system

Abstract

Provide a system. [Solution] A system for automating the detection and reporting of harassment, comprising: A terminal means for capturing interactions such as web conferences, emails, and chats in real time; a server means for receiving and analyzing the text data and voice data transmitted from the terminal means; a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the likelihood of harassment from text and voice data; server means for sending a warning message to a user when potential harassment is detected; a server means for reporting to a harassment hotline when the same user repeatedly ignores warnings; A system including:

Description

[Technical Field]

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

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

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

[0004] Harassment is a serious problem in today's workplaces, and effective measures to prevent it from occurring are needed. However, traditional methods often require time for employees to report harassment, delaying appropriate responses. Furthermore, the criteria for detecting harassment are subjective, making it difficult to respond to individual cases. Furthermore, the inability to detect harassment in real time and respond immediately increases the psychological burden on victims. [Means for solving the problem]

[0005] To address this issue, the present invention provides a system for automating the detection and reporting of harassment. The system includes a terminal device that captures interactions such as web conferences, emails, and chats in real time; a server device that receives and analyzes text and voice data sent from the terminal device; a server device that executes an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data; a server device that sends a warning message to users when possible harassment is detected; and a server device that notifies a harassment hotline if the same user repeatedly ignores the warning. This system enables early detection of harassment and immediate response, potentially improving the workplace environment. Furthermore, the use of a voice recognition engine enables real-time analysis of voice data, resulting in more accurate harassment detection.

[0006] The "terminal means" is a device that captures interactions such as web conferences, e-mails, and chats used by users in real time and transmits the data to a server.

[0007] The "server means" is a computer system that receives text data and voice data sent from the terminal means and performs analysis.

[0008] "Natural language processing" is a technology for processing human language on a computer and analyzing its meaning.

[0009] "Emotion analysis" is a technology that detects emotions from text or audio data and analyzes their content.

[0010] An "AI model" is a model trained by machine learning algorithms to perform tasks such as natural language processing or sentiment analysis.

[0011] A "warning message" is a message that notifies a user when potential harassment is detected.

[0012] "Report data" is information sent to the harassment consultation desk, and includes the content of the problematic remark, the date and time, user information, and the like.

[0013] "Speech recognition engine means" is a technology for converting voice data into text.

[0014] The "Harassment Consultation Desk" is a dedicated desk for reporting and responding to harassment acts. [Brief explanation of the drawings]

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

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

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

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

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

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

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

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

[0023] [First embodiment]

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

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

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

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

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

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

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

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

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

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

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

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

[0036] The present invention relates to a system for automating the detection and reporting of harassment behavior, and specific embodiments thereof will be described below.

[0037] System configuration

[0038] 1. Terminal means

[0039] The terminal means is a device that captures in real time the interactions of users in web conferences, emails, chats, etc. The terminal means uses APIs and hooks to acquire the interactions of each application and send the data to the server.

[0040] 2. Server Means

[0041] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means. The server means processes the text data and voice data separately, and converts the voice data into text using a voice recognition engine means.

[0042] It uses natural language processing and sentiment analysis to run AI models to determine potential harassment from text and audio data.

[0043] Program processing

[0044] Data collection

[0045] The device captures the user's web conference, email, chat, and other interactions in real time and sends the data to a server. For example, it uses the Teams API to obtain the content of comments and sends it to the server, including the audio data.

[0046] Data reception and preprocessing

[0047] The server receives the text data and the voice data and separates the data for analysis. The voice data is converted to text using a speech recognition engine.

[0048] Harassment Detection

[0049] The server uses a natural language processing engine to analyze the text data and extract key keywords and emotional tones, while an emotion analysis engine analyzes the voice data to detect the emotional state of the user.

[0050] The server combines these analysis results and runs an AI model to determine whether there is a possibility of harassment. If the analysis result exceeds a threshold, it is determined to be harassment.

[0051] Sending a warning message

[0052] If the server determines that a message is potentially harassing, it generates a warning message for the user and displays it as a pop-up on the user's screen. For example, a message might say in a Teams chat, "This message may contain harassing content. Please choose your language carefully."

[0053] Report processing

[0054] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[0055] The server generates and automatically sends report data to the harassment hotline, including the content of the offending remark, the date and time, and user information.

[0056] The device will notify the user that a report has been made, for example by displaying a message that reads, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0057] Specific examples

[0058] For example, during a team meeting, someone might say:

[0059] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[0060] The device captures these statements in real time and sends them to the server as text and audio data.

[0061] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[0062] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0063] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[0064] In this way, the system effectively detects harassment and enables immediate action.

[0065] The processing flow will be explained below.

[0066] Step 1:

[0067] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[0068] Step 2:

[0069] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[0070] Step 3:

[0071] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[0072] Step 4:

[0073] The server runs the audio data through a speech recognition engine and converts it into text, or uses a service like Google® Cloud Speech-to-Text to transcribe the audio.

[0074] Step 5:

[0075] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[0076] Step 6:

[0077] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[0078] Step 7:

[0079] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[0080] Step 8:

[0081] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[0082] Step 9:

[0083] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[0084] Step 10:

[0085] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[0086] Step 11:

[0087] The server generates report data, which includes the content of the offending comment, the date and time, and user information. The generated report data is prepared for sending to the harassment hotline.

[0088] Step 12:

[0089] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[0090] Step 13:

[0091] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0092] Through these steps, the system enables automatic detection of harassment behavior and immediate response.

[0093] Example 1

[0094] 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."

[0095] In today's digital communication environment, the possibility of harassment is increasing. However, there are limited means to detect it in a timely manner and take appropriate measures, resulting in a lack of appropriate protection for victims. In particular, for companies and organizations with large user bases, it is difficult to monitor and deal with harassment manually, and there is a growing need for effective automated systems.

[0096] 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.

[0097] In this invention, the server includes device means for capturing user interactions such as meetings, emails, and communications in real time, computer means for receiving and analyzing text and voice data sent from the device means, speech recognition engine means for converting voice data into text data, generative AI model means for analyzing the text and voice data using natural language processing and sentiment analysis to determine the possibility of harassment, means for sending a warning message to a user when the possibility of harassment is detected, and means for automatically reporting to a consultation center if the same user repeatedly ignores warnings. This makes it possible to efficiently and effectively detect harassment even in large-scale communication environments and take timely measures.

[0098] A "user" is an individual or group that communicates using the system.

[0099] "Conference" refers to an activity in which multiple users exchange information in real time using audio and video.

[0100] "Email" is a means of sending and receiving messages electronically over the Internet.

[0101] "Communication" refers to any means of exchanging information, including through text and voice.

[0102] "Real-time" refers to the processing and transmission of data and information instantly, with minimal delay.

[0103] "Device means" refers to a device for capturing a user's communication activity and transmitting that information to another system.

[0104] "Computer Means" refers to a computing device and associated hardware and software for analyzing, storing, and processing received data.

[0105] "Speech recognition engine means" refers to software or hardware for converting voice data into text data.

[0106] "Natural language processing" refers to computer technology that analyzes text data and understands its meaning and sentiment.

[0107] "Sentiment analysis" refers to the technology of detecting emotional tone and state from text or audio data.

[0108] A "generative AI model" refers to an artificial intelligence model that determines the possibility of harassment from input data.

[0109] "Warning message" includes a message that notifies a user in the event of possible harassment.

[0110] A "consultation desk" refers to a department or institution dedicated to receiving reports of harassment and dealing with them.

[0111] "Automatic means for reporting" means a system mechanism for reporting without manual intervention when specified conditions are met.

[0112] This invention is a system for detecting harassing behavior in users' online communications and automating appropriate responses.

[0113] System configuration

[0114] The system includes the following means:

[0115] 1. A device means of capturing user interactions in real time, such as meetings, emails, and correspondence.

[0116] The device (terminal) captures user comments and text messages in real time and transmits the data to a server. For example, it uses APIs such as MICROSOFT (registered trademark) TEAMS (registered trademark) and Zoom to acquire voice data and text chat data during a meeting.

[0117] 2. Computer means for receiving and analyzing the text data and voice data transmitted from the device means

[0118] The server means (computer) receives the data sent from the terminals, classifies and stores the data for analysis. This means includes a high-performance database and data analysis software.

[0119] 3. A speech recognition engine that converts voice data into text data

[0120] The server uses a speech recognition engine such as Google Cloud Speech-to-Text API or IBM Watson® to convert the audio data into text data, which is then used for further analysis.

[0121] 4. Generative AI modeling means for analyzing text and audio data using natural language processing and sentiment analysis to identify potential harassment.

[0122] The server analyzes the text and audio data using natural language processing engines such as Google Cloud Natural Language API and sentiment analysis engines such as IBM Watson Tone Analyzer. The analysis results are fed into a BERT-based generative AI model to identify potential harassment behavior.

[0123] 5. A means of sending warning messages to users when potential harassment is detected

[0124] If the server determines that an activity is likely to be harassing, it generates and sends a warning message to the user, which can appear as a popup on messaging platforms such as Teams or Slack.

[0125] 6. A method for automatically reporting to the helpline if the same user repeatedly ignores warnings

[0126] The server detects users who repeatedly engage in harassing behavior and counts the number of times. If the number of times exceeds a threshold, it automatically generates and sends a report to the harassment hotline. This data includes the content of the offending remarks, the date and time, and user information.

[0127] Specific examples

[0128] For example, during a team meeting, someone might say:

[0129] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[0130] The device captures this speech in real time and sends it to the server as text and audio data.

[0131] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[0132] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0133] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[0134] Prompt Sentence Examples

[0135] The accuracy of harassment detection can be improved by feeding prompt sentences like the following into the AI ​​model:

[0136] "Please determine whether the following conversation constitutes harassment. The statement: 'You're late again. This is annoying because it happens all the time. Can you do it properly?'"

[0137] In this way, the present invention can efficiently and effectively detect harassment behavior even in large-scale communication environments and take countermeasures in a timely manner.

[0138] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0139] Step 1: Data collection

[0140] The device captures users' meetings, emails, chats and other interactions in real time.

[0141] Specifically, the device uses APIs and hooks to retrieve speech and text messages from, for example, web conferencing tools or messaging applications.

[0142] Input: User utterances (voice data) and text messages (text data)

[0143] Data processing: Real-time capture

[0144] Output: Audio and text data

[0145] Step 2: Data reception and preprocessing

[0146] The server receives the text data and voice data sent from the terminal.

[0147] Specifically, the server separates the voice data from the text data for analysis, and converts the voice data into text using a speech recognition engine, such as the Google Cloud Speech-to-Text API.

[0148] Input: Voice and text data sent from the device

[0149] Data processing: Speech to text conversion using a speech recognition engine

[0150] Output: Converted text data

[0151] Step 3: Harassment detection

[0152] The server analyzes the converted text data using a natural language processing engine and a sentiment analysis engine.

[0153] Specifically, the system uses the Google Cloud Natural Language API to extract keywords and emotional tones from text data, and IBM Watson Tone Analyzer to extract emotional tones from audio data.These analysis results are then fed into an AI model (e.g., a BERT-based generative AI model) to determine the likelihood of harassment.

[0154] Input: Converted text data

[0155] Data Processing: Natural Language Processing and Sentiment Analysis

[0156] Output: Harassment possibility assessment result

[0157] Step 4: Sending a warning message

[0158] If the server determines that there is a high possibility of harassment, it generates and sends a warning message to the user.

[0159] Specifically, the server uses the Teams API to generate a warning message, such as "This message may be harassing. Please choose your words carefully," and send it to the user.

[0160] Input: Harassment possibility assessment result

[0161] Data processing: Generate warning messages

[0162] Output: Warning message sent to the user

[0163] Step 5: Report processing

[0164] When the same user repeatedly makes statements that meet the criteria for harassment, the server counts the number of times, and when the number exceeds a threshold, it generates report data and automatically sends it to a consultation desk.

[0165] Specifically, the server generates report data including the content of the offending comment, the date and time, and user information, and sends it to the harassment hotline. It also displays a message on the user's device informing them that a report has been made.

[0166] Input: Count of harassment incidents

[0167] Data processing: Generation of report data

[0168] Output: Report data sent to the consultation desk and message sent to the user

[0169] (Application example 1)

[0170] 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."

[0171] In recent years, harassment between customers and employees in brick-and-mortar stores has become a social problem. However, this type of problem is often difficult to detect and respond to immediately, which often increases the mental burden on employees. Therefore, an effective automatic detection and immediate response system is needed to address this issue.

[0172] 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.

[0173] In this invention, the server includes a terminal means for capturing interactions between customers and employees in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, and a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data. This makes it possible to detect harassment between customers and employees in a physical store in real time and issue an immediate warning or report.

[0174] "Customer" refers to a person who visits a physical store to purchase goods or services.

[0175] "Employee" refers to a staff member who performs duties such as selling products and providing services at a physical store.

[0176] "Interaction" refers to the conversational and behavioral interactions that take place between customers and employees.

[0177] "Real-time capture" refers to instantly recording and acquiring interactions that occur at that moment.

[0178] "Terminal means" refers to a hardware and software system for capturing audio and video data.

[0179] "Text data" refers to data converted from audio into text.

[0180] "Audio data" refers to data that records spoken words or sounds.

[0181] "Server means" refers to a computer system for receiving, analyzing, and storing data.

[0182] "Natural language processing" refers to computer technology for understanding and analyzing human language.

[0183] "Emotion analysis" refers to the technology of analyzing a speaker's emotions and tone from voice and text data.

[0184] An "AI model" refers to a system that uses algorithms based on machine learning and deep learning to analyze specific data and make inferences.

[0185] "Warning message" refers to a warning message that is displayed to a user when harassment behavior is detected.

[0186] A "harassment consultation center" refers to a specialized institution that accepts consultations and takes measures regarding harassment acts.

[0187] "Reporting data" refers to data that records details of harassment when it occurs and sends them to a consultation hotline.

[0188] The present invention is implemented using a system that captures and analyzes interactions between customers and employees in real time. The system includes the following hardware and software:

[0189] Hardware and software used

[0190] Hardware:

[0191] 1. Smartphone (with built-in camera and microphone)

[0192] 2. Cloud-based servers (e.g., AWS EC2)

[0193] software:

[0194] 1. Speech recognition engine (e.g. Google Cloud Speech-to-Text)

[0195] 2. Natural language processing engine (e.g., Google Cloud Natural Language)

[0196] 3. Sentiment analysis engine (e.g., IBM Watson Tone Analyzer)

[0197] Specific processing of the system

[0198] 1. Data Collection:

[0199] The smartphone serves as the terminal device and captures conversations between users (customers and employees) in real time. The smartphone's microphone and camera are used to obtain audio and video data.

[0200] 2. Data reception and preprocessing:

[0201] The server receives the audio and video data sent from the terminal. The audio data is converted into text data via a voice recognition engine. Furthermore, facial expression analysis is performed on the video data to detect changes in emotions.

[0202] 3. Harassment Detection:

[0203] The server uses a natural language processing engine to analyze text data. It also uses an emotion analysis engine to analyze voice and facial expression data to detect the speaker's emotional tone. This data is then combined and a generative AI model is used to determine whether the behavior is likely to be harassing. The determination includes specific keywords and emotional tone.

[0204] 4. Sending warning messages:

[0205] If harassment is detected, the server will pop up a warning message on the terminal in real time, such as "This word may contain elements of harassment. Please use polite language."

[0206] 5. Report Processing:

[0207] If the same customer repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment hotline. The report data includes the content of the remarks, date and time, user information, etc.

[0208] Examples

[0209] For example, in one scenario, a customer might say:

[0210] "Your responses are always slow, can't you do something about it?"

[0211] The smartphone (terminal means) captures this speech with a microphone and transmits it to the server as audio data.

[0212] The server uses a voice recognition engine to convert the voice data into text data and a natural language processing engine to analyze the possibility of harassment.

[0213] A sentiment analysis engine is also used to analyze emotional tone from text and audio data.

[0214] If a harassing act is detected, the server sends a warning message to the terminal means in real time.

[0215] Prompt Sentence Examples

[0216] Capture real-time conversations between customers and store associates and perform text and tone analysis on:

[0217] Text: "You're always slow to respond"

[0218] Emotional Tone: Anger

[0219] Warning message: "This language may be harassing. Please use polite language."

[0220] This will realize a system that can detect harassment between customers and employees in physical stores in real time and issue warnings or reports immediately.

[0221] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0222] Step 1: Data collection

[0223] The device (smartphone) captures the conversation between the customer and the employee using a camera and microphone, thereby obtaining audio and video data as input.

[0224] Step 2: Send data

[0225] The device transmits the captured audio and video data to a server in real time using an internet connection.

[0226] Step 3: Preprocessing the audio data

[0227] The server converts the received voice data into text data using a voice recognition engine (Google Cloud Speech-to-Text). The input is voice data and the output is text data.

[0228] Step 4: Preprocessing the video data

[0229] The server processes the received video data using analysis technology and performs facial expression analysis. The input is video data and the output is emotional tone data.

[0230] Step 5: Natural Language Processing

[0231] The server uses a natural language processing engine (Google Cloud Natural Language) to analyze the text data, extracting important keywords and context, and assessing the likelihood of harassment as a result of the analysis. The input is the text data, and the output is the results of keyword and context analysis.

[0232] Step 6: Sentiment Analysis

[0233] The server uses an emotion analysis engine (IBM Watson Tone Analyzer) to analyze the text data and emotional tone data and determine the emotional tone. The input is the text data and emotional tone data, and the output is the result of the emotional tone analysis.

[0234] Step 7: Determining Harassment

[0235] The server integrates the results of natural language processing and sentiment analysis and uses a generative AI model to determine the likelihood of harassment. The input is keywords, the results of contextual analysis, and the results of emotional tone analysis, and the output is the harassment judgment result.

[0236] Step 8: Sending a warning message

[0237] If the server detects harassment, it will pop up a warning message on the device in real time. The input is the result of the harassment judgment, and the output is the warning message.

[0238] Step 9: Generate and send notification data

[0239] If the same user repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment consultation center. The input is the number of times harassment acts occurred, and the output is the generation and transmission of report data.

[0240] 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.

[0241] The present invention relates to a system for automating the detection and reporting of harassment behavior, and the following describes an embodiment incorporating an emotion engine for analyzing user emotions.

[0242] System configuration

[0243] 1. Terminal means

[0244] The terminal means is a device that captures user interactions such as web conferences, emails, and chats in real time. The terminal means uses APIs and hooks to obtain messages and voice data from each application and saves it in temporary storage. For example, the Teams API can be used to capture what is being said in a meeting.

[0245] 2. Server Means

[0246] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means, classifies the text data and voice data, and converts the voice data into text using a voice recognition engine means.

[0247] It runs AI models that use natural language processing and sentiment analysis to determine potential harassment from text and audio data.

[0248] 3. Emotion Engine

[0249] The emotion engine has the function to analyze the user's emotions and extracts emotions from voice data and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[0250] Program processing

[0251] Data collection

[0252] The device captures users' web conferences, emails, chats, and other interactions in real time and sends the data to a server. For example, it can capture comments made during a Teams meeting in real time and collect them as audio and text data.

[0253] Data reception and preprocessing

[0254] The server receives the data and separates it for analysis: it separates audio data from text data and sends each to a separate processing pipeline.

[0255] The server converts the voice data into text using a speech recognition engine, such as Google Cloud Speech-to-Text.

[0256] Harassment detection and sentiment analysis

[0257] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones.

[0258] The server uses an emotion engine to analyze the user's emotions from the voice data, analyzing the tone, volume, speed, etc. of the voice to detect the user's emotional state.

[0259] The server uses an emotion engine to analyze user emotions from text data, extracting emotions based on the content, structure, and wording of the text.

[0260] Overall rating and warning message sent

[0261] The server evaluates these analysis results comprehensively, using an AI model to score the analysis results and determine the possibility of harassment if a threshold is exceeded.

[0262] If the server detects a potential harassment, it will generate a warning message and display it on the user's screen. For example, a message in chat might say, "This message contains harassing and emotionally inappropriate content. Please choose your words carefully."

[0263] Report processing

[0264] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[0265] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[0266] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[0267] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0268] As a concrete example, suppose the following is said during a meeting:

[0269] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[0270] The device captures these statements in real time and sends them to the server as text and audio data.

[0271] The server receives this speech and converts the speech data into text using a speech recognition engine.

[0272] The server analyzes the text data using a natural language processing engine and extracts keywords such as "It's late," "This is a problem," and "Please do it properly."

[0273] The server uses an emotion engine to analyze the tone of the voice and the emotions in the text content, determining that the message is critical and contains strong emotional elements.

[0274] Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0275] If the user ignores the warning and continues to make critical comments, the server counts the number of times and issues a report if the number exceeds a threshold.

[0276] In this way, the system effectively detects harassment and emotional issues early and enables immediate response.

[0277] The processing flow will be explained below.

[0278] Step 1:

[0279] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[0280] Step 2:

[0281] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[0282] Step 3:

[0283] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[0284] Step 4:

[0285] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[0286] Step 5:

[0287] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[0288] Step 6:

[0289] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[0290] Step 7:

[0291] The server also performs sentiment analysis on text data, extracting emotions based on the content, structure, and wording of the text.

[0292] Step 8:

[0293] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[0294] Step 9:

[0295] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[0296] Step 10:

[0297] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[0298] Step 11:

[0299] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[0300] Step 12:

[0301] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[0302] Step 13:

[0303] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[0304] Step 14:

[0305] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0306] Specific examples

[0307] During a meeting, the following is said:

[0308] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[0309] Specific processing flow

[0310] Step 1:

[0311] The device captures this speech in real time, capturing text and audio data and storing it in temporary storage.

[0312] Step 2:

[0313] The device encrypts the captured data and sends it to the server over a secure channel.

[0314] Step 3:

[0315] The server separates the received data into text and audio data for analysis.

[0316] Step 4:

[0317] The server runs the voice data through a speech recognition engine and converts it into text.

[0318] Step 5:

[0319] The server analyzes the text data using a natural language processing engine and extracts keywords and phrases such as "It's late," "This is a problem," and "Please do it properly."

[0320] Step 6:

[0321] The server analyzes the tone, volume, and speed of the voice using an emotion engine to detect critical emotions.

[0322] Step 7:

[0323] The server analyzes the sentiment of the text from its content and vocabulary, extracting critical and emotional elements.

[0324] Step 8:

[0325] The server combines the analysis results and scores them using an AI model, determining that in this case the harassment threshold has been exceeded.

[0326] Step 9:

[0327] The server generates a warning message to the speaker, saying, "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[0328] Step 10:

[0329] The user is given the opportunity to review the warning message and correct their behavior.

[0330] Step 11:

[0331] If the user ignores the warning again and continues to make critical comments, the server counts the number of times and proceeds to the next step if the number exceeds a threshold.

[0332] Step 12:

[0333] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis.

[0334] Step 13:

[0335] The server automatically sends the generated report data to the harassment consultation center.

[0336] Step 14:

[0337] The device will notify the user that a report has been made, displaying the message, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0338] In this way, the system detects harassment and emotional issues early and enables immediate response.

[0339] Example 2

[0340] 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."

[0341] Conventional harassment detection and reporting systems have problems with insufficient user emotion analysis, making it difficult to detect harassment in real time and respond immediately. Furthermore, recording and reporting repeated harassment behaviors by users who ignore warnings must be done manually, making it difficult to efficiently prevent harassment.

[0342] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0343] In this invention, the server includes means for receiving and classifying text data and voice data transmitted from the terminal means, means for utilizing a voice recognition engine for converting voice data into text, and means for executing an AI model for determining the possibility of harassment from the text and voice data using natural language processing and sentiment analysis, thereby enabling real-time detection of harassment behavior and user sentiment analysis.

[0344] The "terminal means" is a device for capturing exchanges such as web conferences, e-mails, and chats in real time and transmitting them to a server.

[0345] The "server means" is a computer system that receives data sent from the terminal means and classifies and analyzes the data.

[0346] A "voice recognition engine means" is a system that executes voice recognition techniques to convert voice data into text data.

[0347] "Natural language processing" is a technology that analyzes text data and extracts important keywords and phrases.

[0348] "Emotion analysis" is a technology that analyzes a user's emotions from voice data and text data and detects their emotional state.

[0349] The "AI model" is an artificial intelligence technology that determines the possibility of harassment based on the results of natural language processing and sentiment analysis.

[0350] A "warning message" is a notification sent to a user when potential harassment is detected.

[0351] The "reporting mechanism" is a system that records repeated acts of harassment and reports them to a harassment hotline if a threshold is exceeded.

[0352] The present invention is a system for automating the detection and reporting of harassment behavior, and incorporates an emotion engine that analyzes user emotions. Specific embodiments of the present invention will be described below.

[0353] 1. System Configuration

[0354] Terminal means

[0355] The terminal means is a device that captures interactions such as web conferences, emails, and chats in real time. It uses APIs and hooks from the applications used by the user to obtain messages and voice data in real time. As a specific example, the Teams API is used to capture what is said during a meeting.

[0356] Server Means

[0357] The server means is a computer system that receives and analyzes text data and voice data sent from the terminal means. The server means uses a voice recognition engine means to convert voice data into text data. In this case, the Google Cloud Speech-to-Text service can be used.

[0358] The server performs natural language processing and emotion analysis on the text data and voice data, thereby executing an AI model that determines the possibility of harassment. The emotion engine analyzes emotions from the voice data and text data to detect the user's emotional state.

[0359] Emotion Engine

[0360] The emotion engine has the function of analyzing the user's emotions from voice and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[0361] 2. Program Processing Overview

[0362] Data collection

[0363] When a user is engaged in a web conference, email, chat, or other exchange, the device captures this in real time and sends it to the server. For example, remarks made during a Teams meeting can be captured in real time and collected as voice and text data.

[0364] Data reception and preprocessing

[0365] The server receives the data and separates it into voice and text, which is then converted into text using the Google Cloud Speech-to-Text speech recognition engine.

[0366] Harassment detection and sentiment analysis

[0367] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones. The server also analyzes emotions from the audio data, using an emotion engine to analyze voice tone, volume, speed, etc., thereby detecting the user's emotional state.

[0368] 3. Usage example

[0369] As a concrete example, suppose the following is said during a meeting:

[0370] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[0371] System Operation

[0372] 1. The device captures this speech in real time and sends it to the server as text and audio data.

[0373] 2. The server receives this speech and converts the speech data into text using a speech recognition engine.

[0374] 3. The server analyzes the text data using a natural language processing engine and extracts keywords such as "late," "this is a problem," and "please do it properly."

[0375] 4. The emotion engine analyzes the tone of the voice and the emotion of the text content and determines that the message is critical and emotionally charged.

[0376] 5. Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0377] 6. If the user ignores the warning and continues to make critical comments, the server will count the number of times and issue a report if the number exceeds a threshold.

[0378] 4. Example prompts for generative AI models

[0379] You can further improve harassment detection by using prompts like the following to feed your generative AI model:

[0380] "Please analyze the following conversations to detect any harassing behavior and the emotional tone. Please also provide your analysis results based on the specific content of the statements."

[0381] Conversation: "You're late again. This is really annoying. Please be more careful next time."

[0382] Expected output: "This comment contains critical and emotional overtones and may be considered harassment."

[0383] By using such prompts, it becomes possible to perform advanced analysis of both the content and emotions of the statements.

[0384] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0385] Step 1:

[0386] Data collection

[0387] Input: User web conferencing, email, and chat interactions

[0388] Specific operation: The device captures the user's interactions in real time, such as web conferences, emails, and chats. For example, it uses the Teams API to capture what is said during a meeting.

[0389] Output: Audio and text data

[0390] Step 2:

[0391] Data transmission

[0392] Input: Audio and text data

[0393] Specific operation: The device sends the captured audio and text data to the server.

[0394] Output: Audio and text data sent to the server

[0395] Step 3:

[0396] Data Receipt and Classification

[0397] Input: Voice and text data sent from the device

[0398] Specific operation: The server receives data sent from the terminal and classifies it into voice data and text data.

[0399] Output: Classified audio and text data

[0400] Step 4:

[0401] Converting audio data to text

[0402] Input: Classified audio data

[0403] Specific operation: The server uses a speech recognition engine such as Google Cloud Speech-to-Text to convert the voice data into text data.

[0404] Output: Converted text data

[0405] Step 5:

[0406] Natural Language Processing and Keyword Extraction

[0407] Input: Text data

[0408] What it does: The server uses a natural language processing engine to analyze the text data and extract important keywords and phrases.

[0409] Output: Extracted keywords and phrases

[0410] Step 6:

[0411] Emotion analysis (audio data)

[0412] Input: Audio data

[0413] Specific operation: The server uses the emotion engine to analyze the user's emotions from the voice data, specifically analyzing the tone, volume, and speed of the voice.

[0414] Output: Sentiment analysis results of audio data

[0415] Step 7:

[0416] Sentiment analysis (text data)

[0417] Input: Text data and extracted keywords and phrases

[0418] Specific operation: The server uses an emotion engine to analyze the user's emotions from text data, analyzing the content, structure, and wording of the text.

[0419] Output: Sentiment analysis results of text data

[0420] Step 8:

[0421] comprehensive evaluation

[0422] Input: Keywords, phrases, sentiment analysis results of audio data, sentiment analysis results of text data

[0423] Specific operation: The server comprehensively evaluates these analysis results, scores them using an AI model, and determines the possibility of harassment if the threshold is exceeded.

[0424] Output: Evaluation result of potential harassment

[0425] Step 9:

[0426] Generate and send warning messages

[0427] Input: Possible harassment assessment results

[0428] Specific behavior: If the server determines that there is a possibility of harassment, it will generate a warning message and display it as a pop-up on the user's screen.

[0429] Output: A warning message that is displayed to the user.

[0430] Step 10:

[0431] Report processing

[0432] Input: Warning message sending history and user behavior history

[0433] Specific operation: When the same user makes a remark that meets the harassment criteria, the server counts the number of times and generates a report if the number exceeds a set threshold. The generated report data is automatically sent to the harassment consultation desk and a notification is sent to the device that a report has been made.

[0434] Output: Report data to the harassment hotline and a report to the user

[0435] (Application example 2)

[0436] 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."

[0437] In modern corporate environments, harassment behaviors occurring via communication channels such as web conferencing, email, and chat are on the rise. Early detection of such harassment and minimizing damage are essential, but conventional methods often struggle to detect and respond appropriately. Furthermore, if a harassment detection system does not perform emotion analysis, it may miss subtle nuances and make incorrect judgments. Furthermore, the lack of an effective way to generate warning messages for users makes the system ineffective at deterring harassment.

[0438] 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.

[0439] In this invention, the server is a system for automating the detection and reporting of harassment, and includes a terminal means for capturing exchanges such as web conferences, emails, and chats in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a means for sending a warning message to a user when the possibility of harassment is detected, and a means for generating a prompt sentence and inputting it into the generative AI model to optimize the warning message to the user. This makes it possible to detect harassment behavior early and automatically issue appropriate warnings and reports.

[0440] "Harassment" refers to rude, offensive, or critical words or actions directed at others in the workplace or other environment.

[0441] "Detection" is the process by which a system identifies and recognizes targeted behavior or data.

[0442] An "automated reporting system" is a system that detects acts of harassment and automatically notifies the relevant parties of that information.

[0443] A "web conference" is a real-time audio and video conference conducted over the Internet.

[0444] "Email" is a means of sending documents containing digital information over the Internet to specific recipients.

[0445] "Chat" is the exchange of text messages in real time.

[0446] "Capture" is the process of collecting, recording, or copying digital data.

[0447] "Terminal means" refers to a device or application for capturing data and transmitting it to a server.

[0448] "Text data" is digital information expressed in the form of a string of characters.

[0449] "Audio data" refers to data in which audio is recorded in digital format.

[0450] "Server means" is a computer system for receiving, analyzing and otherwise processing data.

[0451] "Natural language processing" is a technology that allows computers to understand, interpret, and generate human language.

[0452] "Sentiment analysis" is a technology that extracts and analyzes emotions from text and audio data.

[0453] An "AI model" is a collection of algorithms designed to perform a specific task using machine learning techniques.

[0454] A "warning message" is a notification message that alerts the user.

[0455] A "prompt" is an instruction that is input into a generative AI model to obtain a result.

[0456] A "generative AI model" is a machine learning model that automatically generates text and images based on given instructions.

[0457] "Optimization" is the process of maximizing performance to achieve a goal.

[0458] This invention relates to a system for automating the detection and reporting of harassment behavior. The system includes a terminal means for capturing interactions such as web conferences, emails, and chats in real time, a server means for receiving and analyzing text and voice data sent from the terminal means, a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a server means for sending a warning message to a user when possible harassment is detected, a server means for reporting to a harassment hotline if the same user repeatedly ignores warnings, and a means for generating prompt sentences and inputting them into the generation AI model to optimize the warning message for the user.

[0459] Hardware and Software Configuration

[0460] 1. Terminal means:

[0461] It captures web conferences, emails, chats, and other interactions in real time on the user's device (PC, smartphone, etc.). This is done using the APIs and hooks of each application. For example, it captures what is being said using the API of a web conference tool.

[0462] 2. Server means:

[0463] A computer system that receives the transmitted data and classifies it for analysis. It converts the voice data into text using a speech recognition engine such as Google Cloud Speech-to-Text.

[0464] It uses a natural language processing engine (such as BERT or GPT-3 (registered trademark)) to analyze text data and extract important keywords and phrases.

[0465] 3. Sentiment Analysis Engine:

[0466] The emotion analysis engine extracts emotions from audio and text data and uses common emotion analysis algorithms (e.g., the Oxyford engine).

[0467] 4. AI model:

[0468] To determine the likelihood of harassment, a machine learning model is used, which inputs the results of sentiment analysis of text and voice and scores the likelihood of harassment.

[0469] Specific examples of processing

[0470] 1. Data capture:

[0471] For example, suppose the following is said during a meeting:

[0472] "Data preparation is too slow. I want you to work more efficiently."

[0473] The device captures these statements in real time and collects them as audio and text data.

[0474] 2. Data reception and preprocessing:

[0475] The server receives the transmitted data, separates it into audio and text data, and sends them to separate processing pipelines.

[0476] 3. Audio to text conversion:

[0477] The server uses Google Cloud Speech-to-Text to convert the voice data into text.

[0478] 4. Natural Language Processing and Sentiment Analysis:

[0479] The server analyzes the text data using a natural language processing engine and extracts keywords such as "too slow" and "efficiently."

[0480] The emotion engine analyzes voice tone, volume, speed, etc. to detect critical and emotional elements.

[0481] 5. Generate warning messages:

[0482] The AI ​​model evaluates these analysis results comprehensively, and if it determines that there is a high possibility of harassment, it generates a warning message for the user, such as the following:

[0483] "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[0484] 6. Prompt generation and input to the AI ​​model:

[0485] The server generates prompts to optimize warning messages and inputs them into the generative AI model. An example of a prompt is as follows:

[0486] "Could this statement constitute workplace harassment (criticism, disrespect, or aggressive behavior)? Please help us determine."

[0487] This allows the system to detect harassing behavior and implement appropriate countermeasures quickly and effectively.

[0488] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0489] Step 1:

[0490] Data Capture

[0491] The terminal captures in real time the user's interactions with web conferences, emails, chats, etc. The terminal uses APIs and hooks to obtain messages and voice data from each application and stores them in temporary storage.

[0492] Input: Real-time data from web conferencing, email, chat, etc.

[0493] Output: Captured text and audio data

[0494] Step 2:

[0495] Data transmission

[0496] The terminal transmits the captured text data and voice data to the server.

[0497] Input: Captured text and audio data

[0498] Output: Data sent from the device to the server

[0499] Step 3:

[0500] Data reception and preprocessing

[0501] The server receives the data sent from the terminal and classifies it into voice data and text data for analysis.

[0502] Input: Text and voice data sent from the device

[0503] Output: Separated audio and text data

[0504] Step 4:

[0505] Converting audio data to text

[0506] The server converts the voice data into text using a speech recognition engine (e.g., Google Cloud Speech-to-Text).

[0507] Input: Audio data

[0508] Output: Converted text data

[0509] Step 5:

[0510] Natural Language Processing

[0511] The server analyzes the converted text data using a natural language processing engine (e.g., BERT or GPT-3) to extract important keywords and phrases.

[0512] Input: Text data

[0513] Output: Analysis results (keywords and phrases)

[0514] Step 6:

[0515] Emotion analysis

[0516] The server uses an emotion engine to analyze the emotion of the text and voice data, analyzing voice tone, volume, and rate to detect emotional states.

[0517] Input: Text and audio data

[0518] Output: Emotion analysis results

[0519] Step 7:

[0520] Determining the possibility of harassment

[0521] The server combines the results of natural language processing and sentiment analysis to use an AI model to score the likelihood of harassment.

[0522] Input: Natural language processing results and sentiment analysis results

[0523] Output: Harassment likelihood score

[0524] Step 8:

[0525] Generate a warning message

[0526] The server generates and sends a warning message to the user when potential harassment is detected. The prompt sentence is used as input to a generative AI model to generate an optimized warning message.

[0527] Input: Harassment likelihood score

[0528] Output: Warning message

[0529] Step 9:

[0530] Reporting and User Notification

[0531] If the user continues to ignore the warning, the server counts the number of times, and if the number exceeds a threshold, it generates and automatically sends report data to the harassment hotline. It also notifies the user that a report has been made.

[0532] Input: Warning ignore count

[0533] Output: Report data and user notification

[0534] 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.

[0535] 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.

[0536] 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.

[0537] [Second embodiment]

[0538] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0539] 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.

[0540] 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).

[0541] 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.

[0542] 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.

[0543] 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).

[0544] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0545] 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.

[0546] 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.

[0547] 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.

[0548] 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.

[0549] 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."

[0550] The present invention relates to a system for automating the detection and reporting of harassment behavior, and specific embodiments thereof will be described below.

[0551] System configuration

[0552] 1. Terminal means

[0553] The terminal means is a device that captures in real time the interactions of users in web conferences, emails, chats, etc. The terminal means uses APIs and hooks to acquire the interactions of each application and send the data to the server.

[0554] 2. Server Means

[0555] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means. The server means processes the text data and voice data separately, and converts the voice data into text using a voice recognition engine means.

[0556] It uses natural language processing and sentiment analysis to run AI models to determine potential harassment from text and audio data.

[0557] Program processing

[0558] Data collection

[0559] The device captures the user's web conference, email, chat, and other interactions in real time and sends the data to a server. For example, it uses the Teams API to obtain the content of comments and sends it to the server, including the audio data.

[0560] Data reception and preprocessing

[0561] The server receives the text data and the voice data and separates the data for analysis. The voice data is converted to text using a speech recognition engine.

[0562] Harassment Detection

[0563] The server uses a natural language processing engine to analyze the text data and extract key keywords and emotional tones, while an emotion analysis engine analyzes the voice data to detect the emotional state of the user.

[0564] The server combines these analysis results and runs an AI model to determine whether there is a possibility of harassment. If the analysis result exceeds a threshold, it is determined to be harassment.

[0565] Sending a warning message

[0566] If the server determines that a message is potentially harassing, it generates a warning message for the user and displays it as a pop-up on the user's screen. For example, a message might say in a Teams chat, "This message may contain harassing content. Please choose your language carefully."

[0567] Report processing

[0568] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[0569] The server generates and automatically sends report data to the harassment hotline, including the content of the offending remark, the date and time, and user information.

[0570] The device will notify the user that a report has been made, for example by displaying a message that reads, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0571] Specific examples

[0572] For example, during a team meeting, someone might say:

[0573] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[0574] The device captures these statements in real time and sends them to the server as text and audio data.

[0575] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[0576] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0577] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[0578] In this way, the system effectively detects harassment and enables immediate action.

[0579] The processing flow will be explained below.

[0580] Step 1:

[0581] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[0582] Step 2:

[0583] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[0584] Step 3:

[0585] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[0586] Step 4:

[0587] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[0588] Step 5:

[0589] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[0590] Step 6:

[0591] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[0592] Step 7:

[0593] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[0594] Step 8:

[0595] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[0596] Step 9:

[0597] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[0598] Step 10:

[0599] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[0600] Step 11:

[0601] The server generates report data, which includes the content of the offending comment, the date and time, and user information. The generated report data is prepared for sending to the harassment hotline.

[0602] Step 12:

[0603] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[0604] Step 13:

[0605] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0606] Through these steps, the system enables automatic detection of harassment behavior and immediate response.

[0607] Example 1

[0608] 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."

[0609] In today's digital communication environment, the possibility of harassment is increasing. However, there are limited means to detect it in a timely manner and take appropriate measures, resulting in a lack of appropriate protection for victims. In particular, for companies and organizations with large user bases, it is difficult to monitor and deal with harassment manually, and there is a growing need for effective automated systems.

[0610] 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.

[0611] In this invention, the server includes device means for capturing user interactions such as meetings, emails, and communications in real time, computer means for receiving and analyzing text and voice data sent from the device means, speech recognition engine means for converting voice data into text data, generative AI model means for analyzing the text and voice data using natural language processing and sentiment analysis to determine the possibility of harassment, means for sending a warning message to a user when the possibility of harassment is detected, and means for automatically reporting to a consultation center if the same user repeatedly ignores warnings. This makes it possible to efficiently and effectively detect harassment even in large-scale communication environments and take timely measures.

[0612] A "user" is an individual or group that communicates using the system.

[0613] "Conference" refers to an activity in which multiple users exchange information in real time using audio and video.

[0614] "Email" is a means of sending and receiving messages electronically over the Internet.

[0615] "Communication" refers to any means of exchanging information, including through text and voice.

[0616] "Real-time" refers to the processing and transmission of data and information instantly, with minimal delay.

[0617] "Device means" refers to a device for capturing a user's communication activity and transmitting that information to another system.

[0618] "Computer Means" refers to a computing device and associated hardware and software for analyzing, storing, and processing received data.

[0619] "Speech recognition engine means" refers to software or hardware for converting voice data into text data.

[0620] "Natural language processing" refers to computer technology that analyzes text data and understands its meaning and sentiment.

[0621] "Sentiment analysis" refers to the technology of detecting emotional tone and state from text or audio data.

[0622] A "generative AI model" refers to an artificial intelligence model that determines the possibility of harassment from input data.

[0623] "Warning message" includes a message that notifies a user in the event of possible harassment.

[0624] A "consultation desk" refers to a department or institution dedicated to receiving reports of harassment and dealing with them.

[0625] "Automatic means for reporting" means a system mechanism for reporting without manual intervention when specified conditions are met.

[0626] This invention is a system for detecting harassing behavior in users' online communications and automating appropriate responses.

[0627] System configuration

[0628] The system includes the following means:

[0629] 1. A device means of capturing user interactions in real time, such as meetings, emails, and correspondence.

[0630] The device (terminal) captures user speech and text messages in real time and sends the data to a server. For example, it uses APIs such as Microsoft Teams and Zoom to acquire voice data and text chat data during a meeting.

[0631] 2. Computer means for receiving and analyzing the text data and voice data transmitted from the device means

[0632] The server means (computer) receives the data sent from the terminals, classifies and stores the data for analysis. This means includes a high-performance database and data analysis software.

[0633] 3. A speech recognition engine that converts voice data into text data

[0634] The server uses a speech recognition engine such as Google Cloud Speech-to-Text API or IBM Watson to convert the audio data into text data, which is then used for further analysis.

[0635] 4. Generative AI modeling means for analyzing text and audio data using natural language processing and sentiment analysis to identify potential harassment.

[0636] The server analyzes the text and audio data using natural language processing engines such as Google Cloud Natural Language API and sentiment analysis engines such as IBM Watson Tone Analyzer. The analysis results are fed into a BERT-based generative AI model to identify potential harassment behavior.

[0637] 5. A means of sending warning messages to users when potential harassment is detected

[0638] If the server determines that an activity is likely to be harassing, it generates and sends a warning message to the user, which can appear as a popup on messaging platforms such as Teams or Slack.

[0639] 6. A method for automatically reporting to the helpline if the same user repeatedly ignores warnings

[0640] The server detects users who repeatedly engage in harassing behavior and counts the number of times. If the number of times exceeds a threshold, it automatically generates and sends a report to the harassment hotline. This data includes the content of the offending remarks, the date and time, and user information.

[0641] Specific examples

[0642] For example, during a team meeting, someone might say:

[0643] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[0644] The device captures this speech in real time and sends it to the server as text and audio data.

[0645] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[0646] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0647] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[0648] Prompt Sentence Examples

[0649] The accuracy of harassment detection can be improved by feeding prompt sentences like the following into the AI ​​model:

[0650] "Please determine whether the following conversation constitutes harassment. The statement: 'You're late again. This is annoying because it happens all the time. Can you do it properly?'"

[0651] In this way, the present invention can efficiently and effectively detect harassment behavior even in large-scale communication environments and take countermeasures in a timely manner.

[0652] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0653] Step 1: Data collection

[0654] The device captures users' meetings, emails, chats and other interactions in real time.

[0655] Specifically, the device uses APIs and hooks to retrieve speech and text messages from, for example, web conferencing tools or messaging applications.

[0656] Input: User utterances (voice data) and text messages (text data)

[0657] Data processing: Real-time capture

[0658] Output: Audio and text data

[0659] Step 2: Data reception and preprocessing

[0660] The server receives the text data and voice data sent from the terminal.

[0661] Specifically, the server separates the voice data from the text data for analysis, and converts the voice data into text using a speech recognition engine, such as the Google Cloud Speech-to-Text API.

[0662] Input: Voice and text data sent from the device

[0663] Data processing: Speech to text conversion using a speech recognition engine

[0664] Output: Converted text data

[0665] Step 3: Harassment detection

[0666] The server analyzes the converted text data using a natural language processing engine and a sentiment analysis engine.

[0667] Specifically, the system uses the Google Cloud Natural Language API to extract keywords and emotional tones from text data, and IBM Watson Tone Analyzer to extract emotional tones from audio data.These analysis results are then fed into an AI model (e.g., a BERT-based generative AI model) to determine the likelihood of harassment.

[0668] Input: Converted text data

[0669] Data Processing: Natural Language Processing and Sentiment Analysis

[0670] Output: Harassment possibility assessment result

[0671] Step 4: Sending a warning message

[0672] If the server determines that there is a high possibility of harassment, it generates and sends a warning message to the user.

[0673] Specifically, the server uses the Teams API to generate a warning message, such as "This message may be harassing. Please choose your words carefully," and send it to the user.

[0674] Input: Harassment possibility assessment result

[0675] Data processing: Generate warning messages

[0676] Output: Warning message sent to the user

[0677] Step 5: Report processing

[0678] When the same user repeatedly makes statements that meet the criteria for harassment, the server counts the number of times, and when the number exceeds a threshold, it generates report data and automatically sends it to a consultation desk.

[0679] Specifically, the server generates report data including the content of the offending comment, the date and time, and user information, and sends it to the harassment hotline. It also displays a message on the user's device informing them that a report has been made.

[0680] Input: Count of harassment incidents

[0681] Data processing: Generation of report data

[0682] Output: Report data sent to the consultation desk and message sent to the user

[0683] (Application example 1)

[0684] 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."

[0685] In recent years, harassment between customers and employees in brick-and-mortar stores has become a social problem. However, this type of problem is often difficult to detect and respond to immediately, which often increases the mental burden on employees. Therefore, an effective automatic detection and immediate response system is needed to address this issue.

[0686] 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.

[0687] In this invention, the server includes a terminal means for capturing interactions between customers and employees in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, and a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data. This makes it possible to detect harassment between customers and employees in a physical store in real time and issue an immediate warning or report.

[0688] "Customer" refers to a person who visits a physical store to purchase goods or services.

[0689] "Employee" refers to a staff member who performs duties such as selling products and providing services at a physical store.

[0690] "Interaction" refers to the conversational and behavioral interactions that take place between customers and employees.

[0691] "Real-time capture" refers to instantly recording and acquiring interactions that occur at that moment.

[0692] "Terminal means" refers to a hardware and software system for capturing audio and video data.

[0693] "Text data" refers to data converted from audio into text.

[0694] "Audio data" refers to data that records spoken words or sounds.

[0695] "Server means" refers to a computer system for receiving, analyzing, and storing data.

[0696] "Natural language processing" refers to computer technology for understanding and analyzing human language.

[0697] "Emotion analysis" refers to the technology of analyzing a speaker's emotions and tone from voice and text data.

[0698] An "AI model" refers to a system that uses algorithms based on machine learning and deep learning to analyze specific data and make inferences.

[0699] "Warning message" refers to a warning message that is displayed to a user when harassment behavior is detected.

[0700] A "harassment consultation center" refers to a specialized institution that accepts consultations and takes measures regarding harassment acts.

[0701] "Reporting data" refers to data that records details of harassment when it occurs and sends them to a consultation hotline.

[0702] The present invention is implemented using a system that captures and analyzes interactions between customers and employees in real time. The system includes the following hardware and software:

[0703] Hardware and software used

[0704] Hardware:

[0705] 1. Smartphone (with built-in camera and microphone)

[0706] 2. Cloud-based servers (e.g. AWS EC2)

[0707] software:

[0708] 1. Speech recognition engine (e.g. Google Cloud Speech-to-Text)

[0709] 2. Natural language processing engine (e.g., Google Cloud Natural Language)

[0710] 3. Sentiment analysis engine (e.g., IBM Watson Tone Analyzer)

[0711] Specific processing of the system

[0712] 1. Data Collection:

[0713] The smartphone serves as the terminal device and captures conversations between users (customers and employees) in real time. The smartphone's microphone and camera are used to obtain audio and video data.

[0714] 2. Data reception and preprocessing:

[0715] The server receives the audio and video data sent from the terminal. The audio data is converted into text data via a voice recognition engine. Furthermore, facial expression analysis is performed on the video data to detect changes in emotions.

[0716] 3. Harassment Detection:

[0717] The server uses a natural language processing engine to analyze text data. It also uses an emotion analysis engine to analyze voice and facial expression data to detect the speaker's emotional tone. This data is then combined and a generative AI model is used to determine whether the behavior is likely to be harassing. The determination includes specific keywords and emotional tone.

[0718] 4. Sending warning messages:

[0719] If harassment is detected, the server will pop up a warning message on the terminal in real time, such as "This word may contain elements of harassment. Please use polite language."

[0720] 5. Report Processing:

[0721] If the same customer repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment hotline. The report data includes the content of the remarks, date and time, user information, etc.

[0722] Examples

[0723] For example, in one scenario, a customer might say:

[0724] "Your responses are always slow, can't you do something about it?"

[0725] The smartphone (terminal means) captures this speech with a microphone and transmits it to the server as audio data.

[0726] The server uses a voice recognition engine to convert the voice data into text data and a natural language processing engine to analyze the possibility of harassment.

[0727] A sentiment analysis engine is also used to analyze emotional tone from text and audio data.

[0728] If a harassing act is detected, the server sends a warning message to the terminal means in real time.

[0729] Prompt Sentence Examples

[0730] Capture real-time conversations between customers and store associates and perform text and tone analysis on:

[0731] Text: "You're always slow to respond"

[0732] Emotional Tone: Anger

[0733] Warning message: "This language may be harassing. Please use polite language."

[0734] This will realize a system that can detect harassment between customers and employees in physical stores in real time and issue warnings or reports immediately.

[0735] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0736] Step 1: Data collection

[0737] The device (smartphone) captures the conversation between the customer and the employee using a camera and microphone, thereby obtaining audio and video data as input.

[0738] Step 2: Send data

[0739] The device transmits the captured audio and video data to a server in real time using an internet connection.

[0740] Step 3: Preprocessing the audio data

[0741] The server converts the received voice data into text data using a voice recognition engine (Google Cloud Speech-to-Text). The input is voice data and the output is text data.

[0742] Step 4: Preprocessing the video data

[0743] The server processes the received video data using analysis technology and performs facial expression analysis. The input is video data and the output is emotional tone data.

[0744] Step 5: Natural Language Processing

[0745] The server uses a natural language processing engine (Google Cloud Natural Language) to analyze the text data, extracting important keywords and context, and assessing the likelihood of harassment as a result of the analysis. The input is the text data, and the output is the results of keyword and context analysis.

[0746] Step 6: Sentiment Analysis

[0747] The server uses an emotion analysis engine (IBM Watson Tone Analyzer) to analyze the text data and emotional tone data and determine the emotional tone. The input is the text data and emotional tone data, and the output is the result of the emotional tone analysis.

[0748] Step 7: Determining Harassment

[0749] The server integrates the results of natural language processing and sentiment analysis and uses a generative AI model to determine the likelihood of harassment. The input is keywords, the results of contextual analysis, and the results of emotional tone analysis, and the output is the harassment judgment result.

[0750] Step 8: Sending a warning message

[0751] If the server detects harassment, it will pop up a warning message on the device in real time. The input is the result of the harassment judgment, and the output is the warning message.

[0752] Step 9: Generate and send notification data

[0753] If the same user repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment consultation center. The input is the number of times harassment acts occurred, and the output is the generation and transmission of report data.

[0754] 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.

[0755] The present invention relates to a system for automating the detection and reporting of harassment behavior, and the following describes an embodiment incorporating an emotion engine for analyzing user emotions.

[0756] System configuration

[0757] 1. Terminal means

[0758] The terminal means is a device that captures user interactions such as web conferences, emails, and chats in real time. The terminal means uses APIs and hooks to obtain messages and voice data from each application and saves it in temporary storage. For example, the Teams API can be used to capture what is being said in a meeting.

[0759] 2. Server Means

[0760] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means, classifies the text data and voice data, and converts the voice data into text using a voice recognition engine means.

[0761] It runs AI models that use natural language processing and sentiment analysis to determine potential harassment from text and audio data.

[0762] 3. Emotion Engine

[0763] The emotion engine has the function to analyze the user's emotions and extracts emotions from voice data and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[0764] Program processing

[0765] Data collection

[0766] The device captures users' web conferences, emails, chats, and other interactions in real time and sends the data to a server. For example, it can capture comments made during a Teams meeting in real time and collect them as audio and text data.

[0767] Data reception and preprocessing

[0768] The server receives the data and separates it for analysis: it separates audio data from text data and sends each to a separate processing pipeline.

[0769] The server converts the voice data into text using a speech recognition engine, such as Google Cloud Speech-to-Text.

[0770] Harassment detection and sentiment analysis

[0771] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones.

[0772] The server uses an emotion engine to analyze the user's emotions from the voice data, analyzing the tone, volume, speed, etc. of the voice to detect the user's emotional state.

[0773] The server uses an emotion engine to analyze user emotions from text data, extracting emotions based on the content, structure, and wording of the text.

[0774] Overall rating and warning message sent

[0775] The server evaluates these analysis results comprehensively, using an AI model to score the analysis results and determine the possibility of harassment if a threshold is exceeded.

[0776] If the server detects a potential harassment, it will generate a warning message and display it on the user's screen. For example, a message in chat might say, "This message contains harassing and emotionally inappropriate content. Please choose your words carefully."

[0777] Report processing

[0778] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[0779] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[0780] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[0781] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0782] As a concrete example, suppose the following is said during a meeting:

[0783] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[0784] The device captures these statements in real time and sends them to the server as text and audio data.

[0785] The server receives this speech and converts the speech data into text using a speech recognition engine.

[0786] The server analyzes the text data using a natural language processing engine and extracts keywords such as "It's late," "This is a problem," and "Please do it properly."

[0787] The server uses an emotion engine to analyze the tone of the voice and the emotions in the text content, determining that the message is critical and contains strong emotional elements.

[0788] Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0789] If the user ignores the warning and continues to make critical comments, the server counts the number of times and issues a report if the number exceeds a threshold.

[0790] In this way, the system effectively detects harassment and emotional issues early and enables immediate response.

[0791] The processing flow will be explained below.

[0792] Step 1:

[0793] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[0794] Step 2:

[0795] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[0796] Step 3:

[0797] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[0798] Step 4:

[0799] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[0800] Step 5:

[0801] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[0802] Step 6:

[0803] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[0804] Step 7:

[0805] The server also performs sentiment analysis on text data, extracting emotions based on the content, structure, and wording of the text.

[0806] Step 8:

[0807] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[0808] Step 9:

[0809] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[0810] Step 10:

[0811] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[0812] Step 11:

[0813] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[0814] Step 12:

[0815] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[0816] Step 13:

[0817] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[0818] Step 14:

[0819] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0820] Specific examples

[0821] During a meeting, the following is said:

[0822] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[0823] Specific processing flow

[0824] Step 1:

[0825] The device captures this speech in real time, capturing text and audio data and storing it in temporary storage.

[0826] Step 2:

[0827] The device encrypts the captured data and sends it to the server over a secure channel.

[0828] Step 3:

[0829] The server separates the received data into text and audio data for analysis.

[0830] Step 4:

[0831] The server runs the voice data through a speech recognition engine and converts it into text.

[0832] Step 5:

[0833] The server analyzes the text data using a natural language processing engine and extracts keywords and phrases such as "It's late," "This is a problem," and "Please do it properly."

[0834] Step 6:

[0835] The server analyzes the tone, volume, and speed of the voice using an emotion engine to detect critical emotions.

[0836] Step 7:

[0837] The server analyzes the sentiment of the text from its content and vocabulary, extracting critical and emotional elements.

[0838] Step 8:

[0839] The server combines the analysis results and scores them using an AI model, determining that in this case the harassment threshold has been exceeded.

[0840] Step 9:

[0841] The server generates a warning message to the speaker, saying, "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[0842] Step 10:

[0843] The user is given the opportunity to review the warning message and correct their behavior.

[0844] Step 11:

[0845] If the user ignores the warning again and continues to make critical comments, the server counts the number of times and proceeds to the next step if the number exceeds a threshold.

[0846] Step 12:

[0847] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis.

[0848] Step 13:

[0849] The server automatically sends the generated report data to the harassment consultation center.

[0850] Step 14:

[0851] The device will notify the user that a report has been made, displaying the message, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[0852] In this way, the system detects harassment and emotional issues early and enables immediate response.

[0853] Example 2

[0854] 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."

[0855] Conventional harassment detection and reporting systems have problems with insufficient user emotion analysis, making it difficult to detect harassment in real time and respond immediately. Furthermore, recording and reporting repeated harassment behaviors by users who ignore warnings must be done manually, making it difficult to efficiently prevent harassment.

[0856] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0857] In this invention, the server includes means for receiving and classifying text data and voice data transmitted from the terminal means, means for utilizing a voice recognition engine for converting voice data into text, and means for executing an AI model for determining the possibility of harassment from the text and voice data using natural language processing and sentiment analysis, thereby enabling real-time detection of harassment behavior and user sentiment analysis.

[0858] The "terminal means" is a device for capturing exchanges such as web conferences, e-mails, and chats in real time and transmitting them to a server.

[0859] The "server means" is a computer system that receives data sent from the terminal means and classifies and analyzes the data.

[0860] A "voice recognition engine means" is a system that executes voice recognition techniques to convert voice data into text data.

[0861] "Natural language processing" is a technology that analyzes text data and extracts important keywords and phrases.

[0862] "Emotion analysis" is a technology that analyzes a user's emotions from voice data and text data and detects their emotional state.

[0863] The "AI model" is an artificial intelligence technology that determines the possibility of harassment based on the results of natural language processing and sentiment analysis.

[0864] A "warning message" is a notification sent to a user when potential harassment is detected.

[0865] The "reporting mechanism" is a system that records repeated acts of harassment and reports them to a harassment hotline if a threshold is exceeded.

[0866] The present invention is a system for automating the detection and reporting of harassment behavior, and incorporates an emotion engine that analyzes user emotions. Specific embodiments of the present invention will be described below.

[0867] 1. System Configuration

[0868] Terminal means

[0869] The terminal means is a device that captures interactions such as web conferences, emails, and chats in real time. It uses APIs and hooks from the applications used by the user to obtain messages and voice data in real time. As a specific example, the Teams API is used to capture what is said during a meeting.

[0870] Server Means

[0871] The server means is a computer system that receives and analyzes text data and voice data sent from the terminal means. The server means uses a voice recognition engine means to convert voice data into text data. In this case, the Google Cloud Speech-to-Text service can be used.

[0872] The server performs natural language processing and emotion analysis on the text data and voice data, thereby executing an AI model that determines the possibility of harassment. The emotion engine analyzes emotions from the voice data and text data to detect the user's emotional state.

[0873] Emotion Engine

[0874] The emotion engine has the function of analyzing the user's emotions from voice and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[0875] 2. Program Processing Overview

[0876] Data collection

[0877] When a user is engaged in a web conference, email, chat, or other exchange, the device captures this in real time and sends it to the server. For example, remarks made during a Teams meeting can be captured in real time and collected as voice and text data.

[0878] Data reception and preprocessing

[0879] The server receives the data and separates it into voice and text, which is then converted into text using the Google Cloud Speech-to-Text speech recognition engine.

[0880] Harassment detection and sentiment analysis

[0881] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones. The server also analyzes emotions from the audio data, using an emotion engine to analyze voice tone, volume, speed, etc., thereby detecting the user's emotional state.

[0882] 3. Usage example

[0883] As a concrete example, suppose the following is said during a meeting:

[0884] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[0885] System Operation

[0886] 1. The device captures this speech in real time and sends it to the server as text and audio data.

[0887] 2. The server receives this speech and converts the speech data into text using a speech recognition engine.

[0888] 3. The server analyzes the text data using a natural language processing engine and extracts keywords such as "late," "this is a problem," and "please do it properly."

[0889] 4. The emotion engine analyzes the tone of the voice and the emotion of the text content and determines that the message is critical and emotionally charged.

[0890] 5. Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[0891] 6. If the user ignores the warning and continues to make critical comments, the server will count the number of times and issue a report if the number exceeds a threshold.

[0892] 4. Example prompts for generative AI models

[0893] You can further improve harassment detection by using prompts like the following to feed your generative AI model:

[0894] "Please analyze the following conversations to detect any harassing behavior and the emotional tone. Please also provide your analysis results based on the specific content of the statements."

[0895] Conversation: "You're late again. This is really annoying. Please be more careful next time."

[0896] Expected output: "This comment contains critical and emotional overtones and may be considered harassment."

[0897] By using such prompts, it becomes possible to perform advanced analysis of both the content and emotions of the statements.

[0898] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0899] Step 1:

[0900] Data collection

[0901] Input: User web conferencing, email, and chat interactions

[0902] Specific operation: The device captures the user's interactions in real time, such as web conferences, emails, and chats. For example, it uses the Teams API to capture what is said during a meeting.

[0903] Output: Audio and text data

[0904] Step 2:

[0905] Data transmission

[0906] Input: Audio and text data

[0907] Specific operation: The device sends the captured audio and text data to the server.

[0908] Output: Audio and text data sent to the server

[0909] Step 3:

[0910] Data Receipt and Classification

[0911] Input: Voice and text data sent from the device

[0912] Specific operation: The server receives data sent from the terminal and classifies it into voice data and text data.

[0913] Output: Classified audio and text data

[0914] Step 4:

[0915] Converting audio data to text

[0916] Input: Classified audio data

[0917] Specific operation: The server uses a speech recognition engine such as Google Cloud Speech-to-Text to convert the voice data into text data.

[0918] Output: Converted text data

[0919] Step 5:

[0920] Natural Language Processing and Keyword Extraction

[0921] Input: Text data

[0922] What it does: The server uses a natural language processing engine to analyze the text data and extract important keywords and phrases.

[0923] Output: Extracted keywords and phrases

[0924] Step 6:

[0925] Emotion analysis (audio data)

[0926] Input: Audio data

[0927] Specific operation: The server uses the emotion engine to analyze the user's emotions from the voice data, specifically analyzing the tone, volume, and speed of the voice.

[0928] Output: Sentiment analysis results of audio data

[0929] Step 7:

[0930] Sentiment analysis (text data)

[0931] Input: Text data and extracted keywords and phrases

[0932] Specific operation: The server uses an emotion engine to analyze the user's emotions from text data, analyzing the content, structure, and wording of the text.

[0933] Output: Sentiment analysis results of text data

[0934] Step 8:

[0935] comprehensive evaluation

[0936] Input: Keywords, phrases, sentiment analysis results of audio data, sentiment analysis results of text data

[0937] Specific operation: The server comprehensively evaluates these analysis results, scores them using an AI model, and determines the possibility of harassment if the threshold is exceeded.

[0938] Output: Evaluation result of potential harassment

[0939] Step 9:

[0940] Generate and send warning messages

[0941] Input: Possible harassment assessment results

[0942] Specific behavior: If the server determines that there is a possibility of harassment, it will generate a warning message and display it as a pop-up on the user's screen.

[0943] Output: A warning message that is displayed to the user.

[0944] Step 10:

[0945] Report processing

[0946] Input: Warning message sending history and user behavior history

[0947] Specific operation: When the same user makes a remark that meets the harassment criteria, the server counts the number of times and generates a report if the number exceeds a set threshold. The generated report data is automatically sent to the harassment consultation desk and a notification is sent to the device that a report has been made.

[0948] Output: Report data to the harassment hotline and a report to the user

[0949] (Application example 2)

[0950] 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."

[0951] In modern corporate environments, harassment behaviors occurring via communication channels such as web conferencing, email, and chat are on the rise. Early detection of such harassment and minimizing damage are essential, but conventional methods often struggle to detect and respond appropriately. Furthermore, if a harassment detection system does not perform emotion analysis, it may miss subtle nuances and make incorrect judgments. Furthermore, the lack of an effective way to generate warning messages for users makes the system ineffective at deterring harassment.

[0952] 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.

[0953] In this invention, the server is a system for automating the detection and reporting of harassment, and includes a terminal means for capturing exchanges such as web conferences, emails, and chats in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a means for sending a warning message to a user when the possibility of harassment is detected, and a means for generating a prompt sentence and inputting it into the generative AI model to optimize the warning message to the user. This makes it possible to detect harassment behavior early and automatically issue appropriate warnings and reports.

[0954] "Harassment" refers to rude, offensive, or critical words or actions directed at others in the workplace or other environment.

[0955] "Detection" is the process by which a system identifies and recognizes targeted behavior or data.

[0956] An "automated reporting system" is a system that detects acts of harassment and automatically notifies the relevant parties of that information.

[0957] A "web conference" is a real-time audio and video conference conducted over the Internet.

[0958] "Email" is a means of sending documents containing digital information over the Internet to specific recipients.

[0959] "Chat" is the exchange of text messages in real time.

[0960] "Capture" is the process of collecting, recording, or copying digital data.

[0961] "Terminal means" refers to a device or application for capturing data and transmitting it to a server.

[0962] "Text data" is digital information expressed in the form of a string of characters.

[0963] "Audio data" refers to data in which audio is recorded in digital format.

[0964] "Server means" is a computer system for receiving, analyzing and otherwise processing data.

[0965] "Natural language processing" is a technology that allows computers to understand, interpret, and generate human language.

[0966] "Sentiment analysis" is a technology that extracts and analyzes emotions from text and audio data.

[0967] An "AI model" is a collection of algorithms designed to perform a specific task using machine learning techniques.

[0968] A "warning message" is a notification message that alerts the user.

[0969] A "prompt" is an instruction that is input into a generative AI model to obtain a result.

[0970] A "generative AI model" is a machine learning model that automatically generates text and images based on given instructions.

[0971] "Optimization" is the process of maximizing performance to achieve a goal.

[0972] This invention relates to a system for automating the detection and reporting of harassment behavior. The system includes a terminal means for capturing interactions such as web conferences, emails, and chats in real time, a server means for receiving and analyzing text and voice data sent from the terminal means, a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a server means for sending a warning message to a user when possible harassment is detected, a server means for reporting to a harassment hotline if the same user repeatedly ignores warnings, and a means for generating prompt sentences and inputting them into the generation AI model to optimize the warning message for the user.

[0973] Hardware and Software Configuration

[0974] 1. Terminal means:

[0975] It captures web conferences, emails, chats, and other interactions in real time on the user's device (PC, smartphone, etc.). This is done using the APIs and hooks of each application. For example, it captures what is being said using the API of a web conference tool.

[0976] 2. Server means:

[0977] A computer system that receives the transmitted data and classifies it for analysis. It converts the voice data into text using a speech recognition engine such as Google Cloud Speech-to-Text.

[0978] Use natural language processing engines (such as BERT or GPT-3) to analyze text data and extract important keywords and phrases.

[0979] 3. Sentiment Analysis Engine:

[0980] The emotion analysis engine extracts emotions from audio and text data and uses common emotion analysis algorithms (e.g., the Oxyford engine).

[0981] 4. AI model:

[0982] To determine the likelihood of harassment, a machine learning model is used, which inputs the results of sentiment analysis of text and voice and scores the likelihood of harassment.

[0983] Specific examples of processing

[0984] 1. Data capture:

[0985] For example, suppose the following is said during a meeting:

[0986] "Data preparation is too slow. I want you to work more efficiently."

[0987] The device captures these statements in real time and collects them as audio and text data.

[0988] 2. Data reception and preprocessing:

[0989] The server receives the transmitted data, separates it into audio and text data, and sends them to separate processing pipelines.

[0990] 3. Audio to text conversion:

[0991] The server uses Google Cloud Speech-to-Text to convert the voice data into text.

[0992] 4. Natural Language Processing and Sentiment Analysis:

[0993] The server analyzes the text data using a natural language processing engine and extracts keywords such as "too slow" and "efficiently."

[0994] The emotion engine analyzes voice tone, volume, speed, etc. to detect critical and emotional elements.

[0995] 5. Generate warning messages:

[0996] The AI ​​model evaluates these analysis results comprehensively, and if it determines that there is a high possibility of harassment, it generates a warning message for the user, such as the following:

[0997] "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[0998] 6. Prompt generation and input to the AI ​​model:

[0999] The server generates prompts to optimize warning messages and inputs them into the generative AI model. An example of a prompt is as follows:

[1000] "Could this statement constitute workplace harassment (criticism, disrespect, or aggressive behavior)? Please help us determine."

[1001] This allows the system to detect harassing behavior and implement appropriate countermeasures quickly and effectively.

[1002] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1003] Step 1:

[1004] Data Capture

[1005] The terminal captures in real time the user's interactions with web conferences, emails, chats, etc. The terminal uses APIs and hooks to obtain messages and voice data from each application and stores them in temporary storage.

[1006] Input: Real-time data from web conferencing, email, chat, etc.

[1007] Output: Captured text and audio data

[1008] Step 2:

[1009] Data transmission

[1010] The terminal transmits the captured text data and voice data to the server.

[1011] Input: Captured text and audio data

[1012] Output: Data sent from the device to the server

[1013] Step 3:

[1014] Data reception and preprocessing

[1015] The server receives the data sent from the terminal and classifies it into voice data and text data for analysis.

[1016] Input: Text and voice data sent from the device

[1017] Output: Separated audio and text data

[1018] Step 4:

[1019] Converting audio data to text

[1020] The server converts the voice data into text using a speech recognition engine (e.g., Google Cloud Speech-to-Text).

[1021] Input: Audio data

[1022] Output: Converted text data

[1023] Step 5:

[1024] Natural Language Processing

[1025] The server analyzes the converted text data using a natural language processing engine (e.g., BERT or GPT-3) to extract important keywords and phrases.

[1026] Input: Text data

[1027] Output: Analysis results (keywords and phrases)

[1028] Step 6:

[1029] Emotion analysis

[1030] The server uses an emotion engine to analyze the emotion of the text and voice data, analyzing voice tone, volume, and rate to detect emotional states.

[1031] Input: Text and audio data

[1032] Output: Emotion analysis results

[1033] Step 7:

[1034] Determining the possibility of harassment

[1035] The server combines the results of natural language processing and sentiment analysis to use an AI model to score the likelihood of harassment.

[1036] Input: Natural language processing results and sentiment analysis results

[1037] Output: Harassment likelihood score

[1038] Step 8:

[1039] Generate a warning message

[1040] The server generates and sends a warning message to the user when potential harassment is detected. The prompt sentence is used as input to a generative AI model to generate an optimized warning message.

[1041] Input: Harassment likelihood score

[1042] Output: Warning message

[1043] Step 9:

[1044] Reporting and User Notification

[1045] If the user continues to ignore the warning, the server counts the number of times, and if the number exceeds a threshold, it generates and automatically sends report data to the harassment hotline. It also notifies the user that a report has been made.

[1046] Input: Warning ignore count

[1047] Output: Report data and user notification

[1048] 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.

[1049] 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.

[1050] 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.

[1051] [Third embodiment]

[1052] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1053] 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.

[1054] 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).

[1055] 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.

[1056] 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.

[1057] 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).

[1058] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1059] 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.

[1060] 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.

[1061] 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.

[1062] 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.

[1063] 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."

[1064] The present invention relates to a system for automating the detection and reporting of harassment behavior, and specific embodiments thereof will be described below.

[1065] System configuration

[1066] 1. Terminal means

[1067] The terminal means is a device that captures in real time the interactions of users in web conferences, emails, chats, etc. The terminal means uses APIs and hooks to acquire the interactions of each application and send the data to the server.

[1068] 2. Server Means

[1069] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means. The server means processes the text data and voice data separately, and converts the voice data into text using a voice recognition engine means.

[1070] It uses natural language processing and sentiment analysis to run AI models to determine potential harassment from text and audio data.

[1071] Program processing

[1072] Data collection

[1073] The device captures the user's web conference, email, chat, and other interactions in real time and sends the data to a server. For example, it uses the Teams API to obtain the content of comments and sends it to the server, including the audio data.

[1074] Data reception and preprocessing

[1075] The server receives the text data and the voice data and separates the data for analysis. The voice data is converted to text using a speech recognition engine.

[1076] Harassment Detection

[1077] The server uses a natural language processing engine to analyze the text data and extract key keywords and emotional tones, while an emotion analysis engine analyzes the voice data to detect the emotional state of the user.

[1078] The server combines these analysis results and runs an AI model to determine whether there is a possibility of harassment. If the analysis result exceeds a threshold, it is determined to be harassment.

[1079] Sending a warning message

[1080] If the server determines that a message is potentially harassing, it generates a warning message for the user and displays it as a pop-up on the user's screen. For example, a message might say in a Teams chat, "This message may contain harassing content. Please choose your language carefully."

[1081] Report processing

[1082] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[1083] The server generates and automatically sends report data to the harassment hotline, including the content of the offending remark, the date and time, and user information.

[1084] The device will notify the user that a report has been made, for example by displaying a message that reads, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1085] Specific examples

[1086] For example, during a team meeting, someone might say:

[1087] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[1088] The device captures these statements in real time and sends them to the server as text and audio data.

[1089] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[1090] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1091] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[1092] In this way, the system effectively detects harassment and enables immediate action.

[1093] The processing flow will be explained below.

[1094] Step 1:

[1095] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[1096] Step 2:

[1097] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[1098] Step 3:

[1099] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[1100] Step 4:

[1101] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[1102] Step 5:

[1103] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[1104] Step 6:

[1105] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[1106] Step 7:

[1107] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[1108] Step 8:

[1109] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[1110] Step 9:

[1111] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[1112] Step 10:

[1113] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[1114] Step 11:

[1115] The server generates report data, which includes the content of the offending comment, the date and time, and user information. The generated report data is prepared for sending to the harassment hotline.

[1116] Step 12:

[1117] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[1118] Step 13:

[1119] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1120] Through these steps, the system enables automatic detection of harassment behavior and immediate response.

[1121] Example 1

[1122] 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."

[1123] In today's digital communication environment, the possibility of harassment is increasing. However, there are limited means to detect it in a timely manner and take appropriate measures, resulting in a lack of appropriate protection for victims. In particular, for companies and organizations with large user bases, it is difficult to monitor and deal with harassment manually, and there is a growing need for effective automated systems.

[1124] 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.

[1125] In this invention, the server includes device means for capturing user interactions such as meetings, emails, and communications in real time, computer means for receiving and analyzing text and voice data sent from the device means, speech recognition engine means for converting voice data into text data, generative AI model means for analyzing the text and voice data using natural language processing and sentiment analysis to determine the possibility of harassment, means for sending a warning message to a user when the possibility of harassment is detected, and means for automatically reporting to a consultation center if the same user repeatedly ignores warnings. This makes it possible to efficiently and effectively detect harassment even in large-scale communication environments and take timely measures.

[1126] A "user" is an individual or group that communicates using the system.

[1127] "Conference" refers to an activity in which multiple users exchange information in real time using audio and video.

[1128] "Email" is a means of sending and receiving messages electronically over the Internet.

[1129] "Communication" refers to any means of exchanging information, including through text and voice.

[1130] "Real-time" refers to the processing and transmission of data and information instantly, with minimal delay.

[1131] "Device means" refers to a device for capturing a user's communication activity and transmitting that information to another system.

[1132] "Computer Means" refers to a computing device and associated hardware and software for analyzing, storing, and processing received data.

[1133] "Speech recognition engine means" refers to software or hardware for converting voice data into text data.

[1134] "Natural language processing" refers to computer technology that analyzes text data and understands its meaning and sentiment.

[1135] "Sentiment analysis" refers to the technology of detecting emotional tone and state from text or audio data.

[1136] A "generative AI model" refers to an artificial intelligence model that determines the possibility of harassment from input data.

[1137] "Warning message" includes a message that notifies a user in the event of possible harassment.

[1138] A "consultation desk" refers to a department or institution dedicated to receiving reports of harassment and dealing with them.

[1139] "Automatic means for reporting" means a system mechanism for reporting without manual intervention when specified conditions are met.

[1140] This invention is a system for detecting harassing behavior in users' online communications and automating appropriate responses.

[1141] System configuration

[1142] The system includes the following means:

[1143] 1. A device means of capturing user interactions in real time, such as meetings, emails, and correspondence.

[1144] The device (terminal) captures user speech and text messages in real time and sends the data to a server. For example, it uses APIs such as Microsoft Teams and Zoom to acquire voice data and text chat data during a meeting.

[1145] 2. Computer means for receiving and analyzing the text data and voice data transmitted from the device means

[1146] The server means (computer) receives the data sent from the terminals, classifies and stores the data for analysis. This means includes a high-performance database and data analysis software.

[1147] 3. A speech recognition engine that converts voice data into text data

[1148] The server uses a speech recognition engine such as Google Cloud Speech-to-Text API or IBM Watson to convert the audio data into text data, which is then used for further analysis.

[1149] 4. Generative AI modeling means for analyzing text and audio data using natural language processing and sentiment analysis to identify potential harassment.

[1150] The server analyzes the text and audio data using natural language processing engines such as Google Cloud Natural Language API and sentiment analysis engines such as IBM Watson Tone Analyzer. The analysis results are fed into a BERT-based generative AI model to identify potential harassment behavior.

[1151] 5. A means of sending warning messages to users when potential harassment is detected

[1152] If the server determines that an activity is likely to be harassing, it generates and sends a warning message to the user, which can appear as a popup on messaging platforms such as Teams or Slack.

[1153] 6. A method for automatically reporting to the helpline if the same user repeatedly ignores warnings

[1154] The server detects users who repeatedly engage in harassing behavior and counts the number of times. If the number of times exceeds a threshold, it automatically generates and sends a report to the harassment hotline. This data includes the content of the offending remarks, the date and time, and user information.

[1155] Specific examples

[1156] For example, during a team meeting, someone might say:

[1157] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[1158] The device captures this speech in real time and sends it to the server as text and audio data.

[1159] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[1160] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1161] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[1162] Prompt Sentence Examples

[1163] The accuracy of harassment detection can be improved by feeding prompt sentences like the following into the AI ​​model:

[1164] "Please determine whether the following conversation constitutes harassment. The statement: 'You're late again. This is annoying because it happens all the time. Can you do it properly?'"

[1165] In this way, the present invention can efficiently and effectively detect harassment behavior even in large-scale communication environments and take countermeasures in a timely manner.

[1166] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1167] Step 1: Data collection

[1168] The device captures users' meetings, emails, chats and other interactions in real time.

[1169] Specifically, the device uses APIs and hooks to retrieve speech and text messages from, for example, web conferencing tools or messaging applications.

[1170] Input: User utterances (voice data) and text messages (text data)

[1171] Data processing: Real-time capture

[1172] Output: Audio and text data

[1173] Step 2: Data reception and preprocessing

[1174] The server receives the text data and voice data sent from the terminal.

[1175] Specifically, the server separates the voice data from the text data for analysis, and converts the voice data into text using a speech recognition engine, such as the Google Cloud Speech-to-Text API.

[1176] Input: Voice and text data sent from the device

[1177] Data processing: Speech to text conversion using a speech recognition engine

[1178] Output: Converted text data

[1179] Step 3: Harassment detection

[1180] The server analyzes the converted text data using a natural language processing engine and a sentiment analysis engine.

[1181] Specifically, the system uses the Google Cloud Natural Language API to extract keywords and emotional tones from text data, and IBM Watson Tone Analyzer to extract emotional tones from audio data.These analysis results are then fed into an AI model (e.g., a BERT-based generative AI model) to determine the likelihood of harassment.

[1182] Input: Converted text data

[1183] Data Processing: Natural Language Processing and Sentiment Analysis

[1184] Output: Harassment possibility assessment result

[1185] Step 4: Sending a warning message

[1186] If the server determines that there is a high possibility of harassment, it generates and sends a warning message to the user.

[1187] Specifically, the server uses the Teams API to generate a warning message, such as "This message may be harassing. Please choose your words carefully," and send it to the user.

[1188] Input: Harassment possibility assessment result

[1189] Data processing: Generate warning messages

[1190] Output: Warning message sent to the user

[1191] Step 5: Report processing

[1192] When the same user repeatedly makes statements that meet the criteria for harassment, the server counts the number of times, and when the number exceeds a threshold, it generates report data and automatically sends it to a consultation desk.

[1193] Specifically, the server generates report data including the content of the offending comment, the date and time, and user information, and sends it to the harassment hotline. It also displays a message on the user's device informing them that a report has been made.

[1194] Input: Count of harassment incidents

[1195] Data processing: Generation of report data

[1196] Output: Report data sent to the consultation desk and message sent to the user

[1197] (Application example 1)

[1198] 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."

[1199] In recent years, harassment between customers and employees in brick-and-mortar stores has become a social problem. However, this type of problem is often difficult to detect and respond to immediately, which often increases the mental burden on employees. Therefore, an effective automatic detection and immediate response system is needed to address this issue.

[1200] 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.

[1201] In this invention, the server includes a terminal means for capturing interactions between customers and employees in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, and a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data. This makes it possible to detect harassment between customers and employees in a physical store in real time and issue an immediate warning or report.

[1202] "Customer" refers to a person who visits a physical store to purchase goods or services.

[1203] "Employee" refers to a staff member who performs duties such as selling products and providing services at a physical store.

[1204] "Interaction" refers to the conversational and behavioral interactions that take place between customers and employees.

[1205] "Real-time capture" refers to instantly recording and acquiring interactions that occur at that moment.

[1206] "Terminal means" refers to a hardware and software system for capturing audio and video data.

[1207] "Text data" refers to data converted from audio into text.

[1208] "Audio data" refers to data that records spoken words or sounds.

[1209] "Server means" refers to a computer system for receiving, analyzing, and storing data.

[1210] "Natural language processing" refers to computer technology for understanding and analyzing human language.

[1211] "Emotion analysis" refers to the technology of analyzing a speaker's emotions and tone from voice and text data.

[1212] An "AI model" refers to a system that uses algorithms based on machine learning and deep learning to analyze specific data and make inferences.

[1213] "Warning message" refers to a warning message that is displayed to a user when harassment behavior is detected.

[1214] A "harassment consultation center" refers to a specialized institution that accepts consultations and takes measures regarding harassment acts.

[1215] "Reporting data" refers to data that records details of harassment when it occurs and sends them to a consultation hotline.

[1216] The present invention is implemented using a system that captures and analyzes interactions between customers and employees in real time. The system includes the following hardware and software:

[1217] Hardware and software used

[1218] Hardware:

[1219] 1. Smartphone (with built-in camera and microphone)

[1220] 2. Cloud-based servers (e.g. AWS EC2)

[1221] software:

[1222] 1. Speech recognition engine (e.g. Google Cloud Speech-to-Text)

[1223] 2. Natural language processing engine (e.g., Google Cloud Natural Language)

[1224] 3. Sentiment analysis engine (e.g., IBM Watson Tone Analyzer)

[1225] Specific processing of the system

[1226] 1. Data Collection:

[1227] The smartphone serves as the terminal device and captures conversations between users (customers and employees) in real time. The smartphone's microphone and camera are used to obtain audio and video data.

[1228] 2. Data reception and preprocessing:

[1229] The server receives the audio and video data sent from the terminal. The audio data is converted into text data via a voice recognition engine. Furthermore, facial expression analysis is performed on the video data to detect changes in emotions.

[1230] 3. Harassment Detection:

[1231] The server uses a natural language processing engine to analyze text data. It also uses an emotion analysis engine to analyze voice and facial expression data to detect the speaker's emotional tone. This data is then combined and a generative AI model is used to determine whether the behavior is likely to be harassing. The determination includes specific keywords and emotional tone.

[1232] 4. Sending warning messages:

[1233] If harassment is detected, the server will pop up a warning message on the terminal in real time, such as "This word may contain elements of harassment. Please use polite language."

[1234] 5. Report Processing:

[1235] If the same customer repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment hotline. The report data includes the content of the remarks, date and time, user information, etc.

[1236] Examples

[1237] For example, in one scenario, a customer might say:

[1238] "Your responses are always slow, can't you do something about it?"

[1239] The smartphone (terminal means) captures this speech with a microphone and transmits it to the server as audio data.

[1240] The server uses a voice recognition engine to convert the voice data into text data and a natural language processing engine to analyze the possibility of harassment.

[1241] A sentiment analysis engine is also used to analyze emotional tone from text and audio data.

[1242] If a harassing act is detected, the server sends a warning message to the terminal means in real time.

[1243] Prompt Sentence Examples

[1244] Capture real-time conversations between customers and store associates and perform text and tone analysis on:

[1245] Text: "You're always slow to respond"

[1246] Emotional Tone: Anger

[1247] Warning message: "This language may be harassing. Please use polite language."

[1248] This will realize a system that can detect harassment between customers and employees in physical stores in real time and issue warnings or reports immediately.

[1249] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1250] Step 1: Data collection

[1251] The device (smartphone) captures the conversation between the customer and the employee using a camera and microphone, thereby obtaining audio and video data as input.

[1252] Step 2: Send data

[1253] The device transmits the captured audio and video data to a server in real time using an internet connection.

[1254] Step 3: Preprocessing the audio data

[1255] The server converts the received voice data into text data using a voice recognition engine (Google Cloud Speech-to-Text). The input is voice data and the output is text data.

[1256] Step 4: Preprocessing the video data

[1257] The server processes the received video data using analysis technology and performs facial expression analysis. The input is video data and the output is emotional tone data.

[1258] Step 5: Natural Language Processing

[1259] The server uses a natural language processing engine (Google Cloud Natural Language) to analyze the text data, extracting important keywords and context, and assessing the likelihood of harassment as a result of the analysis. The input is the text data, and the output is the results of keyword and context analysis.

[1260] Step 6: Sentiment Analysis

[1261] The server uses an emotion analysis engine (IBM Watson Tone Analyzer) to analyze the text data and emotional tone data and determine the emotional tone. The input is the text data and emotional tone data, and the output is the result of the emotional tone analysis.

[1262] Step 7: Determining Harassment

[1263] The server integrates the results of natural language processing and sentiment analysis and uses a generative AI model to determine the likelihood of harassment. The input is keywords, the results of contextual analysis, and the results of emotional tone analysis, and the output is the harassment judgment result.

[1264] Step 8: Sending a warning message

[1265] If the server detects harassment, it will pop up a warning message on the device in real time. The input is the result of the harassment judgment, and the output is the warning message.

[1266] Step 9: Generate and send notification data

[1267] If the same user repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment consultation center. The input is the number of times harassment acts occurred, and the output is the generation and transmission of report data.

[1268] 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.

[1269] The present invention relates to a system for automating the detection and reporting of harassment behavior, and the following describes an embodiment incorporating an emotion engine for analyzing user emotions.

[1270] System configuration

[1271] 1. Terminal means

[1272] The terminal means is a device that captures user interactions such as web conferences, emails, and chats in real time. The terminal means uses APIs and hooks to obtain messages and voice data from each application and saves it in temporary storage. For example, the Teams API can be used to capture what is being said in a meeting.

[1273] 2. Server Means

[1274] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means, classifies the text data and voice data, and converts the voice data into text using a voice recognition engine means.

[1275] It runs AI models that use natural language processing and sentiment analysis to determine potential harassment from text and audio data.

[1276] 3. Emotion Engine

[1277] The emotion engine has the function to analyze the user's emotions and extracts emotions from voice data and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[1278] Program processing

[1279] Data collection

[1280] The device captures users' web conferences, emails, chats, and other interactions in real time and sends the data to a server. For example, it can capture comments made during a Teams meeting in real time and collect them as audio and text data.

[1281] Data reception and preprocessing

[1282] The server receives the data and separates it for analysis: it separates audio data from text data and sends each to a separate processing pipeline.

[1283] The server converts the voice data into text using a speech recognition engine, such as Google Cloud Speech-to-Text.

[1284] Harassment detection and sentiment analysis

[1285] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones.

[1286] The server uses an emotion engine to analyze the user's emotions from the voice data, analyzing the tone, volume, speed, etc. of the voice to detect the user's emotional state.

[1287] The server uses an emotion engine to analyze user emotions from text data, extracting emotions based on the content, structure, and wording of the text.

[1288] Overall rating and warning message sent

[1289] The server evaluates these analysis results comprehensively, using an AI model to score the analysis results and determine the possibility of harassment if a threshold is exceeded.

[1290] If the server detects a potential harassment, it will generate a warning message and display it on the user's screen. For example, a message in chat might say, "This message contains harassing and emotionally inappropriate content. Please choose your words carefully."

[1291] Report processing

[1292] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[1293] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[1294] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[1295] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1296] As a concrete example, suppose the following is said during a meeting:

[1297] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[1298] The device captures these statements in real time and sends them to the server as text and audio data.

[1299] The server receives this speech and converts the speech data into text using a speech recognition engine.

[1300] The server analyzes the text data using a natural language processing engine and extracts keywords such as "late," "this is a problem," and "please do it properly."

[1301] The server uses an emotion engine to analyze the tone of the voice and the emotions in the text content, determining that the message is critical and contains strong emotional elements.

[1302] Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1303] If the user ignores the warning and continues to make critical comments, the server counts the number of times and issues a report if the number exceeds a threshold.

[1304] In this way, the system effectively detects harassment and emotional issues early and enables immediate response.

[1305] The processing flow will be explained below.

[1306] Step 1:

[1307] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[1308] Step 2:

[1309] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[1310] Step 3:

[1311] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[1312] Step 4:

[1313] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[1314] Step 5:

[1315] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[1316] Step 6:

[1317] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[1318] Step 7:

[1319] The server also performs sentiment analysis on text data, extracting emotions based on the content, structure, and wording of the text.

[1320] Step 8:

[1321] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[1322] Step 9:

[1323] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[1324] Step 10:

[1325] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[1326] Step 11:

[1327] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[1328] Step 12:

[1329] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[1330] Step 13:

[1331] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[1332] Step 14:

[1333] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1334] Specific examples

[1335] During a meeting, the following is said:

[1336] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[1337] Specific processing flow

[1338] Step 1:

[1339] The device captures this speech in real time, capturing text and audio data and storing it in temporary storage.

[1340] Step 2:

[1341] The device encrypts the captured data and sends it to the server over a secure channel.

[1342] Step 3:

[1343] The server separates the received data into text and audio data for analysis.

[1344] Step 4:

[1345] The server runs the voice data through a speech recognition engine and converts it into text.

[1346] Step 5:

[1347] The server analyzes the text data using a natural language processing engine and extracts keywords and phrases such as "It's late," "This is a problem," and "Please do it properly."

[1348] Step 6:

[1349] The server analyzes the tone, volume, and speed of the voice using an emotion engine to detect critical emotions.

[1350] Step 7:

[1351] The server analyzes the sentiment of the text from its content and vocabulary, extracting critical and emotional elements.

[1352] Step 8:

[1353] The server combines the analysis results and scores them using an AI model, determining that in this case the harassment threshold has been exceeded.

[1354] Step 9:

[1355] The server generates a warning message to the speaker, saying, "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[1356] Step 10:

[1357] The user is given the opportunity to review the warning message and correct their behavior.

[1358] Step 11:

[1359] If the user ignores the warning again and continues to make critical comments, the server counts the number of times and proceeds to the next step if the number exceeds a threshold.

[1360] Step 12:

[1361] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis.

[1362] Step 13:

[1363] The server automatically sends the generated report data to the harassment consultation center.

[1364] Step 14:

[1365] The device will notify the user that a report has been made, displaying the message, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1366] In this way, the system detects harassment and emotional issues early and enables immediate response.

[1367] Example 2

[1368] 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."

[1369] Conventional harassment detection and reporting systems have problems with insufficient user emotion analysis, making it difficult to detect harassment in real time and respond immediately. Furthermore, recording and reporting repeated harassment behaviors by users who ignore warnings must be done manually, making it difficult to efficiently prevent harassment.

[1370] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1371] In this invention, the server includes means for receiving and classifying text data and voice data transmitted from the terminal means, means for utilizing a voice recognition engine for converting voice data into text, and means for executing an AI model for determining the possibility of harassment from the text and voice data using natural language processing and sentiment analysis, thereby enabling real-time detection of harassment behavior and user sentiment analysis.

[1372] The "terminal means" is a device for capturing exchanges such as web conferences, e-mails, and chats in real time and transmitting them to a server.

[1373] The "server means" is a computer system that receives data sent from the terminal means and classifies and analyzes the data.

[1374] A "voice recognition engine means" is a system that executes voice recognition techniques to convert voice data into text data.

[1375] "Natural language processing" is a technology that analyzes text data and extracts important keywords and phrases.

[1376] "Emotion analysis" is a technology that analyzes a user's emotions from voice data and text data and detects their emotional state.

[1377] The "AI model" is an artificial intelligence technology that determines the possibility of harassment based on the results of natural language processing and sentiment analysis.

[1378] A "warning message" is a notification sent to a user when potential harassment is detected.

[1379] The "reporting mechanism" is a system that records repeated acts of harassment and reports them to a harassment hotline if a threshold is exceeded.

[1380] The present invention is a system for automating the detection and reporting of harassment behavior, and incorporates an emotion engine that analyzes user emotions. Specific embodiments of the present invention will be described below.

[1381] 1. System Configuration

[1382] Terminal means

[1383] The terminal means is a device that captures interactions such as web conferences, emails, and chats in real time. It uses APIs and hooks from the applications used by the user to obtain messages and voice data in real time. As a specific example, the Teams API is used to capture what is said during a meeting.

[1384] Server Means

[1385] The server means is a computer system that receives and analyzes text data and voice data sent from the terminal means. The server means uses a voice recognition engine means to convert voice data into text data. In this case, the Google Cloud Speech-to-Text service can be used.

[1386] The server performs natural language processing and emotion analysis on the text data and voice data, thereby executing an AI model that determines the possibility of harassment. The emotion engine analyzes emotions from the voice data and text data to detect the user's emotional state.

[1387] Emotion Engine

[1388] The emotion engine has the function of analyzing the user's emotions from voice and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[1389] 2. Program Processing Overview

[1390] Data collection

[1391] When a user is engaged in a web conference, email, chat, or other exchange, the device captures this in real time and sends it to the server. For example, remarks made during a Teams meeting can be captured in real time and collected as voice and text data.

[1392] Data reception and preprocessing

[1393] The server receives the data and separates it into voice and text, which is then converted into text using the Google Cloud Speech-to-Text speech recognition engine.

[1394] Harassment detection and sentiment analysis

[1395] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones. The server also analyzes emotions from the audio data, using an emotion engine to analyze voice tone, volume, speed, etc., thereby detecting the user's emotional state.

[1396] 3. Usage example

[1397] As a concrete example, suppose the following is said during a meeting:

[1398] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[1399] System Operation

[1400] 1. The device captures this speech in real time and sends it to the server as text and audio data.

[1401] 2. The server receives this speech and converts the speech data into text using a speech recognition engine.

[1402] 3. The server analyzes the text data using a natural language processing engine and extracts keywords such as "late," "this is a problem," and "please do it properly."

[1403] 4. The emotion engine analyzes the tone of the voice and the emotion of the text content and determines that the message is critical and emotionally charged.

[1404] 5. Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1405] 6. If the user ignores the warning and continues to make critical comments, the server will count the number of times and issue a report if the number exceeds a threshold.

[1406] 4. Example prompts for generative AI models

[1407] You can further improve harassment detection by using prompts like the following to feed your generative AI model:

[1408] "Please analyze the following conversations to detect any harassing behavior and the emotional tone. Please also provide your analysis results based on the specific content of the statements."

[1409] Conversation: "You're late again. This is really annoying. Please be more careful next time."

[1410] Expected output: "This comment contains critical and emotional overtones and may be considered harassment."

[1411] By using such prompts, it becomes possible to perform advanced analysis of both the content and emotions of the statements.

[1412] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1413] Step 1:

[1414] Data collection

[1415] Input: User web conferencing, email, and chat interactions

[1416] Specific operation: The device captures the user's interactions in real time, such as web conferences, emails, and chats. For example, it uses the Teams API to capture what is said during a meeting.

[1417] Output: Audio and text data

[1418] Step 2:

[1419] Data transmission

[1420] Input: Audio and text data

[1421] Specific operation: The device sends the captured audio and text data to the server.

[1422] Output: Audio and text data sent to the server

[1423] Step 3:

[1424] Data Receipt and Classification

[1425] Input: Voice and text data sent from the device

[1426] Specific operation: The server receives data sent from the terminal and classifies it into voice data and text data.

[1427] Output: Classified audio and text data

[1428] Step 4:

[1429] Converting audio data to text

[1430] Input: Classified audio data

[1431] Specific operation: The server uses a speech recognition engine such as Google Cloud Speech-to-Text to convert the voice data into text data.

[1432] Output: Converted text data

[1433] Step 5:

[1434] Natural Language Processing and Keyword Extraction

[1435] Input: Text data

[1436] What it does: The server uses a natural language processing engine to analyze the text data and extract important keywords and phrases.

[1437] Output: Extracted keywords and phrases

[1438] Step 6:

[1439] Emotion analysis (audio data)

[1440] Input: Audio data

[1441] Specific operation: The server uses the emotion engine to analyze the user's emotions from the voice data, specifically analyzing the tone, volume, and speed of the voice.

[1442] Output: Sentiment analysis results of audio data

[1443] Step 7:

[1444] Sentiment analysis (text data)

[1445] Input: Text data and extracted keywords and phrases

[1446] Specific operation: The server uses an emotion engine to analyze the user's emotions from text data, analyzing the content, structure, and wording of the text.

[1447] Output: Sentiment analysis results of text data

[1448] Step 8:

[1449] comprehensive evaluation

[1450] Input: Keywords, phrases, sentiment analysis results of audio data, sentiment analysis results of text data

[1451] Specific operation: The server comprehensively evaluates these analysis results, scores them using an AI model, and determines the possibility of harassment if the threshold is exceeded.

[1452] Output: Evaluation result of potential harassment

[1453] Step 9:

[1454] Generate and send warning messages

[1455] Input: Possible harassment assessment results

[1456] Specific behavior: If the server determines that there is a possibility of harassment, it will generate a warning message and display it as a pop-up on the user's screen.

[1457] Output: A warning message that is displayed to the user.

[1458] Step 10:

[1459] Report processing

[1460] Input: Warning message sending history and user behavior history

[1461] Specific operation: When the same user makes a remark that meets the harassment criteria, the server counts the number of times and generates a report if the number exceeds a set threshold. The generated report data is automatically sent to the harassment consultation desk and a notification is sent to the device that a report has been made.

[1462] Output: Report data to the harassment hotline and a report to the user

[1463] (Application example 2)

[1464] 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."

[1465] In modern corporate environments, harassment behaviors occurring via communication channels such as web conferencing, email, and chat are on the rise. Early detection of such harassment and minimizing damage are essential, but conventional methods often struggle to detect and respond appropriately. Furthermore, if a harassment detection system does not perform emotion analysis, it may miss subtle nuances and make incorrect judgments. Furthermore, the lack of an effective way to generate warning messages for users makes the system ineffective at deterring harassment.

[1466] 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.

[1467] In this invention, the server is a system for automating the detection and reporting of harassment, and includes a terminal means for capturing exchanges such as web conferences, emails, and chats in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a means for sending a warning message to a user when the possibility of harassment is detected, and a means for generating a prompt sentence and inputting it into the generative AI model to optimize the warning message to the user. This makes it possible to detect harassment behavior early and automatically issue appropriate warnings and reports.

[1468] "Harassment" refers to rude, offensive, or critical words or actions directed at others in the workplace or other environment.

[1469] "Detection" is the process by which a system identifies and recognizes targeted behavior or data.

[1470] An "automated reporting system" is a system that detects acts of harassment and automatically notifies the relevant parties of that information.

[1471] A "web conference" is a real-time audio and video conference conducted over the Internet.

[1472] "Email" is a means of sending documents containing digital information over the Internet to specific recipients.

[1473] "Chat" is the exchange of text messages in real time.

[1474] "Capture" is the process of collecting, recording, or copying digital data.

[1475] "Terminal means" refers to a device or application for capturing data and transmitting it to a server.

[1476] "Text data" is digital information expressed in the form of a string of characters.

[1477] "Audio data" refers to data in which audio is recorded in digital format.

[1478] "Server means" is a computer system for receiving, analyzing and otherwise processing data.

[1479] "Natural language processing" is a technology that allows computers to understand, interpret, and generate human language.

[1480] "Sentiment analysis" is a technology that extracts and analyzes emotions from text and audio data.

[1481] An "AI model" is a collection of algorithms designed to perform a specific task using machine learning techniques.

[1482] A "warning message" is a notification message that alerts the user.

[1483] A "prompt" is an instruction that is input into a generative AI model to obtain a result.

[1484] A "generative AI model" is a machine learning model that automatically generates text and images based on given instructions.

[1485] "Optimization" is the process of maximizing performance to achieve a goal.

[1486] This invention relates to a system for automating the detection and reporting of harassment behavior. The system includes a terminal means for capturing interactions such as web conferences, emails, and chats in real time, a server means for receiving and analyzing text and voice data sent from the terminal means, a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a server means for sending a warning message to a user when possible harassment is detected, a server means for reporting to a harassment hotline if the same user repeatedly ignores warnings, and a means for generating prompt sentences and inputting them into the generation AI model to optimize the warning message for the user.

[1487] Hardware and Software Configuration

[1488] 1. Terminal means:

[1489] It captures web conferences, emails, chats, and other interactions in real time on the user's device (PC, smartphone, etc.). This is done using the APIs and hooks of each application. For example, it captures what is being said using the API of a web conference tool.

[1490] 2. Server means:

[1491] A computer system that receives the transmitted data and classifies it for analysis. It converts the voice data into text using a speech recognition engine such as Google Cloud Speech-to-Text.

[1492] Use natural language processing engines (such as BERT or GPT-3) to analyze text data and extract important keywords and phrases.

[1493] 3. Sentiment Analysis Engine:

[1494] The emotion analysis engine extracts emotions from audio and text data and uses common emotion analysis algorithms (e.g., the Oxyford engine).

[1495] 4. AI model:

[1496] To determine the likelihood of harassment, a machine learning model is used, which inputs the results of sentiment analysis of text and voice and scores the likelihood of harassment.

[1497] Specific examples of processing

[1498] 1. Data capture:

[1499] For example, suppose the following is said during a meeting:

[1500] "Data preparation is too slow. I want you to work more efficiently."

[1501] The device captures these statements in real time and collects them as audio and text data.

[1502] 2. Data reception and preprocessing:

[1503] The server receives the transmitted data, separates it into audio and text data, and sends them to separate processing pipelines.

[1504] 3. Audio to text conversion:

[1505] The server uses Google Cloud Speech-to-Text to convert the voice data into text.

[1506] 4. Natural Language Processing and Sentiment Analysis:

[1507] The server analyzes the text data using a natural language processing engine and extracts keywords such as "too slow" and "efficiently."

[1508] The emotion engine analyzes voice tone, volume, speed, etc. to detect critical and emotional elements.

[1509] 5. Generate warning messages:

[1510] The AI ​​model evaluates these analysis results comprehensively, and if it determines that there is a high possibility of harassment, it generates a warning message for the user, such as the following:

[1511] "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[1512] 6. Prompt generation and input to the AI ​​model:

[1513] The server generates prompts to optimize warning messages and inputs them into the generative AI model. An example of a prompt is as follows:

[1514] "Could this statement constitute workplace harassment (criticism, disrespect, or aggressive behavior)? Please help us determine."

[1515] This allows the system to detect harassing behavior and implement appropriate countermeasures quickly and effectively.

[1516] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1517] Step 1:

[1518] Data Capture

[1519] The terminal captures in real time the user's interactions with web conferences, emails, chats, etc. The terminal uses APIs and hooks to obtain messages and voice data from each application and stores them in temporary storage.

[1520] Input: Real-time data from web conferencing, email, chat, etc.

[1521] Output: Captured text and audio data

[1522] Step 2:

[1523] Data transmission

[1524] The terminal transmits the captured text data and voice data to the server.

[1525] Input: Captured text and audio data

[1526] Output: Data sent from the device to the server

[1527] Step 3:

[1528] Data reception and preprocessing

[1529] The server receives the data sent from the terminal and classifies it into voice data and text data for analysis.

[1530] Input: Text and voice data sent from the device

[1531] Output: Separated audio and text data

[1532] Step 4:

[1533] Converting audio data to text

[1534] The server converts the voice data into text using a speech recognition engine (e.g., Google Cloud Speech-to-Text).

[1535] Input: Audio data

[1536] Output: Converted text data

[1537] Step 5:

[1538] Natural Language Processing

[1539] The server analyzes the converted text data using a natural language processing engine (e.g., BERT or GPT-3) to extract important keywords and phrases.

[1540] Input: Text data

[1541] Output: Analysis results (keywords and phrases)

[1542] Step 6:

[1543] Emotion analysis

[1544] The server uses an emotion engine to analyze the emotion of the text and voice data, analyzing voice tone, volume, and rate to detect emotional states.

[1545] Input: Text and audio data

[1546] Output: Emotion analysis results

[1547] Step 7:

[1548] Determining the possibility of harassment

[1549] The server combines the results of natural language processing and sentiment analysis to use an AI model to score the likelihood of harassment.

[1550] Input: Natural language processing results and sentiment analysis results

[1551] Output: Harassment likelihood score

[1552] Step 8:

[1553] Generate a warning message

[1554] The server generates and sends a warning message to the user when potential harassment is detected. The prompt sentence is used as input to a generative AI model to generate an optimized warning message.

[1555] Input: Harassment likelihood score

[1556] Output: Warning message

[1557] Step 9:

[1558] Reporting and User Notification

[1559] If the user continues to ignore the warning, the server counts the number of times, and if the number exceeds a threshold, it generates and automatically sends report data to the harassment hotline. It also notifies the user that a report has been made.

[1560] Input: Warning ignore count

[1561] Output: Report data and user notification

[1562] 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.

[1563] 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.

[1564] 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.

[1565] [Fourth embodiment]

[1566] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1567] 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.

[1568] 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).

[1569] 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.

[1570] 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.

[1571] 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).

[1572] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1573] 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.

[1574] 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.

[1575] 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.

[1576] 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.

[1577] 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.

[1578] 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."

[1579] The present invention relates to a system for automating the detection and reporting of harassment behavior, and specific embodiments thereof will be described below.

[1580] System configuration

[1581] 1. Terminal means

[1582] The terminal means is a device that captures in real time the interactions of users in web conferences, emails, chats, etc. The terminal means uses APIs and hooks to acquire the interactions of each application and send the data to the server.

[1583] 2. Server Means

[1584] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means. The server means processes the text data and voice data separately, and converts the voice data into text using a voice recognition engine means.

[1585] It uses natural language processing and sentiment analysis to run AI models to determine potential harassment from text and audio data.

[1586] Program processing

[1587] Data collection

[1588] The device captures the user's web conference, email, chat, and other interactions in real time and sends the data to a server. For example, it uses the Teams API to obtain the content of comments and sends it to the server, including the audio data.

[1589] Data reception and preprocessing

[1590] The server receives the text data and the voice data and separates the data for analysis. The voice data is converted to text using a speech recognition engine.

[1591] Harassment Detection

[1592] The server uses a natural language processing engine to analyze the text data and extract key keywords and emotional tones, while an emotion analysis engine analyzes the voice data to detect the emotional state of the user.

[1593] The server combines these analysis results and runs an AI model to determine whether there is a possibility of harassment. If the analysis result exceeds a threshold, it is determined to be harassment.

[1594] Sending a warning message

[1595] If the server determines that a message is potentially harassing, it generates a warning message for the user and displays it as a pop-up on the user's screen. For example, a message might say in a Teams chat, "This message may contain harassing content. Please choose your language carefully."

[1596] Report processing

[1597] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[1598] The server generates and automatically sends report data to the harassment hotline, including the content of the offending remark, the date and time, and user information.

[1599] The device will notify the user that a report has been made, for example by displaying a message that reads, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1600] Specific examples

[1601] For example, during a team meeting, someone might say:

[1602] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[1603] The device captures these statements in real time and sends them to the server as text and audio data.

[1604] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[1605] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1606] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[1607] In this way, the system effectively detects harassment and enables immediate action.

[1608] The processing flow will be explained below.

[1609] Step 1:

[1610] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[1611] Step 2:

[1612] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[1613] Step 3:

[1614] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[1615] Step 4:

[1616] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[1617] Step 5:

[1618] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[1619] Step 6:

[1620] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[1621] Step 7:

[1622] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[1623] Step 8:

[1624] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[1625] Step 9:

[1626] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[1627] Step 10:

[1628] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[1629] Step 11:

[1630] The server generates report data, which includes the content of the offending comment, the date and time, and user information. The generated report data is prepared for sending to the harassment hotline.

[1631] Step 12:

[1632] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[1633] Step 13:

[1634] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1635] Through these steps, the system enables automatic detection of harassment behavior and immediate response.

[1636] Example 1

[1637] 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."

[1638] In today's digital communication environment, the possibility of harassment is increasing. However, there are limited means to detect it in a timely manner and take appropriate measures, resulting in a lack of appropriate protection for victims. In particular, for companies and organizations with large user bases, it is difficult to monitor and deal with harassment manually, and there is a growing need for effective automated systems.

[1639] 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.

[1640] In this invention, the server includes device means for capturing user interactions such as meetings, emails, and communications in real time, computer means for receiving and analyzing text and voice data sent from the device means, speech recognition engine means for converting voice data into text data, generative AI model means for analyzing the text and voice data using natural language processing and sentiment analysis to determine the possibility of harassment, means for sending a warning message to a user when the possibility of harassment is detected, and means for automatically reporting to a consultation center if the same user repeatedly ignores warnings. This makes it possible to efficiently and effectively detect harassment even in large-scale communication environments and take timely measures.

[1641] A "user" is an individual or group that communicates using the system.

[1642] "Conference" refers to an activity in which multiple users exchange information in real time using audio and video.

[1643] "Email" is a means of sending and receiving messages electronically over the Internet.

[1644] "Communication" refers to any means of exchanging information, including through text and voice.

[1645] "Real-time" refers to the processing and transmission of data and information instantly, with minimal delay.

[1646] "Device means" refers to a device for capturing a user's communication activity and transmitting that information to another system.

[1647] "Computer Means" refers to a computing device and associated hardware and software for analyzing, storing, and processing received data.

[1648] "Speech recognition engine means" refers to software or hardware for converting voice data into text data.

[1649] "Natural language processing" refers to computer technology that analyzes text data and understands its meaning and sentiment.

[1650] "Sentiment analysis" refers to the technology of detecting emotional tone and state from text or audio data.

[1651] A "generative AI model" refers to an artificial intelligence model that determines the possibility of harassment from input data.

[1652] "Warning message" includes a message that notifies a user in the event of possible harassment.

[1653] A "consultation desk" refers to a department or institution dedicated to receiving reports of harassment and dealing with them.

[1654] "Automatic means for reporting" means a system mechanism for reporting without manual intervention when specified conditions are met.

[1655] This invention is a system for detecting harassing behavior in users' online communications and automating appropriate responses.

[1656] System configuration

[1657] The system includes the following means:

[1658] 1. A device means of capturing user interactions in real time, such as meetings, emails, and correspondence.

[1659] The device (terminal) captures user speech and text messages in real time and sends the data to a server. For example, it uses APIs such as Microsoft Teams and Zoom to acquire voice data and text chat data during a meeting.

[1660] 2. Computer means for receiving and analyzing the text data and voice data transmitted from the device means

[1661] The server means (computer) receives the data sent from the terminals, classifies and stores the data for analysis. This means includes a high-performance database and data analysis software.

[1662] 3. A speech recognition engine that converts voice data into text data

[1663] The server uses a speech recognition engine such as Google Cloud Speech-to-Text API or IBM Watson to convert the audio data into text data, which is then used for further analysis.

[1664] 4. Generative AI modeling means for analyzing text and audio data using natural language processing and sentiment analysis to identify potential harassment.

[1665] The server analyzes the text and audio data using natural language processing engines such as Google Cloud Natural Language API and sentiment analysis engines such as IBM Watson Tone Analyzer. The analysis results are fed into a BERT-based generative AI model to identify potential harassment behavior.

[1666] 5. A means of sending warning messages to users when potential harassment is detected

[1667] If the server determines that an activity is likely to be harassing, it generates and sends a warning message to the user, which can appear as a popup on messaging platforms such as Teams or Slack.

[1668] 6. A method for automatically reporting to the helpline if the same user repeatedly ignores warnings

[1669] The server detects users who repeatedly engage in harassing behavior and counts the number of times. If the number of times exceeds a threshold, it automatically generates and sends a report to the harassment hotline. This data includes the content of the offending remarks, the date and time, and user information.

[1670] Specific examples

[1671] For example, during a team meeting, someone might say:

[1672] "You're late again. It's annoying because you're always like this. Can you do it properly?"

[1673] The device captures this speech in real time and sends it to the server as text and audio data.

[1674] The server analyzes this comment as text data, extracts keywords such as "You're late," "This is a problem," and "Can you do this properly?", and performs sentiment analysis.

[1675] Based on the analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1676] If the user ignores the warning and continues to make critical comments, the server will count them and eventually report them to a harassment hotline.

[1677] Prompt Sentence Examples

[1678] The accuracy of harassment detection can be improved by feeding prompt sentences like the following into the AI ​​model:

[1679] "Please determine whether the following conversation constitutes harassment. The statement: 'You're late again. This is annoying because it happens all the time. Can you do it properly?'"

[1680] In this way, the present invention can efficiently and effectively detect harassment behavior even in large-scale communication environments and take countermeasures in a timely manner.

[1681] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1682] Step 1: Data collection

[1683] The device captures users' meetings, emails, chats and other interactions in real time.

[1684] Specifically, the device uses APIs and hooks to retrieve speech and text messages from, for example, web conferencing tools or messaging applications.

[1685] Input: User utterances (voice data) and text messages (text data)

[1686] Data processing: Real-time capture

[1687] Output: Audio and text data

[1688] Step 2: Data reception and preprocessing

[1689] The server receives the text data and voice data sent from the terminal.

[1690] Specifically, the server separates the voice data from the text data for analysis, and converts the voice data into text using a speech recognition engine, such as the Google Cloud Speech-to-Text API.

[1691] Input: Voice and text data sent from the device

[1692] Data processing: Speech to text conversion using a speech recognition engine

[1693] Output: Converted text data

[1694] Step 3: Harassment detection

[1695] The server analyzes the converted text data using a natural language processing engine and a sentiment analysis engine.

[1696] Specifically, the system uses the Google Cloud Natural Language API to extract keywords and emotional tones from text data, and IBM Watson Tone Analyzer to extract emotional tones from audio data.These analysis results are then fed into an AI model (e.g., a BERT-based generative AI model) to determine the likelihood of harassment.

[1697] Input: Converted text data

[1698] Data Processing: Natural Language Processing and Sentiment Analysis

[1699] Output: Harassment possibility assessment result

[1700] Step 4: Sending a warning message

[1701] If the server determines that there is a high possibility of harassment, it generates and sends a warning message to the user.

[1702] Specifically, the server uses the Teams API to generate a warning message, such as "This message may be harassing. Please choose your words carefully," and send it to the user.

[1703] Input: Harassment possibility assessment result

[1704] Data processing: Generate warning messages

[1705] Output: Warning message sent to the user

[1706] Step 5: Report processing

[1707] When the same user repeatedly makes statements that meet the criteria for harassment, the server counts the number of times, and when the number exceeds a threshold, it generates report data and automatically sends it to a consultation desk.

[1708] Specifically, the server generates report data including the content of the offending comment, the date and time, and user information, and sends it to the harassment hotline. It also displays a message on the user's device informing them that a report has been made.

[1709] Input: Count of harassment incidents

[1710] Data processing: Generation of report data

[1711] Output: Report data sent to the consultation desk and message sent to the user

[1712] (Application example 1)

[1713] 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."

[1714] In recent years, harassment between customers and employees in brick-and-mortar stores has become a social problem. However, this type of problem is often difficult to detect and respond to immediately, which often increases the mental burden on employees. Therefore, an effective automatic detection and immediate response system is needed to address this issue.

[1715] 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.

[1716] In this invention, the server includes a terminal means for capturing interactions between customers and employees in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, and a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data. This makes it possible to detect harassment between customers and employees in a physical store in real time and issue an immediate warning or report.

[1717] "Customer" refers to a person who visits a physical store to purchase goods or services.

[1718] "Employee" refers to a staff member who performs duties such as selling products and providing services at a physical store.

[1719] "Interaction" refers to the conversational and behavioral interactions that take place between customers and employees.

[1720] "Real-time capture" refers to instantly recording and acquiring interactions that occur at that moment.

[1721] "Terminal means" refers to a hardware and software system for capturing audio and video data.

[1722] "Text data" refers to data converted from audio into text.

[1723] "Audio data" refers to data that records spoken words or sounds.

[1724] "Server means" refers to a computer system for receiving, analyzing, and storing data.

[1725] "Natural language processing" refers to computer technology for understanding and analyzing human language.

[1726] "Emotion analysis" refers to the technology of analyzing a speaker's emotions and tone from voice and text data.

[1727] An "AI model" refers to a system that uses algorithms based on machine learning and deep learning to analyze specific data and make inferences.

[1728] "Warning message" refers to a warning message that is displayed to a user when harassment behavior is detected.

[1729] A "harassment consultation center" refers to a specialized institution that accepts consultations and takes measures regarding harassment acts.

[1730] "Reporting data" refers to data that records details of harassment when it occurs and sends them to a consultation hotline.

[1731] The present invention is implemented using a system that captures and analyzes interactions between customers and employees in real time. The system includes the following hardware and software:

[1732] Hardware and software used

[1733] Hardware:

[1734] 1. Smartphone (with built-in camera and microphone)

[1735] 2. Cloud-based servers (e.g. AWS EC2)

[1736] software:

[1737] 1. Speech recognition engine (e.g. Google Cloud Speech-to-Text)

[1738] 2. Natural language processing engine (e.g., Google Cloud Natural Language)

[1739] 3. Sentiment analysis engine (e.g., IBM Watson Tone Analyzer)

[1740] Specific processing of the system

[1741] 1. Data Collection:

[1742] The smartphone serves as the terminal device and captures conversations between users (customers and employees) in real time. The smartphone's microphone and camera are used to obtain audio and video data.

[1743] 2. Data reception and preprocessing:

[1744] The server receives the audio and video data sent from the terminal. The audio data is converted into text data via a voice recognition engine. Furthermore, facial expression analysis is performed on the video data to detect changes in emotions.

[1745] 3. Harassment Detection:

[1746] The server uses a natural language processing engine to analyze text data. It also uses an emotion analysis engine to analyze voice and facial expression data to detect the speaker's emotional tone. This data is then combined and a generative AI model is used to determine whether the behavior is likely to be harassing. The determination includes specific keywords and emotional tone.

[1747] 4. Sending warning messages:

[1748] If harassment is detected, the server will pop up a warning message on the terminal in real time, such as "This word may contain elements of harassment. Please use polite language."

[1749] 5. Report Processing:

[1750] If the same customer repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment hotline. The report data includes the content of the remarks, date and time, user information, etc.

[1751] Examples

[1752] For example, in one scenario, a customer might say:

[1753] "Your responses are always slow, can't you do something about it?"

[1754] The smartphone (terminal means) captures this speech with a microphone and transmits it to the server as audio data.

[1755] The server uses a voice recognition engine to convert the voice data into text data and a natural language processing engine to analyze the possibility of harassment.

[1756] A sentiment analysis engine is also used to analyze emotional tone from text and audio data.

[1757] If a harassing act is detected, the server sends a warning message to the terminal means in real time.

[1758] Prompt Sentence Examples

[1759] Capture real-time conversations between customers and store associates and perform text and tone analysis on:

[1760] Text: "You're always slow to respond"

[1761] Emotional Tone: Anger

[1762] Warning message: "This language may be harassing. Please use polite language."

[1763] This will realize a system that can detect harassment between customers and employees in physical stores in real time and issue warnings or reports immediately.

[1764] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1765] Step 1: Data collection

[1766] The device (smartphone) captures the conversation between the customer and the employee using a camera and microphone, thereby obtaining audio and video data as input.

[1767] Step 2: Send data

[1768] The device transmits the captured audio and video data to a server in real time using an internet connection.

[1769] Step 3: Preprocessing the audio data

[1770] The server converts the received voice data into text data using a voice recognition engine (Google Cloud Speech-to-Text). The input is voice data and the output is text data.

[1771] Step 4: Preprocessing the video data

[1772] The server processes the received video data using analysis technology and performs facial expression analysis. The input is video data and the output is emotional tone data.

[1773] Step 5: Natural Language Processing

[1774] The server uses a natural language processing engine (Google Cloud Natural Language) to analyze the text data, extracting important keywords and context, and assessing the likelihood of harassment as a result of the analysis. The input is the text data, and the output is the results of keyword and context analysis.

[1775] Step 6: Sentiment Analysis

[1776] The server uses an emotion analysis engine (IBM Watson Tone Analyzer) to analyze the text data and emotional tone data and determine the emotional tone. The input is the text data and emotional tone data, and the output is the result of the emotional tone analysis.

[1777] Step 7: Determining Harassment

[1778] The server integrates the results of natural language processing and sentiment analysis and uses a generative AI model to determine the likelihood of harassment. The input is keywords, the results of contextual analysis, and the results of emotional tone analysis, and the output is the harassment judgment result.

[1779] Step 8: Sending a warning message

[1780] If the server detects harassment, it will pop up a warning message on the device in real time. The input is the result of the harassment judgment, and the output is the warning message.

[1781] Step 9: Generate and send notification data

[1782] If the same user repeatedly commits harassment, the server counts the acts and automatically generates and sends report data to the harassment consultation center. The input is the number of times harassment acts occurred, and the output is the generation and transmission of report data.

[1783] 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.

[1784] The present invention relates to a system for automating the detection and reporting of harassment behavior, and the following describes an embodiment incorporating an emotion engine for analyzing user emotions.

[1785] System configuration

[1786] 1. Terminal means

[1787] The terminal means is a device that captures user interactions such as web conferences, emails, and chats in real time. The terminal means uses APIs and hooks to obtain messages and voice data from each application and saves it in temporary storage. For example, the Teams API can be used to capture what is being said in a meeting.

[1788] 2. Server Means

[1789] The server means is a computer system that receives and analyzes the text data and voice data sent from the terminal means, classifies the text data and voice data, and converts the voice data into text using a voice recognition engine means.

[1790] It runs AI models that use natural language processing and sentiment analysis to determine potential harassment from text and audio data.

[1791] 3. Emotion Engine

[1792] The emotion engine has the function to analyze the user's emotions and extracts emotions from voice data and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[1793] Program processing

[1794] Data collection

[1795] The device captures users' web conferences, emails, chats, and other interactions in real time and sends the data to a server. For example, it can capture comments made during a Teams meeting in real time and collect them as audio and text data.

[1796] Data reception and preprocessing

[1797] The server receives the data and separates it for analysis: it separates audio data from text data and sends each to a separate processing pipeline.

[1798] The server converts the voice data into text using a speech recognition engine, such as Google Cloud Speech-to-Text.

[1799] Harassment detection and sentiment analysis

[1800] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones.

[1801] The server uses an emotion engine to analyze the user's emotions from the voice data, analyzing the tone, volume, speed, etc. of the voice to detect the user's emotional state.

[1802] The server uses an emotion engine to analyze user emotions from text data, extracting emotions based on the content, structure, and wording of the text.

[1803] Overall rating and warning message sent

[1804] The server evaluates these analysis results comprehensively, using an AI model to score the analysis results and determine the possibility of harassment if a threshold is exceeded.

[1805] If the server detects a potential harassment, it will generate a warning message and display it on the user's screen. For example, a message in chat might say, "This message contains harassing and emotionally inappropriate content. Please choose your words carefully."

[1806] Report processing

[1807] If the same user makes a remark that meets the criteria for harassment again, the server counts the number of times and proceeds to the next step if the number exceeds the set threshold.

[1808] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[1809] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[1810] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1811] As a concrete example, suppose the following is said during a meeting:

[1812] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[1813] The device captures these statements in real time and sends them to the server as text and audio data.

[1814] The server receives this speech and converts the speech data into text using a speech recognition engine.

[1815] The server analyzes the text data using a natural language processing engine and extracts keywords such as "late," "this is a problem," and "please do it properly."

[1816] The server uses an emotion engine to analyze the tone of the voice and the emotions in the text content, determining that the message is critical and contains strong emotional elements.

[1817] Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1818] If the user ignores the warning and continues to make critical comments, the server counts the number of times and issues a report if the number exceeds a threshold.

[1819] In this way, the system effectively detects harassment and emotional issues early and enables immediate response.

[1820] The processing flow will be explained below.

[1821] Step 1:

[1822] The device captures real-time interactions such as web conferences, emails, and chats, using APIs and hooks to retrieve messages and voice data from each application and store it in temporary storage.

[1823] Step 2:

[1824] The device encrypts the collected text and voice data and transmits it to the server through a secure channel. The data is transmitted using TLS or other protocols to ensure the security of the communication.

[1825] Step 3:

[1826] The server separates the received data for analysis, specifically separating it into audio data and text data, and sends each to a separate processing pipeline.

[1827] Step 4:

[1828] The server runs the audio data through a speech recognition engine to convert it into text, and then uses a service like Google Cloud Speech-to-Text to transcribe the audio.

[1829] Step 5:

[1830] The server analyzes both the converted and directly received text data using a natural language processing (NLP) engine to extract important keywords, phrases, and emotional tones from the text.

[1831] Step 6:

[1832] The server performs emotion analysis on the voice data, analyzing the tone, volume, and speed of the voice to detect emotions such as excitement, anger, and irritation.

[1833] Step 7:

[1834] The server also performs sentiment analysis on text data, extracting emotions based on the content, structure, and wording of the text.

[1835] Step 8:

[1836] The server evaluates the results of natural language processing and sentiment analysis, and uses an AI model to score these analysis results, determining the possibility of harassment if a threshold is exceeded.

[1837] Step 9:

[1838] If the server detects any potential harassment, it will generate a warning message and display it on the user's screen, containing the details of the warning and suggestions for improvement.

[1839] Step 10:

[1840] The user is given the opportunity to review the warning message and correct their behavior. Good user behavior is expected.

[1841] Step 11:

[1842] The server counts the number of times the same user makes a remark that meets the criteria for harassment. If the number of times exceeds the set threshold, the server proceeds to the next step.

[1843] Step 12:

[1844] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis. The generated report data is prepared for sending to the harassment hotline.

[1845] Step 13:

[1846] The server automatically sends the report data to the harassment hotline via email or a dedicated API.

[1847] Step 14:

[1848] The device will notify the user that a report has been made, for example by displaying a message saying, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1849] Specific examples

[1850] During a meeting, the following is said:

[1851] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[1852] Specific processing flow

[1853] Step 1:

[1854] The device captures this speech in real time, capturing text and audio data and storing it in temporary storage.

[1855] Step 2:

[1856] The device encrypts the captured data and sends it to the server over a secure channel.

[1857] Step 3:

[1858] The server separates the received data into text and audio data for analysis.

[1859] Step 4:

[1860] The server runs the voice data through a speech recognition engine and converts it into text.

[1861] Step 5:

[1862] The server analyzes the text data using a natural language processing engine and extracts keywords and phrases such as "It's late," "This is a problem," and "Please do it properly."

[1863] Step 6:

[1864] The server analyzes the tone, volume, and speed of the voice using an emotion engine to detect critical emotions.

[1865] Step 7:

[1866] The server analyzes the sentiment of the text from its content and vocabulary, extracting critical and emotional elements.

[1867] Step 8:

[1868] The server combines the analysis results and scores them using an AI model, determining that in this case the harassment threshold has been exceeded.

[1869] Step 9:

[1870] The server generates a warning message to the speaker, saying, "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[1871] Step 10:

[1872] The user is given the opportunity to review the warning message and correct their behavior.

[1873] Step 11:

[1874] If the user ignores the warning again and continues to make critical comments, the server counts the number of times and proceeds to the next step if the number exceeds a threshold.

[1875] Step 12:

[1876] The server generates report data, which includes the content of the comment in question, the date and time, user information, and the results of sentiment analysis.

[1877] Step 13:

[1878] The server automatically sends the generated report data to the harassment consultation center.

[1879] Step 14:

[1880] The device will notify the user that a report has been made, displaying the message, "Your behavior has been reported to the harassment hotline because you have ignored repeated warnings."

[1881] In this way, the system detects harassment and emotional issues early and enables immediate response.

[1882] Example 2

[1883] 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."

[1884] Conventional harassment detection and reporting systems have problems with insufficient user emotion analysis, making it difficult to detect harassment in real time and respond immediately. Furthermore, recording and reporting repeated harassment behaviors by users who ignore warnings must be done manually, making it difficult to efficiently prevent harassment.

[1885] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1886] In this invention, the server includes means for receiving and classifying text data and voice data transmitted from the terminal means, means for utilizing a voice recognition engine for converting voice data into text, and means for executing an AI model for determining the possibility of harassment from the text and voice data using natural language processing and sentiment analysis, thereby enabling real-time detection of harassment behavior and user sentiment analysis.

[1887] The "terminal means" is a device for capturing exchanges such as web conferences, e-mails, and chats in real time and transmitting them to a server.

[1888] The "server means" is a computer system that receives data sent from the terminal means and classifies and analyzes the data.

[1889] A "voice recognition engine means" is a system that executes voice recognition techniques to convert voice data into text data.

[1890] "Natural language processing" is a technology that analyzes text data and extracts important keywords and phrases.

[1891] "Emotion analysis" is a technology that analyzes a user's emotions from voice data and text data and detects their emotional state.

[1892] The "AI model" is an artificial intelligence technology that determines the possibility of harassment based on the results of natural language processing and sentiment analysis.

[1893] A "warning message" is a notification sent to a user when potential harassment is detected.

[1894] The "reporting mechanism" is a system that records repeated acts of harassment and reports them to a harassment hotline if a threshold is exceeded.

[1895] The present invention is a system for automating the detection and reporting of harassment behavior, and incorporates an emotion engine that analyzes user emotions. Specific embodiments of the present invention will be described below.

[1896] 1. System Configuration

[1897] Terminal means

[1898] The terminal means is a device that captures interactions such as web conferences, emails, and chats in real time. It uses APIs and hooks from the applications used by the user to obtain messages and voice data in real time. As a specific example, the Teams API is used to capture what is said during a meeting.

[1899] Server Means

[1900] The server means is a computer system that receives and analyzes text data and voice data sent from the terminal means. The server means uses a voice recognition engine means to convert voice data into text data. In this case, the Google Cloud Speech-to-Text service can be used.

[1901] The server performs natural language processing and emotion analysis on the text data and voice data, thereby executing an AI model that determines the possibility of harassment. The emotion engine analyzes emotions from the voice data and text data to detect the user's emotional state.

[1902] Emotion Engine

[1903] The emotion engine has the function of analyzing the user's emotions from voice and text data. The results of this emotion analysis are used to determine whether the behavior is harassment.

[1904] 2. Program Processing Overview

[1905] Data collection

[1906] When a user is engaged in a web conference, email, chat, or other exchange, the device captures this in real time and sends it to the server. For example, remarks made during a Teams meeting can be captured in real time and collected as voice and text data.

[1907] Data reception and preprocessing

[1908] The server receives the data and separates it into voice and text, which is then converted into text using the Google Cloud Speech-to-Text speech recognition engine.

[1909] Harassment detection and sentiment analysis

[1910] The server analyzes the text data using a natural language processing engine to extract important keywords, phrases, and emotional tones. The server also analyzes emotions from the audio data, using an emotion engine to analyze voice tone, volume, speed, etc., thereby detecting the user's emotional state.

[1911] 3. Usage example

[1912] As a concrete example, suppose the following is said during a meeting:

[1913] "Are you late again? It's really annoying if you're like this every time. Please do it properly."

[1914] System Operation

[1915] 1. The device captures this speech in real time and sends it to the server as text and audio data.

[1916] 2. The server receives this speech and converts the speech data into text using a speech recognition engine.

[1917] 3. The server analyzes the text data using a natural language processing engine and extracts keywords such as "late," "this is a problem," and "please do it properly."

[1918] 4. The emotion engine analyzes the tone of the voice and the emotion of the text content and determines that the message is critical and emotionally charged.

[1919] 5. Based on these analysis results, the server determines that there is a high possibility of harassment and sends a warning message to the person who made the comment.

[1920] 6. If the user ignores the warning and continues to make critical comments, the server will count the number of times and issue a report if the number exceeds a threshold.

[1921] 4. Example prompts for generative AI models

[1922] You can further improve harassment detection by using prompts like the following to feed your generative AI model:

[1923] "Please analyze the following conversations to detect any harassing behavior and the emotional tone. Please also provide your analysis results based on the specific content of the statements."

[1924] Conversation: "You're late again. This is really annoying. Please be more careful next time."

[1925] Expected output: "This comment contains critical and emotional overtones and may be considered harassment."

[1926] By using such prompts, it becomes possible to perform advanced analysis of both the content and emotions of the statements.

[1927] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1928] Step 1:

[1929] Data collection

[1930] Input: User web conferencing, email, and chat interactions

[1931] Specific operation: The device captures the user's interactions in real time, such as web conferences, emails, and chats. For example, it uses the Teams API to capture what is said during a meeting.

[1932] Output: Audio and text data

[1933] Step 2:

[1934] Data transmission

[1935] Input: Audio and text data

[1936] Specific operation: The device sends the captured audio and text data to the server.

[1937] Output: Audio and text data sent to the server

[1938] Step 3:

[1939] Data Receipt and Classification

[1940] Input: Voice and text data sent from the device

[1941] Specific operation: The server receives data sent from the terminal and classifies it into voice data and text data.

[1942] Output: Classified audio and text data

[1943] Step 4:

[1944] Converting audio data to text

[1945] Input: Classified audio data

[1946] Specific operation: The server uses a speech recognition engine such as Google Cloud Speech-to-Text to convert the voice data into text data.

[1947] Output: Converted text data

[1948] Step 5:

[1949] Natural Language Processing and Keyword Extraction

[1950] Input: Text data

[1951] What it does: The server uses a natural language processing engine to analyze the text data and extract important keywords and phrases.

[1952] Output: Extracted keywords and phrases

[1953] Step 6:

[1954] Emotion analysis (audio data)

[1955] Input: Audio data

[1956] Specific operation: The server uses the emotion engine to analyze the user's emotions from the voice data, specifically analyzing the tone, volume, and speed of the voice.

[1957] Output: Sentiment analysis results of audio data

[1958] Step 7:

[1959] Sentiment analysis (text data)

[1960] Input: Text data and extracted keywords and phrases

[1961] Specific operation: The server uses an emotion engine to analyze the user's emotions from text data, analyzing the content, structure, and wording of the text.

[1962] Output: Sentiment analysis results of text data

[1963] Step 8:

[1964] comprehensive evaluation

[1965] Input: Keywords, phrases, sentiment analysis results of audio data, sentiment analysis results of text data

[1966] Specific operation: The server comprehensively evaluates these analysis results, scores them using an AI model, and determines the possibility of harassment if the threshold is exceeded.

[1967] Output: Evaluation result of potential harassment

[1968] Step 9:

[1969] Generate and send warning messages

[1970] Input: Possible harassment assessment results

[1971] Specific behavior: If the server determines that there is a possibility of harassment, it will generate a warning message and display it as a pop-up on the user's screen.

[1972] Output: A warning message that is displayed to the user.

[1973] Step 10:

[1974] Report processing

[1975] Input: Warning message sending history and user behavior history

[1976] Specific operation: When the same user makes a remark that meets the harassment criteria, the server counts the number of times and generates a report if the number exceeds a set threshold. The generated report data is automatically sent to the harassment consultation desk and a notification is sent to the device that a report has been made.

[1977] Output: Report data to the harassment hotline and a report to the user

[1978] (Application example 2)

[1979] 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."

[1980] In modern corporate environments, harassment behaviors occurring via communication channels such as web conferencing, email, and chat are on the rise. Early detection of such harassment and minimizing damage are essential, but conventional methods often struggle to detect and respond appropriately. Furthermore, if a harassment detection system does not perform emotion analysis, it may miss subtle nuances and make incorrect judgments. Furthermore, the lack of an effective way to generate warning messages for users makes the system ineffective at deterring harassment.

[1981] 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.

[1982] In this invention, the server is a system for automating the detection and reporting of harassment, and includes a terminal means for capturing exchanges such as web conferences, emails, and chats in real time, a means for receiving and analyzing text data and voice data sent from the terminal means, a means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a means for sending a warning message to a user when the possibility of harassment is detected, and a means for generating a prompt sentence and inputting it into the generative AI model to optimize the warning message to the user. This makes it possible to detect harassment behavior early and automatically issue appropriate warnings and reports.

[1983] "Harassment" refers to rude, offensive, or critical words or actions directed at others in the workplace or other environment.

[1984] "Detection" is the process by which a system identifies and recognizes targeted behavior or data.

[1985] An "automated reporting system" is a system that detects acts of harassment and automatically notifies the relevant parties of that information.

[1986] A "web conference" is a real-time audio and video conference conducted over the Internet.

[1987] "Email" is a means of sending documents containing digital information over the Internet to specific recipients.

[1988] "Chat" is the exchange of text messages in real time.

[1989] "Capture" is the process of collecting, recording, or copying digital data.

[1990] "Terminal means" refers to a device or application for capturing data and transmitting it to a server.

[1991] "Text data" is digital information expressed in the form of a string of characters.

[1992] "Audio data" refers to data in which audio is recorded in digital format.

[1993] "Server means" is a computer system for receiving, analyzing and otherwise processing data.

[1994] "Natural language processing" is a technology that allows computers to understand, interpret, and generate human language.

[1995] "Sentiment analysis" is a technology that extracts and analyzes emotions from text and audio data.

[1996] An "AI model" is a collection of algorithms designed to perform a specific task using machine learning techniques.

[1997] A "warning message" is a notification message that alerts the user.

[1998] A "prompt" is an instruction that is input into a generative AI model to obtain a result.

[1999] A "generative AI model" is a machine learning model that automatically generates text and images based on given instructions.

[2000] "Optimization" is the process of maximizing performance to achieve a goal.

[2001] This invention relates to a system for automating the detection and reporting of harassment behavior. The system includes a terminal means for capturing interactions such as web conferences, emails, and chats in real time, a server means for receiving and analyzing text and voice data sent from the terminal means, a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the possibility of harassment from the text and voice data, a server means for sending a warning message to a user when possible harassment is detected, a server means for reporting to a harassment hotline if the same user repeatedly ignores warnings, and a means for generating prompt sentences and inputting them into the generation AI model to optimize the warning message for the user.

[2002] Hardware and Software Configuration

[2003] 1. Terminal means:

[2004] It captures web conferences, emails, chats, and other interactions in real time on the user's device (PC, smartphone, etc.). This is done using the APIs and hooks of each application. For example, it captures what is being said using the API of a web conference tool.

[2005] 2. Server means:

[2006] A computer system that receives the transmitted data and classifies it for analysis. It converts the voice data into text using a speech recognition engine such as Google Cloud Speech-to-Text.

[2007] Use natural language processing engines (such as BERT or GPT-3) to analyze text data and extract important keywords and phrases.

[2008] 3. Sentiment Analysis Engine:

[2009] The emotion analysis engine extracts emotions from audio and text data and uses common emotion analysis algorithms (e.g., the Oxyford engine).

[2010] 4. AI model:

[2011] To determine the likelihood of harassment, a machine learning model is used, which inputs the results of sentiment analysis of text and voice and scores the likelihood of harassment.

[2012] Specific examples of processing

[2013] 1. Data capture:

[2014] For example, suppose the following is said during a meeting:

[2015] "Data preparation is too slow. I want you to work more efficiently."

[2016] The device captures these statements in real time and collects them as audio and text data.

[2017] 2. Data reception and preprocessing:

[2018] The server receives the transmitted data, separates it into audio and text data, and sends them to separate processing pipelines.

[2019] 3. Audio to text conversion:

[2020] The server uses Google Cloud Speech-to-Text to convert the voice data into text.

[2021] 4. Natural Language Processing and Sentiment Analysis:

[2022] The server analyzes the text data using a natural language processing engine and extracts keywords such as "too slow" and "efficiently."

[2023] The emotion engine analyzes voice tone, volume, speed, etc. to detect critical and emotional elements.

[2024] 5. Generate warning messages:

[2025] The AI ​​model evaluates these analysis results comprehensively, and if it determines that there is a high possibility of harassment, it generates a warning message for the user, such as the following:

[2026] "This message contains elements of harassment and emotional inappropriateness. Please choose your words carefully."

[2027] 6. Prompt generation and input to the AI ​​model:

[2028] The server generates prompts to optimize warning messages and inputs them into the generative AI model. An example of a prompt is as follows:

[2029] "Could this statement constitute workplace harassment (criticism, disrespect, or aggressive behavior)? Please help us determine."

[2030] This allows the system to detect harassing behavior and implement appropriate countermeasures quickly and effectively.

[2031] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[2032] Step 1:

[2033] Data Capture

[2034] The terminal captures in real time the user's interactions with web conferences, emails, chats, etc. The terminal uses APIs and hooks to obtain messages and voice data from each application and stores them in temporary storage.

[2035] Input: Real-time data from web conferencing, email, chat, etc.

[2036] Output: Captured text and audio data

[2037] Step 2:

[2038] Data transmission

[2039] The terminal transmits the captured text data and voice data to the server.

[2040] Input: Captured text and audio data

[2041] Output: Data sent from the device to the server

[2042] Step 3:

[2043] Data reception and preprocessing

[2044] The server receives the data sent from the terminal and classifies it into voice data and text data for analysis.

[2045] Input: Text and voice data sent from the device

[2046] Output: Separated audio and text data

[2047] Step 4:

[2048] Converting audio data to text

[2049] The server converts the voice data into text using a speech recognition engine (e.g., Google Cloud Speech-to-Text).

[2050] Input: Audio data

[2051] Output: Converted text data

[2052] Step 5:

[2053] Natural Language Processing

[2054] The server analyzes the converted text data using a natural language processing engine (e.g., BERT or GPT-3) to extract important keywords and phrases.

[2055] Input: Text data

[2056] Output: Analysis results (keywords and phrases)

[2057] Step 6:

[2058] Emotion analysis

[2059] The server uses an emotion engine to analyze the emotion of the text and voice data, analyzing voice tone, volume, and rate to detect emotional states.

[2060] Input: Text and audio data

[2061] Output: Emotion analysis results

[2062] Step 7:

[2063] Determining the possibility of harassment

[2064] The server combines the results of natural language processing and sentiment analysis to use an AI model to score the likelihood of harassment.

[2065] Input: Natural language processing results and sentiment analysis results

[2066] Output: Harassment likelihood score

[2067] Step 8:

[2068] Generate a warning message

[2069] The server generates and sends a warning message to the user when potential harassment is detected. The prompt sentence is used as input to a generative AI model to generate an optimized warning message.

[2070] Input: Harassment likelihood score

[2071] Output: Warning message

[2072] Step 9:

[2073] Reporting and User Notification

[2074] If the user continues to ignore the warning, the server counts the number of times, and if the number exceeds a threshold, it generates and automatically sends report data to the harassment hotline. It also notifies the user that a report has been made.

[2075] Input: Warning ignore count

[2076] Output: Report data and user notification

[2077] 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.

[2078] 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.

[2079] 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.

[2080] 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.

[2081] 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.

[2082] 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.

[2083] 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).

[2084] 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.

[2085] 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."

[2086] 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.

[2087] 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).

[2088] 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.

[2089] 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.

[2090] 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.

[2091] 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.

[2092] 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.

[2093] 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.

[2094] 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.

[2095] 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.

[2096] 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.

[2097] 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.

[2098] The following is further disclosed regarding the above embodiment.

[2099] (Claim 1)

[2100] A system for automating the detection and reporting of harassment, comprising:

[2101] A terminal means for capturing interactions such as web conferences, emails, and chats in real time;

[2102] a server means for receiving and analyzing the text data and voice data transmitted from the terminal means;

[2103] a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the likelihood of harassment from text and voice data;

[2104] server means for sending a warning message to a user when potential harassment is detected;

[2105] a server means for reporting to a harassment hotline when the same user repeatedly ignores warnings;

[2106] A system including:

[2107] (Claim 2)

[2108] 10. The system of claim 1, further comprising a voice recognition engine means for converting voice data into text data in detecting harassment behavior.

[2109] (Claim 3)

[2110] 2. The system according to claim 1, further comprising means for generating and automatically transmitting report data to a harassment hotline.

[2111] "Example 1"

[2112] (Claim 1)

[2113] a device means for capturing user meetings, emails, correspondence, and other interactions in real time;

[2114] computer means for receiving and analyzing the text data and voice data transmitted from the device means;

[2115] a speech recognition engine means for converting speech data into text data;

[2116] A generative AI model means for analyzing text data and voice data using natural language processing and sentiment analysis to determine the possibility of harassment;

[2117] means for sending a warning message to a user when potential harassment is detected;

[2118] a means for automatically reporting to a helpline if the same user repeatedly ignores warnings;

[2119] A system including:

[2120] (Claim 2)

[2121] 10. The system of claim 1, further comprising a speech recognition engine means for converting speech data into text data.

[2122] (Claim 3)

[2123] 10. The system according to claim 1, further comprising means for generating and automatically transmitting report data to a consultation service.

[2124] "Application Example 1"

[2125] (Claim 1)

[2126] A system for automating the detection and reporting of harassment, comprising:

[2127] a terminal means for capturing customer and employee interactions in real time;

[2128] a server means for receiving and analyzing the text data and voice data transmitted from the terminal means;

[2129] a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the likelihood of harassment from text and voice data;

[2130] server means for sending a warning message to a user when potential harassment is detected;

[2131] a server means for reporting to a harassment hotline when the same user repeatedly ignores warnings;

[2132] A system including:

[2133] (Claim 2)

[2134] 10. The system of claim 1, further comprising a speech recognition engine means for converting speech data into text data.

[2135] (Claim 3)

[2136] 10. The system of claim 1, further comprising means for generating and automatically transmitting notification data.

[2137] "Example 2: Combining Emotion Engines"

[2138] (Claim 1)

[2139] A terminal means for capturing interactions such as web conferences, emails, and chats in real time;

[2140] a server means for receiving and classifying text data and voice data transmitted from the terminal means;

[2141] a server means utilizing a speech recognition engine means for converting speech data into text;

[2142] a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the likelihood of harassment from text and voice data;

[2143] emotion engine means for analyzing the emotion of a user from voice data and text data;

[2144] server means for sending a warning message to a user when potential harassment is detected;

[2145] a server means for reporting to a harassment hotline when the same user repeatedly ignores warnings;

[2146] A system including:

[2147] (Claim 2)

[2148] 10. The system of claim 1, wherein emotion engine means is utilized to extract emotions from the audio data and text data.

[2149] (Claim 3)

[2150] 2. The system according to claim 1, further comprising means for generating and automatically transmitting report data to a harassment hotline.

[2151] "Application example 2 when combining emotion engines"

[2152] (Claim 1)

[2153] A system for automating the detection and reporting of harassment, comprising:

[2154] A terminal means for capturing interactions such as web conferences, emails, and chats in real time;

[2155] a server means for receiving and analyzing the text data and voice data transmitted from the terminal means;

[2156] a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the likelihood of harassment from text and voice data;

[2157] server means for sending a warning message to a user when potential harassment is detected;

[2158] a server means for reporting to a harassment hotline when the same user repeatedly ignores warnings;

[2159] A means for generating prompt sentences and inputting the prompt sentences into a generative AI model to optimize warning messages for the user;

[2160] A system including:

[2161] (Claim 2)

[2162] 10. The system of claim 1, further comprising a voice recognition engine means for converting voice data into text data in detecting harassment behavior.

[2163] (Claim 3)

[2164] 2. The system according to claim 1, further comprising means for generating and automatically transmitting report data to a harassment hotline. [Explanation of symbols]

[2165] 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 system for automating the detection and reporting of harassment, comprising: A terminal means for capturing interactions such as web conferences, emails, and chats in real time; a server means for receiving and analyzing the text data and voice data transmitted from the terminal means; a server means for executing an AI model that uses natural language processing and sentiment analysis to determine the likelihood of harassment from text and voice data; server means for sending a warning message to a user when potential harassment is detected; a server means for reporting to a harassment hotline when the same user repeatedly ignores warnings; A system including:

2. 2. The system of claim 1, further comprising a voice recognition engine means for converting voice data into text data in detecting harassment behavior.

3. 2. The system according to claim 1, further comprising means for generating and automatically transmitting report data to a harassment hotline.

Images