] means for receiving student submission data transmitted from the terminal;

[1261] An analysis method using a natural language processing engine that tokenizes the data and analyzes the sentence structure;

[1262] A method to analyze the style of the text based on the analysis results and compare it with the characteristics of the generation AI,

[1263] A means for applying a probabilistic model to calculate a suspected use score of the generating AI based on the analysis results;

[1264] A means for the AI ​​to detect frequently used keywords and phrases and incorporate their frequency into the analysis.

[1265] A means of generating detailed analysis results for each paragraph and generating reports based on them;

[1266] means for returning the generated report to the user;

[1267] A system including:

[1268] (Claim 2)

[1269] 10. The system of claim 1, further comprising means for using a probabilistic model with a machine learning algorithm to determine whether submitted data is human-created or generated by generative AI.

[1270] (Claim 3)

[1271] 10. The system of claim 1, further comprising means for generating an analysis result for each submitted paragraph and highlighting questionable passages.

[1272] "Application Example 1"

[1273] (Claim 1)

[1274] a means for receiving student submitted data;

[1275] A means for analyzing the data using a natural language processing engine;

[1276] A means for applying a probabilistic model to calculate a suspected use score of the generating AI based on the analysis results;

[1277] means for returning the analysis results to the user;

[1278] A means to display a score and suspicious parts to determine whether the text sent is from a generative AI;

[1279] A system including:

[1280] (Claim 2)

[1281] 10. The system of claim 1, further comprising means for comparing the style of the text with human and generative AI characteristics through stylistic analysis.

[1282] (Claim 3)

[1283] 10. The system of claim 1, further comprising: means for generating an analysis result for each submitted paragraph.

[1284] "Example 2: Combining Emotion Engines"

[1285] (Claim 1)

[1286] a means for receiving student submitted data;

[1287] A means for analyzing the data using a natural language processing engine;

[1288] A means for applying a probabilistic model to calculate a suspected use score of the generating AI based on the analysis results;

[1289] A means for recognizing the analysis results and the user's emotions and adjusting the display method;

[1290] means for returning the analysis results to the user;

[1291] A system including:

[1292] (Claim 2)

[1293] 10. The system of claim 1, further comprising means for comparing the style of the text with human and generative AI characteristics through stylistic analysis.

[1294] (Claim 3)

[1295] 10. The system of claim 1, further comprising: means for generating an analysis result for each submitted paragraph.

[1296] "Application example 2 when combining emotion engines"

[1297] (Claim 1)

[1298] a means for receiving student submitted data;

[1299] A means for analyzing the data using a natural language processing engine;

[1300] A means for applying a probabilistic model to calculate a suspected use score of the generating AI based on the analysis results;

[1301] means for returning the analysis results to the user;

[1302] means including an emotion recognition engine for recognizing a user's emotion and adjusting how the analysis results are displayed;

[1303] A system including:

[1304] (Claim 2)

[1305] 10. The system of claim 1, further comprising means for comparing the style of the text with human and generative AI characteristics through stylistic analysis.

[1306] (Claim 3)

[1307] 10. The system of claim 1, further comprising: means for generating an analysis result for each submitted paragraph. [Explanation of symbols]

[1308] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for receiving student submitted data; A means for analyzing the data using a natural language processing engine; A means for applying a probabilistic model to calculate a suspected use score of the generating AI based on the analysis results; means for returning the analysis results to the user; A system including:

2. 10. The system of claim 1, further comprising means for comparing the style of the text with human and generative AI characteristics through stylistic analysis.

3. 10. The system of claim 1, further comprising means for generating an analysis result for each submitted paragraph.

Images