System

Abstract

A system is provided.SOLUTION: An artificial intelligence-based mentoring system for supporting mental health of employees, comprising: means for collecting public information and documents in a company and constructing a dataset for initial learning; terminal means for employees to input basic information; means for causing a generative model to perform initial learning based on the basic information; means for collecting data related to daily conversation and work of employees; means for analyzing the collected data and evaluating emotions and stress levels; A system comprising: means for generating an appropriate alert or advice; means for notifying an employee of the generated alert or advice; means for continuously analyzing data and evaluating effectiveness of the system; and means for retraining a model based on the evaluation results.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] In today's workplace, employees often find it difficult to confide their stress and worries to their superiors, colleagues, or family members. This creates a need for appropriate support for mental health care. However, traditional mental health support typically requires employees to consult directly with a specialist, which poses a high psychological hurdle. In particular, there is a lack of support for overcoming the stress of adapting to changes in the environment, such as new hires, transfers, and mergers. To solve this problem, it is necessary to provide an environment where employees can easily seek advice and to have a system that can detect and address mental health issues early. [Means for solving the problem]

[0005] To address the above-mentioned challenges, the present invention provides an artificial intelligence-based mentoring system for supporting employee mental health. The system includes: means for collecting publicly available information and documents from within a company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training a generative model based on the basic information; means for collecting data related to employees' daily conversations and work; means for analyzing the collected data and assessing their emotions and stress levels; means for generating appropriate alerts and advice based on the assessed stress levels; means for notifying employees of the generated alerts and advice; means for continuously analyzing data and assessing the effectiveness of the system; and means for retraining the model based on the assessment results. This provides an environment in which employees can easily seek mental health advice, enabling early detection of stress and prompt implementation of appropriate measures.

[0006] "Internal public information" refers to documents and data that are shared internally by a company but are not confidential.

[0007] "Documentation" means documentation of business processes, policies, and other administrative materials within an enterprise.

[0008] An "initial training dataset" is a collection of publicly available information and data collected from employees that the system uses to initially train itself.

[0009] A "generative model" is an artificial intelligence trained on collected data, and serves as the basis for mentoring and stress analysis.

[0010] "Terminal" means a device used by an employee to enter information or receive feedback from the system.

[0011] "Daily conversation data" refers to data based on the conversations employees have every day and the communications that occur during work.

[0012] "Analysis" refers to the process of processing collected data and extracting specific features and patterns.

[0013] "Emotion and stress levels" are indicators of an employee's psychological state as inferred from their conversations and behavior.

[0014] An "alert" is a message that notifies the user of a problem or high-stress situation detected by the system.

[0015] "Advice" refers to specific guidelines for action or improvements that the system provides to the user.

[0016] "Means of notification" refers to the function for sending generated alerts and advice to employees' terminals and displaying them.

[0017] "Means for continuous data analysis" means means for periodically analyzing collected data to evaluate the performance and effectiveness of the system.

[0018] "Retraining" is the process of updating a generative model with additional data to improve its accuracy or effectiveness. [Brief explanation of the drawings]

[0019] [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

[0020] 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.

[0021] First, the terms used in the following description will be explained.

[0022] 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).

[0023] 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.

[0024] 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.

[0025] 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.

[0026] 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."

[0027] [First embodiment]

[0028] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0029] 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.

[0030] 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).

[0031] 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.

[0032] 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.

[0033] 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.

[0034] 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.

[0035] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0036] 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.

[0037] 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.

[0038] 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.

[0039] 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."

[0040] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system consists of three main components: a server, a terminal, and a user.

[0041] System Overview

[0042] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[0043] Terminal: A device used by an employee to input information or receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs.

[0044] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through terminals.

[0045] System Operation

[0046] Initial setup and learning phase

[0047] 1. Server: First, collect publicly available information and documents from within the company. This data is used to build an initial learning dataset, including company policies, business processes, and manuals.

[0048] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[0049] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[0050] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[0051] Daily monitoring and stress detection

[0052] 1. Devices: Collect data on employees' daily conversations and work. Methods include periodic surveys and chat-style conversation logs.

[0053] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[0054] 3. Server: Analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[0055] Mentoring and support

[0056] 1. Server: When high stress is detected, it automatically generates an alert and generates appropriate advice, such as "take a short break."

[0057] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[0058] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0059] Continuous evaluation and improvement

[0060] 1. Server: Periodically analyzes data collected from all users to evaluate the effectiveness of the system, for example by graphing fluctuations in stress levels across the entire workforce.

[0061] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[0062] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback.

[0063] Specific examples

[0064] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[0065] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert for Person A saying, "Take a short break," and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0066] This system is designed to provide an environment where employees can easily receive mental health consultations, detect stress early, and promptly implement appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[0067] The processing flow will be explained below.

[0068] Initial setup and learning phase

[0069] Step 1:

[0070] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[0071] Step 2:

[0072] The server creates an initial learning dataset based on the collected data, which prepares the data to be used for initial learning.

[0073] Step 3:

[0074] The terminal provides employees with a screen for entering basic information, and the input form is displayed on a smartphone or PC application.

[0075] Step 4:

[0076] The user enters basic information, for example, "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and sends it from the terminal to the server.

[0077] Step 5:

[0078] The server initially trains the generative model based on the basic information received, thereby preparing a model that corresponds to the company's specific context and needs.

[0079] Daily monitoring and stress detection

[0080] Step 1:

[0081] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style conversation logs.

[0082] Step 2:

[0083] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[0084] Step 3:

[0085] The server analyzes the data sent from the device and uses natural language processing to evaluate emotions and stress levels.

[0086] Step 4:

[0087] The server calculates the stress level based on the analysis results, and generates an alert if signs of high stress are detected.

[0088] Mentoring and support

[0089] Step 1:

[0090] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[0091] Step 2:

[0092] Notify the user of any alerts or advice generated by the device, such as a pop-up notification on the smartphone with the message "Try taking some deep breaths to relax."

[0093] Step 3:

[0094] The user receives a notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0095] Continuous evaluation and improvement

[0096] Step 1:

[0097] The server periodically analyzes the data collected from all users, for example, to graph fluctuations in stress levels across the entire workforce.

[0098] Step 2:

[0099] The device will notify the user of new advice and improvements, for example suggesting new training programs or relaxation techniques.

[0100] Step 3:

[0101] The server retrains the generative model based on the data analysis results, allowing it to provide more effective feedback.

[0102] Through these specific processing steps, a system will be created that continuously supports the mental health of employees and contributes to improving the performance of the entire organization.

[0103] Example 1

[0104] 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."

[0105] Maintaining and improving employee mental health is an important issue in today's workplace. However, many companies lack systems for constantly monitoring employee mental health and providing appropriate support. It is also difficult to detect stress early or provide advice tailored to individual employees, making it difficult to take measures before employee problems become serious.

[0106] 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.

[0107] In this invention, the server includes: means for collecting public information and documents from within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data on employees' daily conversations and work; means for analyzing the collected data and evaluating emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting feedback from employees and using the data for re-learning; means for periodically analyzing data collected from all employees and evaluating the effectiveness of the system; and means for re-learning the generative model based on the evaluation results and generating more effective feedback. This makes it possible to monitor employees' mental health in real time, detect stress early, and provide appropriate support and advice.

[0108] "Public information" refers to information obtained from inside or outside the company, such as company policies, business processes, and manuals.

[0109] "Documents" refers to text data related to specific content, such as company records, manuals, and guides.

[0110] An "initial training dataset" is a dataset constructed from publicly available company information and documents, and refers to a collection of data used to train a generative model.

[0111] "Terminal" refers to the device used by employees to enter basic information and inquiries, including smartphones, tablets, and PCs.

[0112] A "generative model" is an algorithm that learns from data obtained from employees and generates appropriate advice and alerts.

[0113] "Daily conversation and work-related data" refers to text data and voice data generated during employees' daily discussions and work activities.

[0114] "Emotion and stress levels" refers to the psychological state of employees, assessed by analyzing data on their daily conversations and work.

[0115] An "alert" refers to a notification that is generated when an employee's stress level exceeds a certain threshold.

[0116] "Advice" refers to specific suggestions for improving employees' mental health and reducing stress.

[0117] "Feedback" refers to information provided to employees about the advice and results of their actions, and refers to data that is re-entered into the system.

[0118] "Retraining" refers to the process of retraining a generative model using feedback or newly collected data to improve its accuracy.

[0119] "Effectiveness" refers to an indicator that shows how effective the system is in improving employees' mental health.

[0120] This invention relates to an AI-based mentoring system for supporting employee mental health. The system collects publicly available information and documents from within a company, constructs an initial training dataset, and trains a generative model based on the dataset to analyze employees' stress levels and provide appropriate alerts and advice. Feedback from employees is used for retraining, allowing the model to be continuously improved and the effectiveness of the system to be evaluated.

[0121] Server: Initial setup and learning phase

[0122] The server first collects publicly available information and documents from within the company. Specifically, it uses Python scripts to retrieve information such as policies, business processes, and manuals from the company's internal database and file storage, and builds an initial training dataset based on this data. Next, it receives basic employee information and initially trains a generative model specific to the company's context. This process uses Python and TensorFlow to input data into an NLP model (such as the BERT model) to create a company-specific language model.

[0123] Terminal: Data collection and feedback

[0124] The terminal is a device used by employees to input basic information and daily inquiries. To this end, it is designed as a web application using front-end frameworks such as React or Vue.js. The terminal provides an interface for collecting data on employees' daily conversations and work. This includes periodic surveys and interactions with chatbots, which are developed using Dialogflow or the Microsoft Bot Framework. Employee input is provided, for example, through a smartphone application.

[0125] User: System usage

[0126] Employees, who are users, input basic information and everyday concerns through their terminals. For example, an employee may input basic information such as "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and then, during work hours, input concerns such as "Work has been tough and I'm feeling tired lately." This data is sent to the server and used for analysis.

[0127] Server: Data analysis and advice generation

[0128] The server analyzes the data sent from the device and evaluates the employee's emotions and stress levels. Specifically, it performs natural language processing using Python's NLTK library and spaCy to extract keywords from the text data and score the stress level. If an increase in stress level is detected, a generative AI model (e.g., GPT-3) is used to generate specific advice, such as "Take a short break."

[0129] Device: Notifications and feedback collection

[0130] The generated alerts and advice are notified to the user by the device. For example, a message such as "Take a short break" is displayed using the smartphone's push notification function. The user receives the notification, acts according to the instructions, and then enters the results back into the device. For example, feedback is given such as "Took a 5-minute break."

[0131] Server: Continuous evaluation and model retraining

[0132] The server periodically analyzes data collected from all users and performs data analysis using Python's pandas library. Fluctuations in stress levels across employees are graphed using Matplotlib and Seaborn. The generative model is retrained based on the analysis results to provide more effective feedback. Accuracy is improved by retraining the NLP model using feedback data from employees.

[0133] Examples of specific examples and prompts

[0134] For example, when a new employee logs in to the system for the first time and enters basic information (name, department, position), they can then enter "I'm tired because I've already had a lot of work lately" into their device while they're working. The server analyzes this data, detects an increase in stress level, and generates advice to "take a short break" and notifies the device. If the new employee accepts the advice and takes a break, the feedback data is sent to the server and used for relearning.

[0135] Prompt Sentence Examples

[0136] "Tell me about your recent work. Example: Work has been hard and I'm feeling tired lately."

[0137] What's an effective way to relax?

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

[0139] Step 1:

[0140] Server: Collects public information and documents from within the company and builds an initial learning dataset. Specifically, it runs Python scripts to retrieve policies, business processes, manuals, etc. from the company's internal database and file storage, and then formats this into a dataset.

[0141] Input: Public information and documents (policies, business processes, manuals)

[0142] Output: Initial training dataset

[0143] Step 2:

[0144] Devices: Provide an interface for employees to enter basic information. Specifically, create a web form using React or Vue.js and make it available on devices such as smartphones, tablets, and PCs.

[0145] User: For example, a user (employee) enters "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[0146] Input: Employee basic information (name, department, job description)

[0147] Output: Send employee basic information to the server

[0148] Step 3:

[0149] Server: Receives basic employee information and performs initial training on the generative model. Specifically, it uses Python and TensorFlow to input data into an NLP model (such as the BERT model) and builds a company-specific language model.

[0150] Input: Basic employee information (name, department, job description), initial learning dataset

[0151] Output: Company-specific generative model

[0152] Step 4:

[0153] Terminal: Provides an interface for collecting data about employees' daily conversations and work, for example, through periodic surveys or chatbots. Chatbots are developed using Dialogflow or the Microsoft Bot Framework.

[0154] User: During work, input a question such as "Work has been tough and I'm tired lately."

[0155] Input: Daily conversation and business data (text messages)

[0156] Output: Sending daily conversation data to the server

[0157] Step 5:

[0158] Server: Analyzes the data sent from the device and evaluates emotions and stress levels. Specifically, it uses Python's NLTK library and spaCy to analyze text data, extract keywords, and score stress levels.

[0159] Input: Daily conversation and business data

[0160] Output: Emotion and stress level assessment results

[0161] Step 6:

[0162] Server: Based on the evaluation results, it generates appropriate alerts and advice. It uses a generative AI model (e.g., GPT-3) to create specific advice, such as "Take a short break."

[0163] Input: Emotion and stress level assessment results

[0164] Output: Alerts and advice

[0165] Step 7:

[0166] Devices: Notify employees of generated alerts and advice, using push notifications on their smartphones to display messages such as "Take a short break."

[0167] Input: Alerts and Advice

[0168] Output: Employee notification

[0169] Step 8:

[0170] User: Receives alerts and advice and acts accordingly, for example by entering feedback into the device such as "I took a 5-minute break."

[0171] Input: Employee feedback (e.g., took a break)

[0172] Output: Sending feedback to the server

[0173] Step 9:

[0174] Server: Receives employee feedback and uses that data for retraining. Specifically, the NLP model is retrained using the feedback data to improve accuracy. The retraining process is performed using Python and TensorFlow.

[0175] Input: Employee feedback

[0176] Output: Retrained generative model

[0177] Step 10:

[0178] Server: Regularly analyzes data collected from all employees to evaluate the effectiveness of the system. Python's pandas library is used to analyze the data, and Matplotlib and Seaborn are used to graph, for example, fluctuations in stress levels across all employees. Based on the evaluation results, new advice and improvement measures are implemented.

[0179] Input: Feedback data collected from all employees

[0180] Output: System effectiveness evaluation results, new advice and improvements

[0181] (Application example 1)

[0182] 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."

[0183] Maintaining and improving employee mental health is an important issue for companies. However, currently, there are limitations to early detection of employee stress and mental problems and appropriate response. In particular, in factory environments, employees often bear significant physical and mental burdens, requiring rapid and effective responses. Conventional methods rely heavily on self-reporting, making it difficult to grasp mental health conditions in real time or provide appropriate advice. For this reason, a system is needed that provides an environment where employees can easily seek mental health advice, detects stress and mental burden early, and enables appropriate measures to be taken promptly.

[0184] 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.

[0185] In this invention, the server includes means for converting what employees say into text data using a voice recognition system, means for evaluating emotions and stress levels based on the converted text data, means for using a generative AI model to generate appropriate alerts and advice based on the evaluation, means for notifying employees of the generated alerts and advice by voice or text, means for continuously collecting employee feedback and evaluating the effectiveness of the system, and means for retraining the model based on the evaluation results. This makes it possible to monitor employees' mental health in real time and provide appropriate advice and support, thereby reducing the mental burden on employees and providing a healthier work environment.

[0186] A "voice recognition system" is a technology for converting voice data into text data.

[0187] "Emotion and stress level assessment" is the process of analyzing an employee's emotional state and stress level from input data.

[0188] "Generative AI model" refers to an artificial intelligence model that automatically generates advice and alerts for employees based on collected data.

[0189] "Notification of Alerts and Advice" means the means by which an employee is notified of a generated alert or advice, which may be in the form of audio or text.

[0190] "Feedback gathering" is the process of continually gathering responses and reactions from employees.

[0191] "Evaluating system effectiveness" is the process of analyzing collected feedback data to verify the system's performance and effectiveness.

[0192] "Model retraining" is the process of retraining a generative AI model based on evaluation results to improve its accuracy and effectiveness.

[0193] "Daily conversation and work-related data" refers to records of the words and interactions that employees have in the course of their daily work.

[0194] "Early detection of mental health issues" is the process of detecting employee stress and mental health issues as early as possible.

[0195] "Providing appropriate countermeasures" is the means of providing the most appropriate action or assistance for a detected problem.

[0196] These definitions clarify the components of the overall system and their functions.

[0197] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system is composed of three main components: a server, a terminal, and a user.

[0198] System Overview

[0199] Server: A central data processing unit that collects public information and documents from within the company to build an initial learning dataset and trains the generative AI model. It also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[0200] Terminal: The device used by an employee to enter information or receive feedback from the server, including a smartphone, tablet, or PC.

[0201] User: Refers to the employees and managers who use the system. Users input basic information through terminals and carry out daily consultations.

[0202] System behavior and specific functions

[0203] Collecting basic information

[0204] The server first collects publicly available information and documents from the company to build an initial training dataset, including company policies, business processes, and manuals, which are then used to initially train a generative AI model specific to the company.

[0205] The terminal provides a screen where employees can enter basic information (such as name, department, job description, etc.) For example, they can enter information such as "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" on a smartphone application and send it to the server.

[0206] Users enter and submit basic information through a terminal, preparing the system to meet the company's specific needs.

[0207] Daily monitoring and stress detection

[0208] The devices collect data on employees' daily conversations and work activities. They use a voice recognition system to convert the conversations into text data, and collect data through periodic surveys and chat-style dialogue logs.

[0209] The user inputs into the terminal any worries or concerns they have about work. For example, they can send a message such as, "Work has been tough lately and I'm feeling tired."

[0210] The server analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[0211] Mentoring and support

[0212] The server automatically generates an alert when high stress is detected and creates appropriate advice, using a generative AI model to generate specific instructions such as "take a short break."

[0213] The device will then notify the user of any generated alerts or advice, such as a pop-up notification on the smartphone saying, "Try taking some deep breaths to relax."

[0214] The user receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0215] Continuous evaluation and improvement

[0216] The server periodically analyzes the data collected from all users to assess the effectiveness of the system, for example by graphing fluctuations in stress levels across the workforce.

[0217] The device will notify the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[0218] The server retrains the generative AI model based on the analysis results, enabling it to provide more effective feedback.

[0219] Specific examples

[0220] For example, let's consider the case of a new employee. When an employee logs in to the system for the first time, a screen appears where they can enter basic information (name, department, job description). After entering and submitting the information, the server receives it and completes the initial learning process.

[0221] Next, the employee types into the device, "I've already had a lot of work lately and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert to the employee, telling them to "take a short break," and notifies the device. If the employee accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0222] Example prompts for generative AI models

[0223] "If an employee is unhappy or stressed about a recent shift, generate advice that is appropriate to their emotions."

[0224] In this way, the present invention aims to provide an environment where employees can easily seek mental health advice, thereby reducing stress and mental burden and providing a healthier working environment.

[0225] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0226] Step 1:

[0227] The server collects public information and documents within the company.

[0228] Input: Company policies, business processes, manuals, etc.

[0229] Specific operation: The server runs a program that scans documents stored in a specific format and automatically generates a dataset.

[0230] Output: Initial training dataset.

[0231] Step 2:

[0232] The terminal provides a screen for the employee to enter basic information.

[0233] Input: Employee name, department, job description.

[0234] Specific operation: The terminal application displays an input form and sends the data entered by the employee to the server.

[0235] Output: Employee basic information data.

[0236] Step 3:

[0237] The server performs initial training on the generative AI model based on the received basic information.

[0238] Input: Employee basic information data, initial learning dataset.

[0239] How it works: The server uses basic information to train a generative AI model to understand the company's specific context.

[0240] Output: A trained generative AI model.

[0241] Step 4:

[0242] The devices collect data about employees' daily conversations and work.

[0243] Input: Employee voice data.

[0244] Specific operation: The voice data is converted into text data using a voice recognition system and sent to the server.

[0245] Output: Daily conversation text data.

[0246] Step 5:

[0247] The server analyzes the collected data and assesses emotions and stress levels.

[0248] Input: Everyday conversation text data.

[0249] How it works: The server uses a text analysis algorithm to extract keywords and perform sentiment analysis, and then evaluates the stress level based on the results.

[0250] Output: Emotion assessment results and stress level data.

[0251] Step 6:

[0252] The server generates appropriate alerts and advice based on the assessed stress level.

[0253] Input: Emotion assessment results and stress level data.

[0254] What it does: Uses a generative AI model to generate appropriate advice for stressed employees.

[0255] Output: Alerts and advice.

[0256] Step 7:

[0257] The device notifies the employee of any generated alerts or advice.

[0258] Input: Alerts and Advice.

[0259] Specific behavior: The device will display alerts and advice to employees via pop-up notifications and audio notifications.

[0260] Output: Notification to employee.

[0261] Step 8:

[0262] The user receives the notification and acts as instructed.

[0263] Input: Alerts and Advice.

[0264] Specific actions: The user acts according to the notified advice and inputs feedback into the terminal as necessary.

[0265] Output: Feedback data.

[0266] Step 9:

[0267] The server periodically analyzes the data collected from all users to evaluate the effectiveness of the system.

[0268] Input: Feedback data.

[0269] Specific operations: The server analyzes the feedback data using an analytical algorithm and evaluates its effectiveness.

[0270] Output: Effectiveness evaluation results.

[0271] Step 10:

[0272] The server retrains the generative AI model based on the evaluation results to provide more effective feedback.

[0273] Input: Effectiveness evaluation results, feedback data.

[0274] Specific operation: The server uses the evaluation results to retrain the generative AI model and improve the accuracy of the model.

[0275] Output: An improved generative AI model.

[0276] 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.

[0277] As an embodiment of the present invention, we will explain in detail an AI-based mentoring system for supporting employee mental health, which is combined with an emotion engine that recognizes user emotions. This system consists of three main components: a server, a terminal, and a user.

[0278] System Overview

[0279] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice. It also retrains the generative model based on the emotion data detected by the emotion engine.

[0280] Terminal: A device used by employees to input information and receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs. Terminals also use emotion engines to perform voice analysis, facial expression analysis, and text analysis.

[0281] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through their terminals. The terminals are equipped with an emotion engine that analyzes the user's emotions in real time.

[0282] System Operation

[0283] Initial setup and learning phase

[0284] 1. Server: Collects public information and documents from within the company to build an initial learning dataset, including company policies, business processes, and manuals.

[0285] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[0286] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[0287] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[0288] Daily monitoring and stress detection

[0289] 1. Device: Collect data on users' daily conversations and work. Methods for doing so include periodic surveys and chat-style conversation logs.

[0290] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[0291] 3. On the device: The emotion engine recognizes the user's emotions using voice analysis, facial expression analysis, and text analysis, thereby obtaining the user's emotion data in real time.

[0292] 4. Server: Analyzes the data sent from the device. It also analyzes data from the emotion engine to evaluate emotions and stress levels. For example, it determines whether stress levels are rising based on keywords like "tired" or "tough" or changes in facial expressions detected by the emotion engine.

[0293] Mentoring and support

[0294] 1. Server: If high stress is detected, the server generates appropriate alerts and advice for the user. For example, it generates specific instructions such as "take a short break." It can also adjust the advice based on emotional data from the emotion engine.

[0295] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[0296] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0297] Continuous evaluation and improvement

[0298] 1. Server: Periodically analyzes data collected from all users. For example, graphs are created to show fluctuations in the stress levels of all employees. Data from the emotion engine is also analyzed to provide a detailed assessment.

[0299] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[0300] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback. Further adaptively improves the model based on data from the emotion engine.

[0301] Specific examples

[0302] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[0303] Next, Person A types into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects an increase in stress level. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0304] In this way, this embodiment of the present invention provides an environment where employees can easily seek advice about their mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[0305] The processing flow will be explained below.

[0306] Processing flow of a system that combines emotion engines

[0307] Initial setup and learning phase

[0308] Step 1:

[0309] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[0310] Step 2:

[0311] The server builds an initial training dataset based on the collected information, which involves parsing and formatting the collected documents.

[0312] Step 3:

[0313] The device provides employees with a screen to enter basic information, displaying a form that can be accessed on a smartphone or PC.

[0314] Step 4:

[0315] The user enters basic information. For example, "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[0316] Step 5:

[0317] The server initially trains the generative model based on the basic information it receives, helping it understand the company's unique context and business processes.

[0318] Daily monitoring and stress detection

[0319] Step 1:

[0320] The device collects data about the user's daily conversations and work, obtaining information through conversation logs and periodic surveys.

[0321] Step 2:

[0322] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[0323] Step 3:

[0324] The emotion engine built into the device performs voice analysis, facial expression analysis, and text analysis, for example, recognizing emotions by analyzing the user's tone of voice, changes in facial expression, and input text.

[0325] Step 4:

[0326] The server analyzes the conversation data sent from the device and the output data of the emotion engine, which then evaluates emotions and stress levels.

[0327] Step 5:

[0328] The server calculates the stress level based on the analysis results. For example, it determines the stress level based on keywords such as "tired" and "tough" and emotion recognition results such as "sad" and "tired" from the emotion engine.

[0329] Mentoring and support

[0330] Step 1:

[0331] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[0332] Step 2:

[0333] The device notifies the user of generated alerts and advice, using the smartphone's notification function to display a pop-up message saying, "Try taking deep breaths to relax."

[0334] Step 3:

[0335] The user receives a notification and acts on the advice, for example by entering feedback such as "I took a 5-minute break" into the device again.

[0336] Continuous evaluation and improvement

[0337] Step 1:

[0338] The server periodically analyzes the data collected from all users, statistically analyzing stress levels and emotional data to assess their overall mental health.

[0339] Step 2:

[0340] The server retrains the generative model based on the analysis results, for example by adding new data to update the model and improve its accuracy.

[0341] Step 3:

[0342] The device notifies the user of new advice and improvement measures, and suggests new advice and training based on feedback from the results of the exercise.

[0343] Specific examples

[0344] For example, let us explain what happens when a new employee, Mr. A, uses the system.

[0345] Step 1:

[0346] When Mr. A logs in to the system for the first time, a form for entering basic information appears on the smartphone application.

[0347] Step 2:

[0348] Person A enters "Name: Yamada Taro, Department: Sales Department, Job Description: New customer development" and submits.

[0349] Step 3:

[0350] Based on the information received by the server, initial learning is performed to understand the company's specific context and business processes.

[0351] Step 4:

[0352] Person A types into the device, "Work has been tough lately and I'm tired." The device's emotion engine analyzes Person A's voice and facial expressions and detects the emotion "tired."

[0353] Step 5:

[0354] The server analyzes this data and evaluates the stress level. If high stress is detected, an alert is generated telling Person A to take a short break.

[0355] Step 6:

[0356] A pop-up notification appears on A's device with the message, "Try taking some deep breaths to relax." After A takes a break, A sends feedback saying, "You took a 5-minute break."

[0357] Step 7:

[0358] The server periodically analyzes the stress levels and emotional data of all users and retrains the model, improving the accuracy of the next advice.

[0359] In this way, a system that combines an emotion engine continuously supports users' mental health and contributes to improving the performance of the entire organization.

[0360] Example 2

[0361] 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."

[0362] Employee mental health issues have a significant impact on a company's productivity and employee satisfaction, but many existing systems have been inadequate in detecting stress early or providing appropriate support. In particular, there are few systems that can analyze employees' emotional states and stress levels in real time and provide individually customized advice and support. As a result, it has been difficult to detect mental health issues early and take measures.

[0363] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for collecting public information and documents within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data related to employees' daily conversations and work; means for analyzing the collected data using voice analysis, facial expression analysis, and text analysis to evaluate emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting employee feedback and continuously analyzing and evaluating the effectiveness of the system; and means for re-training the generative model based on the evaluation results. This makes it possible to analyze employees' emotional states and stress levels in real time and provide individually customized advice and support.

[0364] "Internal public information" means policies, procedures, manuals, and other documents that are accessible within the company to employees and affiliated organizations.

[0365] "Documents" refer to documents and files used within a company, including data and information such as policies, business processes, and manuals.

[0366] An "initial training dataset" is a collection of data based on a company's policies and business procedures that is used to train a generative model from its initial state.

[0367] "Device" refers to the digital devices used by employees, including smartphones, tablets, and PCs.

[0368] "Generative models" refer to machine learning models created using artificial intelligence algorithms to analyze employees' emotions and stress levels and provide appropriate advice and alerts.

[0369] "Daily conversational and work-related data" refers to all conversational data, chat logs, and work-related information generated by employees in the course of their daily work.

[0370] "Voice analysis" refers to an analytical method for assessing employee emotions and stress levels from voice data.

[0371] "Facial expression analysis" refers to technology that analyzes employees' facial expressions from image or video data to assess their emotional state.

[0372] "Text analysis" refers to a technology that evaluates emotions and stress levels based on text data entered by employees.

[0373] "Assessing emotions and stress levels" means analyzing the collected data and providing a numerical or categorical representation of an employee's current emotional state and stress level.

[0374] An "alert" refers to a notification sent to employees to warn them or instruct them on what to do.

[0375] "Advice" refers to specific suggestions such as guidelines for action and relaxation methods aimed at improving employees' mental health.

[0376] "Feedback" refers to the reactions and information employees provide in response to alerts and advice.

[0377] "Retraining a generative model" means incorporating new data and feedback to update the model so that it can provide more accurate analysis and advice.

[0378] This invention is an AI-based mentoring system for supporting employee mental health. It starts by collecting public information and documents from within the company and building an initial learning dataset. The system consists of three main components: a server, a terminal, and a user.

[0379] server

[0380] The server is the center of the system and has the following main functions:

[0381] Data collection

[0382] The server collects public information and documents from within the company to build an initial learning dataset. This data includes company policies, business processes, manuals, etc. For example, the server obtains data in XML or CSV format from internal file servers and management tools (SharePoint, Confluence, etc.).

[0383] Early Learning

[0384] The server uses the received basic information to initially train a generative model specific to the company's context. This process involves training the generative AI model using a Python program using TensorFlow or PyTorch.

[0385] Terminal

[0386] The terminal provides the interface through which the employee interacts with the system.

[0387] Enter basic information

[0388] Provide a screen for employees to enter basic information (name, department, job description). For example, an input form for "name," "department," and "job description" is displayed on a smartphone application.

[0389] Data collection

[0390] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[0391] Emotion analysis

[0392] The device is equipped with an emotion engine that performs speech, facial expression, and text analysis. Speech data is analyzed using NLTK and DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER and TextBlob).

[0393] User

[0394] Users represent the actions employees take when using the system.

[0395] Input and Feedback

[0396] The user uses the device to input basic information and concerns or questions about daily work. For example, they input text such as "Work has been tough and I'm tired lately." The system also provides feedback in response to alerts and advice from the server and the device.

[0397] Mentoring and support

[0398] Alert Generation

[0399] The server analyzes the data sent from the device and evaluates emotions and stress levels in conjunction with the analysis results of the emotion engine. If high stress is detected, it generates appropriate alerts and advice for the user. For example, it uses Python code to generate instructions such as "Take a short break" using natural language generation tools (such as GPT-3).

[0400] Alert Notifications

[0401] The generated alerts and advice are sent to the user via the device, for example, a pop-up notification on the smartphone displays the message "Try taking deep breaths to relax."

[0402] Continuous evaluation and improvement

[0403] Data analysis and retraining

[0404] The server periodically analyzes data collected from all users to evaluate the effectiveness of the system. This includes graphing fluctuations in stress levels across employees and analyzing them using visualization tools (Tableau and Matplotlib). Based on the analysis results, the generative model is retrained. This incorporates new emotional data and feedback, updating the model to provide more accurate analysis and advice.

[0405] Specific examples

[0406] When new employee A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[0407] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects rising stress levels. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0408] Prompt Sentence Examples

[0409] Here are some examples of prompts for generative AI models:

[0410] Generate initial learning data based on publicly available information within the company and basic employee information to provide appropriate mental health support to newly hired employees. Based on the information entered by the user (Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development), provide examples of specific advice and support.

[0411]

[0412] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0413] Step 1: Data collection

[0414] server:

[0415] The initial learning dataset is constructed by collecting public information and documents from within the company. For example, the server acquires data in XML or CSV format from the company's file server or management tools (SharePoint, Confluence, etc.).

[0416] Input: Internal company public information and documents

[0417] Data processing: Format conversion and organization of collected data

[0418] Output: Initial training dataset

[0419] Step 2: Enter basic information

[0420] Device:

[0421] Provide a screen for employees to enter basic information (name, department, job description). For example, display an input form for "name," "department," and "job description" on a smartphone application.

[0422] Input: Employee basic information

[0423] Data processing: None

[0424] Output: Basic information entered

[0425] User:

[0426] Enter basic information using the terminal and send it to the server. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and press the send button.

[0427] Input: Name, Department, Job Description

[0428] Data processing: None

[0429] Output: Basic information sent

[0430] Step 3: Initial learning

[0431] server:

[0432] Based on the received basic information, a generative model specific to the company's context is initially trained. For example, a Python program can be used to train the generative AI model using TensorFlow or PyTorch.

[0433] Input: Initial training dataset and basic information

[0434] Data Computing: Training generative models (applying machine learning algorithms)

[0435] Output: Initially trained generative model

[0436] Step 4: Routine data collection

[0437] Device:

[0438] Collect data about users' daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[0439] Input: Daily conversation, business data

[0440] Data processing: logging and organization

[0441] Output: Daily data collected

[0442] Step 5: Sentiment Analysis

[0443] Device:

[0444] Using emotion engines, speech analysis, facial expression analysis, and text analysis are performed. For example, speech data is analyzed using NLTK or DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER or TextBlob).

[0445] Input: Collected daily data

[0446] Data Computing: Speech, facial expression, and text emotion analysis

[0447] Output: User emotion data

[0448] Step 6: Stress Assessment

[0449] server:

[0450] The data sent from the device is analyzed and combined with the analysis results of the emotion engine to evaluate emotions and stress levels. For example, stress levels are determined based on keywords such as "tired" or "tough" or detected emotions such as "sad."

[0451] Input: User's daily data and emotional data

[0452] Data calculation: Stress level assessment (analysis of keywords and emotional data)

[0453] Output: Stress level evaluation result

[0454] Step 7: Alert Generation

[0455] server:

[0456] If high stress is detected, appropriate alerts and advice are generated for the user, such as "Take a short break" using natural language generation tools (such as GPT-3) in Python code.

[0457] Input: Stress level assessment result

[0458] Data operations: generating alerts and advice (applying natural language generation tools)

[0459] Output: Alerts and Advice

[0460] Step 8: Alert Notifications

[0461] Device:

[0462] Notify the user of any generated alerts or advice, for example, by displaying a pop-up notification on their smartphone with the message "Try taking some deep breaths to relax."

[0463] Input: Generated alerts and advice

[0464] Data Processing: Message Notification Settings

[0465] Output: User notification

[0466] Step 9: Gather feedback

[0467] User:

[0468] Receive alerts and advice and act on them, for example, take a five-minute break and then enter that feedback back into the device.

[0469] Input: Alert responses and feedback

[0470] Data Processing: Entering and Sending Feedback

[0471] Output: Feedback sent

[0472] Step 10: Continuous data analysis and retraining

[0473] server:

[0474] The data collected from all users is periodically analyzed to evaluate the effectiveness of the system. Fluctuations in the stress levels of all employees are graphed and analyzed using visualization tools (Tableau and Matplotlib). The generative model is retrained based on the analysis results.

[0475] Input: Feedback and daily data collected from each user

[0476] Data Computing: Data Analysis and Generative Model Retraining

[0477] Output: An improved generative model

[0478] (Application example 2)

[0479] 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."

[0480] Mental health issues among employees, especially factory workers, can have a serious impact on productivity and safety. Conventional mentoring systems were unable to analyze workers' emotions or stress levels, making it difficult to respond in real time. Furthermore, due to a lack of means for providing feedback, workers have not received appropriate support. Therefore, there is a need for a system that can detect worker problems early and provide appropriate advice.

[0481] 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.

[0482] In this invention, the server includes means for collecting public information and documents from within the company and building an initial training dataset, terminal means for employees to input basic information, means for initial training a generative model based on the basic information, means for collecting data on workers' daily conversations and work, means for analyzing the collected data and evaluating emotions and stress levels, means including a factory robot that provides feedback in real time based on the emotion data, means for generating appropriate alerts and advice based on the evaluated stress levels, means for notifying the workers of the generated alerts and advice, means for continuously analyzing data and evaluating the effectiveness of the system, and means for re-training the model based on the evaluation results. This makes it possible to monitor workers' emotions and stress levels in real time and provide appropriate feedback and advice.

[0483] "Public information and documents within a company" refers to internal information and publicly available documents such as policies, business processes, and manuals managed by the company.

[0484] An "initial training dataset" is a collection of data based on collected internal company information and documents that is used to train the system's generative model.

[0485] "Terminal means" refers to devices that workers use to input information, including smartphones, tablets, and PCs.

[0486] A "generative model" is an algorithm that uses machine learning and deep learning to assess workers' emotions and stress levels.

[0487] "Daily conversation and work-related data" refers to information related to a worker's daily communications and work activities.

[0488] "Emotion and stress level assessment tools" are techniques and methods for analyzing collected data and determining a worker's emotional state and level of stress.

[0489] A "factory robot" is a device that interacts with workers in a factory and provides real-time feedback based on emotional data.

[0490] "Alert and advice generating means" means a method or tool for generating notifications or instructions to workers based on assessed emotions and stress levels.

[0491] "Means of notification" refers to the means by which generated alerts or advice are communicated to workers, including voice, text message, pop-up notification, etc.

[0492] "Means for continuous data analysis" refers to a method for periodically analyzing the data collected by the system and evaluating the effectiveness of the system.

[0493] "Model retraining" is the process by which the system updates its generative model based on new data to improve its predictive ability and accuracy.

[0494] As an embodiment of the present invention, an AI-based mentoring system for supporting the mental health of workers is described in detail below. This system consists of three main components: a server, a terminal, and a worker.

[0495] System Overview

[0496] Server: Publicly available information and documents from within the company are collected and used to build an initial training dataset. The server then analyzes workers' daily conversation data and work data to evaluate their emotions and stress levels. The generative model is retrained based on the emotion data detected by the emotion engine. The specific software used includes Microsoft Azure's Face API and Speech-to-Text API for emotion analysis. AWS EC2 is used as the data processing server, and AWS RDS is used as the database service.

[0497] Terminal: This refers to the device through which workers input information and receive feedback from the server. This device includes smartphones, tablets, and PCs. Factory robots are also used to collect data on workers' daily conversations and work. The robots are equipped with cameras and microphones, allowing for real-time analysis of facial expressions and voice.

[0498] User (Worker): A factory employee who uses the system. Workers enter basic information and receive daily consultations through a terminal. The terminal's emotion engine analyzes the worker's emotions in real time.

[0499] System Operation

[0500] The server collects public information and documents from within the company to build an initial training dataset. The generative model performs initial training based on this initial dataset. Next, workers enter basic information through their devices, which is sent to the server. The server trains the generative model based on the collected basic information and prepares a model tailored to the company's specific needs. Real-time sentiment analysis is incorporated into this process, enabling more accurate data collection and feedback.

[0501] The terminal (factory robot) collects data on the worker's daily conversations and work. If a worker types into the system, "I've been feeling stressed lately because of the heavy workload," the robot's camera and microphone capture the worker's facial expressions and record their voice. This data is sent to a server for facial expression, voice, and text analysis.

[0502] Continuous evaluation and improvement

[0503] The server periodically analyzes data collected from all workers to assess fluctuations in stress levels across the workforce. It continuously evaluates data from the emotion engine and retrains the generative model based on the results, improving the accuracy and effectiveness of the generated feedback.

[0504] Specific examples

[0505] For example, consider a situation where a factory worker inputs into the system, "I've been working too hard lately and I'm feeling stressed." At this time, the robot's camera captures the worker's facial expressions and its microphone records his / her voice. The worker's emotions and stress level are analyzed, and if high stress is detected, the server generates feedback such as "Take a short break," and the robot notifies the worker.

[0506] Prompt Sentence Examples

[0507] A worker inputs the following into the system: "I've been feeling stressed lately because of the heavy workload." In addition, the camera captures their facial expressions and the microphone records their voice. The system analyzes this data and generates a response for this situation.

[0508] In this way, the present invention provides an environment in which workers can easily seek advice about mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures.

[0509] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0510] Step 1:

[0511] Building the initial dataset

[0512] The server collects publicly available information and documents from within the company, including company policies, work procedures, and business manuals. This collected information is used to create an initial learning dataset. The server stores this dataset in a database and uses it for initial learning.

[0513] Input: Company disclosures and documents.

[0514] Output: Initial training dataset.

[0515] Step 2:

[0516] Enter basic information

[0517] The terminal provides a screen for workers to enter basic information, such as their name, department, and role. The user (worker) uses this screen to enter information and sends it to the server.

[0518] Input: Worker basic information (name, department, role).

[0519] Output: Basic information sent to the server.

[0520] Step 3:

[0521] Early Learning

[0522] The server uses the received basic information to perform initial training of the generative model using an initial training dataset. During this process, the server prepares a generative model tailored to the worker's needs. It uses Microsoft Azure's Face API and Speech-to-Text API.

[0523] Input: Initial training dataset, basic information.

[0524] Output: The initial trained generative model.

[0525] Step 4:

[0526] Daily data collection

[0527] The terminal (factory robot) collects data on the worker's daily conversations and work. The robot uses a camera and microphone to capture the worker's facial expressions and voice in real time. The collected data is sent to a server.

[0528] Input: Data on workers' daily conversations and work (voice, facial expressions).

[0529] Output: Daily data sent to the server.

[0530] Step 5:

[0531] Emotional and stress level assessment

[0532] The server analyzes the daily data sent to it and assesses emotions and stress levels. It uses Microsoft Azure's Face API and Speech-to-Text API to analyze voice and facial expressions. Text data is analyzed using natural language processing techniques.

[0533] Input: Everyday data (voice, facial expressions, text).

[0534] Output: Emotion and stress level assessment results.

[0535] Step 6:

[0536] Generate alerts and advice

[0537] The server generates appropriate alerts and advice based on the evaluation results. For example, if high stress is detected, it will generate feedback such as "Take a short break."

[0538] Input: Emotion and stress level assessment results.

[0539] Output: Alerts and advice.

[0540] Step 7:

[0541] Alerts and advice notifications

[0542] The terminal (factory robot) notifies the worker of the generated alerts and advice, for example, the robot will say in a voice message, "Take a short break."

[0543] Input: Alerts and advice.

[0544] Output: Notice to workers.

[0545] Step 8:

[0546] Continuous data analysis and re-learning

[0547] The server periodically analyzes the data collected from all workers and retrains the generative model, using past data stored in a database to improve the model's accuracy.

[0548] Input: Collected worker data.

[0549] Output: The retrained generative model.

[0550] Examples of concrete examples and prompts

[0551] For example, consider a situation where a factory worker inputs into the system, "I've been working too hard lately and I'm feeling stressed." At this time, the robot's camera captures the worker's facial expressions and its microphone records his / her voice. The worker's emotions and stress level are analyzed, and if high stress is detected, the server generates feedback such as "Take a short break," and the robot notifies the worker.

[0552] Example prompt sentence:

[0553] A worker inputs the following into the system: "I've been feeling stressed lately because of the heavy workload." In addition, the camera captures their facial expressions and the microphone records their voice. The system analyzes this data and generates a response for this situation.

[0554] 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.

[0555] 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.

[0556] 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.

[0557] [Second embodiment]

[0558] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0559] 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.

[0560] 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).

[0561] 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.

[0562] 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.

[0563] 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).

[0564] 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.

[0565] 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.

[0566] 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.

[0567] 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.

[0568] 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.

[0569] 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."

[0570] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system consists of three main components: a server, a terminal, and a user.

[0571] System Overview

[0572] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[0573] Terminal: A device used by an employee to input information or receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs.

[0574] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through terminals.

[0575] System Operation

[0576] Initial setup and learning phase

[0577] 1. Server: First, collect publicly available information and documents from within the company. This data is used to build an initial learning dataset, including company policies, business processes, and manuals.

[0578] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[0579] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[0580] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[0581] Daily monitoring and stress detection

[0582] 1. Devices: Collect data on employees' daily conversations and work. Methods include periodic surveys and chat-style conversation logs.

[0583] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[0584] 3. Server: Analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[0585] Mentoring and support

[0586] 1. Server: When high stress is detected, it automatically generates an alert and generates appropriate advice, such as "take a short break."

[0587] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[0588] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0589] Continuous evaluation and improvement

[0590] 1. Server: Periodically analyzes data collected from all users to evaluate the effectiveness of the system, for example by graphing fluctuations in stress levels across the entire workforce.

[0591] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[0592] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback.

[0593] Specific examples

[0594] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[0595] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert for Person A saying, "Take a short break," and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0596] This system is designed to provide an environment where employees can easily receive mental health consultations, detect stress early, and promptly implement appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[0597] The processing flow will be explained below.

[0598] Initial setup and learning phase

[0599] Step 1:

[0600] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[0601] Step 2:

[0602] The server creates an initial learning dataset based on the collected data, which prepares the data to be used for initial learning.

[0603] Step 3:

[0604] The terminal provides employees with a screen for entering basic information, and the input form is displayed on a smartphone or PC application.

[0605] Step 4:

[0606] The user enters basic information, for example, "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and sends it from the terminal to the server.

[0607] Step 5:

[0608] The server initially trains the generative model based on the basic information received, thereby preparing a model that corresponds to the company's specific context and needs.

[0609] Daily monitoring and stress detection

[0610] Step 1:

[0611] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style conversation logs.

[0612] Step 2:

[0613] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[0614] Step 3:

[0615] The server analyzes the data sent from the device and uses natural language processing to evaluate emotions and stress levels.

[0616] Step 4:

[0617] The server calculates the stress level based on the analysis results, and generates an alert if signs of high stress are detected.

[0618] Mentoring and support

[0619] Step 1:

[0620] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[0621] Step 2:

[0622] Notify the user of any alerts or advice generated by the device, such as a pop-up notification on the smartphone with the message "Try taking some deep breaths to relax."

[0623] Step 3:

[0624] The user receives a notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0625] Continuous evaluation and improvement

[0626] Step 1:

[0627] The server periodically analyzes the data collected from all users, for example, to graph fluctuations in stress levels across the entire workforce.

[0628] Step 2:

[0629] The device will notify the user of new advice and improvements, for example suggesting new training programs or relaxation techniques.

[0630] Step 3:

[0631] The server retrains the generative model based on the data analysis results, allowing it to provide more effective feedback.

[0632] Through these specific processing steps, a system will be created that continuously supports the mental health of employees and contributes to improving the performance of the entire organization.

[0633] Example 1

[0634] 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."

[0635] Maintaining and improving employee mental health is an important issue in today's workplace. However, many companies lack systems for constantly monitoring employee mental health and providing appropriate support. It is also difficult to detect stress early or provide advice tailored to individual employees, making it difficult to take measures before employee problems become serious.

[0636] 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.

[0637] In this invention, the server includes: means for collecting public information and documents from within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data on employees' daily conversations and work; means for analyzing the collected data and evaluating emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting feedback from employees and using the data for re-learning; means for periodically analyzing data collected from all employees and evaluating the effectiveness of the system; and means for re-learning the generative model based on the evaluation results and generating more effective feedback. This makes it possible to monitor employees' mental health in real time, detect stress early, and provide appropriate support and advice.

[0638] "Public information" refers to information obtained from inside or outside the company, such as company policies, business processes, and manuals.

[0639] "Documents" refers to text data related to specific content, such as company records, manuals, and guides.

[0640] An "initial training dataset" is a dataset constructed from publicly available company information and documents, and refers to a collection of data used to train a generative model.

[0641] "Terminal" refers to the device used by employees to enter basic information and inquiries, including smartphones, tablets, and PCs.

[0642] A "generative model" is an algorithm that learns from data obtained from employees and generates appropriate advice and alerts.

[0643] "Daily conversation and work-related data" refers to text data and voice data generated during employees' daily discussions and work activities.

[0644] "Emotion and stress levels" refers to the psychological state of employees, assessed by analyzing data on their daily conversations and work.

[0645] An "alert" refers to a notification that is generated when an employee's stress level exceeds a certain threshold.

[0646] "Advice" refers to specific suggestions for improving employees' mental health and reducing stress.

[0647] "Feedback" refers to information provided to employees about the advice and results of their actions, and refers to data that is re-entered into the system.

[0648] "Retraining" refers to the process of retraining a generative model using feedback or newly collected data to improve its accuracy.

[0649] "Effectiveness" refers to an indicator that shows how effective the system is in improving employees' mental health.

[0650] This invention relates to an AI-based mentoring system for supporting employee mental health. The system collects publicly available information and documents from within a company, constructs an initial training dataset, and trains a generative model based on the dataset to analyze employees' stress levels and provide appropriate alerts and advice. Feedback from employees is used for retraining, allowing the model to be continuously improved and the effectiveness of the system to be evaluated.

[0651] Server: Initial setup and learning phase

[0652] The server first collects publicly available information and documents from within the company. Specifically, it uses Python scripts to retrieve information such as policies, business processes, and manuals from the company's internal database and file storage, and builds an initial training dataset based on this data. Next, it receives basic employee information and initially trains a generative model specific to the company's context. This process uses Python and TensorFlow to input data into an NLP model (such as the BERT model) to create a company-specific language model.

[0653] Terminal: Data collection and feedback

[0654] The terminal is a device used by employees to input basic information and daily inquiries. To this end, it is designed as a web application using front-end frameworks such as React or Vue.js. The terminal provides an interface for collecting data on employees' daily conversations and work. This includes periodic surveys and interactions with chatbots, which are developed using Dialogflow or the Microsoft Bot Framework. Employee input is provided, for example, through a smartphone application.

[0655] User: System usage

[0656] Employees, who are users, input basic information and everyday concerns through their terminals. For example, an employee may input basic information such as "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and then, during work hours, input concerns such as "Work has been tough and I'm feeling tired lately." This data is sent to the server and used for analysis.

[0657] Server: Data analysis and advice generation

[0658] The server analyzes the data sent from the device and evaluates the employee's emotions and stress levels. Specifically, it performs natural language processing using Python's NLTK library and spaCy to extract keywords from the text data and score the stress level. If an increase in stress level is detected, a generative AI model (e.g., GPT-3) is used to generate specific advice, such as "Take a short break."

[0659] Device: Notifications and feedback collection

[0660] The generated alerts and advice are notified to the user by the device. For example, a message such as "Take a short break" is displayed using the smartphone's push notification function. The user receives the notification, acts according to the instructions, and then enters the results back into the device. For example, feedback is given such as "Took a 5-minute break."

[0661] Server: Continuous evaluation and model retraining

[0662] The server periodically analyzes data collected from all users and performs data analysis using Python's pandas library. Fluctuations in stress levels across employees are graphed using Matplotlib and Seaborn. The generative model is retrained based on the analysis results to provide more effective feedback. Accuracy is improved by retraining the NLP model using feedback data from employees.

[0663] Examples of specific examples and prompts

[0664] For example, when a new employee logs in to the system for the first time and enters basic information (name, department, position), they can then enter "I'm tired because I've already had a lot of work lately" into their device while they're working. The server analyzes this data, detects an increase in stress level, and generates advice to "take a short break" and notifies the device. If the new employee accepts the advice and takes a break, the feedback data is sent to the server and used for relearning.

[0665] Prompt Sentence Examples

[0666] "Tell me about your recent work. Example: Work has been hard and I'm feeling tired lately."

[0667] What's an effective way to relax?

[0668] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0669] Step 1:

[0670] Server: Collects public information and documents from within the company and builds an initial learning dataset. Specifically, it runs Python scripts to retrieve policies, business processes, manuals, etc. from the company's internal database and file storage, and then formats this into a dataset.

[0671] Input: Public information and documents (policies, business processes, manuals)

[0672] Output: Initial training dataset

[0673] Step 2:

[0674] Devices: Provide an interface for employees to enter basic information. Specifically, create a web form using React or Vue.js and make it available on devices such as smartphones, tablets, and PCs.

[0675] User: For example, a user (employee) enters "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[0676] Input: Employee basic information (name, department, job description)

[0677] Output: Send employee basic information to the server

[0678] Step 3:

[0679] Server: Receives basic employee information and performs initial training on the generative model. Specifically, it uses Python and TensorFlow to input data into an NLP model (such as the BERT model) and builds a company-specific language model.

[0680] Input: Basic employee information (name, department, job description), initial learning dataset

[0681] Output: Company-specific generative model

[0682] Step 4:

[0683] Terminal: Provides an interface for collecting data about employees' daily conversations and work, for example, through periodic surveys or chatbots. Chatbots are developed using Dialogflow or the Microsoft Bot Framework.

[0684] User: During work, input a question such as "Work has been tough and I'm tired lately."

[0685] Input: Daily conversation and business data (text messages)

[0686] Output: Sending daily conversation data to the server

[0687] Step 5:

[0688] Server: Analyzes the data sent from the device and evaluates emotions and stress levels. Specifically, it uses Python's NLTK library and spaCy to analyze text data, extract keywords, and score stress levels.

[0689] Input: Daily conversation and business data

[0690] Output: Emotion and stress level assessment results

[0691] Step 6:

[0692] Server: Based on the evaluation results, it generates appropriate alerts and advice. It uses a generative AI model (e.g., GPT-3) to create specific advice, such as "Take a short break."

[0693] Input: Emotion and stress level assessment results

[0694] Output: Alerts and advice

[0695] Step 7:

[0696] Devices: Notify employees of generated alerts and advice, using push notifications on their smartphones to display messages such as "Take a short break."

[0697] Input: Alerts and Advice

[0698] Output: Employee notification

[0699] Step 8:

[0700] User: Receives alerts and advice and acts accordingly, for example by entering feedback into the device such as "I took a 5-minute break."

[0701] Input: Employee feedback (e.g., took a break)

[0702] Output: Sending feedback to the server

[0703] Step 9:

[0704] Server: Receives employee feedback and uses that data for retraining. Specifically, the NLP model is retrained using the feedback data to improve accuracy. The retraining process is performed using Python and TensorFlow.

[0705] Input: Employee feedback

[0706] Output: Retrained generative model

[0707] Step 10:

[0708] Server: Regularly analyzes data collected from all employees to evaluate the effectiveness of the system. Python's pandas library is used to analyze the data, and Matplotlib and Seaborn are used to graph, for example, fluctuations in stress levels across all employees. Based on the evaluation results, new advice and improvement measures are implemented.

[0709] Input: Feedback data collected from all employees

[0710] Output: System effectiveness evaluation results, new advice and improvements

[0711] (Application example 1)

[0712] 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."

[0713] Maintaining and improving employee mental health is an important issue for companies. However, currently, there are limitations to early detection of employee stress and mental problems and appropriate response. In particular, in factory environments, employees often bear significant physical and mental burdens, requiring rapid and effective responses. Conventional methods rely heavily on self-reporting, making it difficult to grasp mental health conditions in real time or provide appropriate advice. For this reason, a system is needed that provides an environment where employees can easily seek mental health advice, detects stress and mental burden early, and enables appropriate measures to be taken promptly.

[0714] 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.

[0715] In this invention, the server includes means for converting what employees say into text data using a voice recognition system, means for evaluating emotions and stress levels based on the converted text data, means for using a generative AI model to generate appropriate alerts and advice based on the evaluation, means for notifying employees of the generated alerts and advice by voice or text, means for continuously collecting employee feedback and evaluating the effectiveness of the system, and means for retraining the model based on the evaluation results. This makes it possible to monitor employees' mental health in real time and provide appropriate advice and support, thereby reducing the mental burden on employees and providing a healthier work environment.

[0716] A "voice recognition system" is a technology for converting voice data into text data.

[0717] "Emotion and stress level assessment" is the process of analyzing an employee's emotional state and stress level from input data.

[0718] "Generative AI model" refers to an artificial intelligence model that automatically generates advice and alerts for employees based on collected data.

[0719] "Notification of Alerts and Advice" means the means by which an employee is notified of a generated alert or advice, which may be in the form of audio or text.

[0720] "Feedback gathering" is the process of continually gathering responses and reactions from employees.

[0721] "Evaluating system effectiveness" is the process of analyzing collected feedback data to verify the system's performance and effectiveness.

[0722] "Model retraining" is the process of retraining a generative AI model based on evaluation results to improve its accuracy and effectiveness.

[0723] "Daily conversation and work-related data" refers to records of the words and interactions that employees have in the course of their daily work.

[0724] "Early detection of mental health issues" is the process of detecting employee stress and mental health issues as early as possible.

[0725] "Providing appropriate countermeasures" is the means of providing the most appropriate action or assistance for a detected problem.

[0726] These definitions clarify the components of the overall system and their functions.

[0727] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system is composed of three main components: a server, a terminal, and a user.

[0728] System Overview

[0729] Server: A central data processing unit that collects public information and documents from within the company to build an initial learning dataset and trains the generative AI model. It also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[0730] Terminal: The device used by an employee to enter information or receive feedback from the server, including a smartphone, tablet, or PC.

[0731] User: Refers to the employees and managers who use the system. Users input basic information through terminals and carry out daily consultations.

[0732] System behavior and specific functions

[0733] Collecting basic information

[0734] The server first collects publicly available information and documents from the company to build an initial training dataset, including company policies, business processes, and manuals, which are then used to initially train a generative AI model specific to the company.

[0735] The terminal provides a screen where employees can enter basic information (such as name, department, job description, etc.) For example, they can enter information such as "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" on a smartphone application and send it to the server.

[0736] Users enter and submit basic information through a terminal, preparing the system to meet the company's specific needs.

[0737] Daily monitoring and stress detection

[0738] The devices collect data on employees' daily conversations and work activities. They use a voice recognition system to convert the conversations into text data, and collect data through periodic surveys and chat-style dialogue logs.

[0739] The user inputs into the terminal any worries or concerns they have about work. For example, they can send a message such as, "Work has been tough lately and I'm feeling tired."

[0740] The server analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[0741] Mentoring and support

[0742] The server automatically generates an alert when high stress is detected and creates appropriate advice, using a generative AI model to generate specific instructions such as "take a short break."

[0743] The device will then notify the user of any generated alerts or advice, such as a pop-up notification on the smartphone saying, "Try taking some deep breaths to relax."

[0744] The user receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0745] Continuous evaluation and improvement

[0746] The server periodically analyzes the data collected from all users to assess the effectiveness of the system, for example by graphing fluctuations in stress levels across the workforce.

[0747] The device will notify the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[0748] The server retrains the generative AI model based on the analysis results, enabling it to provide more effective feedback.

[0749] Specific examples

[0750] For example, let's consider the case of a new employee. When an employee logs in to the system for the first time, a screen appears where they can enter basic information (name, department, job description). After entering and submitting the information, the server receives it and completes the initial learning process.

[0751] Next, the employee types into the device, "I've already had a lot of work lately and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert to the employee, telling them to "take a short break," and notifies the device. If the employee accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0752] Example prompts for generative AI models

[0753] "If an employee is unhappy or stressed about a recent shift, generate advice that is appropriate to their emotions."

[0754] In this way, the present invention aims to provide an environment where employees can easily seek mental health advice, thereby reducing stress and mental burden and providing a healthier working environment.

[0755] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0756] Step 1:

[0757] The server collects public information and documents within the company.

[0758] Input: Company policies, business processes, manuals, etc.

[0759] Specific operation: The server runs a program that scans documents stored in a specific format and automatically generates a dataset.

[0760] Output: Initial training dataset.

[0761] Step 2:

[0762] The terminal provides a screen for the employee to enter basic information.

[0763] Input: Employee name, department, job description.

[0764] Specific operation: The terminal application displays an input form and sends the data entered by the employee to the server.

[0765] Output: Employee basic information data.

[0766] Step 3:

[0767] The server performs initial training on the generative AI model based on the received basic information.

[0768] Input: Employee basic information data, initial learning dataset.

[0769] How it works: The server uses basic information to train a generative AI model to understand the company's specific context.

[0770] Output: A trained generative AI model.

[0771] Step 4:

[0772] The devices collect data about employees' daily conversations and work.

[0773] Input: Employee voice data.

[0774] Specific operation: The voice data is converted into text data using a voice recognition system and sent to the server.

[0775] Output: Daily conversation text data.

[0776] Step 5:

[0777] The server analyzes the collected data and assesses emotions and stress levels.

[0778] Input: Everyday conversation text data.

[0779] How it works: The server uses a text analysis algorithm to extract keywords and perform sentiment analysis, and then evaluates the stress level based on the results.

[0780] Output: Emotion assessment results and stress level data.

[0781] Step 6:

[0782] The server generates appropriate alerts and advice based on the assessed stress level.

[0783] Input: Emotion assessment results and stress level data.

[0784] What it does: Uses a generative AI model to generate appropriate advice for stressed employees.

[0785] Output: Alerts and advice.

[0786] Step 7:

[0787] The device notifies the employee of any generated alerts or advice.

[0788] Input: Alerts and Advice.

[0789] Specific behavior: The device will display alerts and advice to employees via pop-up notifications and audio notifications.

[0790] Output: Notification to employee.

[0791] Step 8:

[0792] The user receives the notification and acts as instructed.

[0793] Input: Alerts and Advice.

[0794] Specific actions: The user acts according to the notified advice and inputs feedback into the terminal as necessary.

[0795] Output: Feedback data.

[0796] Step 9:

[0797] The server periodically analyzes the data collected from all users to evaluate the effectiveness of the system.

[0798] Input: Feedback data.

[0799] Specific operations: The server analyzes the feedback data using an analytical algorithm and evaluates its effectiveness.

[0800] Output: Effectiveness evaluation results.

[0801] Step 10:

[0802] The server retrains the generative AI model based on the evaluation results to provide more effective feedback.

[0803] Input: Effectiveness evaluation results, feedback data.

[0804] Specific operation: The server uses the evaluation results to retrain the generative AI model and improve the accuracy of the model.

[0805] Output: An improved generative AI model.

[0806] 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.

[0807] As an embodiment of the present invention, we will explain in detail an AI-based mentoring system for supporting employee mental health, which is combined with an emotion engine that recognizes user emotions. This system consists of three main components: a server, a terminal, and a user.

[0808] System Overview

[0809] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice. It also retrains the generative model based on the emotion data detected by the emotion engine.

[0810] Terminal: A device used by employees to input information and receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs. Terminals also use emotion engines to perform voice analysis, facial expression analysis, and text analysis.

[0811] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through their terminals. The terminals are equipped with an emotion engine that analyzes the user's emotions in real time.

[0812] System Operation

[0813] Initial setup and learning phase

[0814] 1. Server: Collects public information and documents from within the company to build an initial learning dataset, including company policies, business processes, and manuals.

[0815] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[0816] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[0817] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[0818] Daily monitoring and stress detection

[0819] 1. Device: Collect data on users' daily conversations and work. Methods for doing so include periodic surveys and chat-style conversation logs.

[0820] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[0821] 3. On the device: The emotion engine recognizes the user's emotions using voice analysis, facial expression analysis, and text analysis, thereby obtaining the user's emotion data in real time.

[0822] 4. Server: Analyzes the data sent from the device. It also analyzes data from the emotion engine to evaluate emotions and stress levels. For example, it determines whether stress levels are rising based on keywords like "tired" or "tough" or changes in facial expressions detected by the emotion engine.

[0823] Mentoring and support

[0824] 1. Server: If high stress is detected, the server generates appropriate alerts and advice for the user. For example, it generates specific instructions such as "take a short break." It can also adjust the advice based on emotional data from the emotion engine.

[0825] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[0826] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[0827] Continuous evaluation and improvement

[0828] 1. Server: Periodically analyzes data collected from all users. For example, graphs are created to show fluctuations in the stress levels of all employees. Data from the emotion engine is also analyzed to provide a detailed assessment.

[0829] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[0830] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback. Further adaptively improves the model based on data from the emotion engine.

[0831] Specific examples

[0832] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[0833] Next, Person A types into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects an increase in stress level. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0834] In this way, this embodiment of the present invention provides an environment where employees can easily seek advice about their mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[0835] The processing flow will be explained below.

[0836] Processing flow of a system that combines emotion engines

[0837] Initial setup and learning phase

[0838] Step 1:

[0839] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[0840] Step 2:

[0841] The server builds an initial training dataset based on the collected information, which involves parsing and formatting the collected documents.

[0842] Step 3:

[0843] The device provides employees with a screen to enter basic information, displaying a form that can be accessed on a smartphone or PC.

[0844] Step 4:

[0845] The user enters basic information. For example, "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[0846] Step 5:

[0847] The server initially trains the generative model based on the basic information it receives, helping it understand the company's unique context and business processes.

[0848] Daily monitoring and stress detection

[0849] Step 1:

[0850] The device collects data about the user's daily conversations and work, obtaining information through conversation logs and periodic surveys.

[0851] Step 2:

[0852] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[0853] Step 3:

[0854] The emotion engine built into the device performs voice analysis, facial expression analysis, and text analysis, for example, recognizing emotions by analyzing the user's tone of voice, changes in facial expression, and input text.

[0855] Step 4:

[0856] The server analyzes the conversation data sent from the device and the output data of the emotion engine, which then evaluates emotions and stress levels.

[0857] Step 5:

[0858] The server calculates the stress level based on the analysis results. For example, it determines the stress level based on keywords such as "tired" and "tough" and emotion recognition results such as "sad" and "tired" from the emotion engine.

[0859] Mentoring and support

[0860] Step 1:

[0861] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[0862] Step 2:

[0863] The device notifies the user of generated alerts and advice, using the smartphone's notification function to display a pop-up message saying, "Try taking deep breaths to relax."

[0864] Step 3:

[0865] The user receives a notification and acts on the advice, for example by entering feedback such as "I took a 5-minute break" into the device again.

[0866] Continuous evaluation and improvement

[0867] Step 1:

[0868] The server periodically analyzes the data collected from all users, statistically analyzing stress levels and emotional data to assess their overall mental health.

[0869] Step 2:

[0870] The server retrains the generative model based on the analysis results, for example by adding new data to update the model and improve its accuracy.

[0871] Step 3:

[0872] The device notifies the user of new advice and improvement measures, and suggests new advice and training based on feedback from the results of the exercise.

[0873] Specific examples

[0874] For example, let us explain what happens when a new employee, Mr. A, uses the system.

[0875] Step 1:

[0876] When Mr. A logs in to the system for the first time, a form for entering basic information appears on the smartphone application.

[0877] Step 2:

[0878] Person A enters "Name: Yamada Taro, Department: Sales Department, Job Description: New customer development" and submits.

[0879] Step 3:

[0880] Based on the information received by the server, initial learning is performed to understand the company's specific context and business processes.

[0881] Step 4:

[0882] Person A types into the device, "Work has been tough lately and I'm tired." The device's emotion engine analyzes Person A's voice and facial expressions and detects the emotion "tired."

[0883] Step 5:

[0884] The server analyzes this data and evaluates the stress level. If high stress is detected, an alert is generated telling Person A to take a short break.

[0885] Step 6:

[0886] A pop-up notification appears on A's device with the message, "Try taking some deep breaths to relax." After A takes a break, A sends feedback saying, "You took a 5-minute break."

[0887] Step 7:

[0888] The server periodically analyzes the stress levels and emotional data of all users and retrains the model, improving the accuracy of the next advice.

[0889] In this way, a system that combines an emotion engine continuously supports users' mental health and contributes to improving the performance of the entire organization.

[0890] Example 2

[0891] 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."

[0892] Employee mental health issues have a significant impact on a company's productivity and employee satisfaction, but many existing systems have been inadequate in detecting stress early or providing appropriate support. In particular, there are few systems that can analyze employees' emotional states and stress levels in real time and provide individually customized advice and support. As a result, it has been difficult to detect mental health issues early and take measures.

[0893] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for collecting public information and documents within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data related to employees' daily conversations and work; means for analyzing the collected data using voice analysis, facial expression analysis, and text analysis to evaluate emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting employee feedback and continuously analyzing and evaluating the effectiveness of the system; and means for re-training the generative model based on the evaluation results. This makes it possible to analyze employees' emotional states and stress levels in real time and provide individually customized advice and support.

[0894] "Internal public information" means policies, procedures, manuals, and other documents that are accessible within the company to employees and affiliated organizations.

[0895] "Documents" refer to documents and files used within a company, including data and information such as policies, business processes, and manuals.

[0896] An "initial training dataset" is a collection of data based on a company's policies and business procedures that is used to train a generative model from its initial state.

[0897] "Device" refers to the digital devices used by employees, including smartphones, tablets, and PCs.

[0898] "Generative models" refer to machine learning models created using artificial intelligence algorithms to analyze employees' emotions and stress levels and provide appropriate advice and alerts.

[0899] "Daily conversational and work-related data" refers to all conversational data, chat logs, and work-related information generated by employees in the course of their daily work.

[0900] "Voice analysis" refers to an analytical method for assessing employee emotions and stress levels from voice data.

[0901] "Facial expression analysis" refers to technology that analyzes employees' facial expressions from image or video data to assess their emotional state.

[0902] "Text analysis" refers to a technology that evaluates emotions and stress levels based on text data entered by employees.

[0903] "Assessing emotions and stress levels" means analyzing the collected data and providing a numerical or categorical representation of an employee's current emotional state and stress level.

[0904] An "alert" refers to a notification sent to employees to warn them or instruct them on what to do.

[0905] "Advice" refers to specific suggestions such as guidelines for action and relaxation methods aimed at improving employees' mental health.

[0906] "Feedback" refers to the reactions and information employees provide in response to alerts and advice.

[0907] "Retraining a generative model" means incorporating new data and feedback to update the model so that it can provide more accurate analysis and advice.

[0908] This invention is an AI-based mentoring system for supporting employee mental health. It starts by collecting public information and documents from within the company and building an initial learning dataset. The system consists of three main components: a server, a terminal, and a user.

[0909] server

[0910] The server is the center of the system and has the following main functions:

[0911] Data collection

[0912] The server collects public information and documents from within the company to build an initial learning dataset. This data includes company policies, business processes, manuals, etc. For example, the server obtains data in XML or CSV format from internal file servers and management tools (SharePoint, Confluence, etc.).

[0913] Early Learning

[0914] The server uses the received basic information to initially train a generative model specific to the company's context. This process involves training the generative AI model using a Python program using TensorFlow or PyTorch.

[0915] Terminal

[0916] The terminal provides the interface through which the employee interacts with the system.

[0917] Enter basic information

[0918] Provide a screen for employees to enter basic information (name, department, job description). For example, an input form for "name," "department," and "job description" is displayed on a smartphone application.

[0919] Data collection

[0920] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[0921] Emotion analysis

[0922] The device is equipped with an emotion engine that performs speech, facial expression, and text analysis. Speech data is analyzed using NLTK and DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER and TextBlob).

[0923] User

[0924] Users represent the actions employees take when using the system.

[0925] Input and Feedback

[0926] The user uses the device to input basic information and concerns or questions about daily work. For example, they input text such as "Work has been tough and I'm tired lately." The system also provides feedback in response to alerts and advice from the server and the device.

[0927] Mentoring and support

[0928] Alert Generation

[0929] The server analyzes the data sent from the device and evaluates emotions and stress levels in conjunction with the analysis results of the emotion engine. If high stress is detected, it generates appropriate alerts and advice for the user. For example, it uses Python code to generate instructions such as "Take a short break" using natural language generation tools (such as GPT-3).

[0930] Alert Notifications

[0931] The generated alerts and advice are sent to the user via the device, for example, a pop-up notification on the smartphone displays the message "Try taking deep breaths to relax."

[0932] Continuous evaluation and improvement

[0933] Data analysis and retraining

[0934] The server periodically analyzes data collected from all users to evaluate the effectiveness of the system. This includes graphing fluctuations in stress levels across employees and analyzing them using visualization tools (Tableau and Matplotlib). Based on the analysis results, the generative model is retrained. This incorporates new emotional data and feedback, updating the model to provide more accurate analysis and advice.

[0935] Specific examples

[0936] When new employee A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[0937] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects rising stress levels. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[0938] Prompt Sentence Examples

[0939] Here are some examples of prompts for generative AI models:

[0940] Generate initial learning data based on publicly available information within the company and basic employee information to provide appropriate mental health support to newly hired employees. Based on the information entered by the user (Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development), provide examples of specific advice and support.

[0941]

[0942] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0943] Step 1: Data collection

[0944] server:

[0945] The initial learning dataset is constructed by collecting public information and documents from within the company. For example, the server acquires data in XML or CSV format from the company's file server or management tools (SharePoint, Confluence, etc.).

[0946] Input: Internal company public information and documents

[0947] Data processing: Format conversion and organization of collected data

[0948] Output: Initial training dataset

[0949] Step 2: Enter basic information

[0950] Device:

[0951] Provide a screen for employees to enter basic information (name, department, job description). For example, display an input form for "name," "department," and "job description" on a smartphone application.

[0952] Input: Employee basic information

[0953] Data processing: None

[0954] Output: Basic information entered

[0955] User:

[0956] Enter basic information using the terminal and send it to the server. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and press the send button.

[0957] Input: Name, Department, Job Description

[0958] Data processing: None

[0959] Output: Basic information sent

[0960] Step 3: Initial learning

[0961] server:

[0962] Based on the received basic information, a generative model specific to the company's context is initially trained. For example, a Python program can be used to train the generative AI model using TensorFlow or PyTorch.

[0963] Input: Initial training dataset and basic information

[0964] Data Computing: Training generative models (applying machine learning algorithms)

[0965] Output: Initially trained generative model

[0966] Step 4: Routine data collection

[0967] Device:

[0968] Collect data about users' daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[0969] Input: Daily conversation, business data

[0970] Data processing: logging and organization

[0971] Output: Daily data collected

[0972] Step 5: Sentiment Analysis

[0973] Device:

[0974] Using emotion engines, speech analysis, facial expression analysis, and text analysis are performed. For example, speech data is analyzed using NLTK or DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER or TextBlob).

[0975] Input: Collected daily data

[0976] Data Computing: Speech, facial expression, and text emotion analysis

[0977] Output: User emotion data

[0978] Step 6: Stress Assessment

[0979] server:

[0980] The data sent from the device is analyzed and combined with the analysis results of the emotion engine to evaluate emotions and stress levels. For example, stress levels are determined based on keywords such as "tired" or "tough" or detected emotions such as "sad."

[0981] Input: User's daily data and emotional data

[0982] Data calculation: Stress level assessment (analysis of keywords and emotional data)

[0983] Output: Stress level evaluation result

[0984] Step 7: Alert Generation

[0985] server:

[0986] If high stress is detected, appropriate alerts and advice are generated for the user, such as "Take a short break" using natural language generation tools (such as GPT-3) in Python code.

[0987] Input: Stress level assessment result

[0988] Data operations: generating alerts and advice (applying natural language generation tools)

[0989] Output: Alerts and Advice

[0990] Step 8: Alert Notifications

[0991] Device:

[0992] Notify the user of any generated alerts or advice, for example, by displaying a pop-up notification on their smartphone with the message "Try taking some deep breaths to relax."

[0993] Input: Generated alerts and advice

[0994] Data Processing: Message Notification Settings

[0995] Output: User notification

[0996] Step 9: Gather feedback

[0997] User:

[0998] Receive alerts and advice and act on them, for example, take a five-minute break and then enter that feedback back into the device.

[0999] Input: Alert responses and feedback

[1000] Data Processing: Entering and Sending Feedback

[1001] Output: Feedback sent

[1002] Step 10: Continuous data analysis and retraining

[1003] server:

[1004] The data collected from all users is periodically analyzed to evaluate the effectiveness of the system. Fluctuations in the stress levels of all employees are graphed and analyzed using visualization tools (Tableau and Matplotlib). The generative model is retrained based on the analysis results.

[1005] Input: Feedback and daily data collected from each user

[1006] Data Computing: Data Analysis and Generative Model Retraining

[1007] Output: An improved generative model

[1008] (Application example 2)

[1009] 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."

[1010] Mental health issues among employees, especially factory workers, can have a serious impact on productivity and safety. Conventional mentoring systems were unable to analyze workers' emotions or stress levels, making it difficult to respond in real time. Furthermore, due to a lack of means for providing feedback, workers have not received appropriate support. Therefore, there is a need for a system that can detect worker problems early and provide appropriate advice.

[1011] 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.

[1012] In this invention, the server includes means for collecting public information and documents from within the company and building an initial training dataset, terminal means for employees to input basic information, means for initial training a generative model based on the basic information, means for collecting data on workers' daily conversations and work, means for analyzing the collected data and evaluating emotions and stress levels, means including a factory robot that provides feedback in real time based on the emotion data, means for generating appropriate alerts and advice based on the evaluated stress levels, means for notifying the workers of the generated alerts and advice, means for continuously analyzing data and evaluating the effectiveness of the system, and means for re-training the model based on the evaluation results. This makes it possible to monitor workers' emotions and stress levels in real time and provide appropriate feedback and advice.

[1013] "Public information and documents within a company" refers to internal information and publicly available documents such as policies, business processes, and manuals managed by the company.

[1014] An "initial training dataset" is a collection of data based on collected internal company information and documents that is used to train the system's generative model.

[1015] "Terminal means" refers to devices that workers use to input information, including smartphones, tablets, and PCs.

[1016] A "generative model" is an algorithm that uses machine learning and deep learning to assess workers' emotions and stress levels.

[1017] "Daily conversation and work-related data" refers to information related to a worker's daily communications and work activities.

[1018] "Emotion and stress level assessment tools" are techniques and methods for analyzing collected data and determining a worker's emotional state and level of stress.

[1019] A "factory robot" is a device that interacts with workers in a factory and provides real-time feedback based on emotional data.

[1020] "Alert and advice generating means" means a method or tool for generating notifications or instructions to workers based on assessed emotions and stress levels.

[1021] "Means of notification" refers to the means by which generated alerts or advice are communicated to workers, including voice, text message, pop-up notification, etc.

[1022] "Means for continuous data analysis" refers to a method for periodically analyzing the data collected by the system and evaluating the effectiveness of the system.

[1023] "Model retraining" is the process by which the system updates its generative model based on new data to improve its predictive ability and accuracy.

[1024] As an embodiment of the present invention, an AI-based mentoring system for supporting the mental health of workers is described in detail below. This system consists of three main components: a server, a terminal, and a worker.

[1025] System Overview

[1026] Server: Publicly available information and documents from within the company are collected and used to build an initial training dataset. The server then analyzes workers' daily conversation data and work data to evaluate their emotions and stress levels. The generative model is retrained based on the emotion data detected by the emotion engine. The specific software used includes Microsoft Azure's Face API and Speech-to-Text API for emotion analysis. AWS EC2 is used as the data processing server, and AWS RDS is used as the database service.

[1027] Terminal: This refers to the device through which workers input information and receive feedback from the server. This device includes smartphones, tablets, and PCs. Factory robots are also used to collect data on workers' daily conversations and work. The robots are equipped with cameras and microphones, allowing for real-time analysis of facial expressions and voice.

[1028] User (Worker): A factory employee who uses the system. Workers enter basic information and receive daily consultations through a terminal. The terminal's emotion engine analyzes the worker's emotions in real time.

[1029] System Operation

[1030] The server collects public information and documents from within the company to build an initial training dataset. The generative model performs initial training based on this initial dataset. Next, workers enter basic information through their devices, which is sent to the server. The server trains the generative model based on the collected basic information and prepares a model tailored to the company's specific needs. Real-time sentiment analysis is incorporated into this process, enabling more accurate data collection and feedback.

[1031] The terminal (factory robot) collects data on the worker's daily conversations and work. If a worker types into the system, "I've been feeling stressed lately because of the heavy workload," the robot's camera and microphone capture the worker's facial expressions and record their voice. This data is sent to a server for facial expression, voice, and text analysis.

[1032] Continuous evaluation and improvement

[1033] The server periodically analyzes data collected from all workers to assess fluctuations in stress levels across the workforce. It continuously evaluates data from the emotion engine and retrains the generative model based on the results, improving the accuracy and effectiveness of the generated feedback.

[1034] Specific examples

[1035] For example, consider a situation where a factory worker inputs into the system, "I've been working too hard lately and I'm feeling stressed." At this time, the robot's camera captures the worker's facial expressions and its microphone records his / her voice. The worker's emotions and stress level are analyzed, and if high stress is detected, the server generates feedback such as "Take a short break," and the robot notifies the worker.

[1036] Prompt Sentence Examples

[1037] A worker inputs the following into the system: "I've been feeling stressed lately because of the heavy workload." In addition, the camera captures their facial expressions and the microphone records their voice. The system analyzes this data and generates a response for this situation.

[1038] In this way, the present invention provides an environment in which workers can easily seek advice about mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures.

[1039] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1040] Step 1:

[1041] Building the initial dataset

[1042] The server collects publicly available information and documents from within the company, including company policies, work procedures, and business manuals. This collected information is used to create an initial learning dataset. The server stores this dataset in a database and uses it for initial learning.

[1043] Input: Company disclosures and documents.

[1044] Output: Initial training dataset.

[1045] Step 2:

[1046] Enter basic information

[1047] The terminal provides a screen for workers to enter basic information, such as their name, department, and role. The user (worker) uses this screen to enter information and sends it to the server.

[1048] Input: Worker basic information (name, department, role).

[1049] Output: Basic information sent to the server.

[1050] Step 3:

[1051] Early Learning

[1052] The server uses the received basic information to perform initial training of the generative model using an initial training dataset. During this process, the server prepares a generative model tailored to the worker's needs. It uses Microsoft Azure's Face API and Speech-to-Text API.

[1053] Input: Initial training dataset, basic information.

[1054] Output: The initial trained generative model.

[1055] Step 4:

[1056] Daily data collection

[1057] The terminal (factory robot) collects data on the worker's daily conversations and work. The robot uses a camera and microphone to capture the worker's facial expressions and voice in real time. The collected data is sent to a server.

[1058] Input: Data on workers' daily conversations and work (voice, facial expressions).

[1059] Output: Daily data sent to the server.

[1060] Step 5:

[1061] Emotional and stress level assessment

[1062] The server analyzes the daily data sent to it and assesses emotions and stress levels. It uses Microsoft Azure's Face API and Speech-to-Text API to analyze voice and facial expressions. Text data is analyzed using natural language processing techniques.

[1063] Input: Everyday data (voice, facial expressions, text).

[1064] Output: Emotion and stress level assessment results.

[1065] Step 6:

[1066] Generate alerts and advice

[1067] The server generates appropriate alerts and advice based on the evaluation results. For example, if high stress is detected, it will generate feedback such as "Take a short break."

[1068] Input: Emotion and stress level assessment results.

[1069] Output: Alerts and advice.

[1070] Step 7:

[1071] Alerts and advice notifications

[1072] The terminal (factory robot) notifies the worker of the generated alerts and advice, for example, the robot will say in a voice message, "Take a short break."

[1073] Input: Alerts and advice.

[1074] Output: Notice to workers.

[1075] Step 8:

[1076] Continuous data analysis and re-learning

[1077] The server periodically analyzes the data collected from all workers and retrains the generative model, using past data stored in a database to improve the model's accuracy.

[1078] Input: Collected worker data.

[1079] Output: The retrained generative model.

[1080] Examples of concrete examples and prompts

[1081] For example, consider a situation where a factory worker inputs into the system, "I've been working too hard lately and I'm feeling stressed." At this time, the robot's camera captures the worker's facial expressions and its microphone records his / her voice. The worker's emotions and stress level are analyzed, and if high stress is detected, the server generates feedback such as "Take a short break," and the robot notifies the worker.

[1082] Example prompt sentence:

[1083] A worker inputs the following into the system: "I've been feeling stressed lately because of the heavy workload." In addition, the camera captures their facial expressions and the microphone records their voice. The system analyzes this data and generates a response for this situation.

[1084] 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.

[1085] 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.

[1086] 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.

[1087] [Third embodiment]

[1088] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1089] 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.

[1090] 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).

[1091] 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.

[1092] 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.

[1093] 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).

[1094] 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.

[1095] 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.

[1096] 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.

[1097] 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.

[1098] 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.

[1099] 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."

[1100] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system consists of three main components: a server, a terminal, and a user.

[1101] System Overview

[1102] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[1103] Terminal: A device used by an employee to input information or receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs.

[1104] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through terminals.

[1105] System Operation

[1106] Initial setup and learning phase

[1107] 1. Server: First, collect publicly available information and documents from within the company. This data is used to build an initial learning dataset, including company policies, business processes, and manuals.

[1108] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[1109] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[1110] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[1111] Daily monitoring and stress detection

[1112] 1. Devices: Collect data on employees' daily conversations and work. Methods include periodic surveys and chat-style conversation logs.

[1113] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[1114] 3. Server: Analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[1115] Mentoring and support

[1116] 1. Server: When high stress is detected, it automatically generates an alert and generates appropriate advice, such as "take a short break."

[1117] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[1118] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1119] Continuous evaluation and improvement

[1120] 1. Server: Periodically analyzes data collected from all users to evaluate the effectiveness of the system, for example by graphing fluctuations in stress levels across the entire workforce.

[1121] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[1122] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback.

[1123] Specific examples

[1124] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[1125] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert for Person A saying, "Take a short break," and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1126] This system is designed to provide an environment where employees can easily receive mental health consultations, detect stress early, and promptly implement appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[1127] The processing flow will be explained below.

[1128] Initial setup and learning phase

[1129] Step 1:

[1130] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[1131] Step 2:

[1132] The server creates an initial learning dataset based on the collected data, which prepares the data to be used for initial learning.

[1133] Step 3:

[1134] The terminal provides employees with a screen for entering basic information, and the input form is displayed on a smartphone or PC application.

[1135] Step 4:

[1136] The user enters basic information, for example, "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and sends it from the terminal to the server.

[1137] Step 5:

[1138] The server initially trains the generative model based on the basic information received, thereby preparing a model that corresponds to the company's specific context and needs.

[1139] Daily monitoring and stress detection

[1140] Step 1:

[1141] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style conversation logs.

[1142] Step 2:

[1143] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[1144] Step 3:

[1145] The server analyzes the data sent from the device and uses natural language processing to evaluate emotions and stress levels.

[1146] Step 4:

[1147] The server calculates the stress level based on the analysis results, and generates an alert if signs of high stress are detected.

[1148] Mentoring and support

[1149] Step 1:

[1150] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[1151] Step 2:

[1152] Notify the user of any alerts or advice generated by the device, such as a pop-up notification on the smartphone with the message "Try taking some deep breaths to relax."

[1153] Step 3:

[1154] The user receives a notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1155] Continuous evaluation and improvement

[1156] Step 1:

[1157] The server periodically analyzes the data collected from all users, for example, to graph fluctuations in stress levels across the entire workforce.

[1158] Step 2:

[1159] The device will notify the user of new advice and improvements, for example suggesting new training programs or relaxation techniques.

[1160] Step 3:

[1161] The server retrains the generative model based on the data analysis results, allowing it to provide more effective feedback.

[1162] Through these specific processing steps, a system will be created that continuously supports the mental health of employees and contributes to improving the performance of the entire organization.

[1163] Example 1

[1164] 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."

[1165] Maintaining and improving employee mental health is an important issue in today's workplace. However, many companies lack systems for constantly monitoring employee mental health and providing appropriate support. It is also difficult to detect stress early or provide advice tailored to individual employees, making it difficult to take measures before employee problems become serious.

[1166] 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.

[1167] In this invention, the server includes: means for collecting public information and documents from within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data on employees' daily conversations and work; means for analyzing the collected data and evaluating emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting feedback from employees and using the data for re-learning; means for periodically analyzing data collected from all employees and evaluating the effectiveness of the system; and means for re-learning the generative model based on the evaluation results and generating more effective feedback. This makes it possible to monitor employees' mental health in real time, detect stress early, and provide appropriate support and advice.

[1168] "Public information" refers to information obtained from inside or outside the company, such as company policies, business processes, and manuals.

[1169] "Documents" refers to text data related to specific content, such as company records, manuals, and guides.

[1170] An "initial training dataset" is a dataset constructed from publicly available company information and documents, and refers to a collection of data used to train a generative model.

[1171] "Terminal" refers to the device used by employees to enter basic information and inquiries, including smartphones, tablets, and PCs.

[1172] A "generative model" is an algorithm that learns from data obtained from employees and generates appropriate advice and alerts.

[1173] "Daily conversation and work-related data" refers to text data and voice data generated during employees' daily discussions and work activities.

[1174] "Emotion and stress levels" refers to the psychological state of employees, assessed by analyzing data on their daily conversations and work.

[1175] An "alert" refers to a notification that is generated when an employee's stress level exceeds a certain threshold.

[1176] "Advice" refers to specific suggestions for improving employees' mental health and reducing stress.

[1177] "Feedback" refers to information provided to employees about the advice and results of their actions, and refers to data that is re-entered into the system.

[1178] "Retraining" refers to the process of retraining a generative model using feedback or newly collected data to improve its accuracy.

[1179] "Effectiveness" refers to an indicator that shows how effective the system is in improving employees' mental health.

[1180] This invention relates to an AI-based mentoring system for supporting employee mental health. The system collects publicly available information and documents from within a company, constructs an initial training dataset, and trains a generative model based on the dataset to analyze employees' stress levels and provide appropriate alerts and advice. Feedback from employees is used for retraining, allowing the model to be continuously improved and the effectiveness of the system to be evaluated.

[1181] Server: Initial setup and learning phase

[1182] The server first collects publicly available information and documents from within the company. Specifically, it uses Python scripts to retrieve information such as policies, business processes, and manuals from the company's internal database and file storage, and builds an initial training dataset based on this data. Next, it receives basic employee information and initially trains a generative model specific to the company's context. This process uses Python and TensorFlow to input data into an NLP model (such as the BERT model) to create a company-specific language model.

[1183] Terminal: Data collection and feedback

[1184] The terminal is a device used by employees to input basic information and daily inquiries. To this end, it is designed as a web application using front-end frameworks such as React or Vue.js. The terminal provides an interface for collecting data on employees' daily conversations and work. This includes periodic surveys and interactions with chatbots, which are developed using Dialogflow or the Microsoft Bot Framework. Employee input is provided, for example, through a smartphone application.

[1185] User: System usage

[1186] Employees, who are users, input basic information and everyday concerns through their terminals. For example, an employee may input basic information such as "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and then, during work hours, input concerns such as "Work has been tough and I'm feeling tired lately." This data is sent to the server and used for analysis.

[1187] Server: Data analysis and advice generation

[1188] The server analyzes the data sent from the device and evaluates the employee's emotions and stress levels. Specifically, it performs natural language processing using Python's NLTK library and spaCy to extract keywords from the text data and score the stress level. If an increase in stress level is detected, a generative AI model (e.g., GPT-3) is used to generate specific advice, such as "Take a short break."

[1189] Device: Notifications and feedback collection

[1190] The generated alerts and advice are notified to the user by the device. For example, a message such as "Take a short break" is displayed using the smartphone's push notification function. The user receives the notification, acts according to the instructions, and then enters the results back into the device. For example, feedback is given such as "Took a 5-minute break."

[1191] Server: Continuous evaluation and model retraining

[1192] The server periodically analyzes data collected from all users and performs data analysis using Python's pandas library. Fluctuations in stress levels across employees are graphed using Matplotlib and Seaborn. The generative model is retrained based on the analysis results to provide more effective feedback. Accuracy is improved by retraining the NLP model using feedback data from employees.

[1193] Examples of specific examples and prompts

[1194] For example, when a new employee logs in to the system for the first time and enters basic information (name, department, position), they can then enter "I'm tired because I've already had a lot of work lately" into their device while they're working. The server analyzes this data, detects an increase in stress level, and generates advice to "take a short break" and notifies the device. If the new employee accepts the advice and takes a break, the feedback data is sent to the server and used for relearning.

[1195] Prompt Sentence Examples

[1196] "Tell me about your recent work. Example: Work has been hard and I'm feeling tired lately."

[1197] What's an effective way to relax?

[1198] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1199] Step 1:

[1200] Server: Collects public information and documents from within the company and builds an initial learning dataset. Specifically, it runs Python scripts to retrieve policies, business processes, manuals, etc. from the company's internal database and file storage, and then formats this into a dataset.

[1201] Input: Public information and documents (policies, business processes, manuals)

[1202] Output: Initial training dataset

[1203] Step 2:

[1204] Devices: Provide an interface for employees to enter basic information. Specifically, create a web form using React or Vue.js and make it available on devices such as smartphones, tablets, and PCs.

[1205] User: For example, a user (employee) enters "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[1206] Input: Employee basic information (name, department, job description)

[1207] Output: Send employee basic information to the server

[1208] Step 3:

[1209] Server: Receives basic employee information and performs initial training on the generative model. Specifically, it uses Python and TensorFlow to input data into an NLP model (such as the BERT model) and builds a company-specific language model.

[1210] Input: Basic employee information (name, department, job description), initial learning dataset

[1211] Output: Company-specific generative model

[1212] Step 4:

[1213] Terminal: Provides an interface for collecting data about employees' daily conversations and work, for example, through periodic surveys or chatbots. Chatbots are developed using Dialogflow or the Microsoft Bot Framework.

[1214] User: During work, input a question such as "Work has been tough and I'm tired lately."

[1215] Input: Daily conversation and business data (text messages)

[1216] Output: Sending daily conversation data to the server

[1217] Step 5:

[1218] Server: Analyzes the data sent from the device and evaluates emotions and stress levels. Specifically, it uses Python's NLTK library and spaCy to analyze text data, extract keywords, and score stress levels.

[1219] Input: Daily conversation and business data

[1220] Output: Emotion and stress level assessment results

[1221] Step 6:

[1222] Server: Based on the evaluation results, it generates appropriate alerts and advice. It uses a generative AI model (e.g., GPT-3) to create specific advice, such as "Take a short break."

[1223] Input: Emotion and stress level assessment results

[1224] Output: Alerts and advice

[1225] Step 7:

[1226] Devices: Notify employees of generated alerts and advice, using push notifications on their smartphones to display messages such as "Take a short break."

[1227] Input: Alerts and Advice

[1228] Output: Employee notification

[1229] Step 8:

[1230] User: Receives alerts and advice and acts accordingly, for example by entering feedback into the device such as "I took a 5-minute break."

[1231] Input: Employee feedback (e.g., took a break)

[1232] Output: Sending feedback to the server

[1233] Step 9:

[1234] Server: Receives employee feedback and uses that data for retraining. Specifically, the NLP model is retrained using the feedback data to improve accuracy. The retraining process is performed using Python and TensorFlow.

[1235] Input: Employee feedback

[1236] Output: Retrained generative model

[1237] Step 10:

[1238] Server: Regularly analyzes data collected from all employees to evaluate the effectiveness of the system. Python's pandas library is used to analyze the data, and Matplotlib and Seaborn are used to graph, for example, fluctuations in stress levels across all employees. Based on the evaluation results, new advice and improvement measures are implemented.

[1239] Input: Feedback data collected from all employees

[1240] Output: System effectiveness evaluation results, new advice and improvements

[1241] (Application example 1)

[1242] 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."

[1243] Maintaining and improving employee mental health is an important issue for companies. However, currently, there are limitations to early detection of employee stress and mental problems and appropriate response. In particular, in factory environments, employees often bear significant physical and mental burdens, requiring rapid and effective responses. Conventional methods rely heavily on self-reporting, making it difficult to grasp mental health conditions in real time or provide appropriate advice. For this reason, a system is needed that provides an environment where employees can easily seek mental health advice, detects stress and mental burden early, and enables appropriate measures to be taken promptly.

[1244] 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.

[1245] In this invention, the server includes means for converting what employees say into text data using a voice recognition system, means for evaluating emotions and stress levels based on the converted text data, means for using a generative AI model to generate appropriate alerts and advice based on the evaluation, means for notifying employees of the generated alerts and advice by voice or text, means for continuously collecting employee feedback and evaluating the effectiveness of the system, and means for retraining the model based on the evaluation results. This makes it possible to monitor employees' mental health in real time and provide appropriate advice and support, thereby reducing the mental burden on employees and providing a healthier work environment.

[1246] A "voice recognition system" is a technology for converting voice data into text data.

[1247] "Emotion and stress level assessment" is the process of analyzing an employee's emotional state and stress level from input data.

[1248] "Generative AI model" refers to an artificial intelligence model that automatically generates advice and alerts for employees based on collected data.

[1249] "Notification of Alerts and Advice" means the means by which an employee is notified of a generated alert or advice, which may be in the form of audio or text.

[1250] "Feedback gathering" is the process of continually gathering responses and reactions from employees.

[1251] "Evaluating system effectiveness" is the process of analyzing collected feedback data to verify the system's performance and effectiveness.

[1252] "Model retraining" is the process of retraining a generative AI model based on evaluation results to improve its accuracy and effectiveness.

[1253] "Daily conversation and work-related data" refers to records of the words and interactions that employees have in the course of their daily work.

[1254] "Early detection of mental health issues" is the process of detecting employee stress and mental health issues as early as possible.

[1255] "Providing appropriate countermeasures" is the means of providing the most appropriate action or assistance for a detected problem.

[1256] These definitions clarify the components of the overall system and their functions.

[1257] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system is composed of three main components: a server, a terminal, and a user.

[1258] System Overview

[1259] Server: A central data processing unit that collects public information and documents from within the company to build an initial learning dataset and trains the generative AI model. It also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[1260] Terminal: The device used by an employee to enter information or receive feedback from the server, including a smartphone, tablet, or PC.

[1261] User: Refers to the employees and managers who use the system. Users input basic information through terminals and carry out daily consultations.

[1262] System behavior and specific functions

[1263] Collecting basic information

[1264] The server first collects publicly available information and documents from the company to build an initial training dataset, including company policies, business processes, and manuals, which are then used to initially train a generative AI model specific to the company.

[1265] The terminal provides a screen where employees can enter basic information (such as name, department, job description, etc.) For example, they can enter information such as "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" on a smartphone application and send it to the server.

[1266] Users enter and submit basic information through a terminal, preparing the system to meet the company's specific needs.

[1267] Daily monitoring and stress detection

[1268] The devices collect data on employees' daily conversations and work activities. They use a voice recognition system to convert the conversations into text data, and collect data through periodic surveys and chat-style dialogue logs.

[1269] The user inputs into the terminal any worries or concerns they have about work. For example, they can send a message such as, "Work has been tough lately and I'm feeling tired."

[1270] The server analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[1271] Mentoring and support

[1272] The server automatically generates an alert when high stress is detected and creates appropriate advice, using a generative AI model to generate specific instructions such as "take a short break."

[1273] The device will then notify the user of any generated alerts or advice, such as a pop-up notification on the smartphone saying, "Try taking some deep breaths to relax."

[1274] The user receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1275] Continuous evaluation and improvement

[1276] The server periodically analyzes the data collected from all users to assess the effectiveness of the system, for example by graphing fluctuations in stress levels across the workforce.

[1277] The device will notify the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[1278] The server retrains the generative AI model based on the analysis results, enabling it to provide more effective feedback.

[1279] Specific examples

[1280] For example, let's consider the case of a new employee. When an employee logs in to the system for the first time, a screen appears where they can enter basic information (name, department, job description). After entering and submitting the information, the server receives it and completes the initial learning process.

[1281] Next, the employee types into the device, "I've already had a lot of work lately and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert to the employee, telling them to "take a short break," and notifies the device. If the employee accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1282] Example prompts for generative AI models

[1283] "If an employee is unhappy or stressed about a recent shift, generate advice that is appropriate to their emotions."

[1284] In this way, the present invention aims to provide an environment where employees can easily seek mental health advice, thereby reducing stress and mental burden and providing a healthier working environment.

[1285] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1286] Step 1:

[1287] The server collects public information and documents within the company.

[1288] Input: Company policies, business processes, manuals, etc.

[1289] Specific operation: The server runs a program that scans documents stored in a specific format and automatically generates a dataset.

[1290] Output: Initial training dataset.

[1291] Step 2:

[1292] The terminal provides a screen for the employee to enter basic information.

[1293] Input: Employee name, department, job description.

[1294] Specific operation: The terminal application displays an input form and sends the data entered by the employee to the server.

[1295] Output: Employee basic information data.

[1296] Step 3:

[1297] The server performs initial training on the generative AI model based on the received basic information.

[1298] Input: Employee basic information data, initial learning dataset.

[1299] How it works: The server uses basic information to train a generative AI model to understand the company's specific context.

[1300] Output: A trained generative AI model.

[1301] Step 4:

[1302] The devices collect data about employees' daily conversations and work.

[1303] Input: Employee voice data.

[1304] Specific operation: The voice data is converted into text data using a voice recognition system and sent to the server.

[1305] Output: Daily conversation text data.

[1306] Step 5:

[1307] The server analyzes the collected data and assesses emotions and stress levels.

[1308] Input: Everyday conversation text data.

[1309] How it works: The server uses a text analysis algorithm to extract keywords and perform sentiment analysis, and then evaluates the stress level based on the results.

[1310] Output: Emotion assessment results and stress level data.

[1311] Step 6:

[1312] The server generates appropriate alerts and advice based on the assessed stress level.

[1313] Input: Emotion assessment results and stress level data.

[1314] What it does: Uses a generative AI model to generate appropriate advice for stressed employees.

[1315] Output: Alerts and advice.

[1316] Step 7:

[1317] The device notifies the employee of any generated alerts or advice.

[1318] Input: Alerts and Advice.

[1319] Specific behavior: The device will display alerts and advice to employees via pop-up notifications and audio notifications.

[1320] Output: Notification to employee.

[1321] Step 8:

[1322] The user receives the notification and acts as instructed.

[1323] Input: Alerts and Advice.

[1324] Specific actions: The user acts according to the notified advice and inputs feedback into the terminal as necessary.

[1325] Output: Feedback data.

[1326] Step 9:

[1327] The server periodically analyzes the data collected from all users to evaluate the effectiveness of the system.

[1328] Input: Feedback data.

[1329] Specific operations: The server analyzes the feedback data using an analytical algorithm and evaluates its effectiveness.

[1330] Output: Effectiveness evaluation results.

[1331] Step 10:

[1332] The server retrains the generative AI model based on the evaluation results to provide more effective feedback.

[1333] Input: Effectiveness evaluation results, feedback data.

[1334] Specific operation: The server uses the evaluation results to retrain the generative AI model and improve the accuracy of the model.

[1335] Output: An improved generative AI model.

[1336] 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.

[1337] As an embodiment of the present invention, we will explain in detail an AI-based mentoring system for supporting employee mental health, which is combined with an emotion engine that recognizes user emotions. This system consists of three main components: a server, a terminal, and a user.

[1338] System Overview

[1339] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice. It also retrains the generative model based on the emotion data detected by the emotion engine.

[1340] Terminal: A device used by employees to input information and receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs. Terminals also use emotion engines to perform voice analysis, facial expression analysis, and text analysis.

[1341] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through their terminals. The terminals are equipped with an emotion engine that analyzes the user's emotions in real time.

[1342] System Operation

[1343] Initial setup and learning phase

[1344] 1. Server: Collects public information and documents from within the company to build an initial learning dataset, including company policies, business processes, and manuals.

[1345] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[1346] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[1347] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[1348] Daily monitoring and stress detection

[1349] 1. Device: Collect data on users' daily conversations and work. Methods for doing so include periodic surveys and chat-style conversation logs.

[1350] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[1351] 3. On the device: The emotion engine recognizes the user's emotions using voice analysis, facial expression analysis, and text analysis, thereby obtaining the user's emotion data in real time.

[1352] 4. Server: Analyzes the data sent from the device. It also analyzes data from the emotion engine to evaluate emotions and stress levels. For example, it determines whether stress levels are rising based on keywords like "tired" or "tough" or changes in facial expressions detected by the emotion engine.

[1353] Mentoring and support

[1354] 1. Server: If high stress is detected, the server generates appropriate alerts and advice for the user. For example, it generates specific instructions such as "take a short break." It can also adjust the advice based on emotional data from the emotion engine.

[1355] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[1356] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1357] Continuous evaluation and improvement

[1358] 1. Server: Periodically analyzes data collected from all users. For example, graphs are created to show fluctuations in the stress levels of all employees. Data from the emotion engine is also analyzed to provide a detailed assessment.

[1359] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[1360] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback. Further adaptively improves the model based on data from the emotion engine.

[1361] Specific examples

[1362] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[1363] Next, Person A types into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects an increase in stress level. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1364] In this way, this embodiment of the present invention provides an environment where employees can easily seek advice about their mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[1365] The processing flow will be explained below.

[1366] Processing flow of a system that combines emotion engines

[1367] Initial setup and learning phase

[1368] Step 1:

[1369] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[1370] Step 2:

[1371] The server builds an initial training dataset based on the collected information, which involves parsing and formatting the collected documents.

[1372] Step 3:

[1373] The device provides employees with a screen to enter basic information, displaying a form that can be accessed on a smartphone or PC.

[1374] Step 4:

[1375] The user enters basic information. For example, "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[1376] Step 5:

[1377] The server initially trains the generative model based on the basic information it receives, helping it understand the company's unique context and business processes.

[1378] Daily monitoring and stress detection

[1379] Step 1:

[1380] The device collects data about the user's daily conversations and work, obtaining information through conversation logs and periodic surveys.

[1381] Step 2:

[1382] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[1383] Step 3:

[1384] The emotion engine built into the device performs voice analysis, facial expression analysis, and text analysis, for example, recognizing emotions by analyzing the user's tone of voice, changes in facial expression, and input text.

[1385] Step 4:

[1386] The server analyzes the conversation data sent from the device and the output data of the emotion engine, which then evaluates emotions and stress levels.

[1387] Step 5:

[1388] The server calculates the stress level based on the analysis results. For example, it determines the stress level based on keywords such as "tired" and "tough" and emotion recognition results such as "sad" and "tired" from the emotion engine.

[1389] Mentoring and support

[1390] Step 1:

[1391] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[1392] Step 2:

[1393] The device notifies the user of generated alerts and advice, using the smartphone's notification function to display a pop-up message saying, "Try taking deep breaths to relax."

[1394] Step 3:

[1395] The user receives a notification and acts on the advice, for example by entering feedback such as "I took a 5-minute break" into the device again.

[1396] Continuous evaluation and improvement

[1397] Step 1:

[1398] The server periodically analyzes the data collected from all users, statistically analyzing stress levels and emotional data to assess their overall mental health.

[1399] Step 2:

[1400] The server retrains the generative model based on the analysis results, for example by adding new data to update the model and improve its accuracy.

[1401] Step 3:

[1402] The device notifies the user of new advice and improvement measures, and suggests new advice and training based on feedback from the results of the exercise.

[1403] Specific examples

[1404] For example, let us explain what happens when a new employee, Mr. A, uses the system.

[1405] Step 1:

[1406] When Mr. A logs in to the system for the first time, a form for entering basic information appears on the smartphone application.

[1407] Step 2:

[1408] Person A enters "Name: Yamada Taro, Department: Sales Department, Job Description: New customer development" and submits.

[1409] Step 3:

[1410] Based on the information received by the server, initial learning is performed to understand the company's specific context and business processes.

[1411] Step 4:

[1412] Person A types into the device, "Work has been tough lately and I'm tired." The device's emotion engine analyzes Person A's voice and facial expressions and detects the emotion "tired."

[1413] Step 5:

[1414] The server analyzes this data and evaluates the stress level. If high stress is detected, an alert is generated telling Person A to take a short break.

[1415] Step 6:

[1416] A pop-up notification appears on A's device with the message, "Try taking some deep breaths to relax." After A takes a break, A sends feedback saying, "You took a 5-minute break."

[1417] Step 7:

[1418] The server periodically analyzes the stress levels and emotional data of all users and retrains the model, improving the accuracy of the next advice.

[1419] In this way, a system that combines an emotion engine continuously supports users' mental health and contributes to improving the performance of the entire organization.

[1420] Example 2

[1421] 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."

[1422] Employee mental health issues have a significant impact on a company's productivity and employee satisfaction, but many existing systems have been inadequate in detecting stress early or providing appropriate support. In particular, there are few systems that can analyze employees' emotional states and stress levels in real time and provide individually customized advice and support. As a result, it has been difficult to detect mental health issues early and take measures.

[1423] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for collecting public information and documents within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data related to employees' daily conversations and work; means for analyzing the collected data using voice analysis, facial expression analysis, and text analysis to evaluate emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting employee feedback and continuously analyzing and evaluating the effectiveness of the system; and means for re-training the generative model based on the evaluation results. This makes it possible to analyze employees' emotional states and stress levels in real time and provide individually customized advice and support.

[1424] "Internal public information" means policies, procedures, manuals, and other documents that are accessible within the company to employees and affiliated organizations.

[1425] "Documents" refer to documents and files used within a company, including data and information such as policies, business processes, and manuals.

[1426] An "initial training dataset" is a collection of data based on a company's policies and business procedures that is used to train a generative model from its initial state.

[1427] "Device" refers to the digital devices used by employees, including smartphones, tablets, and PCs.

[1428] "Generative models" refer to machine learning models created using artificial intelligence algorithms to analyze employees' emotions and stress levels and provide appropriate advice and alerts.

[1429] "Daily conversational and work-related data" refers to all conversational data, chat logs, and work-related information generated by employees in the course of their daily work.

[1430] "Voice analysis" refers to an analytical method for assessing employee emotions and stress levels from voice data.

[1431] "Facial expression analysis" refers to technology that analyzes employees' facial expressions from image or video data to assess their emotional state.

[1432] "Text analysis" refers to a technology that evaluates emotions and stress levels based on text data entered by employees.

[1433] "Assessing emotions and stress levels" means analyzing the collected data and providing a numerical or categorical representation of an employee's current emotional state and stress level.

[1434] An "alert" refers to a notification sent to employees to warn them or instruct them on what to do.

[1435] "Advice" refers to specific suggestions such as guidelines for action and relaxation methods aimed at improving employees' mental health.

[1436] "Feedback" refers to the reactions and information employees provide in response to alerts and advice.

[1437] "Retraining a generative model" means incorporating new data and feedback to update the model so that it can provide more accurate analysis and advice.

[1438] This invention is an AI-based mentoring system for supporting employee mental health. It starts by collecting public information and documents from within the company and building an initial learning dataset. The system consists of three main components: a server, a terminal, and a user.

[1439] server

[1440] The server is the center of the system and has the following main functions:

[1441] Data collection

[1442] The server collects public information and documents from within the company to build an initial learning dataset. This data includes company policies, business processes, manuals, etc. For example, the server obtains data in XML or CSV format from internal file servers and management tools (SharePoint, Confluence, etc.).

[1443] Early Learning

[1444] The server uses the received basic information to initially train a generative model specific to the company's context. This process involves training the generative AI model using a Python program using TensorFlow or PyTorch.

[1445] Terminal

[1446] The terminal provides the interface through which the employee interacts with the system.

[1447] Enter basic information

[1448] Provide a screen for employees to enter basic information (name, department, job description). For example, an input form for "name," "department," and "job description" is displayed on a smartphone application.

[1449] Data collection

[1450] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[1451] Emotion analysis

[1452] The device is equipped with an emotion engine that performs speech, facial expression, and text analysis. Speech data is analyzed using NLTK and DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER and TextBlob).

[1453] User

[1454] Users represent the actions employees take when using the system.

[1455] Input and Feedback

[1456] The user uses the device to input basic information and concerns or questions about daily work. For example, they input text such as "Work has been tough and I'm tired lately." The system also provides feedback in response to alerts and advice from the server and the device.

[1457] Mentoring and support

[1458] Alert Generation

[1459] The server analyzes the data sent from the device and evaluates emotions and stress levels in conjunction with the analysis results of the emotion engine. If high stress is detected, it generates appropriate alerts and advice for the user. For example, it uses Python code to generate instructions such as "Take a short break" using natural language generation tools (such as GPT-3).

[1460] Alert Notifications

[1461] The generated alerts and advice are sent to the user via the device, for example, a pop-up notification on the smartphone displays the message "Try taking deep breaths to relax."

[1462] Continuous evaluation and improvement

[1463] Data analysis and retraining

[1464] The server periodically analyzes data collected from all users to evaluate the effectiveness of the system. This includes graphing fluctuations in stress levels across employees and analyzing them using visualization tools (Tableau and Matplotlib). Based on the analysis results, the generative model is retrained. This incorporates new emotional data and feedback, updating the model to provide more accurate analysis and advice.

[1465] Specific examples

[1466] When new employee A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[1467] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects rising stress levels. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1468] Prompt Sentence Examples

[1469] Here are some examples of prompts for generative AI models:

[1470] Generate initial learning data based on publicly available information within the company and basic employee information to provide appropriate mental health support to newly hired employees. Based on the information entered by the user (Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development), provide examples of specific advice and support.

[1471]

[1472] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1473] Step 1: Data collection

[1474] server:

[1475] The initial learning dataset is constructed by collecting public information and documents from within the company. For example, the server acquires data in XML or CSV format from the company's file server or management tools (SharePoint, Confluence, etc.).

[1476] Input: Internal company public information and documents

[1477] Data processing: Format conversion and organization of collected data

[1478] Output: Initial training dataset

[1479] Step 2: Enter basic information

[1480] Device:

[1481] Provide a screen for employees to enter basic information (name, department, job description). For example, display an input form for "name," "department," and "job description" on a smartphone application.

[1482] Input: Employee basic information

[1483] Data processing: None

[1484] Output: Basic information entered

[1485] User:

[1486] Enter basic information using the terminal and send it to the server. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and press the send button.

[1487] Input: Name, Department, Job Description

[1488] Data processing: None

[1489] Output: Basic information sent

[1490] Step 3: Initial learning

[1491] server:

[1492] Based on the received basic information, a generative model specific to the company's context is initially trained. For example, a Python program can be used to train the generative AI model using TensorFlow or PyTorch.

[1493] Input: Initial training dataset and basic information

[1494] Data Computing: Training generative models (applying machine learning algorithms)

[1495] Output: Initially trained generative model

[1496] Step 4: Routine data collection

[1497] Device:

[1498] Collect data about users' daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[1499] Input: Daily conversation, business data

[1500] Data processing: logging and organization

[1501] Output: Daily data collected

[1502] Step 5: Sentiment Analysis

[1503] Device:

[1504] Using emotion engines, speech analysis, facial expression analysis, and text analysis are performed. For example, speech data is analyzed using NLTK or DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER or TextBlob).

[1505] Input: Collected daily data

[1506] Data Computing: Speech, facial expression, and text emotion analysis

[1507] Output: User emotion data

[1508] Step 6: Stress Assessment

[1509] server:

[1510] The data sent from the device is analyzed and combined with the analysis results of the emotion engine to evaluate emotions and stress levels. For example, stress levels are determined based on keywords such as "tired" or "tough" or detected emotions such as "sad."

[1511] Input: User's daily data and emotional data

[1512] Data calculation: Stress level assessment (analysis of keywords and emotional data)

[1513] Output: Stress level evaluation result

[1514] Step 7: Alert Generation

[1515] server:

[1516] If high stress is detected, appropriate alerts and advice are generated for the user, such as "Take a short break" using natural language generation tools (such as GPT-3) in Python code.

[1517] Input: Stress level assessment result

[1518] Data operations: generating alerts and advice (applying natural language generation tools)

[1519] Output: Alerts and Advice

[1520] Step 8: Alert Notifications

[1521] Device:

[1522] Notify the user of any generated alerts or advice, for example, by displaying a pop-up notification on their smartphone with the message "Try taking some deep breaths to relax."

[1523] Input: Generated alerts and advice

[1524] Data Processing: Message Notification Settings

[1525] Output: User notification

[1526] Step 9: Gather feedback

[1527] User:

[1528] Receive alerts and advice and act on them, for example, take a five-minute break and then enter that feedback back into the device.

[1529] Input: Alert responses and feedback

[1530] Data Processing: Entering and Sending Feedback

[1531] Output: Feedback sent

[1532] Step 10: Continuous data analysis and retraining

[1533] server:

[1534] The data collected from all users is periodically analyzed to evaluate the effectiveness of the system. Fluctuations in the stress levels of all employees are graphed and analyzed using visualization tools (Tableau and Matplotlib). The generative model is retrained based on the analysis results.

[1535] Input: Feedback and daily data collected from each user

[1536] Data Computing: Data Analysis and Generative Model Retraining

[1537] Output: An improved generative model

[1538] (Application example 2)

[1539] 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."

[1540] Mental health issues among employees, especially factory workers, can have a serious impact on productivity and safety. Conventional mentoring systems were unable to analyze workers' emotions or stress levels, making it difficult to respond in real time. Furthermore, due to a lack of means for providing feedback, workers have not received appropriate support. Therefore, there is a need for a system that can detect worker problems early and provide appropriate advice.

[1541] 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.

[1542] In this invention, the server includes means for collecting public information and documents from within the company and building an initial training dataset, terminal means for employees to input basic information, means for initial training a generative model based on the basic information, means for collecting data on workers' daily conversations and work, means for analyzing the collected data and evaluating emotions and stress levels, means including a factory robot that provides feedback in real time based on the emotion data, means for generating appropriate alerts and advice based on the evaluated stress levels, means for notifying the workers of the generated alerts and advice, means for continuously analyzing data and evaluating the effectiveness of the system, and means for re-training the model based on the evaluation results. This makes it possible to monitor workers' emotions and stress levels in real time and provide appropriate feedback and advice.

[1543] "Public information and documents within a company" refers to internal information and publicly available documents such as policies, business processes, and manuals managed by the company.

[1544] An "initial training dataset" is a collection of data based on collected internal company information and documents that is used to train the system's generative model.

[1545] "Terminal means" refers to devices that workers use to input information, including smartphones, tablets, and PCs.

[1546] A "generative model" is an algorithm that uses machine learning and deep learning to assess workers' emotions and stress levels.

[1547] "Daily conversation and work-related data" refers to information related to a worker's daily communications and work activities.

[1548] "Emotion and stress level assessment tools" are techniques and methods for analyzing collected data and determining a worker's emotional state and level of stress.

[1549] A "factory robot" is a device that interacts with workers in a factory and provides real-time feedback based on emotional data.

[1550] "Alert and advice generating means" means a method or tool for generating notifications or instructions to workers based on assessed emotions and stress levels.

[1551] "Means of notification" refers to the means by which generated alerts or advice are communicated to workers, including voice, text message, pop-up notification, etc.

[1552] "Means for continuous data analysis" refers to a method for periodically analyzing the data collected by the system and evaluating the effectiveness of the system.

[1553] "Model retraining" is the process by which the system updates its generative model based on new data to improve its predictive ability and accuracy.

[1554] As an embodiment of the present invention, an AI-based mentoring system for supporting the mental health of workers is described in detail below. This system consists of three main components: a server, a terminal, and a worker.

[1555] System Overview

[1556] Server: Publicly available information and documents from within the company are collected and used to build an initial training dataset. The server then analyzes workers' daily conversation data and work data to evaluate their emotions and stress levels. The generative model is retrained based on the emotion data detected by the emotion engine. The specific software used includes Microsoft Azure's Face API and Speech-to-Text API for emotion analysis. AWS EC2 is used as the data processing server, and AWS RDS is used as the database service.

[1557] Terminal: This refers to the device through which workers input information and receive feedback from the server. This device includes smartphones, tablets, and PCs. Factory robots are also used to collect data on workers' daily conversations and work. The robots are equipped with cameras and microphones, allowing for real-time analysis of facial expressions and voice.

[1558] User (Worker): A factory employee who uses the system. Workers enter basic information and receive daily consultations through a terminal. The terminal's emotion engine analyzes the worker's emotions in real time.

[1559] System Operation

[1560] The server collects public information and documents from within the company to build an initial training dataset. The generative model performs initial training based on this initial dataset. Next, workers enter basic information through their devices, which is sent to the server. The server trains the generative model based on the collected basic information and prepares a model tailored to the company's specific needs. Real-time sentiment analysis is incorporated into this process, enabling more accurate data collection and feedback.

[1561] The terminal (factory robot) collects data on the worker's daily conversations and work. If a worker types into the system, "I've been feeling stressed lately because of the heavy workload," the robot's camera and microphone capture the worker's facial expressions and record their voice. This data is sent to a server for facial expression, voice, and text analysis.

[1562] Continuous evaluation and improvement

[1563] The server periodically analyzes data collected from all workers to assess fluctuations in stress levels across the workforce. It continuously evaluates data from the emotion engine and retrains the generative model based on the results, improving the accuracy and effectiveness of the generated feedback.

[1564] Specific examples

[1565] For example, consider a situation where a factory worker inputs into the system, "I've been working too hard lately and I'm feeling stressed." At this time, the robot's camera captures the worker's facial expressions and its microphone records his / her voice. The worker's emotions and stress level are analyzed, and if high stress is detected, the server generates feedback such as "Take a short break," and the robot notifies the worker.

[1566] Prompt Sentence Examples

[1567] A worker inputs the following into the system: "I've been feeling stressed lately because of the heavy workload." In addition, the camera captures their facial expressions and the microphone records their voice. The system analyzes this data and generates a response for this situation.

[1568] In this way, the present invention provides an environment in which workers can easily seek advice about mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures.

[1569] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1570] Step 1:

[1571] Building the initial dataset

[1572] The server collects publicly available information and documents from within the company, including company policies, work procedures, and business manuals. This collected information is used to create an initial learning dataset. The server stores this dataset in a database and uses it for initial learning.

[1573] Input: Company disclosures and documents.

[1574] Output: Initial training dataset.

[1575] Step 2:

[1576] Enter basic information

[1577] The terminal provides a screen for workers to enter basic information, such as their name, department, and role. The user (worker) uses this screen to enter information and sends it to the server.

[1578] Input: Worker basic information (name, department, role).

[1579] Output: Basic information sent to the server.

[1580] Step 3:

[1581] Early Learning

[1582] The server uses the received basic information to perform initial training of the generative model using an initial training dataset. During this process, the server prepares a generative model tailored to the worker's needs. It uses Microsoft Azure's Face API and Speech-to-Text API.

[1583] Input: Initial training dataset, basic information.

[1584] Output: The initial trained generative model.

[1585] Step 4:

[1586] Daily data collection

[1587] The terminal (factory robot) collects data on the worker's daily conversations and work. The robot uses a camera and microphone to capture the worker's facial expressions and voice in real time. The collected data is sent to a server.

[1588] Input: Data on workers' daily conversations and work (voice, facial expressions).

[1589] Output: Daily data sent to the server.

[1590] Step 5:

[1591] Emotional and stress level assessment

[1592] The server analyzes the daily data sent to it and assesses emotions and stress levels. It uses Microsoft Azure's Face API and Speech-to-Text API to analyze voice and facial expressions. Text data is analyzed using natural language processing techniques.

[1593] Input: Everyday data (voice, facial expressions, text).

[1594] Output: Emotion and stress level assessment results.

[1595] Step 6:

[1596] Generate alerts and advice

[1597] The server generates appropriate alerts and advice based on the evaluation results. For example, if high stress is detected, it will generate feedback such as "Take a short break."

[1598] Input: Emotion and stress level assessment results.

[1599] Output: Alerts and advice.

[1600] Step 7:

[1601] Alerts and advice notifications

[1602] The terminal (factory robot) notifies the worker of the generated alerts and advice, for example, the robot will say in a voice message, "Take a short break."

[1603] Input: Alerts and advice.

[1604] Output: Notice to workers.

[1605] Step 8:

[1606] Continuous data analysis and re-learning

[1607] The server periodically analyzes the data collected from all workers and retrains the generative model, using past data stored in a database to improve the model's accuracy.

[1608] Input: Collected worker data.

[1609] Output: The retrained generative model.

[1610] Examples of concrete examples and prompts

[1611] For example, consider a situation where a factory worker inputs into the system, "I've been working too hard lately and I'm feeling stressed." At this time, the robot's camera captures the worker's facial expressions and its microphone records his / her voice. The worker's emotions and stress level are analyzed, and if high stress is detected, the server generates feedback such as "Take a short break," and the robot notifies the worker.

[1612] Example prompt sentence:

[1613] A worker inputs the following into the system: "I've been feeling stressed lately because of the heavy workload." In addition, the camera captures their facial expressions and the microphone records their voice. The system analyzes this data and generates a response for this situation.

[1614] 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.

[1615] 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.

[1616] 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.

[1617] [Fourth embodiment]

[1618] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1619] 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.

[1620] 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).

[1621] 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.

[1622] 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.

[1623] 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).

[1624] 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.

[1625] 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.

[1626] 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.

[1627] 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.

[1628] 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.

[1629] 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.

[1630] 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."

[1631] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system consists of three main components: a server, a terminal, and a user.

[1632] System Overview

[1633] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[1634] Terminal: A device used by an employee to input information or receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs.

[1635] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through terminals.

[1636] System Operation

[1637] Initial setup and learning phase

[1638] 1. Server: First, collect publicly available information and documents from within the company. This data is used to build an initial learning dataset, including company policies, business processes, and manuals.

[1639] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[1640] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[1641] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[1642] Daily monitoring and stress detection

[1643] 1. Devices: Collect data on employees' daily conversations and work. Methods include periodic surveys and chat-style conversation logs.

[1644] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[1645] 3. Server: Analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[1646] Mentoring and support

[1647] 1. Server: When high stress is detected, it automatically generates an alert and generates appropriate advice, such as "take a short break."

[1648] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[1649] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1650] Continuous evaluation and improvement

[1651] 1. Server: Periodically analyzes data collected from all users to evaluate the effectiveness of the system, for example by graphing fluctuations in stress levels across the entire workforce.

[1652] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[1653] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback.

[1654] Specific examples

[1655] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[1656] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert for Person A saying, "Take a short break," and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1657] This system is designed to provide an environment where employees can easily receive mental health consultations, detect stress early, and promptly implement appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[1658] The processing flow will be explained below.

[1659] Initial setup and learning phase

[1660] Step 1:

[1661] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[1662] Step 2:

[1663] The server creates an initial learning dataset based on the collected data, which prepares the data to be used for initial learning.

[1664] Step 3:

[1665] The terminal provides employees with a screen for entering basic information, and the input form is displayed on a smartphone or PC application.

[1666] Step 4:

[1667] The user enters basic information, for example, "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and sends it from the terminal to the server.

[1668] Step 5:

[1669] The server initially trains the generative model based on the basic information received, thereby preparing a model that corresponds to the company's specific context and needs.

[1670] Daily monitoring and stress detection

[1671] Step 1:

[1672] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style conversation logs.

[1673] Step 2:

[1674] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[1675] Step 3:

[1676] The server analyzes the data sent from the device and uses natural language processing to evaluate emotions and stress levels.

[1677] Step 4:

[1678] The server calculates the stress level based on the analysis results, and generates an alert if signs of high stress are detected.

[1679] Mentoring and support

[1680] Step 1:

[1681] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[1682] Step 2:

[1683] Notify the user of any alerts or advice generated by the device, such as a pop-up notification on the smartphone with the message "Try taking some deep breaths to relax."

[1684] Step 3:

[1685] The user receives a notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1686] Continuous evaluation and improvement

[1687] Step 1:

[1688] The server periodically analyzes the data collected from all users, for example, to graph fluctuations in stress levels across the entire workforce.

[1689] Step 2:

[1690] The device will notify the user of new advice and improvements, for example suggesting new training programs or relaxation techniques.

[1691] Step 3:

[1692] The server retrains the generative model based on the data analysis results, allowing it to provide more effective feedback.

[1693] Through these specific processing steps, a system will be created that continuously supports the mental health of employees and contributes to improving the performance of the entire organization.

[1694] Example 1

[1695] 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."

[1696] Maintaining and improving employee mental health is an important issue in today's workplace. However, many companies lack systems for constantly monitoring employee mental health and providing appropriate support. It is also difficult to detect stress early or provide advice tailored to individual employees, making it difficult to take measures before employee problems become serious.

[1697] 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.

[1698] In this invention, the server includes: means for collecting public information and documents from within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data on employees' daily conversations and work; means for analyzing the collected data and evaluating emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting feedback from employees and using the data for re-learning; means for periodically analyzing data collected from all employees and evaluating the effectiveness of the system; and means for re-learning the generative model based on the evaluation results and generating more effective feedback. This makes it possible to monitor employees' mental health in real time, detect stress early, and provide appropriate support and advice.

[1699] "Public information" refers to information obtained from inside or outside the company, such as company policies, business processes, and manuals.

[1700] "Documents" refers to text data related to specific content, such as company records, manuals, and guides.

[1701] An "initial training dataset" is a dataset constructed from publicly available company information and documents, and refers to a collection of data used to train a generative model.

[1702] "Terminal" refers to the device used by employees to enter basic information and inquiries, including smartphones, tablets, and PCs.

[1703] A "generative model" is an algorithm that learns from data obtained from employees and generates appropriate advice and alerts.

[1704] "Daily conversation and work-related data" refers to text data and voice data generated during employees' daily discussions and work activities.

[1705] "Emotion and stress levels" refers to the psychological state of employees, assessed by analyzing data on their daily conversations and work.

[1706] An "alert" refers to a notification that is generated when an employee's stress level exceeds a certain threshold.

[1707] "Advice" refers to specific suggestions for improving employees' mental health and reducing stress.

[1708] "Feedback" refers to information provided to employees about the advice and results of their actions, and refers to data that is re-entered into the system.

[1709] "Retraining" refers to the process of retraining a generative model using feedback or newly collected data to improve its accuracy.

[1710] "Effectiveness" refers to an indicator that shows how effective the system is in improving employees' mental health.

[1711] This invention relates to an AI-based mentoring system for supporting employee mental health. The system collects publicly available information and documents from within a company, constructs an initial training dataset, and trains a generative model based on the dataset to analyze employees' stress levels and provide appropriate alerts and advice. Feedback from employees is used for retraining, allowing the model to be continuously improved and the effectiveness of the system to be evaluated.

[1712] Server: Initial setup and learning phase

[1713] The server first collects publicly available information and documents from within the company. Specifically, it uses Python scripts to retrieve information such as policies, business processes, and manuals from the company's internal database and file storage, and builds an initial training dataset based on this data. Next, it receives basic employee information and initially trains a generative model specific to the company's context. This process uses Python and TensorFlow to input data into an NLP model (such as the BERT model) to create a company-specific language model.

[1714] Terminal: Data collection and feedback

[1715] The terminal is a device used by employees to input basic information and daily inquiries. To this end, it is designed as a web application using front-end frameworks such as React or Vue.js. The terminal provides an interface for collecting data on employees' daily conversations and work. This includes periodic surveys and interactions with chatbots, which are developed using Dialogflow or the Microsoft Bot Framework. Employee input is provided, for example, through a smartphone application.

[1716] User: System usage

[1717] Employees, who are users, input basic information and everyday concerns through their terminals. For example, an employee may input basic information such as "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development," and then, during work hours, input concerns such as "Work has been tough and I'm feeling tired lately." This data is sent to the server and used for analysis.

[1718] Server: Data analysis and advice generation

[1719] The server analyzes the data sent from the device and evaluates the employee's emotions and stress levels. Specifically, it performs natural language processing using Python's NLTK library and spaCy to extract keywords from the text data and score the stress level. If an increase in stress level is detected, a generative AI model (e.g., GPT-3) is used to generate specific advice, such as "Take a short break."

[1720] Device: Notifications and feedback collection

[1721] The generated alerts and advice are notified to the user by the device. For example, a message such as "Take a short break" is displayed using the smartphone's push notification function. The user receives the notification, acts according to the instructions, and then enters the results back into the device. For example, feedback is given such as "Took a 5-minute break."

[1722] Server: Continuous evaluation and model retraining

[1723] The server periodically analyzes data collected from all users and performs data analysis using Python's pandas library. Fluctuations in stress levels across employees are graphed using Matplotlib and Seaborn. The generative model is retrained based on the analysis results to provide more effective feedback. Accuracy is improved by retraining the NLP model using feedback data from employees.

[1724] Examples of specific examples and prompts

[1725] For example, when a new employee logs in to the system for the first time and enters basic information (name, department, position), they can then enter "I'm tired because I've already had a lot of work lately" into their device while they're working. The server analyzes this data, detects an increase in stress level, and generates advice to "take a short break" and notifies the device. If the new employee accepts the advice and takes a break, the feedback data is sent to the server and used for relearning.

[1726] Prompt Sentence Examples

[1727] "Tell me about your recent work. Example: Work has been hard and I'm feeling tired lately."

[1728] What's an effective way to relax?

[1729] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1730] Step 1:

[1731] Server: Collects public information and documents from within the company and builds an initial learning dataset. Specifically, it runs Python scripts to retrieve policies, business processes, manuals, etc. from the company's internal database and file storage, and then formats this into a dataset.

[1732] Input: Public information and documents (policies, business processes, manuals)

[1733] Output: Initial training dataset

[1734] Step 2:

[1735] Devices: Provide an interface for employees to enter basic information. Specifically, create a web form using React or Vue.js and make it available on devices such as smartphones, tablets, and PCs.

[1736] User: For example, a user (employee) enters "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[1737] Input: Employee basic information (name, department, job description)

[1738] Output: Send employee basic information to the server

[1739] Step 3:

[1740] Server: Receives basic employee information and performs initial training on the generative model. Specifically, it uses Python and TensorFlow to input data into an NLP model (such as the BERT model) and builds a company-specific language model.

[1741] Input: Basic employee information (name, department, job description), initial learning dataset

[1742] Output: Company-specific generative model

[1743] Step 4:

[1744] Terminal: Provides an interface for collecting data about employees' daily conversations and work, for example, through periodic surveys or chatbots. Chatbots are developed using Dialogflow or the Microsoft Bot Framework.

[1745] User: During work, input a question such as "Work has been tough and I'm tired lately."

[1746] Input: Daily conversation and business data (text messages)

[1747] Output: Sending daily conversation data to the server

[1748] Step 5:

[1749] Server: Analyzes the data sent from the device and evaluates emotions and stress levels. Specifically, it uses Python's NLTK library and spaCy to analyze text data, extract keywords, and score stress levels.

[1750] Input: Daily conversation and business data

[1751] Output: Emotion and stress level assessment results

[1752] Step 6:

[1753] Server: Based on the evaluation results, it generates appropriate alerts and advice. It uses a generative AI model (e.g., GPT-3) to create specific advice, such as "Take a short break."

[1754] Input: Emotion and stress level assessment results

[1755] Output: Alerts and advice

[1756] Step 7:

[1757] Devices: Notify employees of generated alerts and advice, using push notifications on their smartphones to display messages such as "Take a short break."

[1758] Input: Alerts and Advice

[1759] Output: Employee notification

[1760] Step 8:

[1761] User: Receives alerts and advice and acts accordingly, for example by entering feedback into the device such as "I took a 5-minute break."

[1762] Input: Employee feedback (e.g., took a break)

[1763] Output: Sending feedback to the server

[1764] Step 9:

[1765] Server: Receives employee feedback and uses that data for retraining. Specifically, the NLP model is retrained using the feedback data to improve accuracy. The retraining process is performed using Python and TensorFlow.

[1766] Input: Employee feedback

[1767] Output: Retrained generative model

[1768] Step 10:

[1769] Server: Regularly analyzes data collected from all employees to evaluate the effectiveness of the system. Python's pandas library is used to analyze the data, and Matplotlib and Seaborn are used to graph, for example, fluctuations in stress levels across all employees. Based on the evaluation results, new advice and improvement measures are implemented.

[1770] Input: Feedback data collected from all employees

[1771] Output: System effectiveness evaluation results, new advice and improvements

[1772] (Application example 1)

[1773] 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."

[1774] Maintaining and improving employee mental health is an important issue for companies. However, currently, there are limitations to early detection of employee stress and mental problems and appropriate response. In particular, in factory environments, employees often bear significant physical and mental burdens, requiring rapid and effective responses. Conventional methods rely heavily on self-reporting, making it difficult to grasp mental health conditions in real time or provide appropriate advice. For this reason, a system is needed that provides an environment where employees can easily seek mental health advice, detects stress and mental burden early, and enables appropriate measures to be taken promptly.

[1775] 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.

[1776] In this invention, the server includes means for converting what employees say into text data using a voice recognition system, means for evaluating emotions and stress levels based on the converted text data, means for using a generative AI model to generate appropriate alerts and advice based on the evaluation, means for notifying employees of the generated alerts and advice by voice or text, means for continuously collecting employee feedback and evaluating the effectiveness of the system, and means for retraining the model based on the evaluation results. This makes it possible to monitor employees' mental health in real time and provide appropriate advice and support, thereby reducing the mental burden on employees and providing a healthier work environment.

[1777] A "voice recognition system" is a technology for converting voice data into text data.

[1778] "Emotion and stress level assessment" is the process of analyzing an employee's emotional state and stress level from input data.

[1779] "Generative AI model" refers to an artificial intelligence model that automatically generates advice and alerts for employees based on collected data.

[1780] "Notification of Alerts and Advice" means the means by which an employee is notified of a generated alert or advice, which may be in the form of audio or text.

[1781] "Feedback gathering" is the process of continually gathering responses and reactions from employees.

[1782] "Evaluating system effectiveness" is the process of analyzing collected feedback data to verify the system's performance and effectiveness.

[1783] "Model retraining" is the process of retraining a generative AI model based on evaluation results to improve its accuracy and effectiveness.

[1784] "Daily conversation and work-related data" refers to records of the words and interactions that employees have in the course of their daily work.

[1785] "Early detection of mental health issues" is the process of detecting employee stress and mental health issues as early as possible.

[1786] "Providing appropriate countermeasures" is the means of providing the most appropriate action or assistance for a detected problem.

[1787] These definitions clarify the components of the overall system and their functions.

[1788] As an embodiment of the present invention, an artificial intelligence-based mentoring system for supporting the mental health of employees will be described. This system is composed of three main components: a server, a terminal, and a user.

[1789] System Overview

[1790] Server: A central data processing unit that collects public information and documents from within the company to build an initial learning dataset and trains the generative AI model. It also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice.

[1791] Terminal: The device used by an employee to enter information or receive feedback from the server, including a smartphone, tablet, or PC.

[1792] User: Refers to the employees and managers who use the system. Users input basic information through terminals and carry out daily consultations.

[1793] System behavior and specific functions

[1794] Collecting basic information

[1795] The server first collects publicly available information and documents from the company to build an initial training dataset, including company policies, business processes, and manuals, which are then used to initially train a generative AI model specific to the company.

[1796] The terminal provides a screen where employees can enter basic information (such as name, department, job description, etc.) For example, they can enter information such as "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" on a smartphone application and send it to the server.

[1797] Users enter and submit basic information through a terminal, preparing the system to meet the company's specific needs.

[1798] Daily monitoring and stress detection

[1799] The devices collect data on employees' daily conversations and work activities. They use a voice recognition system to convert the conversations into text data, and collect data through periodic surveys and chat-style dialogue logs.

[1800] The user inputs into the terminal any worries or concerns they have about work. For example, they can send a message such as, "Work has been tough lately and I'm feeling tired."

[1801] The server analyzes the data sent from the device and evaluates emotions and stress levels. For example, it detects keywords such as "tired" and "tough" and determines whether stress levels are rising.

[1802] Mentoring and support

[1803] The server automatically generates an alert when high stress is detected and creates appropriate advice, using a generative AI model to generate specific instructions such as "take a short break."

[1804] The device will then notify the user of any generated alerts or advice, such as a pop-up notification on the smartphone saying, "Try taking some deep breaths to relax."

[1805] The user receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1806] Continuous evaluation and improvement

[1807] The server periodically analyzes the data collected from all users to assess the effectiveness of the system, for example by graphing fluctuations in stress levels across the workforce.

[1808] The device will notify the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[1809] The server retrains the generative AI model based on the analysis results, enabling it to provide more effective feedback.

[1810] Specific examples

[1811] For example, let's consider the case of a new employee. When an employee logs in to the system for the first time, a screen appears where they can enter basic information (name, department, job description). After entering and submitting the information, the server receives it and completes the initial learning process.

[1812] Next, the employee types into the device, "I've already had a lot of work lately and I'm tired." The server analyzes this information and detects an increase in stress level. The server generates an alert to the employee, telling them to "take a short break," and notifies the device. If the employee accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1813] Example prompts for generative AI models

[1814] "If an employee is unhappy or stressed about a recent shift, generate advice that is appropriate to their emotions."

[1815] In this way, the present invention aims to provide an environment where employees can easily seek mental health advice, thereby reducing stress and mental burden and providing a healthier working environment.

[1816] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1817] Step 1:

[1818] The server collects public information and documents within the company.

[1819] Input: Company policies, business processes, manuals, etc.

[1820] Specific operation: The server runs a program that scans documents stored in a specific format and automatically generates a dataset.

[1821] Output: Initial training dataset.

[1822] Step 2:

[1823] The terminal provides a screen for the employee to enter basic information.

[1824] Input: Employee name, department, job description.

[1825] Specific operation: The terminal application displays an input form and sends the data entered by the employee to the server.

[1826] Output: Employee basic information data.

[1827] Step 3:

[1828] The server performs initial training on the generative AI model based on the received basic information.

[1829] Input: Employee basic information data, initial learning dataset.

[1830] How it works: The server uses basic information to train a generative AI model to understand the company's specific context.

[1831] Output: A trained generative AI model.

[1832] Step 4:

[1833] The devices collect data about employees' daily conversations and work.

[1834] Input: Employee voice data.

[1835] Specific operation: The voice data is converted into text data using a voice recognition system and sent to the server.

[1836] Output: Daily conversation text data.

[1837] Step 5:

[1838] The server analyzes the collected data and assesses emotions and stress levels.

[1839] Input: Everyday conversation text data.

[1840] How it works: The server uses a text analysis algorithm to extract keywords and perform sentiment analysis, and then evaluates the stress level based on the results.

[1841] Output: Emotion assessment results and stress level data.

[1842] Step 6:

[1843] The server generates appropriate alerts and advice based on the assessed stress level.

[1844] Input: Emotion assessment results and stress level data.

[1845] What it does: Uses a generative AI model to generate appropriate advice for stressed employees.

[1846] Output: Alerts and advice.

[1847] Step 7:

[1848] The device notifies the employee of any generated alerts or advice.

[1849] Input: Alerts and Advice.

[1850] Specific behavior: The device will display alerts and advice to employees via pop-up notifications and audio notifications.

[1851] Output: Notification to employee.

[1852] Step 8:

[1853] The user receives the notification and acts as instructed.

[1854] Input: Alerts and Advice.

[1855] Specific actions: The user acts according to the notified advice and inputs feedback into the terminal as necessary.

[1856] Output: Feedback data.

[1857] Step 9:

[1858] The server periodically analyzes the data collected from all users to evaluate the effectiveness of the system.

[1859] Input: Feedback data.

[1860] Specific operations: The server analyzes the feedback data using an analytical algorithm and evaluates its effectiveness.

[1861] Output: Effectiveness evaluation results.

[1862] Step 10:

[1863] The server retrains the generative AI model based on the evaluation results to provide more effective feedback.

[1864] Input: Effectiveness evaluation results, feedback data.

[1865] Specific operation: The server uses the evaluation results to retrain the generative AI model and improve the accuracy of the model.

[1866] Output: An improved generative AI model.

[1867] 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.

[1868] As an embodiment of the present invention, we will explain in detail an AI-based mentoring system for supporting employee mental health, which is combined with an emotion engine that recognizes user emotions. This system consists of three main components: a server, a terminal, and a user.

[1869] System Overview

[1870] Server: A central system that collects public information and documents from within the company, builds an initial training dataset, and trains the generative model. The server also analyzes users' daily conversation data, evaluates their emotions and stress levels, and generates appropriate alerts and advice. It also retrains the generative model based on the emotion data detected by the emotion engine.

[1871] Terminal: A device used by employees to input information and receive feedback from the server. This includes digital devices such as smartphones, tablets, and PCs. Terminals also use emotion engines to perform voice analysis, facial expression analysis, and text analysis.

[1872] Users: Employees and managers who use the system. Users enter basic information and receive daily consultations through their terminals. The terminals are equipped with an emotion engine that analyzes the user's emotions in real time.

[1873] System Operation

[1874] Initial setup and learning phase

[1875] 1. Server: Collects public information and documents from within the company to build an initial learning dataset, including company policies, business processes, and manuals.

[1876] 2. Terminal: A screen is provided where employees can enter basic information (name, department, job description). For example, an input form is displayed on a smartphone application.

[1877] 3. User: Enter basic information through the terminal. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and send it to the server.

[1878] 4. Server: Based on the received basic information, the server initially trains a generative model specific to the company's context. This initial training prepares the server to respond to the company's specific needs.

[1879] Daily monitoring and stress detection

[1880] 1. Device: Collect data on users' daily conversations and work. Methods for doing so include periodic surveys and chat-style conversation logs.

[1881] 2. User: Enters into the terminal any worries or concerns that arise during daily work. For example, the user sends a message such as, "Work has been tough and I'm feeling tired lately."

[1882] 3. On the device: The emotion engine recognizes the user's emotions using voice analysis, facial expression analysis, and text analysis, thereby obtaining the user's emotion data in real time.

[1883] 4. Server: Analyzes the data sent from the device. It also analyzes data from the emotion engine to evaluate emotions and stress levels. For example, it determines whether stress levels are rising based on keywords like "tired" or "tough" or changes in facial expressions detected by the emotion engine.

[1884] Mentoring and support

[1885] 1. Server: If high stress is detected, the server generates appropriate alerts and advice for the user. For example, it generates specific instructions such as "take a short break." It can also adjust the advice based on emotional data from the emotion engine.

[1886] 2. Device: Notify the user of the generated alert or advice, for example, a pop-up notification on a smartphone saying, "Try taking some deep breaths to relax."

[1887] 3. User: Receives the notification and acts according to the instructions, for example, by entering feedback into the device again, such as "I took a 5-minute break."

[1888] Continuous evaluation and improvement

[1889] 1. Server: Periodically analyzes data collected from all users. For example, graphs are created to show fluctuations in the stress levels of all employees. Data from the emotion engine is also analyzed to provide a detailed assessment.

[1890] 2. Device: Notifying the user of new advice and improvements, for example, suggesting new training programs or relaxation techniques.

[1891] 3. Server: Retrains the generative model based on the analysis results to provide more effective feedback. Further adaptively improves the model based on data from the emotion engine.

[1892] Specific examples

[1893] For example, let's consider the case where new employee A uses the system. When A logs in to the system for the first time, a screen appears where he can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[1894] Next, Person A types into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects an increase in stress level. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1895] In this way, this embodiment of the present invention provides an environment where employees can easily seek advice about their mental health, and by combining it with an emotion engine, it is designed to enable early detection of stress and prompt implementation of appropriate measures. By providing customized mentoring for each employee, this system is expected to improve the mental health of the entire workplace.

[1896] The processing flow will be explained below.

[1897] Processing flow of a system that combines emotion engines

[1898] Initial setup and learning phase

[1899] Step 1:

[1900] The server collects public information and documents from within the company, such as company policies, business processes, and manuals.

[1901] Step 2:

[1902] The server builds an initial training dataset based on the collected information, which involves parsing and formatting the collected documents.

[1903] Step 3:

[1904] The device provides employees with a screen to enter basic information, displaying a form that can be accessed on a smartphone or PC.

[1905] Step 4:

[1906] The user enters basic information. For example, "Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development" and submits the form.

[1907] Step 5:

[1908] The server initially trains the generative model based on the basic information it receives, helping it understand the company's unique context and business processes.

[1909] Daily monitoring and stress detection

[1910] Step 1:

[1911] The device collects data about the user's daily conversations and work, obtaining information through conversation logs and periodic surveys.

[1912] Step 2:

[1913] Users input into the terminal any worries or concerns they may have during their daily work. For example, they could send a message such as, "Work has been tough and I'm feeling tired lately."

[1914] Step 3:

[1915] The emotion engine built into the device performs voice analysis, facial expression analysis, and text analysis, for example, recognizing emotions by analyzing the user's tone of voice, changes in facial expression, and input text.

[1916] Step 4:

[1917] The server analyzes the conversation data sent from the device and the output data of the emotion engine, which then evaluates emotions and stress levels.

[1918] Step 5:

[1919] The server calculates the stress level based on the analysis results. For example, it determines the stress level based on keywords such as "tired" and "tough" and emotion recognition results such as "sad" and "tired" from the emotion engine.

[1920] Mentoring and support

[1921] Step 1:

[1922] If the server detects high stress, it generates appropriate alerts and advice for the user, such as specific instructions like "take a short break."

[1923] Step 2:

[1924] The device notifies the user of generated alerts and advice, using the smartphone's notification function to display a pop-up message saying, "Try taking deep breaths to relax."

[1925] Step 3:

[1926] The user receives a notification and acts on the advice, for example by entering feedback such as "I took a 5-minute break" into the device again.

[1927] Continuous evaluation and improvement

[1928] Step 1:

[1929] The server periodically analyzes the data collected from all users, statistically analyzing stress levels and emotional data to assess their overall mental health.

[1930] Step 2:

[1931] The server retrains the generative model based on the analysis results, for example by adding new data to update the model and improve its accuracy.

[1932] Step 3:

[1933] The device notifies the user of new advice and improvement measures, and suggests new advice and training based on feedback from the results of the exercise.

[1934] Specific examples

[1935] For example, let us explain what happens when a new employee, Mr. A, uses the system.

[1936] Step 1:

[1937] When Mr. A logs in to the system for the first time, a form for entering basic information appears on the smartphone application.

[1938] Step 2:

[1939] Person A enters "Name: Yamada Taro, Department: Sales Department, Job Description: New customer development" and submits.

[1940] Step 3:

[1941] Based on the information received by the server, initial learning is performed to understand the company's specific context and business processes.

[1942] Step 4:

[1943] Person A types into the device, "Work has been tough lately and I'm tired." The device's emotion engine analyzes Person A's voice and facial expressions and detects the emotion "tired."

[1944] Step 5:

[1945] The server analyzes this data and evaluates the stress level. If high stress is detected, an alert is generated telling Person A to take a short break.

[1946] Step 6:

[1947] A pop-up notification appears on A's device with the message, "Try taking some deep breaths to relax." After A takes a break, A sends feedback saying, "You took a 5-minute break."

[1948] Step 7:

[1949] The server periodically analyzes the stress levels and emotional data of all users and retrains the model, improving the accuracy of the next advice.

[1950] In this way, a system that combines an emotion engine continuously supports users' mental health and contributes to improving the performance of the entire organization.

[1951] Example 2

[1952] 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."

[1953] Employee mental health issues have a significant impact on a company's productivity and employee satisfaction, but many existing systems have been inadequate in detecting stress early or providing appropriate support. In particular, there are few systems that can analyze employees' emotional states and stress levels in real time and provide individually customized advice and support. As a result, it has been difficult to detect mental health issues early and take measures.

[1954] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for collecting public information and documents within the company and constructing an initial training dataset; terminal means for employees to input basic information; means for initial training of a generative model based on the basic information; means for collecting data related to employees' daily conversations and work; means for analyzing the collected data using voice analysis, facial expression analysis, and text analysis to evaluate emotions and stress levels; means for generating appropriate alerts and advice based on the evaluated stress levels; means for notifying employees of the generated alerts and advice; means for collecting employee feedback and continuously analyzing and evaluating the effectiveness of the system; and means for re-training the generative model based on the evaluation results. This makes it possible to analyze employees' emotional states and stress levels in real time and provide individually customized advice and support.

[1955] "Internal public information" means policies, procedures, manuals, and other documents that are accessible within the company to employees and affiliated organizations.

[1956] "Documents" refer to documents and files used within a company, including data and information such as policies, business processes, and manuals.

[1957] An "initial training dataset" is a collection of data based on a company's policies and business procedures that is used to train a generative model from its initial state.

[1958] "Device" refers to the digital devices used by employees, including smartphones, tablets, and PCs.

[1959] "Generative models" refer to machine learning models created using artificial intelligence algorithms to analyze employees' emotions and stress levels and provide appropriate advice and alerts.

[1960] "Daily conversational and work-related data" refers to all conversational data, chat logs, and work-related information generated by employees in the course of their daily work.

[1961] "Voice analysis" refers to an analytical method for assessing employee emotions and stress levels from voice data.

[1962] "Facial expression analysis" refers to technology that analyzes employees' facial expressions from image or video data to assess their emotional state.

[1963] "Text analysis" refers to a technology that evaluates emotions and stress levels based on text data entered by employees.

[1964] "Assessing emotions and stress levels" means analyzing the collected data and providing a numerical or categorical representation of an employee's current emotional state and stress level.

[1965] An "alert" refers to a notification sent to employees to warn them or instruct them on what to do.

[1966] "Advice" refers to specific suggestions such as guidelines for action and relaxation methods aimed at improving employees' mental health.

[1967] "Feedback" refers to the reactions and information employees provide in response to alerts and advice.

[1968] "Retraining a generative model" means incorporating new data and feedback to update the model so that it can provide more accurate analysis and advice.

[1969] This invention is an AI-based mentoring system for supporting employee mental health. It starts by collecting public information and documents from within the company and building an initial learning dataset. The system consists of three main components: a server, a terminal, and a user.

[1970] server

[1971] The server is the center of the system and has the following main functions:

[1972] Data collection

[1973] The server collects public information and documents from within the company to build an initial learning dataset. This data includes company policies, business processes, manuals, etc. For example, the server obtains data in XML or CSV format from internal file servers and management tools (SharePoint, Confluence, etc.).

[1974] Early Learning

[1975] The server uses the received basic information to initially train a generative model specific to the company's context. This process involves training the generative AI model using a Python program using TensorFlow or PyTorch.

[1976] Terminal

[1977] The terminal provides the interface through which the employee interacts with the system.

[1978] Enter basic information

[1979] Provide a screen for employees to enter basic information (name, department, job description). For example, an input form for "name," "department," and "job description" is displayed on a smartphone application.

[1980] Data collection

[1981] The device collects data about the user's daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[1982] Emotion analysis

[1983] The device is equipped with an emotion engine that performs speech, facial expression, and text analysis. Speech data is analyzed using NLTK and DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER and TextBlob).

[1984] User

[1985] Users represent the actions employees take when using the system.

[1986] Input and Feedback

[1987] The user uses the device to input basic information and concerns or questions about daily work. For example, they input text such as "Work has been tough and I'm tired lately." The system also provides feedback in response to alerts and advice from the server and the device.

[1988] Mentoring and support

[1989] Alert Generation

[1990] The server analyzes the data sent from the device and evaluates emotions and stress levels in conjunction with the analysis results of the emotion engine. If high stress is detected, it generates appropriate alerts and advice for the user. For example, it uses Python code to generate instructions such as "Take a short break" using natural language generation tools (such as GPT-3).

[1991] Alert Notifications

[1992] The generated alerts and advice are sent to the user via the device, for example, a pop-up notification on the smartphone displays the message "Try taking deep breaths to relax."

[1993] Continuous evaluation and improvement

[1994] Data analysis and retraining

[1995] The server periodically analyzes data collected from all users to evaluate the effectiveness of the system. This includes graphing fluctuations in stress levels across employees and analyzing them using visualization tools (Tableau and Matplotlib). Based on the analysis results, the generative model is retrained. This incorporates new emotional data and feedback, updating the model to provide more accurate analysis and advice.

[1996] Specific examples

[1997] When new employee A logs in to the system for the first time, a screen appears where he or she can enter basic information (name, department, position). When A enters and submits the information, the server receives it and completes the initial learning process.

[1998] Next, Person A enters into the device, "I've already had a lot of work to do recently and I'm tired." Meanwhile, the device's emotion engine analyzes Person A's facial expressions and detects emotions such as "sad" or "tired." The server analyzes this information and detects rising stress levels. The server generates an alert telling Person A to "take a short break" and notifies the device. If Person A accepts the advice and takes a break, the feedback is sent back to the server, and the system continues to learn from the data.

[1999] Prompt Sentence Examples

[2000] Here are some examples of prompts for generative AI models:

[2001] Generate initial learning data based on publicly available information within the company and basic employee information to provide appropriate mental health support to newly hired employees. Based on the information entered by the user (Name: Taro Yamada, Department: Sales Department, Job Description: New Customer Development), provide examples of specific advice and support.

[2002]

[2003] The flow of the identification process in the second embodiment will be described with reference to FIG.

[2004] Step 1: Data collection

[2005] server:

[2006] The initial learning dataset is constructed by collecting public information and documents from within the company. For example, the server acquires data in XML or CSV format from the company's file server or management tools (SharePoint, Confluence, etc.).

[2007] Input: Internal company public information and documents

[2008] Data processing: Format conversion and organization of collected data

[2009] Output: Initial training dataset

[2010] Step 2: Enter basic information

[2011] Device:

[2012] Provide a screen for employees to enter basic information (name, department, job description). For example, display an input form for "name," "department," and "job description" on a smartphone application.

[2013] Input: Employee basic information

[2014] Data processing: None

[2015] Output: Basic information entered

[2016] User:

[2017] Enter basic information using the terminal and send it to the server. For example, enter "Name: Yamada Taro, Department: Sales Department, Job Description: New Customer Development" and press the send button.

[2018] Input: Name, Department, Job Description

[2019] Data processing: None

[2020] Output: Basic information sent

[2021] Step 3: Initial learning

[2022] server:

[2023] Based on the received basic information, a generative model specific to the company's context is initially trained. For example, a Python program can be used to train the generative AI model using TensorFlow or PyTorch.

[2024] Input: Initial training dataset and basic information

[2025] Data Computing: Training generative models (applying machine learning algorithms)

[2026] Output: Initially trained generative model

[2027] Step 4: Routine data collection

[2028] Device:

[2029] Collect data about users' daily conversations and work, including periodic surveys and chat-style interaction logs (e.g., Slack, Microsoft Teams logs).

[2030] Input: Daily conversation, business data

[2031] Data processing: logging and organization

[2032] Output: Daily data collected

[2033] Step 5: Sentiment Analysis

[2034] Device:

[2035] Using emotion engines, speech analysis, facial expression analysis, and text analysis are performed. For example, speech data is analyzed using NLTK or DeepSpeech, and text data is evaluated using sentiment analysis tools (VADER or TextBlob).

[2036] Input: Collected daily data

[2037] Data Computing: Speech, facial expression, and text emotion analysis

[2038] Output: User emotion data

[2039] Step 6: Stress Assessment

[2040] server:

[2041] The data sent from the device is analyzed and combined with the analysis results of the emotion engine to evaluate emotions and stress levels. For example, stress levels are determined based on keywords such as "tired" or "tough" or detected emotions such as "sad."

[2042] Input: User's daily data and emotional data

[2043] Data calculation: Stress level assessment (analysis of keywords and emotional data)

[2044] Output: Stress level evaluation result

[2045] Step 7: Alert Generation

[2046] server:

[2047] If high stress is detected, appropriate alerts and advice are generated for the user, such as "Take a short break" using natural language generation tools (such as GPT-3) in Python code.

[2048] Input: Stress level assessment result

[2049] Data operations: generating alerts and advice (applying natural language generation tools)

[2050] Output: Alerts and Advice

[2051] Step 8: Alert Notifications

[2052] Device:

[2053] Notify the user of any generated alerts or advice, for example, by displaying a pop-up notification on their smartphone with the message "Try taking some deep breaths to relax."

[2054] Input: Generated alerts and advice

[2055] Data Processing: Message Notification Settings

[2056] Output: User notification

[2057] Step 9: Gather feedback

[2058] User:

[2059] Receive alerts and advice and act on them, for example, take a five-minute break and then enter that feedback back into the device.

[2060] Input: Alert responses and feedback

[2061] Data Processing: Entering and Sending Feedback

[2062] Output: Feedback sent

[2063] Step 10: Continuous data analysis and retraining

[2064] server:

[2065] The data collected from all users is periodically analyzed to evaluate the effectiveness of the system. Fluctuations in the stress levels of all employees are graphed and analyzed using visualization tools (Tableau and Matplotlib). The generative model is retrained based on the analysis results.

[2066] Input: Feedback and daily data collected fr...

Claims

1. An artificial intelligence-based mentoring system for supporting employee mental health, comprising: A means of collecting public information and documents from within the company and building an initial learning dataset; a terminal means for employees to input basic information; A means for initially training a generative model based on the basic information; A means of collecting data on employees' daily conversations and work; means for analyzing the collected data and assessing emotions and stress levels; means for generating appropriate alerts and / or advice based on the assessed stress level; means for notifying employees of the generated alerts and advice; A means of continually analyzing data and evaluating the effectiveness of the system; means for re-learning the model based on the evaluation results; A system including:

2. A means of collecting data on employees' daily conversations and work activities to assess their stress levels; means for early detection and alert generation of potential mental health issues based on the assessed stress level; a means for notifying employees of the alert and providing appropriate measures; The system of claim 1 further comprising:

3. A means to continually evaluate the effectiveness of the alerts and advice generated and improve the model based on feedback; and means for generating new advice based on the improved model and notifying the employee; The system of claim 1 further comprising:

Images