System

Abstract

A system is provided.SOLUTION: A system comprising: means for collecting voice data of a workplace in real time and transcribing the voice data; means for analyzing the transcribed data and performing emotion analysis; means for determining a degree of unpleasantness based on a result of the emotion analysis; and means for storing related data and notifying a compliance department when the degree of unpleasantness exceeds a specific threshold.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] Power harassment and other types of harassment in the workplace are serious problems for victims. Victims typically hesitate to report them for fear of retaliation or unfair evaluation. As a result, harassment can go unnoticed, potentially having a negative impact on morale and productivity throughout the organization. Furthermore, companies face legal risks. This invention aims to solve these problems by providing a system that automatically detects harassment and reports it promptly and appropriately. [Means for solving the problem]

[0005] The present invention provides the following means: a system including means for collecting workplace voice data in real time and transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, means for determining the level of discomfort based on the results of the sentiment analysis, and means for saving related data and notifying a compliance department when the level of discomfort exceeds a specific threshold. The system also includes a reporting means for automatically executing the reporting. This allows for the rapid and effective detection of workplace harassment and enables appropriate responses.

[0006] "Audio data" refers to data that has been converted from an audio signal into a format that can be recorded, stored, and played back digitally.

[0007] "Real-time" refers to the method and timing in which input data is processed almost instantly.

[0008] "Transcription" is the process of converting audio data into text form.

[0009] A "means" is a device, method, or process used to accomplish a particular function or purpose.

[0010] "Analysis" is the process of breaking down data or information and revealing its structure and meaning.

[0011] "Sentiment analysis" is the process of extracting a speaker's feelings and intentions from text data and identifying emotions such as joy, anger, sadness, and displeasure.

[0012] "Offensiveness" is a criterion for quantitatively assessing how offensive a particular statement or behavior is to the recipient.

[0013] A "threshold" is a reference value for determining the occurrence of a certain phenomenon or state.

[0014] "Related data" is a collection of information related to a specific event, and includes voice data, text data, sentiment analysis results, and the like.

[0015] "Reporting" is the act of conveying information to relevant parties about a specific event or situation.

[0016] A "compliance department" is a department within a company or organization that monitors and manages regulatory compliance and ethical behavior.

[0017] "Natural language processing" is a general term for technologies and methods that allow computers to understand, generate, and respond to human language. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0019] 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.

[0020] First, the terms used in the following description will be explained.

[0021] 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).

[0022] 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.

[0023] 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.

[0024] 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.

[0025] 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."

[0026] [First embodiment]

[0027] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0028] 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.

[0029] 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).

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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.

[0034] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0035] 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.

[0036] 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.

[0037] 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.

[0038] 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."

[0039] The AI ​​Guardian system of the present invention collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports the incident to the compliance department as necessary. The following describes the program processing of this system in detail.

[0040] 1. Collection and transmission of voice data

[0041] The device is set up to constantly capture audio data from meetings and daily work. The microphones on the device collect surrounding sounds as digital data.

[0042] Once the voice data is collected, the device transcribes it in real time by converting speech to text using speech recognition software (e.g., open source or commercial speech recognition APIs) built into the device.

[0043] The transcribed text data is sent from the device to a server using HTTP requests or other communication protocols.

[0044] 2. Sentiment analysis and unpleasantness judgment

[0045] The server receives the text data sent from the terminal, and the received text data is processed within the server for analysis.

[0046] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis.

[0047] Based on the results of natural language processing, the server performs sentiment analysis by calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[0048] Based on the emotion score, the server determines the degree of annoyance, and if the degree of annoyance exceeds a certain threshold, the text data is identified as potentially harassing.

[0049] 3. Identify potential harassment

[0050] The server identifies text data that it determines to be highly offensive as potentially harassing, and stores the identified data together with audio data, transcription data, and sentiment analysis data.

[0051] The stored data is retained for later verification and use as evidence.

[0052] 4. Report to the Compliance Department

[0053] The server automatically notifies the compliance department when potentially harassing data is stored, in the form of a report of the stored data, including audio data, transcription data, and sentiment analysis results.

[0054] Once a report is made, compliance users receive a notification via the software dashboard and email.

[0055] Specific examples

[0056] For example, if someone says "Work more efficiently" during a work meeting, this comment may cause offense. In that case, the following process will occur:

[0057] 1. The device captures the audio of the conversation, transcribes it in real time into text data such as "Work more efficiently," and sends it to the server.

[0058] 2. The server analyzes the text data and performs sentiment analysis, determining that the comment is likely to cause high levels of offense.

[0059] 3. The server will store the data as potentially harassing data and report the relevant data to the compliance department.

[0060] 4. Compliance users receive a notification, view detailed information on a dashboard, and take appropriate action.

[0061] As described above, the system of the present invention automatically detects and notifies workplace harassment, enabling prompt and appropriate responses, thereby contributing to protecting victims and reducing legal risks for companies.

[0062] The processing flow will be explained below.

[0063] Step 1:

[0064] The device uses a microphone to capture surrounding sounds in order to constantly collect voice data within the workplace.

[0065] Step 2:

[0066] The device transcribes the captured audio data in real time by converting the audio into text using a speech recognition API (e.g., a speech recognition engine).

[0067] Step 3:

[0068] The terminal transmits the generated text data to the server using a communication protocol such as an HTTP request.

[0069] Step 4:

[0070] To analyze the received text data, the server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text and perform grammatical analysis.

[0071] Step 5:

[0072] The server performs sentiment analysis based on the analyzed text data, calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[0073] Step 6:

[0074] The server evaluates the emotion score to determine the level of annoyance, and identifies potential harassment if the level of annoyance exceeds a certain threshold.

[0075] Step 7:

[0076] The server stores audio data, transcription data, and sentiment analysis data related to text data determined to be potentially harassing in a database.

[0077] Step 8:

[0078] The server generates reports based on the stored data and sends them to the compliance department via email or a dedicated dashboard.

[0079] Step 9:

[0080] The user (compliance department officer) receives the report, checks the detailed information on the provided dashboard, takes necessary measures based on the content of the notification, and, in some cases, provides support to the victim.

[0081] Example 1

[0082] 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."

[0083] Workplace harassment is a serious problem that causes psychological stress for employees and reduces productivity throughout an organization. However, detecting and responding to harassment on the spot is difficult, and victims often miss opportunities to receive appropriate support. Furthermore, traditional manual monitoring and reporting systems lack immediacy, making it difficult to respond quickly. Therefore, there is a need for the development of a system that can analyze workplace audio data in real time, automatically detect inappropriate remarks, and appropriately report them.

[0084] 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.

[0085] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing natural language processing to analyze the structure of text data, means for performing sentiment analysis based on the analysis results and obtaining a sentiment score, means for determining the degree of discomfort based on the sentiment score and storing related data when a specific threshold is exceeded, means for reporting to a compliance department based on the determination result, and means for automating the reporting. This makes it possible to automatically and quickly detect workplace harassment and immediately report it to the appropriate department, thereby protecting victims and improving the soundness of the organization.

[0086] "Voice data" refers to any audio information generated in the workplace and is collected digitally using a microphone.

[0087] "Transcription means" refers to the technology or software used to convert voice data into text data, and uses voice recognition technology to convert voice into text in real time.

[0088] "Natural language processing means" refers to technology and software that analyzes text data and understands its grammatical structure and meaning, and performs tokenization and syntactic analysis.

[0089] "Sentiment analysis" refers to the technology of analyzing the content of text data and quantifying the emotions expressed therein, obtaining emotional scores such as joy, sadness, anger, and displeasure.

[0090] "Offensiveness assessment means" refers to a technology or method for assessing the offensiveness of text data based on the results of sentiment analysis, and determining that the content is inappropriate if it exceeds a certain threshold.

[0091] "Related data" refers to all data related to the detected inappropriate remarks, including audio data, transcription data, sentiment analysis results, the date and time of the remarks, and information about the speaker.

[0092] "Reporting tools" refers to technology or software that automatically notifies the compliance department of any relevant data when inappropriate comments or potential harassment are identified.

[0093] A "compliance department" refers to the department within a company that oversees compliance with laws and regulations and ethical issues, and is responsible for receiving reports and taking appropriate action.

[0094] This invention, "AI Guardian," is a system that collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports them to the compliance department as necessary. The program processing of this system is explained in detail below.

[0095] Collection and transmission of voice data

[0096] The devices are installed at each company's bases and are always on. A high-sensitivity microphone installed in the device digitally collects audio data from workplace meetings and daily work. The collected audio data is temporarily stored on the device and converted into text data in real time using speech recognition APIs such as Google Cloud Speech-to-Text and IBM Watson Speech to Text. The converted text data is sent from the device to the server using an HTTP request.

[0097] Text data analysis

[0098] The server receives the text data sent from the device and analyzes it using a natural language processing library (e.g., spaCy or nltk). Specifically, the server tokenizes the text data and performs grammatical analysis to extract sentence structure and keywords. This analysis clarifies how the text data is structured.

[0099] Sentiment analysis and discomfort judgment

[0100] The server then performs sentiment analysis on the analyzed text data, calling sentiment analysis APIs such as the Google Natural Language API and IBM Watson Tone Analyzer to obtain a sentiment score for the text data. The sentiment score quantifies emotions such as joy, sadness, anger, and displeasure, and the server uses this score to evaluate the emotional tendency of the remarks. Based on the obtained sentiment score, the server determines the degree of discomfort, and if the degree of discomfort exceeds a certain threshold, it identifies the text data as potentially harassing.

[0101] Data storage and reporting

[0102] The server stores text data that is judged to be highly offensive in a database along with related data (audio data, transcription data, sentiment analysis results, etc.). The stored data can be referenced later for verification or as evidence. The server then immediately reports this data to the compliance department. Reports are made in the form of reports via HTTP POST requests or email, and the reports include the audio data, transcription data, and sentiment analysis results. Once a report is made, the compliance department user receives a notification, can view details on a dashboard, and can take appropriate action.

[0103] Specific examples

[0104] For example, if someone says "Work more efficiently" during a work meeting, the system will do the following:

[0105] 1. The device collects the audio of the conversation and converts it into a digital signal through a highly sensitive microphone.

[0106] 2. The device transcribes the voice data in real time, generates text data such as "Work more efficiently," and sends it to the server.

[0107] 3. The server tokenizes the received text data and analyzes its grammatical structure.

[0108] 4. The server uses the sentiment analysis API to obtain a sentiment score for the text data. In this case, it determines that "Work efficiently" is likely to indicate high levels of displeasure.

[0109] 5. The server identifies this data as potentially harassing and stores it in a database along with the audio data, transcription data, and sentiment analysis results.

[0110] 6. The server sends the generated report to the compliance department via HTTP POST request and / or email, and a compliance department user is notified.

[0111] Prompt Sentence Examples

[0112] For example, the following prompt sentence is used for a generative AI model:

[0113] If someone says "Work more efficiently" during a work meeting, determine whether this statement is potentially harassing. Tokenize the text data, perform sentiment analysis, and if a high level of discomfort is detected, explain why.

[0114] By introducing this system, companies will be able to quickly and accurately detect workplace harassment and take appropriate measures, making it a powerful tool for ensuring the psychological safety of employees and improving the work environment.

[0115] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0116] Step 1:

[0117] The device collects workplace audio data in real time. Specifically, it uses a high-sensitivity microphone to digitally capture the audio of meetings and daily work. The input is ambient analog audio data, and the output is audio data converted into a digital signal.

[0118] Step 2:

[0119] The device transcribes the collected voice data. Specifically, it converts the voice data into text data using voice recognition software (e.g., Google Cloud Speech-to-Text or IBM Watson Speech to Text). The input is digital voice data, and the output is transcribed text data.

[0120] Step 3:

[0121] The device sends the transcribed text data to the server. Specifically, it sends the text data to the server using an HTTP request or other communication protocol. The input is the transcribed text data, and the output is the text data sent to the server.

[0122] Step 4:

[0123] The server analyzes the received text data. Specifically, it uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis. The input is text data, and the output is grammatically analyzed text data.

[0124] Step 5:

[0125] The server performs sentiment analysis on the analyzed text data. Specifically, it calls a sentiment analysis API (e.g., Google Natural Language API or IBM Watson Tone Analyzer) to obtain a sentiment score. The input is the grammatically analyzed text data, and the output is a sentiment score.

[0126] Step 6:

[0127] The server determines the degree of discomfort based on the emotion score. Specifically, it evaluates whether the text data is unpleasant by comparing the acquired emotion score with a set threshold. The input is the emotion score, and the output is the evaluation result of the discomfort degree.

[0128] Step 7:

[0129] The server stores the text data judged to be highly unpleasant together with related data. Specifically, it stores the text data, emotion scores, audio data, etc. in a database. The input is the text data judged to be highly unpleasant and the related data, and the output is the data stored in the database.

[0130] Step 8:

[0131] The server then sends a report to the compliance department based on the stored data. Specifically, it sends the report in the form of a report via email or HTTP POST request. The input is the data stored in the database, and the output is a report to the compliance department.

[0132] Step 9:

[0133] Compliance department users review the received reports and take appropriate action. Specifically, they receive notifications via the software dashboard or email, check the details, and take appropriate action. The input is the report data from the server, and the output is the action taken.

[0134] (Application example 1)

[0135] 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."

[0136] In modern workplaces, especially in factories, harassment can occur during communication with workers. Such harassment not only worsens the working environment, but also increases legal risks for companies. Furthermore, factory robots and automated systems mean that on-site monitoring of working conditions is insufficient, making it difficult to protect workers.

[0137] 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.

[0138] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, means for determining the level of discomfort based on the results of the sentiment analysis, means for saving related data and notifying the management department when the level of discomfort exceeds a specific threshold, and means for collecting and analyzing voice data using factory robots to monitor physical labor and protect workers, thereby enabling early detection of harassment in the workplace and protection of workers.

[0139] "Voice data" means data for storing or processing acoustic signals in digital form, including, in particular, speech and other sounds in the workplace or on a factory floor.

[0140] "Transcription means" refers to technology or equipment for converting voice data into text data, and primarily refers to voice recognition software.

[0141] "Sentiment analysis means" refers to natural language processing techniques that analyze text data and determine the emotions and intensity of emotions contained therein.

[0142] "Discomfort level" is a measure of how much a certain statement or action makes the other person feel uncomfortable, and is calculated from the results of emotion analysis.

[0143] An "administrative department" is a department within a company or organization that is responsible for managing compliance and the working environment.

[0144] A "factory robot" is a mechanical device that performs automated tasks in a factory and may have a variety of functions, such as collecting and analyzing voice data.

[0145] "Worker protection" refers to measures and efforts to ensure that workers can work in a safe and comfortable environment.

[0146] The present invention relates to an AI Guardian system for detecting harassment and protecting workers in workplaces and factory environments. The system collects voice data in real time and analyzes the data for sentiment analysis. Specifically, the following detailed embodiments are described.

[0147] First, a microphone built into the factory robot collects voice data. This voice data is transcribed in real time using voice recognition software (e.g., Google's voice recognition API). The transcribed data is sent to a server and analyzed using a natural language processing library (e.g., spaCy or nltk).

[0148] Next, the server performs sentiment analysis. This analysis uses a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer). Sentiment analysis calculates an emotional score for the text data. The level of discomfort is determined based on this score, and if it exceeds a certain threshold, the relevant data is automatically saved and a report is sent to the management department.

[0149] The reporting function is implemented using the Twilio API, and reports are sent via SMS or email, and include collected voice data, transcription data, and sentiment analysis results.

[0150] A specific example is shown below. If someone says "You're completely useless" in a conversation within a factory, this speech is collected by a microphone. Speech recognition software transcribes this speech and sends it to a server. The server performs sentiment analysis on this text data, and if it determines that the sentiment score is high, it uses the Twilio API to notify management along with the message "You're completely useless."

[0151] Prompt Sentence Examples

[0152] Perform a sentiment analysis on the following text and print the results:

[0153] "You're completely useless"

[0154] This system will enable early detection of workplace harassment and rapid response, strengthening worker protection and reducing legal risks for companies.

[0155] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0156] Step 1:

[0157] The terminal collects the audio data. Specifically, microphones built into the factory robots capture the surrounding sounds and store them digitally. The terminal collects the audio data in real time.

[0158] Input: Ambient audio (analog data)

[0159] Output: Audio data (digital data)

[0160] Step 2:

[0161] The device transcribes the collected voice data in real time and uses speech recognition software to convert the collected voice data into text data.

[0162] Input: Audio data (digital data)

[0163] Output: Transcription data (text data)

[0164] Specific operation: The device sends voice data to the voice recognition API and receives the returned text data.

[0165] Step 3:

[0166] The device sends the transcription data to the server. The text data is sent to the server using a communication protocol such as an HTTP request.

[0167] Input: Transcription data (text data)

[0168] Output: Text data sent to the server

[0169] Specific operation: The device sends text data to the server via a communication protocol.

[0170] Step 4:

[0171] The server receives the text data and analyzes it using a natural language processing library, tokenizing the text data and performing grammatical analysis.

[0172] Input: Text data (transcription data)

[0173] Output: Analyzed text data

[0174] Specific operation: The server analyzes the text data using a natural language processing library (e.g. spaCy, nltk).

[0175] Step 5:

[0176] The server calls the sentiment analysis API to obtain the sentiment score of the text data, and determines the level of discomfort based on the sentiment score.

[0177] Input: Parsed text data

[0178] Output: Emotion score, unpleasantness judgment result

[0179] Specific operation: The server obtains the sentiment score using a sentiment analysis API (e.g., Google Natural Language API, IBM Watson Tone Analyzer).

[0180] Step 6:

[0181] If the server detects that the level of discomfort exceeds a certain threshold, it will store the relevant data and notify the management department via SMS or email.

[0182] Input: Emotion score, discomfort level judgment result

[0183] Output: Saved data, notification to management department

[0184] Specific operation: The server saves the relevant data in a database and sends a notification using the Twilio API.

[0185] This concludes the detailed explanation of each processing step in the system of the application example. This flow enables real-time monitoring of voice data, and the detection and immediate reporting of harassment.

[0186] 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.

[0187] The "AI Guardian" system of the present invention collects workplace voice data in real time and automatically detects harassing remarks. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the situation to the compliance department if necessary. The following describes the program processing of this system in detail.

[0188] 1. Collection and transmission of voice data

[0189] The device continuously captures audio data from meetings and daily work. The device's built-in microphone collects surrounding sounds as digital data.

[0190] After collecting the audio data, the device transcribes it in real time, using voice recognition software to convert the audio data into text format.

[0191] The converted text data is then sent from the device to the server using an HTTP request or other communication protocol.

[0192] 2. Sentiment analysis and unpleasantness judgment

[0193] The server receives the text data sent from the terminal, and processes the received text data for analysis.

[0194] The server uses a natural language processing library to tokenize the text data and perform grammatical analysis.

[0195] After grammar analysis, the server performs sentiment analysis on the text data using an emotion engine, which extracts sentiment scores from the text, such as joy, anger, sadness, and displeasure.

[0196] At the same time, the server also performs sentiment analysis based on the voice data, combining the results of voice recognition with the emotion engine analysis to more accurately estimate the user's emotions.

[0197] The server determines the degree of discomfort based on the results of these sentiment analyses. If the sentiment score exceeds a certain threshold, the corresponding speech is deemed to cause discomfort.

[0198] 3. Identify potential harassment

[0199] The server identifies text and audio data that is deemed highly offensive as potentially harassing, and stores the identified data along with sentiment analysis results and related data.

[0200] The stored data will be retained for future verification and use as evidence.

[0201] 4. Report to the Compliance Department

[0202] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[0203] The report includes relevant audio data, text data, and sentiment analysis results.

[0204] Specific examples

[0205] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[0206] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[0207] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[0208] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[0209] 4. The server stores the relevant data and automatically notifies the compliance department.

[0210] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[0211] In this way, the system of the present invention performs emotion analysis from both voice and text, thereby improving the accuracy of identifying harassment, enabling prompt and appropriate responses, improving the work environment, and reducing legal risks.

[0212] The processing flow will be explained below.

[0213] Step 1:

[0214] The device uses microphones installed in the workplace to capture audio data from meetings and daily work in real time.

[0215] Step 2:

[0216] The device uses speech recognition software (e.g., a speech recognition engine) to transcribe the captured voice data in real time, which is then converted into text data.

[0217] Step 3:

[0218] The terminal transmits the generated text data to the server using an applicable communication protocol such as an HTTP request.

[0219] Step 4:

[0220] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize and parse the received text data.

[0221] Step 5:

[0222] The server performs sentiment analysis on the text data using an emotion engine, which extracts emotion scores from the text, such as happiness, anger, sadness, displeasure, etc.

[0223] Step 6:

[0224] The server also performs sentiment analysis of the voice data using an emotion engine, which analyzes the tone, stress, and intonation of the voice to identify emotions.

[0225] Step 7:

[0226] The server combines the emotion scores from both the text and the audio to determine the annoyance level, and evaluates whether the annoyance level exceeds a certain threshold.

[0227] Step 8:

[0228] If the server determines that the offense level is high, it identifies the data as potentially harassing and stores the associated audio data, transcription data, and sentiment analysis data in a database.

[0229] Step 9:

[0230] The server generates detailed reports based on the stored data and automatically reports them to the compliance department via email or a dedicated dashboard.

[0231] Step 10:

[0232] The user (compliance department staff) receives the report and checks the notified report on the dashboard. Based on the detailed information, they take necessary measures and, in some cases, provide support to the victim.

[0233] Example 2

[0234] 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."

[0235] Detecting and responding to workplace harassment is extremely important, but traditional methods rely primarily on human monitoring and reporting, making it difficult to respond quickly and objectively. This can lead to a worsening work environment, causing mental stress and reduced productivity for employees. Furthermore, traditional methods make it difficult to preserve and manage evidence of harassment, which can delay necessary responses.

[0236] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0237] In this invention, the server includes: means for collecting voice data in real time; means for transcribing the voice data; means for analyzing the transcription data and the voice data using a natural language processing library; means for performing sentiment analysis of the transcription data and the voice data using an emotion engine; means for determining the level of discomfort based on the results of the sentiment analysis; and means for storing related data, generating data for reporting, and reporting to a compliance department when the level of discomfort exceeds a specific threshold. This enables real-time detection and rapid response to workplace harassment. Furthermore, by appropriately storing evidence of highly offensive remarks and taking action based on objective data, the system can improve the work environment and reduce employee mental stress.

[0238] "Audio data" refers to data that records audio in digital format.

[0239] "Transcription" is the process of converting audio data into text form.

[0240] A "natural language processing library" is a set of software tools for analyzing text data, including tokenization and grammatical analysis.

[0241] "Emotion Engine" refers to algorithms and techniques for extracting emotion scores from text and audio data.

[0242] The "discomfort level" is an index that indicates the degree of a particular emotional state calculated based on the emotional score.

[0243] A "threshold" is a boundary value for triggering a specific action when an indicator such as discomfort level or emotion score exceeds a certain standard.

[0244] "Related data" refers to all data related to determining harassment, such as the results of sentiment analysis, the content of statements, and audio data.

[0245] A "compliance department" is a specialized department within a company that monitors and promotes compliance with laws and regulations and ethical behavior.

[0246] "Notification" refers to the process of reporting specific events or data.

[0247] "Data storage" refers to storing acquired data for future use.

[0248] The system of this invention, "AI Guardian," collects workplace voice data in real time and automatically detects harassing comments. It also incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the incident to the compliance department if necessary. This system is primarily implemented using the following hardware and software:

[0249] Audio data collection and transcription

[0250] The device is equipped with a microphone to continuously capture audio data during meetings and daily work. It is also equipped with speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson. This software converts the collected audio data into text in real time. The text data is then sent to a server using a communication protocol such as an HTTP request.

[0251] Text data analysis and sentiment analysis

[0252] When the server receives the text data sent from the device, it analyzes the data using a natural language processing library (such as NLTK or spaCy). First, the text data is tokenized and grammatically analyzed. Then, an emotion engine (such as the Text Analytics API from Azure Cognitive Services) is used to perform a sentiment analysis of the text and audio data. This emotion engine extracts an emotion score from the text and audio, quantifying emotions such as joy, anger, sadness, and displeasure.

[0253] Determining discomfort and identifying harassment

[0254] The server analyzes the emotion scores obtained from the emotion engine and determines the level of offense. If an emotion score exceeding a certain threshold is detected, the remark is deemed to be offensive. This highly offensive remark is identified as potentially harassing, and the related data (audio data, text data, and emotion analysis results) is stored.

[0255] Report to the Compliance Department

[0256] The server generates a report based on data judged to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard. The generated report includes related audio data, text data, and sentiment analysis results.

[0257] Specific examples

[0258] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[0259] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[0260] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[0261] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[0262] 4. The server stores the relevant data and automatically notifies the compliance department.

[0263] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[0264] Prompt Sentence Examples

[0265] For example, this process begins by inputting a prompt such as, "Please analyze the statement 'Work more efficiently' made during a meeting." into the generative AI model.

[0266] In this way, this system, "AI Guardian," performs emotion analysis from both voice and text, improving the accuracy of identifying harassment and enabling prompt and appropriate responses, thereby improving the work environment and reducing legal risks.

[0267] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0268] Step 1:

[0269] The device uses a microphone to collect real-time audio data from workplace meetings and daily work.

[0270] Input: Ambient conversation

[0271] Output: Digital audio data

[0272] Specific operation: The microphone installed in the device captures audio and stores it as digital data.

[0273] Step 2:

[0274] The device transcribes the collected voice data using speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson.

[0275] Input: Audio data

[0276] Output: Transcribed text data

[0277] Specific operation: Voice data is input into voice recognition software, and the conversion result is obtained as text data.

[0278] Step 3:

[0279] The device sends the generated text data to the server using an HTTP request, using HTTP or HTTPS as the communication protocol.

[0280] Input: Transcription data

[0281] Output: Text data sent to the server

[0282] Specific operation: Include text data in the body of an HTTP request and send it to the API endpoint.

[0283] Step 4:

[0284] The server uses a natural language processing library (such as NLTK or spaCy) to tokenize and grammatically analyze the received text data.

[0285] Input: Received text data

[0286] Output: Tokenized and parsed text data

[0287] Specific operation: The text data is broken down into tokens and the grammatical structure is analyzed.

[0288] Step 5:

[0289] The server uses an emotion engine (such as the Text Analytics API from Azure Cognitive Services) to perform emotion analysis on the text and audio data, and extracts an emotion score.

[0290] Input: Tokenized and grammatically analyzed text data, audio data

[0291] Output: Sentiment score

[0292] Specific operation: The analyzed text data and voice data are input into the emotion engine to calculate emotion scores such as joy, anger, sadness, and displeasure.

[0293] Step 6:

[0294] The server determines the degree of discomfort based on the extracted emotion score. If the degree of discomfort exceeds a certain threshold, the comment is deemed to be unpleasant.

[0295] Input: Sentiment score

[0296] Output: Discomfort level judgment result

[0297] Specific operation: The emotion score is compared with a specific threshold, and if the threshold is exceeded, the unpleasantness level is determined to be high.

[0298] Step 7:

[0299] The server identifies text and audio data that is determined to be highly offensive as potentially harassing and stores the associated data.

[0300] Input: Discomfort level judgment result

[0301] Output: Saved text and audio data

[0302] Specific action: Potentially harassing data will be stored in a database.

[0303] Step 8:

[0304] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[0305] Input: Stored text and audio data

[0306] Output: Generated reports, notification

[0307] Specific Action: Generates a report and sends an email using SMTP or displays a notification on the dashboard.

[0308] Step 9:

[0309] The user (compliance department officer) receives a notification, checks detailed information on the dashboard, and takes the necessary action.

[0310] Input: Report notification, detailed information

[0311] Output: Response Action

[0312] Specific Action: Review reports on the dashboard and take appropriate action.

[0313] (Application example 2)

[0314] 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."

[0315] Conventional harassment detection systems are limited to the workplace, making it difficult to effectively monitor harassment and inappropriate comments in public places or the wider society. Furthermore, manual reporting and response is required, resulting in delayed real-time responses. Therefore, there is a need for a system that can detect inappropriate comments in real time, even in public places, and respond quickly.

[0316] 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.

[0317] In this invention, the server includes means for collecting voice data from workplaces and public places in real time, transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, and means for determining the level of discomfort based on the results of the sentiment analysis. This makes it possible to detect harassment and inappropriate remarks in real time not only in the workplace but also in public places, and to report them promptly.

[0318] "Workplaces and public places" refers not only to workplaces but also to environments that include places widely used by the general public, such as public transportation, parks, and commercial facilities.

[0319] "Audio data" refers to data that digitally represents people's conversations and environmental sounds collected via a voice input device such as a microphone.

[0320] "Transcription" is the process of analyzing audio data and converting its content into corresponding text data.

[0321] "Sentiment analysis" is a technology that detects emotions from text data or audio data and evaluates the content as an emotional score such as joy, anger, sadness, or displeasure.

[0322] "Offensiveness" is an index calculated based on sentiment analysis that assesses the degree to which a certain statement or event causes offense.

[0323] A "security department" refers to a specialized department or team that responds when inappropriate remarks or harassment are detected.

[0324] "Reporting" is the process by which the system automatically notifies the security department of relevant data when it detects potentially offensive speech.

[0325] "Speech recognition" is a technology that analyzes voice data and converts what is being said into text.

[0326] "Natural language processing" is a general term for artificial intelligence technology that analyzes text data and understands and analyzes its meaning and emotions.

[0327] "Harassment likelihood" refers to an assessment of the likelihood that a particular statement constitutes harassment based on sentiment analysis.

[0328] The system for implementing this invention, "AI Harassment Guardian," has the function of collecting voice data in real time in the workplace and public places, detecting and reporting inappropriate remarks and harassment. Specifically, it consists of the following elements:

[0329] 1. System Configuration

[0330] Hardware

[0331] Smartphones: Equipped with microphones for collecting audio data.

[0332] software

[0333] Python: Used as a programming language.

[0334] pydub: A library used to manipulate and process audio data.

[0335] vosk: An open-source speech recognition model library.

[0336] transformers: A natural language processing toolkit for sentiment analysis.

[0337] 2. System Operation

[0338] Collection and processing of voice data

[0339] The device (smartphone) captures conversations in meetings and public places in real time via a microphone. The captured audio data is processed by the pydub library and standardized to a 16 kHz sample rate.

[0340] Voice Recognition

[0341] The server transcribes the processed audio data using the vosk library, which converts the audio data into text data.

[0342] sentiment analysis

[0343] The server performs sentiment analysis on the converted text data using the transformers library, and as a result, the text data is assigned an emotional score such as joy, anger, sadness, or displeasure.

[0344] Determining discomfort and reporting

[0345] The server determines the level of discomfort based on the emotional score, and if the level of discomfort exceeds a certain threshold, it is deemed to be harassment. This information is automatically reported to the security department via email or a dedicated dashboard.

[0346] Specific examples

[0347] For example, if someone says "Go away!" on a public bus, the following will happen:

[0348] 1. The device (smartphone) captures the audio and transcribes what is said.

[0349] 2. The server performs sentiment analysis on the transcribed data and determines that the level of unpleasantness is high.

[0350] 3. The server stores this data and notifies the security department.

[0351] Prompt Sentence Examples

[0352] Prompt statement:

[0353] "Audio data is captured, transcribed in real time, and then sentiment analysis is performed. If a specific negative sentiment is indicated with a high score, the remark is determined to be harassment and a report is sent."

[0354] This system will enable real-time monitoring and rapid response to harassment and inappropriate comments not only in the workplace but also in public places.

[0355] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0356] Step 1:

[0357] The device (smartphone) collects audio data in real time during meetings and public places. The microphone on the device captures the surrounding sounds and saves them as audio data. The audio data is then stored digitally on the device.

[0358] Step 2:

[0359] The device converts the collected audio data to a 16kHz sample rate using the pydub library, which processes the audio data into a format suitable for voice recognition.

[0360] Step 3:

[0361] The device transcribes the processed voice data using the vosk library, recognizes the input voice data, and outputs the corresponding text data. The resulting conversation is saved in text format on the device.

[0362] Step 4:

[0363] The terminal sends the generated text data to the server using a communication protocol such as an HTTP request. At this time, the text data is transferred to the server.

[0364] Step 5:

[0365] The server receives the text data and, to analyze it, tokenizes it using a natural language processing (NLP) library, converting the text data into a format that can be analyzed.

[0366] Step 6:

[0367] The server uses the transformers library for sentiment analysis on the tokenized text data: it computes sentiment scores from the tokenized data as input, evaluates the sentiment (e.g., joy, anger, sadness, displeasure, etc.) for each text fragment, and produces a sentiment score as output.

[0368] Step 7:

[0369] The server determines the offensiveness based on the sentiment score. It analyzes each sentiment score to determine whether the offensiveness exceeds a certain threshold. If the offensiveness is high, the text fragment is determined to be likely to be harassing.

[0370] Step 8:

[0371] The server stores text fragments that are judged to be highly offensive and their emotion scores as related data, which can be used for later verification or as evidence.

[0372] Step 9:

[0373] The server automatically sends alerts to security teams based on relevant data, either via email or a dedicated dashboard, and includes sentiment analysis results along with relevant data.

[0374] Step 10:

[0375] The user (security department staff) receives a notification, checks the details on the dashboard, and takes immediate action if necessary.

[0376] 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.

[0377] 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.

[0378] 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.

[0379] [Second embodiment]

[0380] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0381] 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.

[0382] 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).

[0383] 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.

[0384] 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.

[0385] 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).

[0386] 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.

[0387] 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.

[0388] 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.

[0389] 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.

[0390] 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.

[0391] 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."

[0392] The AI ​​Guardian system of the present invention collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports the incident to the compliance department as necessary. The following describes the program processing of this system in detail.

[0393] 1. Collection and transmission of voice data

[0394] The device is set up to constantly capture audio data from meetings and daily work. The microphones on the device collect surrounding sounds as digital data.

[0395] Once the voice data is collected, the device transcribes it in real time by converting speech to text using speech recognition software (e.g., open source or commercial speech recognition APIs) built into the device.

[0396] The transcribed text data is sent from the device to a server using HTTP requests or other communication protocols.

[0397] 2. Sentiment analysis and unpleasantness judgment

[0398] The server receives the text data sent from the terminal, and the received text data is processed within the server for analysis.

[0399] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis.

[0400] Based on the results of natural language processing, the server performs sentiment analysis by calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[0401] Based on the emotion score, the server determines the degree of annoyance, and if the degree of annoyance exceeds a certain threshold, the text data is identified as potentially harassing.

[0402] 3. Identify potential harassment

[0403] The server identifies text data that it determines to be highly offensive as potentially harassing, and stores the identified data together with audio data, transcription data, and sentiment analysis data.

[0404] The stored data is retained for later verification and use as evidence.

[0405] 4. Report to the Compliance Department

[0406] The server automatically notifies the compliance department when potentially harassing data is stored, in the form of a report of the stored data, including audio data, transcription data, and sentiment analysis results.

[0407] Once a report is made, compliance users receive a notification via the software dashboard and email.

[0408] Specific examples

[0409] For example, if someone says "Work more efficiently" during a work meeting, this comment may cause offense. In that case, the following process will occur:

[0410] 1. The device captures the audio of the conversation, transcribes it in real time into text data such as "Work more efficiently," and sends it to the server.

[0411] 2. The server analyzes the text data and performs sentiment analysis, determining that the comment is likely to cause high levels of offense.

[0412] 3. The server will store the data as potentially harassing data and report the relevant data to the compliance department.

[0413] 4. Compliance users receive a notification, view detailed information on a dashboard, and take appropriate action.

[0414] As described above, the system of the present invention automatically detects and notifies workplace harassment, enabling prompt and appropriate responses, thereby contributing to protecting victims and reducing legal risks for companies.

[0415] The processing flow will be explained below.

[0416] Step 1:

[0417] The device uses a microphone to capture surrounding sounds in order to constantly collect voice data within the workplace.

[0418] Step 2:

[0419] The device transcribes the captured audio data in real time by converting the audio into text using a speech recognition API (e.g., a speech recognition engine).

[0420] Step 3:

[0421] The terminal transmits the generated text data to the server using a communication protocol such as an HTTP request.

[0422] Step 4:

[0423] To analyze the received text data, the server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text and perform grammatical analysis.

[0424] Step 5:

[0425] The server performs sentiment analysis based on the analyzed text data, calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[0426] Step 6:

[0427] The server evaluates the emotion score to determine the level of annoyance, and identifies potential harassment if the level of annoyance exceeds a certain threshold.

[0428] Step 7:

[0429] The server stores audio data, transcription data, and sentiment analysis data related to text data determined to be potentially harassing in a database.

[0430] Step 8:

[0431] The server generates reports based on the stored data and sends them to the compliance department via email or a dedicated dashboard.

[0432] Step 9:

[0433] The user (compliance department officer) receives the report, checks the detailed information on the provided dashboard, takes necessary measures based on the content of the notification, and, in some cases, provides support to the victim.

[0434] Example 1

[0435] 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."

[0436] Workplace harassment is a serious problem that causes psychological stress for employees and reduces productivity throughout an organization. However, detecting and responding to harassment on the spot is difficult, and victims often miss opportunities to receive appropriate support. Furthermore, traditional manual monitoring and reporting systems lack immediacy, making it difficult to respond quickly. Therefore, there is a need for the development of a system that can analyze workplace audio data in real time, automatically detect inappropriate remarks, and appropriately report them.

[0437] 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.

[0438] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing natural language processing to analyze the structure of text data, means for performing sentiment analysis based on the analysis results and obtaining a sentiment score, means for determining the degree of discomfort based on the sentiment score and storing related data when a specific threshold is exceeded, means for reporting to a compliance department based on the determination result, and means for automating the reporting. This makes it possible to automatically and quickly detect workplace harassment and immediately report it to the appropriate department, thereby protecting victims and improving the soundness of the organization.

[0439] "Voice data" refers to any audio information generated in the workplace and is collected digitally using a microphone.

[0440] "Transcription means" refers to the technology or software used to convert voice data into text data, and uses voice recognition technology to convert voice into text in real time.

[0441] "Natural language processing means" refers to technology and software that analyzes text data and understands its grammatical structure and meaning, and performs tokenization and syntactic analysis.

[0442] "Sentiment analysis" refers to the technology of analyzing the content of text data and quantifying the emotions expressed therein, obtaining emotional scores such as joy, sadness, anger, and displeasure.

[0443] "Offensiveness assessment means" refers to a technology or method for assessing the offensiveness of text data based on the results of sentiment analysis, and determining that the content is inappropriate if it exceeds a certain threshold.

[0444] "Related data" refers to all data related to the detected inappropriate remarks, including audio data, transcription data, sentiment analysis results, the date and time of the remarks, and information about the speaker.

[0445] "Reporting tools" refers to technology or software that automatically notifies the compliance department of any relevant data when inappropriate comments or potential harassment are identified.

[0446] A "compliance department" refers to the department within a company that oversees compliance with laws and regulations and ethical issues, and is responsible for receiving reports and taking appropriate action.

[0447] This invention, "AI Guardian," is a system that collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports them to the compliance department as necessary. The program processing of this system is explained in detail below.

[0448] Collection and transmission of voice data

[0449] The devices are installed at each company's bases and are always on. A high-sensitivity microphone installed in the device digitally collects audio data from workplace meetings and daily work. The collected audio data is temporarily stored on the device and converted into text data in real time using speech recognition APIs such as Google Cloud Speech-to-Text and IBM Watson Speech to Text. The converted text data is sent from the device to the server using an HTTP request.

[0450] Text data analysis

[0451] The server receives the text data sent from the device and analyzes it using a natural language processing library (e.g., spaCy or nltk). Specifically, the server tokenizes the text data and performs grammatical analysis to extract sentence structure and keywords. This analysis clarifies how the text data is structured.

[0452] Sentiment analysis and discomfort judgment

[0453] The server then performs sentiment analysis on the analyzed text data, calling sentiment analysis APIs such as the Google Natural Language API and IBM Watson Tone Analyzer to obtain a sentiment score for the text data. The sentiment score quantifies emotions such as joy, sadness, anger, and displeasure, and the server uses this score to evaluate the emotional tendency of the remarks. Based on the obtained sentiment score, the server determines the degree of discomfort, and if the degree of discomfort exceeds a certain threshold, it identifies the text data as potentially harassing.

[0454] Data storage and reporting

[0455] The server stores text data that is judged to be highly offensive in a database along with related data (audio data, transcription data, sentiment analysis results, etc.). The stored data can be referenced later for verification or as evidence. The server then immediately reports this data to the compliance department. Reports are made in the form of reports via HTTP POST requests or email, and the reports include the audio data, transcription data, and sentiment analysis results. Once a report is made, the compliance department user receives a notification, can view details on a dashboard, and can take appropriate action.

[0456] Specific examples

[0457] For example, if someone says "Work more efficiently" during a work meeting, the system will do the following:

[0458] 1. The device collects the audio of the conversation and converts it into a digital signal through a highly sensitive microphone.

[0459] 2. The device transcribes the voice data in real time, generates text data such as "Work more efficiently," and sends it to the server.

[0460] 3. The server tokenizes the received text data and analyzes its grammatical structure.

[0461] 4. The server uses the sentiment analysis API to obtain a sentiment score for the text data. In this case, it determines that "Work efficiently" is likely to indicate high levels of displeasure.

[0462] 5. The server identifies this data as potentially harassing and stores it in a database along with the audio data, transcription data, and sentiment analysis results.

[0463] 6. The server sends the generated report to the compliance department via HTTP POST request and / or email, and a compliance department user is notified.

[0464] Prompt Sentence Examples

[0465] For example, the following prompt sentence is used for a generative AI model:

[0466] If someone says "Work more efficiently" during a work meeting, determine whether this statement is potentially harassing. Tokenize the text data, perform sentiment analysis, and if a high level of discomfort is detected, explain why.

[0467] By introducing this system, companies will be able to quickly and accurately detect workplace harassment and take appropriate measures, making it a powerful tool for ensuring the psychological safety of employees and improving the work environment.

[0468] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0469] Step 1:

[0470] The device collects workplace audio data in real time. Specifically, it uses a high-sensitivity microphone to digitally capture the audio of meetings and daily work. The input is ambient analog audio data, and the output is audio data converted into a digital signal.

[0471] Step 2:

[0472] The device transcribes the collected voice data. Specifically, it converts the voice data into text data using voice recognition software (e.g., Google Cloud Speech-to-Text or IBM Watson Speech to Text). The input is digital voice data, and the output is transcribed text data.

[0473] Step 3:

[0474] The device sends the transcribed text data to the server. Specifically, it sends the text data to the server using an HTTP request or other communication protocol. The input is the transcribed text data, and the output is the text data sent to the server.

[0475] Step 4:

[0476] The server analyzes the received text data. Specifically, it uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis. The input is text data, and the output is grammatically analyzed text data.

[0477] Step 5:

[0478] The server performs sentiment analysis on the analyzed text data. Specifically, it calls a sentiment analysis API (e.g., Google Natural Language API or IBM Watson Tone Analyzer) to obtain a sentiment score. The input is the grammatically analyzed text data, and the output is a sentiment score.

[0479] Step 6:

[0480] The server determines the degree of discomfort based on the emotion score. Specifically, it evaluates whether the text data is unpleasant by comparing the acquired emotion score with a set threshold. The input is the emotion score, and the output is the evaluation result of the discomfort degree.

[0481] Step 7:

[0482] The server stores the text data judged to be highly unpleasant together with related data. Specifically, it stores the text data, emotion scores, audio data, etc. in a database. The input is the text data judged to be highly unpleasant and the related data, and the output is the data stored in the database.

[0483] Step 8:

[0484] The server then sends a report to the compliance department based on the stored data. Specifically, it sends the report in the form of a report via email or HTTP POST request. The input is the data stored in the database, and the output is a report to the compliance department.

[0485] Step 9:

[0486] Compliance department users review the received reports and take appropriate action. Specifically, they receive notifications via the software dashboard or email, check the details, and take appropriate action. The input is the report data from the server, and the output is the action taken.

[0487] (Application example 1)

[0488] 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."

[0489] In modern workplaces, especially in factories, harassment can occur during communication with workers. Such harassment not only worsens the working environment, but also increases legal risks for companies. Furthermore, factory robots and automated systems mean that on-site monitoring of working conditions is insufficient, making it difficult to protect workers.

[0490] 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.

[0491] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, means for determining the level of discomfort based on the results of the sentiment analysis, means for saving related data and notifying the management department when the level of discomfort exceeds a specific threshold, and means for collecting and analyzing voice data using factory robots to monitor physical labor and protect workers, thereby enabling early detection of harassment in the workplace and protection of workers.

[0492] "Voice data" means data for storing or processing acoustic signals in digital form, including, in particular, speech and other sounds in the workplace or on a factory floor.

[0493] "Transcription means" refers to technology or equipment for converting voice data into text data, and primarily refers to voice recognition software.

[0494] "Sentiment analysis means" refers to natural language processing techniques that analyze text data and determine the emotions and intensity of emotions contained therein.

[0495] "Discomfort level" is a measure of how much a certain statement or action makes the other person feel uncomfortable, and is calculated from the results of emotion analysis.

[0496] An "administrative department" is a department within a company or organization that is responsible for managing compliance and the working environment.

[0497] A "factory robot" is a mechanical device that performs automated tasks in a factory and may have a variety of functions, such as collecting and analyzing voice data.

[0498] "Worker protection" refers to measures and efforts to ensure that workers can work in a safe and comfortable environment.

[0499] The present invention relates to an AI Guardian system for detecting harassment and protecting workers in workplaces and factory environments. The system collects voice data in real time and analyzes the data for sentiment analysis. Specifically, the following detailed embodiments are described.

[0500] First, a microphone built into the factory robot collects voice data. This voice data is transcribed in real time using voice recognition software (e.g., Google's voice recognition API). The transcribed data is sent to a server and analyzed using a natural language processing library (e.g., spaCy or nltk).

[0501] Next, the server performs sentiment analysis. This analysis uses a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer). Sentiment analysis calculates an emotional score for the text data. The level of discomfort is determined based on this score, and if it exceeds a certain threshold, the relevant data is automatically saved and a report is sent to the management department.

[0502] The reporting function is implemented using the Twilio API, and reports are sent via SMS or email, and include collected voice data, transcription data, and sentiment analysis results.

[0503] A specific example is shown below. If someone says "You're completely useless" in a conversation within a factory, this speech is collected by a microphone. Speech recognition software transcribes this speech and sends it to a server. The server performs sentiment analysis on this text data, and if it determines that the sentiment score is high, it uses the Twilio API to notify management along with the message "You're completely useless."

[0504] Prompt Sentence Examples

[0505] Perform a sentiment analysis on the following text and print the results:

[0506] "You're completely useless"

[0507] This system will enable early detection of workplace harassment and rapid response, strengthening worker protection and reducing legal risks for companies.

[0508] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0509] Step 1:

[0510] The terminal collects the audio data. Specifically, microphones built into the factory robots capture the surrounding sounds and store them digitally. The terminal collects the audio data in real time.

[0511] Input: Ambient audio (analog data)

[0512] Output: Audio data (digital data)

[0513] Step 2:

[0514] The device transcribes the collected voice data in real time and uses speech recognition software to convert the collected voice data into text data.

[0515] Input: Audio data (digital data)

[0516] Output: Transcription data (text data)

[0517] Specific operation: The device sends voice data to the voice recognition API and receives the returned text data.

[0518] Step 3:

[0519] The device sends the transcription data to the server. The text data is sent to the server using a communication protocol such as an HTTP request.

[0520] Input: Transcription data (text data)

[0521] Output: Text data sent to the server

[0522] Specific operation: The device sends text data to the server via a communication protocol.

[0523] Step 4:

[0524] The server receives the text data and analyzes it using a natural language processing library, tokenizing the text data and performing grammatical analysis.

[0525] Input: Text data (transcription data)

[0526] Output: Analyzed text data

[0527] Specific operation: The server analyzes the text data using a natural language processing library (e.g. spaCy, nltk).

[0528] Step 5:

[0529] The server calls the sentiment analysis API to obtain the sentiment score of the text data, and determines the level of discomfort based on the sentiment score.

[0530] Input: Parsed text data

[0531] Output: Emotion score, unpleasantness judgment result

[0532] Specific operation: The server obtains the sentiment score using a sentiment analysis API (e.g., Google Natural Language API, IBM Watson Tone Analyzer).

[0533] Step 6:

[0534] If the server detects that the level of discomfort exceeds a certain threshold, it will store the relevant data and notify the management department via SMS or email.

[0535] Input: Emotion score, discomfort level judgment result

[0536] Output: Saved data, notification to management department

[0537] Specific operation: The server saves the relevant data in a database and sends a notification using the Twilio API.

[0538] This concludes the detailed explanation of each processing step in the system of the application example. This flow enables real-time monitoring of voice data, and the detection and immediate reporting of harassment.

[0539] 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.

[0540] The "AI Guardian" system of the present invention collects workplace voice data in real time and automatically detects harassing remarks. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the situation to the compliance department if necessary. The following describes the program processing of this system in detail.

[0541] 1. Collection and transmission of voice data

[0542] The device continuously captures audio data from meetings and daily work. The device's built-in microphone collects surrounding sounds as digital data.

[0543] After collecting the audio data, the device transcribes it in real time, using voice recognition software to convert the audio data into text format.

[0544] The converted text data is then sent from the device to the server using an HTTP request or other communication protocol.

[0545] 2. Sentiment analysis and unpleasantness judgment

[0546] The server receives the text data sent from the terminal, and processes the received text data for analysis.

[0547] The server uses a natural language processing library to tokenize the text data and perform grammatical analysis.

[0548] After grammar analysis, the server performs sentiment analysis on the text data using an emotion engine, which extracts sentiment scores from the text, such as joy, anger, sadness, and displeasure.

[0549] At the same time, the server also performs sentiment analysis based on the voice data, combining the results of voice recognition with the emotion engine analysis to more accurately estimate the user's emotions.

[0550] The server determines the degree of discomfort based on the results of these sentiment analyses. If the sentiment score exceeds a certain threshold, the corresponding speech is deemed to cause discomfort.

[0551] 3. Identify potential harassment

[0552] The server identifies text and audio data that is deemed highly offensive as potentially harassing, and stores the identified data along with sentiment analysis results and related data.

[0553] The stored data will be retained for future verification and use as evidence.

[0554] 4. Report to the Compliance Department

[0555] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[0556] The report includes relevant audio data, text data, and sentiment analysis results.

[0557] Specific examples

[0558] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[0559] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[0560] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[0561] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[0562] 4. The server stores the relevant data and automatically notifies the compliance department.

[0563] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[0564] In this way, the system of the present invention performs emotion analysis from both voice and text, thereby improving the accuracy of identifying harassment, enabling prompt and appropriate responses, improving the work environment, and reducing legal risks.

[0565] The processing flow will be explained below.

[0566] Step 1:

[0567] The device uses microphones installed in the workplace to capture audio data from meetings and daily work in real time.

[0568] Step 2:

[0569] The device uses speech recognition software (e.g., a speech recognition engine) to transcribe the captured voice data in real time, which is then converted into text data.

[0570] Step 3:

[0571] The terminal transmits the generated text data to the server using an applicable communication protocol such as an HTTP request.

[0572] Step 4:

[0573] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize and parse the received text data.

[0574] Step 5:

[0575] The server performs sentiment analysis on the text data using an emotion engine, which extracts emotion scores from the text, such as happiness, anger, sadness, displeasure, etc.

[0576] Step 6:

[0577] The server also performs sentiment analysis of the voice data using an emotion engine, which analyzes the tone, stress, and intonation of the voice to identify emotions.

[0578] Step 7:

[0579] The server combines the emotion scores from both the text and the audio to determine the annoyance level, and evaluates whether the annoyance level exceeds a certain threshold.

[0580] Step 8:

[0581] If the server determines that the offense level is high, it identifies the data as potentially harassing and stores the associated audio data, transcription data, and sentiment analysis data in a database.

[0582] Step 9:

[0583] The server generates detailed reports based on the stored data and automatically reports them to the compliance department via email or a dedicated dashboard.

[0584] Step 10:

[0585] The user (compliance department staff) receives the report and checks the notified report on the dashboard. Based on the detailed information, they take necessary measures and, in some cases, provide support to the victim.

[0586] Example 2

[0587] 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."

[0588] Detecting and responding to workplace harassment is extremely important, but traditional methods rely primarily on human monitoring and reporting, making it difficult to respond quickly and objectively. This can lead to a worsening work environment, causing mental stress and reduced productivity for employees. Furthermore, traditional methods make it difficult to preserve and manage evidence of harassment, which can delay necessary responses.

[0589] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0590] In this invention, the server includes: means for collecting voice data in real time; means for transcribing the voice data; means for analyzing the transcription data and the voice data using a natural language processing library; means for performing sentiment analysis of the transcription data and the voice data using an emotion engine; means for determining the level of discomfort based on the results of the sentiment analysis; and means for storing related data, generating data for reporting, and reporting to a compliance department when the level of discomfort exceeds a specific threshold. This enables real-time detection and rapid response to workplace harassment. Furthermore, by appropriately storing evidence of highly offensive remarks and taking action based on objective data, the system can improve the work environment and reduce employee mental stress.

[0591] "Audio data" refers to data that records audio in digital format.

[0592] "Transcription" is the process of converting audio data into text form.

[0593] A "natural language processing library" is a set of software tools for analyzing text data, including tokenization and grammatical analysis.

[0594] "Emotion Engine" refers to algorithms and techniques for extracting emotion scores from text and audio data.

[0595] The "discomfort level" is an index that indicates the degree of a particular emotional state calculated based on the emotional score.

[0596] A "threshold" is a boundary value for triggering a specific action when an indicator such as discomfort level or emotion score exceeds a certain standard.

[0597] "Related data" refers to all data related to determining harassment, such as the results of sentiment analysis, the content of statements, and audio data.

[0598] A "compliance department" is a specialized department within a company that monitors and promotes compliance with laws and regulations and ethical behavior.

[0599] "Notification" refers to the process of reporting specific events or data.

[0600] "Data storage" refers to storing acquired data for future use.

[0601] The system of this invention, "AI Guardian," collects workplace voice data in real time and automatically detects harassing comments. It also incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the incident to the compliance department if necessary. This system is primarily implemented using the following hardware and software:

[0602] Audio data collection and transcription

[0603] The device is equipped with a microphone to continuously capture audio data during meetings and daily work. It is also equipped with speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson. This software converts the collected audio data into text in real time. The text data is then sent to a server using a communication protocol such as an HTTP request.

[0604] Text data analysis and sentiment analysis

[0605] When the server receives the text data sent from the device, it analyzes the data using a natural language processing library (such as NLTK or spaCy). First, the text data is tokenized and grammatically analyzed. Then, an emotion engine (such as the Text Analytics API from Azure Cognitive Services) is used to perform a sentiment analysis of the text and audio data. This emotion engine extracts an emotion score from the text and audio, quantifying emotions such as joy, anger, sadness, and displeasure.

[0606] Determining discomfort and identifying harassment

[0607] The server analyzes the emotion scores obtained from the emotion engine and determines the level of offense. If an emotion score exceeding a certain threshold is detected, the remark is deemed to be offensive. This highly offensive remark is identified as potentially harassing, and the related data (audio data, text data, and emotion analysis results) is stored.

[0608] Report to the Compliance Department

[0609] The server generates a report based on data judged to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard. The generated report includes related audio data, text data, and sentiment analysis results.

[0610] Specific examples

[0611] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[0612] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[0613] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[0614] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[0615] 4. The server stores the relevant data and automatically notifies the compliance department.

[0616] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[0617] Prompt Sentence Examples

[0618] For example, this process begins by inputting a prompt such as, "Please analyze the statement 'Work more efficiently' made during a meeting." into the generative AI model.

[0619] In this way, this system, "AI Guardian," performs emotion analysis from both voice and text, improving the accuracy of identifying harassment and enabling prompt and appropriate responses, thereby improving the work environment and reducing legal risks.

[0620] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0621] Step 1:

[0622] The device uses a microphone to collect real-time audio data from workplace meetings and daily work.

[0623] Input: Ambient conversation

[0624] Output: Digital audio data

[0625] Specific operation: The microphone installed in the device captures audio and stores it as digital data.

[0626] Step 2:

[0627] The device transcribes the collected voice data using speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson.

[0628] Input: Audio data

[0629] Output: Transcribed text data

[0630] Specific operation: Voice data is input into voice recognition software, and the conversion result is obtained as text data.

[0631] Step 3:

[0632] The device sends the generated text data to the server using an HTTP request, using HTTP or HTTPS as the communication protocol.

[0633] Input: Transcription data

[0634] Output: Text data sent to the server

[0635] Specific operation: Include text data in the body of an HTTP request and send it to the API endpoint.

[0636] Step 4:

[0637] The server uses a natural language processing library (such as NLTK or spaCy) to tokenize and grammatically analyze the received text data.

[0638] Input: Received text data

[0639] Output: Tokenized and parsed text data

[0640] Specific operation: The text data is broken down into tokens and the grammatical structure is analyzed.

[0641] Step 5:

[0642] The server uses an emotion engine (such as the Text Analytics API from Azure Cognitive Services) to perform emotion analysis on the text and audio data, and extracts an emotion score.

[0643] Input: Tokenized and grammatically analyzed text data, audio data

[0644] Output: Sentiment score

[0645] Specific operation: The analyzed text data and voice data are input into the emotion engine to calculate emotion scores such as joy, anger, sadness, and displeasure.

[0646] Step 6:

[0647] The server determines the degree of discomfort based on the extracted emotion score. If the degree of discomfort exceeds a certain threshold, the comment is deemed to be unpleasant.

[0648] Input: Sentiment score

[0649] Output: Discomfort level judgment result

[0650] Specific operation: The emotion score is compared with a specific threshold, and if the threshold is exceeded, the unpleasantness level is determined to be high.

[0651] Step 7:

[0652] The server identifies text and audio data that is determined to be highly offensive as potentially harassing and stores the associated data.

[0653] Input: Discomfort level judgment result

[0654] Output: Saved text and audio data

[0655] Specific action: Potentially harassing data will be stored in a database.

[0656] Step 8:

[0657] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[0658] Input: Stored text and audio data

[0659] Output: Generated reports, notification

[0660] Specific Action: Generates a report and sends an email using SMTP or displays a notification on the dashboard.

[0661] Step 9:

[0662] The user (compliance department officer) receives a notification, checks detailed information on the dashboard, and takes the necessary action.

[0663] Input: Report notification, detailed information

[0664] Output: Response Action

[0665] Specific Action: Review reports on the dashboard and take appropriate action.

[0666] (Application example 2)

[0667] 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."

[0668] Conventional harassment detection systems are limited to the workplace, making it difficult to effectively monitor harassment and inappropriate comments in public places or the wider society. Furthermore, manual reporting and response is required, resulting in delayed real-time responses. Therefore, there is a need for a system that can detect inappropriate comments in real time, even in public places, and respond quickly.

[0669] 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.

[0670] In this invention, the server includes means for collecting voice data from workplaces and public places in real time, transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, and means for determining the level of discomfort based on the results of the sentiment analysis. This makes it possible to detect harassment and inappropriate remarks in real time not only in the workplace but also in public places, and to report them promptly.

[0671] "Workplaces and public places" refers not only to workplaces but also to environments that include places widely used by the general public, such as public transportation, parks, and commercial facilities.

[0672] "Audio data" refers to data that digitally represents people's conversations and environmental sounds collected via a voice input device such as a microphone.

[0673] "Transcription" is the process of analyzing audio data and converting its content into corresponding text data.

[0674] "Sentiment analysis" is a technology that detects emotions from text data or audio data and evaluates the content as an emotional score such as joy, anger, sadness, or displeasure.

[0675] "Offensiveness" is an index calculated based on sentiment analysis that assesses the degree to which a certain statement or event causes offense.

[0676] A "security department" refers to a specialized department or team that responds when inappropriate remarks or harassment are detected.

[0677] "Reporting" is the process by which the system automatically notifies the security department of relevant data when it detects potentially offensive speech.

[0678] "Speech recognition" is a technology that analyzes voice data and converts what is being said into text.

[0679] "Natural language processing" is a general term for artificial intelligence technology that analyzes text data and understands and analyzes its meaning and emotions.

[0680] "Harassment likelihood" refers to an assessment of the likelihood that a particular statement constitutes harassment based on sentiment analysis.

[0681] The system for implementing this invention, "AI Harassment Guardian," has the function of collecting voice data in real time in the workplace and public places, detecting and reporting inappropriate remarks and harassment. Specifically, it consists of the following elements:

[0682] 1. System Configuration

[0683] Hardware

[0684] Smartphones: Equipped with microphones for collecting audio data.

[0685] software

[0686] Python: Used as a programming language.

[0687] pydub: A library used to manipulate and process audio data.

[0688] vosk: An open-source speech recognition model library.

[0689] transformers: A natural language processing toolkit for sentiment analysis.

[0690] 2. System Operation

[0691] Collection and processing of voice data

[0692] The device (smartphone) captures conversations in meetings and public places in real time via a microphone. The captured audio data is processed by the pydub library and standardized to a 16 kHz sample rate.

[0693] Voice Recognition

[0694] The server transcribes the processed audio data using the vosk library, which converts the audio data into text data.

[0695] sentiment analysis

[0696] The server performs sentiment analysis on the converted text data using the transformers library, and as a result, the text data is assigned an emotional score such as joy, anger, sadness, or displeasure.

[0697] Determining discomfort and reporting

[0698] The server determines the level of discomfort based on the emotional score, and if the level of discomfort exceeds a certain threshold, it is deemed to be harassment. This information is automatically reported to the security department via email or a dedicated dashboard.

[0699] Specific examples

[0700] For example, if someone says "Go away!" on a public bus, the following will happen:

[0701] 1. The device (smartphone) captures the audio and transcribes what is said.

[0702] 2. The server performs sentiment analysis on the transcribed data and determines that the level of unpleasantness is high.

[0703] 3. The server stores this data and notifies the security department.

[0704] Prompt Sentence Examples

[0705] Prompt statement:

[0706] "Audio data is captured, transcribed in real time, and then sentiment analysis is performed. If a specific negative sentiment is indicated with a high score, the remark is determined to be harassment and a report is sent."

[0707] This system will enable real-time monitoring and rapid response to harassment and inappropriate comments not only in the workplace but also in public places.

[0708] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0709] Step 1:

[0710] The device (smartphone) collects audio data in real time during meetings and public places. The microphone on the device captures the surrounding sounds and saves them as audio data. The audio data is then stored digitally on the device.

[0711] Step 2:

[0712] The device converts the collected audio data to a 16kHz sample rate using the pydub library, which processes the audio data into a format suitable for voice recognition.

[0713] Step 3:

[0714] The device transcribes the processed voice data using the vosk library, recognizes the input voice data, and outputs the corresponding text data. The resulting conversation is saved in text format on the device.

[0715] Step 4:

[0716] The terminal sends the generated text data to the server using a communication protocol such as an HTTP request. At this time, the text data is transferred to the server.

[0717] Step 5:

[0718] The server receives the text data and, to analyze it, tokenizes it using a natural language processing (NLP) library, converting the text data into a format that can be analyzed.

[0719] Step 6:

[0720] The server uses the transformers library for sentiment analysis on the tokenized text data: it computes sentiment scores from the tokenized data as input, evaluates the sentiment (e.g., joy, anger, sadness, displeasure, etc.) for each text fragment, and produces a sentiment score as output.

[0721] Step 7:

[0722] The server determines the offensiveness based on the sentiment score. It analyzes each sentiment score to determine whether the offensiveness exceeds a certain threshold. If the offensiveness is high, the text fragment is determined to be likely to be harassing.

[0723] Step 8:

[0724] The server stores text fragments that are judged to be highly offensive and their emotion scores as related data, which can be used for later verification or as evidence.

[0725] Step 9:

[0726] The server automatically sends alerts to security teams based on relevant data, either via email or a dedicated dashboard, and includes sentiment analysis results along with relevant data.

[0727] Step 10:

[0728] The user (security department staff) receives a notification, checks the details on the dashboard, and takes immediate action if necessary.

[0729] 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.

[0730] 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.

[0731] 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.

[0732] [Third embodiment]

[0733] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0734] 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.

[0735] 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).

[0736] 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.

[0737] 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.

[0738] 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).

[0739] 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.

[0740] 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.

[0741] 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.

[0742] 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.

[0743] 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.

[0744] 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."

[0745] The AI ​​Guardian system of the present invention collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports the incident to the compliance department as necessary. The following describes the program processing of this system in detail.

[0746] 1. Collection and transmission of voice data

[0747] The device is set up to constantly capture audio data from meetings and daily work. The microphones on the device collect surrounding sounds as digital data.

[0748] Once the voice data is collected, the device transcribes it in real time by converting speech to text using speech recognition software (e.g., open source or commercial speech recognition APIs) built into the device.

[0749] The transcribed text data is sent from the device to a server using HTTP requests or other communication protocols.

[0750] 2. Sentiment analysis and unpleasantness judgment

[0751] The server receives the text data sent from the terminal, and the received text data is processed within the server for analysis.

[0752] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis.

[0753] Based on the results of natural language processing, the server performs sentiment analysis by calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[0754] Based on the emotion score, the server determines the degree of annoyance, and if the degree of annoyance exceeds a certain threshold, the text data is identified as potentially harassing.

[0755] 3. Identify potential harassment

[0756] The server identifies text data that it determines to be highly offensive as potentially harassing, and stores the identified data together with audio data, transcription data, and sentiment analysis data.

[0757] The stored data is retained for later verification and use as evidence.

[0758] 4. Report to the Compliance Department

[0759] The server automatically notifies the compliance department when potentially harassing data is stored, in the form of a report of the stored data, including audio data, transcription data, and sentiment analysis results.

[0760] Once a report is made, compliance users receive a notification via the software dashboard and email.

[0761] Specific examples

[0762] For example, if someone says "Work more efficiently" during a work meeting, this comment may cause offense. In that case, the following process will occur:

[0763] 1. The device captures the audio of the conversation, transcribes it in real time into text data such as "Work more efficiently," and sends it to the server.

[0764] 2. The server analyzes the text data and performs sentiment analysis, determining that the comment is likely to cause high levels of offense.

[0765] 3. The server will store the data as potentially harassing data and report the relevant data to the compliance department.

[0766] 4. Compliance users receive a notification, view detailed information on a dashboard, and take appropriate action.

[0767] As described above, the system of the present invention automatically detects and notifies workplace harassment, enabling prompt and appropriate responses, thereby contributing to protecting victims and reducing legal risks for companies.

[0768] The processing flow will be explained below.

[0769] Step 1:

[0770] The device uses a microphone to capture surrounding sounds in order to constantly collect voice data within the workplace.

[0771] Step 2:

[0772] The device transcribes the captured audio data in real time by converting the audio into text using a speech recognition API (e.g., a speech recognition engine).

[0773] Step 3:

[0774] The terminal transmits the generated text data to the server using a communication protocol such as an HTTP request.

[0775] Step 4:

[0776] To analyze the received text data, the server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text and perform grammatical analysis.

[0777] Step 5:

[0778] The server performs sentiment analysis based on the analyzed text data, calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[0779] Step 6:

[0780] The server evaluates the emotion score to determine the level of annoyance, and identifies potential harassment if the level of annoyance exceeds a certain threshold.

[0781] Step 7:

[0782] The server stores audio data, transcription data, and sentiment analysis data related to text data determined to be potentially harassing in a database.

[0783] Step 8:

[0784] The server generates reports based on the stored data and sends them to the compliance department via email or a dedicated dashboard.

[0785] Step 9:

[0786] The user (compliance department officer) receives the report, checks the detailed information on the provided dashboard, takes necessary measures based on the content of the notification, and, in some cases, provides support to the victim.

[0787] Example 1

[0788] 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."

[0789] Workplace harassment is a serious problem that causes psychological stress for employees and reduces productivity throughout an organization. However, detecting and responding to harassment on the spot is difficult, and victims often miss opportunities to receive appropriate support. Furthermore, traditional manual monitoring and reporting systems lack immediacy, making it difficult to respond quickly. Therefore, there is a need for the development of a system that can analyze workplace audio data in real time, automatically detect inappropriate remarks, and appropriately report them.

[0790] 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.

[0791] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing natural language processing to analyze the structure of text data, means for performing sentiment analysis based on the analysis results and obtaining a sentiment score, means for determining the degree of discomfort based on the sentiment score and storing related data when a specific threshold is exceeded, means for reporting to a compliance department based on the determination result, and means for automating the reporting. This makes it possible to automatically and quickly detect workplace harassment and immediately report it to the appropriate department, thereby protecting victims and improving the soundness of the organization.

[0792] "Voice data" refers to any audio information generated in the workplace and is collected digitally using a microphone.

[0793] "Transcription means" refers to the technology or software used to convert voice data into text data, and uses voice recognition technology to convert voice into text in real time.

[0794] "Natural language processing means" refers to technology and software that analyzes text data and understands its grammatical structure and meaning, and performs tokenization and syntactic analysis.

[0795] "Sentiment analysis" refers to the technology of analyzing the content of text data and quantifying the emotions expressed therein, obtaining emotional scores such as joy, sadness, anger, and displeasure.

[0796] "Offensiveness assessment means" refers to a technology or method for assessing the offensiveness of text data based on the results of sentiment analysis, and determining that the content is inappropriate if it exceeds a certain threshold.

[0797] "Related data" refers to all data related to the detected inappropriate remarks, including audio data, transcription data, sentiment analysis results, the date and time of the remarks, and information about the speaker.

[0798] "Reporting tools" refers to technology or software that automatically notifies the compliance department of any relevant data when inappropriate comments or potential harassment are identified.

[0799] A "compliance department" refers to the department within a company that oversees compliance with laws and regulations and ethical issues, and is responsible for receiving reports and taking appropriate action.

[0800] This invention, "AI Guardian," is a system that collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports them to the compliance department as necessary. The program processing of this system is explained in detail below.

[0801] Collection and transmission of voice data

[0802] The devices are installed at each company's bases and are always on. A high-sensitivity microphone installed in the device digitally collects audio data from workplace meetings and daily work. The collected audio data is temporarily stored on the device and converted into text data in real time using speech recognition APIs such as Google Cloud Speech-to-Text and IBM Watson Speech to Text. The converted text data is sent from the device to the server using an HTTP request.

[0803] Text data analysis

[0804] The server receives the text data sent from the device and analyzes it using a natural language processing library (e.g., spaCy or nltk). Specifically, the server tokenizes the text data and performs grammatical analysis to extract sentence structure and keywords. This analysis clarifies how the text data is structured.

[0805] Sentiment analysis and discomfort judgment

[0806] The server then performs sentiment analysis on the analyzed text data, calling sentiment analysis APIs such as the Google Natural Language API and IBM Watson Tone Analyzer to obtain a sentiment score for the text data. The sentiment score quantifies emotions such as joy, sadness, anger, and displeasure, and the server uses this score to evaluate the emotional tendency of the remarks. Based on the obtained sentiment score, the server determines the degree of discomfort, and if the degree of discomfort exceeds a certain threshold, it identifies the text data as potentially harassing.

[0807] Data storage and reporting

[0808] The server stores text data that is judged to be highly offensive in a database along with related data (audio data, transcription data, sentiment analysis results, etc.). The stored data can be referenced later for verification or as evidence. The server then immediately reports this data to the compliance department. Reports are made in the form of reports via HTTP POST requests or email, and the reports include the audio data, transcription data, and sentiment analysis results. Once a report is made, the compliance department user receives a notification, can view details on a dashboard, and can take appropriate action.

[0809] Specific examples

[0810] For example, if someone says "Work more efficiently" during a work meeting, the system will do the following:

[0811] 1. The device collects the audio of the conversation and converts it into a digital signal through a highly sensitive microphone.

[0812] 2. The device transcribes the voice data in real time, generates text data such as "Work more efficiently," and sends it to the server.

[0813] 3. The server tokenizes the received text data and analyzes its grammatical structure.

[0814] 4. The server uses the sentiment analysis API to obtain a sentiment score for the text data. In this case, it determines that "Work efficiently" is likely to indicate high levels of displeasure.

[0815] 5. The server identifies this data as potentially harassing and stores it in a database along with the audio data, transcription data, and sentiment analysis results.

[0816] 6. The server sends the generated report to the compliance department via HTTP POST request and / or email, and a compliance department user is notified.

[0817] Prompt Sentence Examples

[0818] For example, the following prompt sentence is used for a generative AI model:

[0819] If someone says "Work more efficiently" during a work meeting, determine whether this statement is potentially harassing. Tokenize the text data, perform sentiment analysis, and if a high level of discomfort is detected, explain why.

[0820] By introducing this system, companies will be able to quickly and accurately detect workplace harassment and take appropriate measures, making it a powerful tool for ensuring the psychological safety of employees and improving the work environment.

[0821] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0822] Step 1:

[0823] The device collects workplace audio data in real time. Specifically, it uses a high-sensitivity microphone to digitally capture the audio of meetings and daily work. The input is ambient analog audio data, and the output is audio data converted into a digital signal.

[0824] Step 2:

[0825] The device transcribes the collected voice data. Specifically, it converts the voice data into text data using voice recognition software (e.g., Google Cloud Speech-to-Text or IBM Watson Speech to Text). The input is digital voice data, and the output is transcribed text data.

[0826] Step 3:

[0827] The device sends the transcribed text data to the server. Specifically, it sends the text data to the server using an HTTP request or other communication protocol. The input is the transcribed text data, and the output is the text data sent to the server.

[0828] Step 4:

[0829] The server analyzes the received text data. Specifically, it uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis. The input is text data, and the output is grammatically analyzed text data.

[0830] Step 5:

[0831] The server performs sentiment analysis on the analyzed text data. Specifically, it calls a sentiment analysis API (e.g., Google Natural Language API or IBM Watson Tone Analyzer) to obtain a sentiment score. The input is the grammatically analyzed text data, and the output is a sentiment score.

[0832] Step 6:

[0833] The server determines the degree of discomfort based on the emotion score. Specifically, it evaluates whether the text data is unpleasant by comparing the acquired emotion score with a set threshold. The input is the emotion score, and the output is the evaluation result of the discomfort degree.

[0834] Step 7:

[0835] The server stores the text data judged to be highly unpleasant together with related data. Specifically, it stores the text data, emotion scores, audio data, etc. in a database. The input is the text data judged to be highly unpleasant and the related data, and the output is the data stored in the database.

[0836] Step 8:

[0837] The server then sends a report to the compliance department based on the stored data. Specifically, it sends the report in the form of a report via email or HTTP POST request. The input is the data stored in the database, and the output is a report to the compliance department.

[0838] Step 9:

[0839] Compliance department users review the received reports and take appropriate action. Specifically, they receive notifications via the software dashboard or email, check the details, and take appropriate action. The input is the report data from the server, and the output is the action taken.

[0840] (Application example 1)

[0841] 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."

[0842] In modern workplaces, especially in factories, harassment can occur during communication with workers. Such harassment not only worsens the working environment, but also increases legal risks for companies. Furthermore, factory robots and automated systems mean that on-site monitoring of working conditions is insufficient, making it difficult to protect workers.

[0843] 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.

[0844] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, means for determining the level of discomfort based on the results of the sentiment analysis, means for saving related data and notifying the management department when the level of discomfort exceeds a specific threshold, and means for collecting and analyzing voice data using factory robots to monitor physical labor and protect workers, thereby enabling early detection of harassment in the workplace and protection of workers.

[0845] "Voice data" means data for storing or processing acoustic signals in digital form, including, in particular, speech and other sounds in the workplace or on a factory floor.

[0846] "Transcription means" refers to technology or equipment for converting voice data into text data, and primarily refers to voice recognition software.

[0847] "Sentiment analysis means" refers to natural language processing techniques that analyze text data and determine the emotions and intensity of emotions contained therein.

[0848] "Discomfort level" is a measure of how much a certain statement or action makes the other person feel uncomfortable, and is calculated from the results of emotion analysis.

[0849] An "administrative department" is a department within a company or organization that is responsible for managing compliance and the working environment.

[0850] A "factory robot" is a mechanical device that performs automated tasks in a factory and may have a variety of functions, such as collecting and analyzing voice data.

[0851] "Worker protection" refers to measures and efforts to ensure that workers can work in a safe and comfortable environment.

[0852] The present invention relates to an AI Guardian system for detecting harassment and protecting workers in workplaces and factory environments. The system collects voice data in real time and analyzes the data for sentiment analysis. Specifically, the following detailed embodiments are described.

[0853] First, a microphone built into the factory robot collects voice data. This voice data is transcribed in real time using voice recognition software (e.g., Google's voice recognition API). The transcribed data is sent to a server and analyzed using a natural language processing library (e.g., spaCy or nltk).

[0854] Next, the server performs sentiment analysis. This analysis uses a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer). Sentiment analysis calculates an emotional score for the text data. The level of discomfort is determined based on this score, and if it exceeds a certain threshold, the relevant data is automatically saved and a report is sent to the management department.

[0855] The reporting function is implemented using the Twilio API, and reports are sent via SMS or email, and include collected voice data, transcription data, and sentiment analysis results.

[0856] A specific example is shown below. If someone says "You're completely useless" in a conversation within a factory, this speech is collected by a microphone. Speech recognition software transcribes this speech and sends it to a server. The server performs sentiment analysis on this text data, and if it determines that the sentiment score is high, it uses the Twilio API to notify management along with the message "You're completely useless."

[0857] Prompt Sentence Examples

[0858] Perform a sentiment analysis on the following text and print the results:

[0859] "You're completely useless"

[0860] This system will enable early detection of workplace harassment and rapid response, strengthening worker protection and reducing legal risks for companies.

[0861] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0862] Step 1:

[0863] The terminal collects the audio data. Specifically, microphones built into the factory robots capture the surrounding sounds and store them digitally. The terminal collects the audio data in real time.

[0864] Input: Ambient audio (analog data)

[0865] Output: Audio data (digital data)

[0866] Step 2:

[0867] The device transcribes the collected voice data in real time and uses speech recognition software to convert the collected voice data into text data.

[0868] Input: Audio data (digital data)

[0869] Output: Transcription data (text data)

[0870] Specific operation: The device sends voice data to the voice recognition API and receives the returned text data.

[0871] Step 3:

[0872] The device sends the transcription data to the server. The text data is sent to the server using a communication protocol such as an HTTP request.

[0873] Input: Transcription data (text data)

[0874] Output: Text data sent to the server

[0875] Specific operation: The device sends text data to the server via a communication protocol.

[0876] Step 4:

[0877] The server receives the text data and analyzes it using a natural language processing library, tokenizing the text data and performing grammatical analysis.

[0878] Input: Text data (transcription data)

[0879] Output: Analyzed text data

[0880] Specific operation: The server analyzes the text data using a natural language processing library (e.g. spaCy, nltk).

[0881] Step 5:

[0882] The server calls the sentiment analysis API to obtain the sentiment score of the text data, and determines the level of discomfort based on the sentiment score.

[0883] Input: Parsed text data

[0884] Output: Emotion score, unpleasantness judgment result

[0885] Specific operation: The server obtains the sentiment score using a sentiment analysis API (e.g., Google Natural Language API, IBM Watson Tone Analyzer).

[0886] Step 6:

[0887] If the server detects that the level of discomfort exceeds a certain threshold, it will store the relevant data and notify the management department via SMS or email.

[0888] Input: Emotion score, discomfort level judgment result

[0889] Output: Saved data, notification to management department

[0890] Specific operation: The server saves the relevant data in a database and sends a notification using the Twilio API.

[0891] This concludes the detailed explanation of each processing step in the system of the application example. This flow enables real-time monitoring of voice data, and the detection and immediate reporting of harassment.

[0892] 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.

[0893] The "AI Guardian" system of the present invention collects workplace voice data in real time and automatically detects harassing remarks. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the situation to the compliance department if necessary. The following describes the program processing of this system in detail.

[0894] 1. Collection and transmission of voice data

[0895] The device continuously captures audio data from meetings and daily work. The device's built-in microphone collects surrounding sounds as digital data.

[0896] After collecting the audio data, the device transcribes it in real time, using voice recognition software to convert the audio data into text format.

[0897] The converted text data is then sent from the device to the server using an HTTP request or other communication protocol.

[0898] 2. Sentiment analysis and unpleasantness judgment

[0899] The server receives the text data sent from the terminal, and processes the received text data for analysis.

[0900] The server uses a natural language processing library to tokenize the text data and perform grammatical analysis.

[0901] After grammar analysis, the server performs sentiment analysis on the text data using an emotion engine, which extracts sentiment scores from the text, such as joy, anger, sadness, and displeasure.

[0902] At the same time, the server also performs sentiment analysis based on the voice data, combining the results of voice recognition with the emotion engine analysis to more accurately estimate the user's emotions.

[0903] The server determines the degree of discomfort based on the results of these sentiment analyses. If the sentiment score exceeds a certain threshold, the corresponding speech is deemed to cause discomfort.

[0904] 3. Identify potential harassment

[0905] The server identifies text and audio data that is deemed highly offensive as potentially harassing, and stores the identified data along with sentiment analysis results and related data.

[0906] The stored data will be retained for future verification and use as evidence.

[0907] 4. Report to the Compliance Department

[0908] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[0909] The report includes relevant audio data, text data, and sentiment analysis results.

[0910] Specific examples

[0911] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[0912] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[0913] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[0914] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[0915] 4. The server stores the relevant data and automatically notifies the compliance department.

[0916] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[0917] In this way, the system of the present invention performs emotion analysis from both voice and text, thereby improving the accuracy of identifying harassment, enabling prompt and appropriate responses, improving the work environment, and reducing legal risks.

[0918] The processing flow will be explained below.

[0919] Step 1:

[0920] The device uses microphones installed in the workplace to capture audio data from meetings and daily work in real time.

[0921] Step 2:

[0922] The device uses speech recognition software (e.g., a speech recognition engine) to transcribe the captured voice data in real time, which is then converted into text data.

[0923] Step 3:

[0924] The terminal transmits the generated text data to the server using an applicable communication protocol such as an HTTP request.

[0925] Step 4:

[0926] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize and parse the received text data.

[0927] Step 5:

[0928] The server performs sentiment analysis on the text data using an emotion engine, which extracts emotion scores from the text, such as happiness, anger, sadness, displeasure, etc.

[0929] Step 6:

[0930] The server also performs sentiment analysis of the voice data using an emotion engine, which analyzes the tone, stress, and intonation of the voice to identify emotions.

[0931] Step 7:

[0932] The server combines the emotion scores from both the text and the audio to determine the annoyance level, and evaluates whether the annoyance level exceeds a certain threshold.

[0933] Step 8:

[0934] If the server determines that the offense level is high, it identifies the data as potentially harassing and stores the associated audio data, transcription data, and sentiment analysis data in a database.

[0935] Step 9:

[0936] The server generates detailed reports based on the stored data and automatically reports them to the compliance department via email or a dedicated dashboard.

[0937] Step 10:

[0938] The user (compliance department staff) receives the report and checks the notified report on the dashboard. Based on the detailed information, they take necessary measures and, in some cases, provide support to the victim.

[0939] Example 2

[0940] 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."

[0941] Detecting and responding to workplace harassment is extremely important, but traditional methods rely primarily on human monitoring and reporting, making it difficult to respond quickly and objectively. This can lead to a worsening work environment, causing mental stress and reduced productivity for employees. Furthermore, traditional methods make it difficult to preserve and manage evidence of harassment, which can delay necessary responses.

[0942] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0943] In this invention, the server includes: means for collecting voice data in real time; means for transcribing the voice data; means for analyzing the transcription data and the voice data using a natural language processing library; means for performing sentiment analysis of the transcription data and the voice data using an emotion engine; means for determining the level of discomfort based on the results of the sentiment analysis; and means for storing related data, generating data for reporting, and reporting to a compliance department when the level of discomfort exceeds a specific threshold. This enables real-time detection and rapid response to workplace harassment. Furthermore, by appropriately storing evidence of highly offensive remarks and taking action based on objective data, the system can improve the work environment and reduce employee mental stress.

[0944] "Audio data" refers to data that records audio in digital format.

[0945] "Transcription" is the process of converting audio data into text form.

[0946] A "natural language processing library" is a set of software tools for analyzing text data, including tokenization and grammatical analysis.

[0947] "Emotion Engine" refers to algorithms and techniques for extracting emotion scores from text and audio data.

[0948] The "discomfort level" is an index that indicates the degree of a particular emotional state calculated based on the emotional score.

[0949] A "threshold" is a boundary value for triggering a specific action when an indicator such as discomfort level or emotion score exceeds a certain standard.

[0950] "Related data" refers to all data related to determining harassment, such as the results of sentiment analysis, the content of statements, and audio data.

[0951] A "compliance department" is a specialized department within a company that monitors and promotes compliance with laws and regulations and ethical behavior.

[0952] "Notification" refers to the process of reporting specific events or data.

[0953] "Data storage" refers to storing acquired data for future use.

[0954] The system of this invention, "AI Guardian," collects workplace voice data in real time and automatically detects harassing comments. It also incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the incident to the compliance department if necessary. This system is primarily implemented using the following hardware and software:

[0955] Audio data collection and transcription

[0956] The device is equipped with a microphone to continuously capture audio data during meetings and daily work. It is also equipped with speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson. This software converts the collected audio data into text in real time. The text data is then sent to a server using a communication protocol such as an HTTP request.

[0957] Text data analysis and sentiment analysis

[0958] When the server receives the text data sent from the device, it analyzes the data using a natural language processing library (such as NLTK or spaCy). First, the text data is tokenized and grammatically analyzed. Then, an emotion engine (such as the Text Analytics API from Azure Cognitive Services) is used to perform a sentiment analysis of the text and audio data. This emotion engine extracts an emotion score from the text and audio, quantifying emotions such as joy, anger, sadness, and displeasure.

[0959] Determining discomfort and identifying harassment

[0960] The server analyzes the emotion scores obtained from the emotion engine and determines the level of offense. If an emotion score exceeding a certain threshold is detected, the remark is deemed to be offensive. This highly offensive remark is identified as potentially harassing, and the related data (audio data, text data, and emotion analysis results) is stored.

[0961] Report to the Compliance Department

[0962] The server generates a report based on data judged to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard. The generated report includes related audio data, text data, and sentiment analysis results.

[0963] Specific examples

[0964] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[0965] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[0966] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[0967] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[0968] 4. The server stores the relevant data and automatically notifies the compliance department.

[0969] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[0970] Prompt Sentence Examples

[0971] For example, this process begins by inputting a prompt such as, "Please analyze the statement 'Work more efficiently' made during a meeting." into the generative AI model.

[0972] In this way, this system, "AI Guardian," performs emotion analysis from both voice and text, improving the accuracy of identifying harassment and enabling prompt and appropriate responses, thereby improving the work environment and reducing legal risks.

[0973] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0974] Step 1:

[0975] The device uses a microphone to collect real-time audio data from workplace meetings and daily work.

[0976] Input: Ambient conversation

[0977] Output: Digital audio data

[0978] Specific operation: The microphone installed in the device captures audio and stores it as digital data.

[0979] Step 2:

[0980] The device transcribes the collected voice data using speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson.

[0981] Input: Audio data

[0982] Output: Transcribed text data

[0983] Specific operation: Voice data is input into voice recognition software, and the conversion result is obtained as text data.

[0984] Step 3:

[0985] The device sends the generated text data to the server using an HTTP request, using HTTP or HTTPS as the communication protocol.

[0986] Input: Transcription data

[0987] Output: Text data sent to the server

[0988] Specific operation: Include text data in the body of an HTTP request and send it to the API endpoint.

[0989] Step 4:

[0990] The server uses a natural language processing library (such as NLTK or spaCy) to tokenize and grammatically analyze the received text data.

[0991] Input: Received text data

[0992] Output: Tokenized and parsed text data

[0993] Specific operation: The text data is broken down into tokens and the grammatical structure is analyzed.

[0994] Step 5:

[0995] The server uses an emotion engine (such as the Text Analytics API from Azure Cognitive Services) to perform emotion analysis on the text and audio data, and extracts an emotion score.

[0996] Input: Tokenized and grammatically analyzed text data, audio data

[0997] Output: Sentiment score

[0998] Specific operation: The analyzed text data and voice data are input into the emotion engine to calculate emotion scores such as joy, anger, sadness, and displeasure.

[0999] Step 6:

[1000] The server determines the degree of discomfort based on the extracted emotion score. If the degree of discomfort exceeds a certain threshold, the comment is deemed to be unpleasant.

[1001] Input: Sentiment score

[1002] Output: Discomfort level judgment result

[1003] Specific operation: The emotion score is compared with a specific threshold, and if the threshold is exceeded, the unpleasantness level is determined to be high.

[1004] Step 7:

[1005] The server identifies text and audio data that is determined to be highly offensive as potentially harassing and stores the associated data.

[1006] Input: Discomfort level judgment result

[1007] Output: Saved text and audio data

[1008] Specific action: Potentially harassing data will be stored in a database.

[1009] Step 8:

[1010] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[1011] Input: Stored text and audio data

[1012] Output: Generated reports, notification

[1013] Specific Action: Generates a report and sends an email using SMTP or displays a notification on the dashboard.

[1014] Step 9:

[1015] The user (compliance department officer) receives a notification, checks detailed information on the dashboard, and takes the necessary action.

[1016] Input: Report notification, detailed information

[1017] Output: Response Action

[1018] Specific Action: Review reports on the dashboard and take appropriate action.

[1019] (Application example 2)

[1020] 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."

[1021] Conventional harassment detection systems are limited to the workplace, making it difficult to effectively monitor harassment and inappropriate comments in public places or the wider society. Furthermore, manual reporting and response is required, resulting in delayed real-time responses. Therefore, there is a need for a system that can detect inappropriate comments in real time, even in public places, and respond quickly.

[1022] 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.

[1023] In this invention, the server includes means for collecting voice data from workplaces and public places in real time, transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, and means for determining the level of discomfort based on the results of the sentiment analysis. This makes it possible to detect harassment and inappropriate remarks in real time not only in the workplace but also in public places, and to report them promptly.

[1024] "Workplaces and public places" refers not only to workplaces but also to environments that include places widely used by the general public, such as public transportation, parks, and commercial facilities.

[1025] "Audio data" refers to data that digitally represents people's conversations and environmental sounds collected via a voice input device such as a microphone.

[1026] "Transcription" is the process of analyzing audio data and converting its content into corresponding text data.

[1027] "Sentiment analysis" is a technology that detects emotions from text data or audio data and evaluates the content as an emotional score such as joy, anger, sadness, or displeasure.

[1028] "Offensiveness" is an index calculated based on sentiment analysis that assesses the degree to which a certain statement or event causes offense.

[1029] A "security department" refers to a specialized department or team that responds when inappropriate remarks or harassment are detected.

[1030] "Reporting" is the process by which the system automatically notifies the security department of relevant data when it detects potentially offensive speech.

[1031] "Speech recognition" is a technology that analyzes voice data and converts what is being said into text.

[1032] "Natural language processing" is a general term for artificial intelligence technology that analyzes text data and understands and analyzes its meaning and emotions.

[1033] "Harassment likelihood" refers to an assessment of the likelihood that a particular statement constitutes harassment based on sentiment analysis.

[1034] The system for implementing this invention, "AI Harassment Guardian," has the function of collecting voice data in real time in the workplace and public places, detecting and reporting inappropriate remarks and harassment. Specifically, it consists of the following elements:

[1035] 1. System Configuration

[1036] Hardware

[1037] Smartphones: Equipped with microphones for collecting audio data.

[1038] software

[1039] Python: Used as a programming language.

[1040] pydub: A library used to manipulate and process audio data.

[1041] vosk: An open-source speech recognition model library.

[1042] transformers: A natural language processing toolkit for sentiment analysis.

[1043] 2. System Operation

[1044] Collection and processing of voice data

[1045] The device (smartphone) captures conversations in meetings and public places in real time via a microphone. The captured audio data is processed by the pydub library and standardized to a 16 kHz sample rate.

[1046] Voice Recognition

[1047] The server transcribes the processed audio data using the vosk library, which converts the audio data into text data.

[1048] sentiment analysis

[1049] The server performs sentiment analysis on the converted text data using the transformers library, and as a result, the text data is assigned an emotional score such as joy, anger, sadness, or displeasure.

[1050] Determining discomfort and reporting

[1051] The server determines the level of discomfort based on the emotional score, and if the level of discomfort exceeds a certain threshold, it is deemed to be harassment. This information is automatically reported to the security department via email or a dedicated dashboard.

[1052] Specific examples

[1053] For example, if someone says "Go away!" on a public bus, the following will happen:

[1054] 1. The device (smartphone) captures the audio and transcribes what is said.

[1055] 2. The server performs sentiment analysis on the transcribed data and determines that the level of unpleasantness is high.

[1056] 3. The server stores this data and notifies the security department.

[1057] Prompt Sentence Examples

[1058] Prompt statement:

[1059] "Audio data is captured, transcribed in real time, and then sentiment analysis is performed. If a specific negative sentiment is indicated with a high score, the remark is determined to be harassment and a report is sent."

[1060] This system will enable real-time monitoring and rapid response to harassment and inappropriate comments not only in the workplace but also in public places.

[1061] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1062] Step 1:

[1063] The device (smartphone) collects audio data in real time during meetings and public places. The microphone on the device captures the surrounding sounds and saves them as audio data. The audio data is then stored digitally on the device.

[1064] Step 2:

[1065] The device converts the collected audio data to a 16kHz sample rate using the pydub library, which processes the audio data into a format suitable for voice recognition.

[1066] Step 3:

[1067] The device transcribes the processed voice data using the vosk library, recognizes the input voice data, and outputs the corresponding text data. The resulting conversation is saved in text format on the device.

[1068] Step 4:

[1069] The terminal sends the generated text data to the server using a communication protocol such as an HTTP request. At this time, the text data is transferred to the server.

[1070] Step 5:

[1071] The server receives the text data and, to analyze it, tokenizes it using a natural language processing (NLP) library, converting the text data into a format that can be analyzed.

[1072] Step 6:

[1073] The server uses the transformers library for sentiment analysis on the tokenized text data: it computes sentiment scores from the tokenized data as input, evaluates the sentiment (e.g., joy, anger, sadness, displeasure, etc.) for each text fragment, and produces a sentiment score as output.

[1074] Step 7:

[1075] The server determines the offensiveness based on the sentiment score. It analyzes each sentiment score to determine whether the offensiveness exceeds a certain threshold. If the offensiveness is high, the text fragment is determined to be likely to be harassing.

[1076] Step 8:

[1077] The server stores text fragments that are judged to be highly offensive and their emotion scores as related data, which can be used for later verification or as evidence.

[1078] Step 9:

[1079] The server automatically sends alerts to security teams based on relevant data, either via email or a dedicated dashboard, and includes sentiment analysis results along with relevant data.

[1080] Step 10:

[1081] The user (security department staff) receives a notification, checks the details on the dashboard, and takes immediate action if necessary.

[1082] 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.

[1083] 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.

[1084] 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.

[1085] [Fourth embodiment]

[1086] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1087] 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.

[1088] 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).

[1089] 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.

[1090] 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.

[1091] 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).

[1092] 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.

[1093] 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.

[1094] 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.

[1095] 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.

[1096] 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.

[1097] 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.

[1098] 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."

[1099] The AI ​​Guardian system of the present invention collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports the incident to the compliance department as necessary. The following describes the program processing of this system in detail.

[1100] 1. Collection and transmission of voice data

[1101] The device is set up to constantly capture audio data from meetings and daily work. The microphones on the device collect surrounding sounds as digital data.

[1102] Once the voice data is collected, the device transcribes it in real time by converting speech to text using speech recognition software (e.g., open source or commercial speech recognition APIs) built into the device.

[1103] The transcribed text data is sent from the device to a server using HTTP requests or other communication protocols.

[1104] 2. Sentiment analysis and unpleasantness judgment

[1105] The server receives the text data sent from the terminal, and the received text data is processed within the server for analysis.

[1106] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis.

[1107] Based on the results of natural language processing, the server performs sentiment analysis by calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[1108] Based on the emotion score, the server determines the degree of annoyance, and if the degree of annoyance exceeds a certain threshold, the text data is identified as potentially harassing.

[1109] 3. Identify potential harassment

[1110] The server identifies text data that it determines to be highly offensive as potentially harassing, and stores the identified data together with audio data, transcription data, and sentiment analysis data.

[1111] The stored data is retained for later verification and use as evidence.

[1112] 4. Report to the Compliance Department

[1113] The server automatically notifies the compliance department when potentially harassing data is stored, in the form of a report of the stored data, including audio data, transcription data, and sentiment analysis results.

[1114] Once a report is made, compliance users receive a notification via the software dashboard and email.

[1115] Specific examples

[1116] For example, if someone says "Work more efficiently" during a work meeting, this comment may cause offense. In that case, the following process will occur:

[1117] 1. The device captures the audio of the conversation, transcribes it in real time into text data such as "Work more efficiently," and sends it to the server.

[1118] 2. The server analyzes the text data and performs sentiment analysis, determining that the comment is likely to cause high levels of offense.

[1119] 3. The server will store the data as potentially harassing data and report the relevant data to the compliance department.

[1120] 4. Compliance users receive a notification, view detailed information on a dashboard, and take appropriate action.

[1121] As described above, the system of the present invention automatically detects and notifies workplace harassment, enabling prompt and appropriate responses, thereby contributing to protecting victims and reducing legal risks for companies.

[1122] The processing flow will be explained below.

[1123] Step 1:

[1124] The device uses a microphone to capture surrounding sounds in order to constantly collect voice data within the workplace.

[1125] Step 2:

[1126] The device transcribes the captured audio data in real time by converting the audio into text using a speech recognition API (e.g., a speech recognition engine).

[1127] Step 3:

[1128] The terminal transmits the generated text data to the server using a communication protocol such as an HTTP request.

[1129] Step 4:

[1130] To analyze the received text data, the server uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text and perform grammatical analysis.

[1131] Step 5:

[1132] The server performs sentiment analysis based on the analyzed text data, calling a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer) to obtain a sentiment score for the text data.

[1133] Step 6:

[1134] The server evaluates the emotion score to determine the level of annoyance, and identifies potential harassment if the level of annoyance exceeds a certain threshold.

[1135] Step 7:

[1136] The server stores audio data, transcription data, and sentiment analysis data related to text data determined to be potentially harassing in a database.

[1137] Step 8:

[1138] The server generates reports based on the stored data and sends them to the compliance department via email or a dedicated dashboard.

[1139] Step 9:

[1140] The user (compliance department officer) receives the report, checks the detailed information on the provided dashboard, takes necessary measures based on the content of the notification, and, in some cases, provides support to the victim.

[1141] Example 1

[1142] 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."

[1143] Workplace harassment is a serious problem that causes psychological stress for employees and reduces productivity throughout an organization. However, detecting and responding to harassment on the spot is difficult, and victims often miss opportunities to receive appropriate support. Furthermore, traditional manual monitoring and reporting systems lack immediacy, making it difficult to respond quickly. Therefore, there is a need for the development of a system that can analyze workplace audio data in real time, automatically detect inappropriate remarks, and appropriately report them.

[1144] 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.

[1145] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing natural language processing to analyze the structure of text data, means for performing sentiment analysis based on the analysis results and obtaining a sentiment score, means for determining the degree of discomfort based on the sentiment score and storing related data when a specific threshold is exceeded, means for reporting to a compliance department based on the determination result, and means for automating the reporting. This makes it possible to automatically and quickly detect workplace harassment and immediately report it to the appropriate department, thereby protecting victims and improving the soundness of the organization.

[1146] "Voice data" refers to any audio information generated in the workplace and is collected digitally using a microphone.

[1147] "Transcription means" refers to the technology or software used to convert voice data into text data, and uses voice recognition technology to convert voice into text in real time.

[1148] "Natural language processing means" refers to technology and software that analyzes text data and understands its grammatical structure and meaning, and performs tokenization and syntactic analysis.

[1149] "Sentiment analysis" refers to the technology of analyzing the content of text data and quantifying the emotions expressed therein, obtaining emotional scores such as joy, sadness, anger, and displeasure.

[1150] "Offensiveness assessment means" refers to a technology or method for assessing the offensiveness of text data based on the results of sentiment analysis, and determining that the content is inappropriate if it exceeds a certain threshold.

[1151] "Related data" refers to all data related to the detected inappropriate remarks, including audio data, transcription data, sentiment analysis results, the date and time of the remarks, and information about the speaker.

[1152] "Reporting tools" refers to technology or software that automatically notifies the compliance department of any relevant data when inappropriate comments or potential harassment are identified.

[1153] A "compliance department" refers to the department within a company that oversees compliance with laws and regulations and ethical issues, and is responsible for receiving reports and taking appropriate action.

[1154] This invention, "AI Guardian," is a system that collects workplace voice data in real time, automatically detects harassing comments, performs sentiment analysis, and reports them to the compliance department as necessary. The program processing of this system is explained in detail below.

[1155] Collection and transmission of voice data

[1156] The devices are installed at each company's bases and are always on. A high-sensitivity microphone installed in the device digitally collects audio data from workplace meetings and daily work. The collected audio data is temporarily stored on the device and converted into text data in real time using speech recognition APIs such as Google Cloud Speech-to-Text and IBM Watson Speech to Text. The converted text data is sent from the device to the server using an HTTP request.

[1157] Text data analysis

[1158] The server receives the text data sent from the device and analyzes it using a natural language processing library (e.g., spaCy or nltk). Specifically, the server tokenizes the text data and performs grammatical analysis to extract sentence structure and keywords. This analysis clarifies how the text data is structured.

[1159] Sentiment analysis and discomfort judgment

[1160] The server then performs sentiment analysis on the analyzed text data, calling sentiment analysis APIs such as the Google Natural Language API and IBM Watson Tone Analyzer to obtain a sentiment score for the text data. The sentiment score quantifies emotions such as joy, sadness, anger, and displeasure, and the server uses this score to evaluate the emotional tendency of the remarks. Based on the obtained sentiment score, the server determines the degree of discomfort, and if the degree of discomfort exceeds a certain threshold, it identifies the text data as potentially harassing.

[1161] Data storage and reporting

[1162] The server stores text data that is judged to be highly offensive in a database along with related data (audio data, transcription data, sentiment analysis results, etc.). The stored data can be referenced later for verification or as evidence. The server then immediately reports this data to the compliance department. Reports are made in the form of reports via HTTP POST requests or email, and the reports include the audio data, transcription data, and sentiment analysis results. Once a report is made, the compliance department user receives a notification, can view details on a dashboard, and can take appropriate action.

[1163] Specific examples

[1164] For example, if someone says "Work more efficiently" during a work meeting, the system will do the following:

[1165] 1. The device collects the audio of the conversation and converts it into a digital signal through a highly sensitive microphone.

[1166] 2. The device transcribes the voice data in real time, generates text data such as "Work more efficiently," and sends it to the server.

[1167] 3. The server tokenizes the received text data and analyzes its grammatical structure.

[1168] 4. The server uses the sentiment analysis API to obtain a sentiment score for the text data. In this case, it determines that "Work efficiently" is likely to indicate high levels of displeasure.

[1169] 5. The server identifies this data as potentially harassing and stores it in a database along with the audio data, transcription data, and sentiment analysis results.

[1170] 6. The server sends the generated report to the compliance department via HTTP POST request and / or email, and a compliance department user is notified.

[1171] Prompt Sentence Examples

[1172] For example, the following prompt sentence is used for a generative AI model:

[1173] If someone says "Work more efficiently" during a work meeting, determine whether this statement is potentially harassing. Tokenize the text data, perform sentiment analysis, and if a high level of discomfort is detected, explain why.

[1174] By introducing this system, companies will be able to quickly and accurately detect workplace harassment and take appropriate measures, making it a powerful tool for ensuring the psychological safety of employees and improving the work environment.

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

[1176] Step 1:

[1177] The device collects workplace audio data in real time. Specifically, it uses a high-sensitivity microphone to digitally capture the audio of meetings and daily work. The input is ambient analog audio data, and the output is audio data converted into a digital signal.

[1178] Step 2:

[1179] The device transcribes the collected voice data. Specifically, it converts the voice data into text data using voice recognition software (e.g., Google Cloud Speech-to-Text or IBM Watson Speech to Text). The input is digital voice data, and the output is transcribed text data.

[1180] Step 3:

[1181] The device sends the transcribed text data to the server. Specifically, it sends the text data to the server using an HTTP request or other communication protocol. The input is the transcribed text data, and the output is the text data sent to the server.

[1182] Step 4:

[1183] The server analyzes the received text data. Specifically, it uses a natural language processing library (e.g., spaCy or nltk) to tokenize the text data and perform grammatical analysis. The input is text data, and the output is grammatically analyzed text data.

[1184] Step 5:

[1185] The server performs sentiment analysis on the analyzed text data. Specifically, it calls a sentiment analysis API (e.g., Google Natural Language API or IBM Watson Tone Analyzer) to obtain a sentiment score. The input is the grammatically analyzed text data, and the output is a sentiment score.

[1186] Step 6:

[1187] The server determines the degree of discomfort based on the emotion score. Specifically, it evaluates whether the text data is unpleasant by comparing the acquired emotion score with a set threshold. The input is the emotion score, and the output is the evaluation result of the discomfort degree.

[1188] Step 7:

[1189] The server stores the text data judged to be highly unpleasant together with related data. Specifically, it stores the text data, emotion scores, audio data, etc. in a database. The input is the text data judged to be highly unpleasant and the related data, and the output is the data stored in the database.

[1190] Step 8:

[1191] The server then sends a report to the compliance department based on the stored data. Specifically, it sends the report in the form of a report via email or HTTP POST request. The input is the data stored in the database, and the output is a report to the compliance department.

[1192] Step 9:

[1193] Compliance department users review the received reports and take appropriate action. Specifically, they receive notifications via the software dashboard or email, check the details, and take appropriate action. The input is the report data from the server, and the output is the action taken.

[1194] (Application example 1)

[1195] 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."

[1196] In modern workplaces, especially in factories, harassment can occur during communication with workers. Such harassment not only worsens the working environment, but also increases legal risks for companies. Furthermore, factory robots and automated systems mean that on-site monitoring of working conditions is insufficient, making it difficult to protect workers.

[1197] 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.

[1198] In this invention, the server includes means for collecting voice data from the workplace in real time and transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, means for determining the level of discomfort based on the results of the sentiment analysis, means for saving related data and notifying the management department when the level of discomfort exceeds a specific threshold, and means for collecting and analyzing voice data using factory robots to monitor physical labor and protect workers, thereby enabling early detection of harassment in the workplace and protection of workers.

[1199] "Voice data" means data for storing or processing acoustic signals in digital form, including, in particular, speech and other sounds in the workplace or on a factory floor.

[1200] "Transcription means" refers to technology or equipment for converting voice data into text data, and primarily refers to voice recognition software.

[1201] "Sentiment analysis means" refers to natural language processing techniques that analyze text data and determine the emotions and intensity of emotions contained therein.

[1202] "Discomfort level" is a measure of how much a certain statement or action makes the other person feel uncomfortable, and is calculated from the results of emotion analysis.

[1203] An "administrative department" is a department within a company or organization that is responsible for managing compliance and the working environment.

[1204] A "factory robot" is a mechanical device that performs automated tasks in a factory and may have a variety of functions, such as collecting and analyzing voice data.

[1205] "Worker protection" refers to measures and efforts to ensure that workers can work in a safe and comfortable environment.

[1206] The present invention relates to an AI Guardian system for detecting harassment and protecting workers in workplaces and factory environments. The system collects voice data in real time and analyzes the data for sentiment analysis. Specifically, the following detailed embodiments are described.

[1207] First, a microphone built into the factory robot collects voice data. This voice data is transcribed in real time using voice recognition software (e.g., Google's voice recognition API). The transcribed data is sent to a server and analyzed using a natural language processing library (e.g., spaCy or nltk).

[1208] Next, the server performs sentiment analysis. This analysis uses a sentiment analysis API (e.g., Google Natural Language API or IBM Watson's Tone Analyzer). Sentiment analysis calculates an emotional score for the text data. The level of discomfort is determined based on this score, and if it exceeds a certain threshold, the relevant data is automatically saved and a report is sent to the management department.

[1209] The reporting function is implemented using the Twilio API, and reports are sent via SMS or email, and include collected voice data, transcription data, and sentiment analysis results.

[1210] A specific example is shown below. If someone says "You're completely useless" in a conversation within a factory, this speech is collected by a microphone. Speech recognition software transcribes this speech and sends it to a server. The server performs sentiment analysis on this text data, and if it determines that the sentiment score is high, it uses the Twilio API to notify management along with the message "You're completely useless."

[1211] Prompt Sentence Examples

[1212] Perform a sentiment analysis on the following text and print the results:

[1213] "You're completely useless"

[1214] This system will enable early detection of workplace harassment and rapid response, strengthening worker protection and reducing legal risks for companies.

[1215] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1216] Step 1:

[1217] The terminal collects the audio data. Specifically, microphones built into the factory robots capture the surrounding sounds and store them digitally. The terminal collects the audio data in real time.

[1218] Input: Ambient audio (analog data)

[1219] Output: Audio data (digital data)

[1220] Step 2:

[1221] The device transcribes the collected voice data in real time and uses speech recognition software to convert the collected voice data into text data.

[1222] Input: Audio data (digital data)

[1223] Output: Transcription data (text data)

[1224] Specific operation: The device sends voice data to the voice recognition API and receives the returned text data.

[1225] Step 3:

[1226] The device sends the transcription data to the server. The text data is sent to the server using a communication protocol such as an HTTP request.

[1227] Input: Transcription data (text data)

[1228] Output: Text data sent to the server

[1229] Specific operation: The device sends text data to the server via a communication protocol.

[1230] Step 4:

[1231] The server receives the text data and analyzes it using a natural language processing library, tokenizing the text data and performing grammatical analysis.

[1232] Input: Text data (transcription data)

[1233] Output: Analyzed text data

[1234] Specific operation: The server analyzes the text data using a natural language processing library (e.g. spaCy, nltk).

[1235] Step 5:

[1236] The server calls the sentiment analysis API to obtain the sentiment score of the text data, and determines the level of discomfort based on the sentiment score.

[1237] Input: Parsed text data

[1238] Output: Emotion score, unpleasantness judgment result

[1239] Specific operation: The server obtains the sentiment score using a sentiment analysis API (e.g., Google Natural Language API, IBM Watson Tone Analyzer).

[1240] Step 6:

[1241] If the server detects that the level of discomfort exceeds a certain threshold, it will store the relevant data and notify the management department via SMS or email.

[1242] Input: Emotion score, discomfort level judgment result

[1243] Output: Saved data, notification to management department

[1244] Specific operation: The server saves the relevant data in a database and sends a notification using the Twilio API.

[1245] This concludes the detailed explanation of each processing step in the system of the application example. This flow enables real-time monitoring of voice data, and the detection and immediate reporting of harassment.

[1246] 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.

[1247] The "AI Guardian" system of the present invention collects workplace voice data in real time and automatically detects harassing remarks. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the situation to the compliance department if necessary. The following describes the program processing of this system in detail.

[1248] 1. Collection and transmission of voice data

[1249] The device continuously captures audio data from meetings and daily work. The device's built-in microphone collects surrounding sounds as digital data.

[1250] After collecting the audio data, the device transcribes it in real time, using voice recognition software to convert the audio data into text format.

[1251] The converted text data is then sent from the device to the server using an HTTP request or other communication protocol.

[1252] 2. Sentiment analysis and unpleasantness judgment

[1253] The server receives the text data sent from the terminal, and processes the received text data for analysis.

[1254] The server uses a natural language processing library to tokenize the text data and perform grammatical analysis.

[1255] After grammar analysis, the server performs sentiment analysis on the text data using an emotion engine, which extracts sentiment scores from the text, such as joy, anger, sadness, and displeasure.

[1256] At the same time, the server also performs sentiment analysis based on the voice data, combining the results of voice recognition with the emotion engine analysis to more accurately estimate the user's emotions.

[1257] The server determines the degree of discomfort based on the results of these sentiment analyses. If the sentiment score exceeds a certain threshold, the corresponding speech is deemed to cause discomfort.

[1258] 3. Identify potential harassment

[1259] The server identifies text and audio data that is deemed highly offensive as potentially harassing, and stores the identified data along with sentiment analysis results and related data.

[1260] The stored data will be retained for future verification and use as evidence.

[1261] 4. Report to the Compliance Department

[1262] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[1263] The report includes relevant audio data, text data, and sentiment analysis results.

[1264] Specific examples

[1265] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[1266] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[1267] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[1268] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[1269] 4. The server stores the relevant data and automatically notifies the compliance department.

[1270] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[1271] In this way, the system of the present invention performs emotion analysis from both voice and text, thereby improving the accuracy of identifying harassment, enabling prompt and appropriate responses, improving the work environment, and reducing legal risks.

[1272] The processing flow will be explained below.

[1273] Step 1:

[1274] The device uses microphones installed in the workplace to capture audio data from meetings and daily work in real time.

[1275] Step 2:

[1276] The device uses speech recognition software (e.g., a speech recognition engine) to transcribe the captured voice data in real time, which is then converted into text data.

[1277] Step 3:

[1278] The terminal transmits the generated text data to the server using an applicable communication protocol such as an HTTP request.

[1279] Step 4:

[1280] The server uses a natural language processing library (e.g., spaCy or nltk) to tokenize and parse the received text data.

[1281] Step 5:

[1282] The server performs sentiment analysis on the text data using an emotion engine, which extracts emotion scores from the text, such as happiness, anger, sadness, displeasure, etc.

[1283] Step 6:

[1284] The server also performs sentiment analysis of the voice data using an emotion engine, which analyzes the tone, stress, and intonation of the voice to identify emotions.

[1285] Step 7:

[1286] The server combines the emotion scores from both the text and the audio to determine the annoyance level, and evaluates whether the annoyance level exceeds a certain threshold.

[1287] Step 8:

[1288] If the server determines that the offense level is high, it identifies the data as potentially harassing and stores the associated audio data, transcription data, and sentiment analysis data in a database.

[1289] Step 9:

[1290] The server generates detailed reports based on the stored data and automatically reports them to the compliance department via email or a dedicated dashboard.

[1291] Step 10:

[1292] The user (compliance department staff) receives the report and checks the notified report on the dashboard. Based on the detailed information, they take necessary measures and, in some cases, provide support to the victim.

[1293] Example 2

[1294] 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."

[1295] Detecting and responding to workplace harassment is extremely important, but traditional methods rely primarily on human monitoring and reporting, making it difficult to respond quickly and objectively. This can lead to a worsening work environment, causing mental stress and reduced productivity for employees. Furthermore, traditional methods make it difficult to preserve and manage evidence of harassment, which can delay necessary responses.

[1296] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1297] In this invention, the server includes: means for collecting voice data in real time; means for transcribing the voice data; means for analyzing the transcription data and the voice data using a natural language processing library; means for performing sentiment analysis of the transcription data and the voice data using an emotion engine; means for determining the level of discomfort based on the results of the sentiment analysis; and means for storing related data, generating data for reporting, and reporting to a compliance department when the level of discomfort exceeds a specific threshold. This enables real-time detection and rapid response to workplace harassment. Furthermore, by appropriately storing evidence of highly offensive remarks and taking action based on objective data, the system can improve the work environment and reduce employee mental stress.

[1298] "Audio data" refers to data that records audio in digital format.

[1299] "Transcription" is the process of converting audio data into text form.

[1300] A "natural language processing library" is a set of software tools for analyzing text data, including tokenization and grammatical analysis.

[1301] "Emotion Engine" refers to algorithms and techniques for extracting emotion scores from text and audio data.

[1302] The "discomfort level" is an index that indicates the degree of a particular emotional state calculated based on the emotional score.

[1303] A "threshold" is a boundary value for triggering a specific action when an indicator such as discomfort level or emotion score exceeds a certain standard.

[1304] "Related data" refers to all data related to determining harassment, such as the results of sentiment analysis, the content of statements, and audio data.

[1305] A "compliance department" is a specialized department within a company that monitors and promotes compliance with laws and regulations and ethical behavior.

[1306] "Notification" refers to the process of reporting specific events or data.

[1307] "Data storage" refers to storing acquired data for future use.

[1308] The system of this invention, "AI Guardian," collects workplace voice data in real time and automatically detects harassing comments. It also incorporates an emotion engine that recognizes the user's emotions, evaluates the level of discomfort based on the emotion analysis results, and can report the incident to the compliance department if necessary. This system is primarily implemented using the following hardware and software:

[1309] Audio data collection and transcription

[1310] The device is equipped with a microphone to continuously capture audio data during meetings and daily work. It is also equipped with speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson. This software converts the collected audio data into text in real time. The text data is then sent to a server using a communication protocol such as an HTTP request.

[1311] Text data analysis and sentiment analysis

[1312] When the server receives the text data sent from the device, it analyzes the data using a natural language processing library (such as NLTK or spaCy). First, the text data is tokenized and grammatically analyzed. Then, an emotion engine (such as the Text Analytics API from Azure Cognitive Services) is used to perform a sentiment analysis of the text and audio data. This emotion engine extracts an emotion score from the text and audio, quantifying emotions such as joy, anger, sadness, and displeasure.

[1313] Determining discomfort and identifying harassment

[1314] The server analyzes the emotion scores obtained from the emotion engine and determines the level of offense. If an emotion score exceeding a certain threshold is detected, the remark is deemed to be offensive. This highly offensive remark is identified as potentially harassing, and the related data (audio data, text data, and emotion analysis results) is stored.

[1315] Report to the Compliance Department

[1316] The server generates a report based on data judged to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard. The generated report includes related audio data, text data, and sentiment analysis results.

[1317] Specific examples

[1318] For example, if someone says "Work more efficiently" during a meeting, the following happens:

[1319] 1. The device captures the audio, transcribes the speech, generates text data such as "Work more efficiently," and sends it to the server.

[1320] 2. The server analyzes the received text data and calculates the emotion score using the emotion engine. At the same time, the voice data is also analyzed by the emotion engine.

[1321] 3. The server combines the text and speech sentiment scores and determines that the statement "Work more efficiently" evokes high annoyance.

[1322] 4. The server stores the relevant data and automatically notifies the compliance department.

[1323] 5. The user (compliance department officer) receives a notification, checks the detailed information on the dashboard, and takes the necessary action.

[1324] Prompt Sentence Examples

[1325] For example, this process begins by inputting a prompt such as, "Please analyze the statement 'Work more efficiently' made during a meeting." into the generative AI model.

[1326] In this way, this system, "AI Guardian," performs emotion analysis from both voice and text, improving the accuracy of identifying harassment and enabling prompt and appropriate responses, thereby improving the work environment and reducing legal risks.

[1327] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1328] Step 1:

[1329] The device uses a microphone to collect real-time audio data from workplace meetings and daily work.

[1330] Input: Ambient conversation

[1331] Output: Digital audio data

[1332] Specific operation: The microphone installed in the device captures audio and stores it as digital data.

[1333] Step 2:

[1334] The device transcribes the collected voice data using speech recognition software, such as Google Cloud Speech-to-Text API or IBM Watson.

[1335] Input: Audio data

[1336] Output: Transcribed text data

[1337] Specific operation: Voice data is input into voice recognition software, and the conversion result is obtained as text data.

[1338] Step 3:

[1339] The device sends the generated text data to the server using an HTTP request, using HTTP or HTTPS as the communication protocol.

[1340] Input: Transcription data

[1341] Output: Text data sent to the server

[1342] Specific operation: Include text data in the body of an HTTP request and send it to the API endpoint.

[1343] Step 4:

[1344] The server uses a natural language processing library (such as NLTK or spaCy) to tokenize and grammatically analyze the received text data.

[1345] Input: Received text data

[1346] Output: Tokenized and parsed text data

[1347] Specific operation: The text data is broken down into tokens and the grammatical structure is analyzed.

[1348] Step 5:

[1349] The server uses an emotion engine (such as the Text Analytics API from Azure Cognitive Services) to perform emotion analysis on the text and audio data, and extracts an emotion score.

[1350] Input: Tokenized and grammatically analyzed text data, audio data

[1351] Output: Sentiment score

[1352] Specific operation: The analyzed text data and voice data are input into the emotion engine to calculate emotion scores such as joy, anger, sadness, and displeasure.

[1353] Step 6:

[1354] The server determines the degree of discomfort based on the extracted emotion score. If the degree of discomfort exceeds a certain threshold, the comment is deemed to be unpleasant.

[1355] Input: Sentiment score

[1356] Output: Discomfort level judgment result

[1357] Specific operation: The emotion score is compared with a specific threshold, and if the threshold is exceeded, the unpleasantness level is determined to be high.

[1358] Step 7:

[1359] The server identifies text and audio data that is determined to be highly offensive as potentially harassing and stores the associated data.

[1360] Input: Discomfort level judgment result

[1361] Output: Saved text and audio data

[1362] Specific action: Potentially harassing data will be stored in a database.

[1363] Step 8:

[1364] The server generates a report based on data that is deemed to be highly offensive and automatically notifies the compliance department via email or a dedicated dashboard.

[1365] Input: Stored text and audio data

[1366] Output: Generated reports, notification

[1367] Specific Action: Generates a report and sends an email using SMTP or displays a notification on the dashboard.

[1368] Step 9:

[1369] The user (compliance department officer) receives a notification, checks detailed information on the dashboard, and takes the necessary action.

[1370] Input: Report notification, detailed information

[1371] Output: Response Action

[1372] Specific Action: Review reports on the dashboard and take appropriate action.

[1373] (Application example 2)

[1374] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1375] Conventional harassment detection systems are limited to the workplace, making it difficult to effectively monitor harassment and inappropriate comments in public places or the wider society. Furthermore, manual reporting and response is required, resulting in delayed real-time responses. Therefore, there is a need for a system that can detect inappropriate comments in real time, even in public places, and respond quickly.

[1376] 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.

[1377] In this invention, the server includes means for collecting voice data from workplaces and public places in real time, transcribing the voice data, means for analyzing the transcribed data and performing sentiment analysis, and means for determining the level of discomfort based on the results of the sentiment analysis. This makes it possible to detect harassment and inappropriate remarks in real time not only in the workplace but also in public places, and to report them promptly.

[1378] "Workplaces and public places" refers not only to workplaces but also to environments that include places widely used by the general public, such as public transportation, parks, and commercial facilities.

[1379] "Audio data" refers to data that digitally represents people's conversations and environmental sounds collected via a voice input device such as a microphone.

[1380] "Transcription" is the process of analyzing audio data and converting its content into corresponding text data.

[1381] "Sentiment analysis" is a technology that detects emotions from text data or audio data and evaluates the content as an emotional score such as joy, anger, sadness, or displeasure.

[1382] "Offensiveness" is an index calculated based on sentiment analysis that assesses the degree to which a certain statement or event causes offense.

[1383] A "security department" refers to a specialized department or team that responds when inappropriate remarks or harassment are detected.

[1384] "Reporting" is the process by which the system automatically notifies the security department of relevant data when it detects potentially offensive speech.

[1385] "Speech recognition" is a technology that analyzes voice data and converts what is being said into text.

[1386] "Natural language processing" is a general term for artificial intelligence technology that analyzes text data and understands and analyzes its meaning and emotions.

[1387] "Harassment likelihood" refers to an assessment of the likelihood that a particular statement constitutes harassment based on sentiment analysis.

[1388] The system for implementing this invention, "AI Harassment Guardian," has the function of collecting voice data in real time in the workplace and public places, detecting and reporting inappropriate remarks and harassment. Specifically, it consists of the following elements:

[1389] 1. System Configuration

[1390] Hardware

[1391] Smartphones: Equipped with microphones for collecting audio data.

[1392] software

[1393] Python: Used as a programming language.

[1394] pydub: A library used to manipulate and process audio data.

[1395] vosk: An open-source speech recognition model library.

[1396] transformers: A natural language processing toolkit for sentiment analysis.

[1397] 2. System Operation

[1398] Collection and processing of voice data

[1399] The device (smartphone) captures conversations in meetings and public places in real time via a microphone. The captured audio data is processed by the pydub library and standardized to a 16 kHz sample rate.

[1400] Voice Recognition

[1401] The server transcribes the processed audio data using the vosk library, which converts the audio data into text data.

[1402] sentiment analysis

[1403] The server performs sentiment analysis on the converted text data using the transformers library, and as a result, the text data is assigned an emotional score such as joy, anger, sadness, or displeasure.

[1404] Determining discomfort and reporting

[1405] The server determines the level of discomfort based on the emotional score, and if the level of discomfort exceeds a certain threshold, it is deemed to be harassment. This information is automatically reported to the security department via email or a dedicated dashboard.

[1406] Specific examples

[1407] For example, if someone says "Go away!" on a public bus, the following will happen:

[1408] 1. The device (smartphone) captures the audio and transcribes what is said.

[1409] 2. The server performs sentiment analysis on the transcribed data and determines that the level of unpleasantness is high.

[1410] 3. The server stores this data and notifies the security department.

[1411] Prompt Sentence Examples

[1412] Prompt statement:

[1413] "Audio data is captured, transcribed in real time, and then sentiment analysis is performed. If a specific negative sentiment is indicated with a high score, the remark is determined to be harassment and a report is sent."

[1414] This system will enable real-time monitoring and rapid response to harassment and inappropriate comments not only in the workplace but also in public places.

[1415] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1416] Step 1:

[1417] The device (smartphone) collects audio data in real time during meetings and public places. The microphone on the device captures the surrounding sounds and saves them as audio data. The audio data is then stored digitally on the device.

[1418] Step 2:

[1419] The device converts the collected audio data to a 16kHz sample rate using the pydub library, which processes the audio data into a format suitable for voice recognition.

[1420] Step 3:

[1421] The device transcribes the processed voice data using the vosk library, recognizes the input voice data, and outputs the corresponding text data. The resulting conversation is saved in text format on the device.

[1422] Step 4:

[1423] The terminal sends the generated text data to the server using a communication protocol such as an HTTP request. At this time, the text data is transferred to the server.

[1424] Step 5:

[1425] The server receives the text data and, to analyze it, tokenizes it using a natural language processing (NLP) library, converting the text data into a format that can be analyzed.

[1426] Step 6:

[1427] The server uses the transformers library for sentiment analysis on the tokenized text data: it computes sentiment scores from the tokenized data as input, evaluates the sentiment (e.g., joy, anger, sadness, displeasure, etc.) for each text fragment, and produces a sentiment score as output.

[1428] Step 7:

[1429] The server determines the offensiveness based on the sentiment score. It analyzes each sentiment score to determine whether the offensiveness exceeds a certain threshold. If the offensiveness is high, the text fragment is determined to be likely to be harassing.

[1430] Step 8:

[1431] The server stores text fragments that are judged to be highly offensive and their emotion scores as related data, which can be used for later verification or as evidence.

[1432] Step 9:

[1433] The server automatically sends alerts to security teams based on relevant data, either via email or a dedicated dashboard, and includes sentiment analysis results along with relevant data.

[1434] Step 10:

[1435] The user (security department staff) receives a notification, checks the details on the dashboard, and takes immediate action if necessary.

[1436] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1437] 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.

[1438] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1439] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1440] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1441] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1442] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1443] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1444] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1445] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1446] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1447] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1448] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1449] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1450] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1451] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1452] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1453] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1454] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1455] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1456] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1457] The following is further disclosed regarding the above embodiment.

[1458] (Claim 1)

[1459] A means for collecting workplace audio data in real time and transcribing the audio data;

[1460] means for analyzing the transcription data and performing sentiment analysis;

[1461] means for determining a degree of discomfort based on the emotion analysis result;

[1462] means for storing relevant data and notifying a compliance department if the discomfort level exceeds a certain threshold;

[1463] A system including:

[1464] (Claim 2)

[1465] a speech recognition means for real-time transcription of speech data collected at the workplace;

[1466] natural language processing means for performing sentiment analysis using the transcription data;

[1467] means for determining the possibility of harassment based on the emotion analysis results;

[1468] 2. The system according to claim 1, further comprising means for storing relevant data and executing a notification based on the determination result.

[1469] (Claim 3)

[1470] 2. The system according to claim 1, further comprising a notification means for automatically executing said notification.

[1471] "Example 1"

[1472] (Claim 1)

[1473] A means for collecting workplace audio data in real time and transcribing the audio data;

[1474] means for analyzing the transcription data and performing natural language processing to analyze the structure of the text data;

[1475] means for performing sentiment analysis based on the analysis result and obtaining a sentiment score;

[1476] means for determining a degree of discomfort based on the emotion score and storing relevant data when the degree of discomfort exceeds a certain threshold;

[1477] a means for reporting to a compliance department based on the determination result;

[1478] means for automating said notification;

[1479] A system including:

[1480] (Claim 2)

[1481] a speech recognition means for real-time transcription of speech data collected at the workplace;

[1482] natural language processing means for performing sentiment analysis using the transcription data;

[1483] means for determining the possibility of harassment based on the emotion analysis results;

[1484] 2. The system according to claim 1, further comprising means for storing relevant data and executing a notification based on the determination result.

[1485] (Claim 3)

[1486] 2. The system according to claim 1, further comprising a notification means for automatically executing said notification.

[1487] "Application Example 1"

[1488] (Claim 1)

[1489] A means for collecting workplace audio data in real time and transcribing the audio data;

[1490] means for analyzing the transcription data and performing sentiment analysis;

[1491] means for determining a degree of discomfort based on the emotion analysis result;

[1492] means for storing relevant data and notifying a management department when the discomfort level exceeds a certain threshold;

[1493] A means of collecting and analyzing voice data from factory robots to monitor physical labor and protect workers;

[1494] A system including:

[1495] (Claim 2)

[1496] a speech recognition means for real-time transcription of speech data collected at the workplace;

[1497] natural language processing means for performing sentiment analysis using the transcription data;

[1498] means for determining the possibility of harassment based on the emotion analysis results;

[1499] and means for storing relevant data and executing a notification based on the determination result.

[1500] 10. The system of claim 1.

[1501] (Claim 3)

[1502] Including a notification means for automatically executing the notification.

[1503] 10. The system of claim 1.

[1504] "Example 2: Combining Emotion Engines"

[1505] (Claim 1)

[1506] a means for collecting audio data in real time;

[1507] means for transcribing the audio data;

[1508] means for analyzing the transcription data and audio data using a natural language processing library;

[1509] means for performing sentiment analysis of the transcription data and audio data using a sentiment engine;

[1510] means for determining a degree of discomfort based on the emotion analysis result;

[1511] means for storing relevant data, generating data for reporting, and reporting to a compliance department when the discomfort level exceeds a certain threshold;

[1512] A system including:

[1513] (Claim 2)

[1514] means for real-time transcription using speech recognition means;

[1515] means for analyzing the transcription data and the audio data to perform sentiment analysis;

[1516] means for determining a degree of discomfort based on the emotion analysis result;

[1517] and means for storing relevant data and executing a notification based on the determination result.

[1518] 10. The system of claim 1.

[1519] (Claim 3)

[1520] and means for automatically executing the notification.

[1521] 10. The system of claim 1.

[1522] "Application example 2 when combining emotion engines"

[1523] (Claim 1)

[1524] A means for collecting voice data from a workplace or public place in real time and transcribing the voice data;

[1525] means for analyzing the transcription data and performing sentiment analysis;

[1526] means for determining a degree of discomfort based on the emotion analysis result;

[1527] means for storing relevant data and notifying a security department if the discomfort level exceeds a certain threshold;

[1528] A system including:

[1529] (Claim 2)

[1530] a speech recognition means for real-time transcription of speech data collected in the workplace or in public places;

[1531] natural language processing means for performing sentiment analysis using the transcription data;

[1532] means for determining the possibility of harassment based on the emotion analysis results;

[1533] 2. The system according to claim 1, further comprising means for storing relevant data and executing a notification based on the determination result.

[1534] (Claim 3)

[1535] 2. The system according to claim 1, further comprising a notification means for automatically executing said notification. [Explanation of symbols]

[1536] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means for collecting workplace audio data in real time and transcribing the audio data; means for analyzing the transcription data and performing sentiment analysis; means for determining a degree of discomfort based on the emotion analysis result; means for storing relevant data and notifying a compliance department if the discomfort level exceeds a certain threshold; A system including:

2. a speech recognition means for real-time transcription of speech data collected at the workplace; natural language processing means for performing sentiment analysis using the transcription data; means for determining the possibility of harassment based on the emotion analysis results; 2. The system according to claim 1, further comprising means for storing relevant data and executing a notification based on the result of said determination.

3. 2. The system according to claim 1, further comprising a notification means for automatically executing said notification.

Images