System

Abstract

To improve response accuracy of a chatbot.SOLUTION: A system including means for collecting learning data, means for initially filtering the collected learning data, means for cleaning the initially filtered data, means for presenting the cleaned data to a user and receiving confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, and means for implementing the relearned artificial intelligence model and operating as a chatbot.SELECTED DRAWING: Figure 11

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] With typical AI chatbots, the training data format varies, so manual data cleaning is required to prevent overfitting and the leakage of personal and confidential information. This requires a lot of man-hours to collect, clean, and retrain the training data. Furthermore, a decline in data quality and response accuracy can lead to inconsistent chatbot performance. [Means for solving the problem]

[0005] The present invention solves the above-mentioned problems by providing a system including: means for collecting training data; means for initially filtering the collected training data; means for cleaning the initially filtered data; means for presenting the cleaned data to a user and accepting confirmation and correction by the user; means for relearning an AI model based on the data confirmed and corrected by the user; and means for implementing the re-trained AI model and operating it as a chatbot. The present invention enables automatic collection and cleaning of training data, minimizing human labor and enabling efficient and highly accurate operation of chatbots.

[0006] "Training data" is text data collected from users that is used to train the chatbot.

[0007] "Means of collection" refers to the means for obtaining training data from designated websites or databases.

[0008] The "initial filtering means" is a means for removing unnecessary elements from the collected data and standardizing the format.

[0009] "Artificial intelligence cleaning methods" are AI technologies that organize initially filtered data, correct input errors, and mask personal information.

[0010] The "means for presenting to the user" refers to a means for displaying the cleaned data to the user and allowing the user to confirm or correct the data.

[0011] "Means for retraining" refers to means for retraining an AI model based on data corrected by the user.

[0012] "Implementation means" refers to the means for incorporating the retrained artificial intelligence model into a chatbot and making it operational.

[0013] "Means for collecting feedback" refers to means for collecting evaluations and comments from users regarding the chatbot's responses.

[0014] "Means of monitoring" refers to the means for monitoring the chatbot's response accuracy and performance, and identifying areas for improvement as needed. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram 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

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

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

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

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

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

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

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

[0023] [First embodiment]

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

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

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

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

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

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

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

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

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

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

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

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

[0036] The purpose of this invention is to automate the development and operation process of an AI chatbot system. The specific operation of the system's program is explained below.

[0037] First, the device collects learning data. The user provides the device with information about the websites and databases to be collected. The device uses the specified information to send a data collection command to the server. The server then accesses the specified websites and databases and retrieves the necessary text data. For example, the data needed for learning can be collected from FAQ pages or user comments.

[0038] Next, the server performs initial filtering on the collected data to standardize the data format and remove unnecessary information, such as removing HTML tags and formatting line breaks and whitespace.

[0039] Next, the cleaning AI cleans the filtered data. It uses natural language processing technology to correct typos and inconsistencies in the data. It also automatically recognizes and masks personal and confidential information. For example, it appropriately conceals personal information such as email addresses and credit card information.

[0040] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This checking and correction is an important step in maintaining the accuracy of the data.

[0041] With the corrected data, the server passes it to the retraining AI, which uses the new data to retrain the chatbot's model. This process ensures that the AI ​​model is updated with the latest information.

[0042] The new model generated through retraining is sent from the server to the device. The device then installs this model into the chatbot and begins operation. Users ask questions using the chatbot, and the chatbot responds based on the data it has learned.

[0043] While the chatbot is running, the server monitors the quality of responses, collects user feedback, and uses it for the next retraining phase, allowing the chatbot to continually improve its accuracy.

[0044] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. A new model is generated and installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[0045] In this way, the present invention effectively automates the collection, cleaning, relearning, and operation of learning data, enabling the construction and operation of a highly accurate chatbot system.

[0046] The processing flow will be explained below.

[0047] Step 1:

[0048] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[0049] Step 2:

[0050] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[0051] Step 3:

[0052] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[0053] Step 4:

[0054] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0055] Step 5:

[0056] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[0057] Step 6:

[0058] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[0059] Step 7:

[0060] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[0061] Step 8:

[0062] A user asks a question using the chatbot, the chatbot generates an appropriate response based on the new model, and the server monitors and quality-controls this response.

[0063] Step 9:

[0064] The server collects user feedback and evaluates the quality of responses. This feedback is used during the next retraining phase. This continuous feedback loop helps improve the chatbot's performance.

[0065] Example 1

[0066] 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."

[0067] Conventional AI chatbot systems required a great deal of time and effort, from collecting and filtering training data, cleaning the data, to retraining the model, and then operating it. Furthermore, manual verification work was required to protect personal information and maintain data accuracy, and automation was insufficient, resulting in low efficiency. Furthermore, there was an insufficient mechanism for continuously improving the quality of chatbot responses during operation, making it difficult to appropriately incorporate customer feedback.

[0068] 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.

[0069] In this invention, the server includes: means for a user to provide information to be collected and a terminal to send a data collection instruction to the server based on that information; means for the server to access specified websites and databases to collect training data; means for the server to initially filter the training data and remove unnecessary information; means for a cleaning AI to correct input errors and inconsistencies in the filtered data using natural language processing technology and mask personal information; means for presenting the cleaned data to the user and accepting confirmation and correction by the user; means for a re-training AI to re-train an AI model based on the data confirmed and corrected by the user; means for sending the AI ​​model generated by the re-training to the terminal and operating it as a chatbot; and means for monitoring the quality of the chatbot's responses, collecting user feedback, and using it for the next re-training. This automates the process from collecting training data to re-training the model, operation, and continuous improvement of response quality, enabling the construction and operation of an efficient and highly accurate AI chatbot system.

[0070] "User" means a person or organization that uses the system to specify the targets for collecting learning data and to review and correct the data.

[0071] A "terminal" is a hardware or software component that allows a user to input information, send data collection instructions to a server, install a retrained model, and operate it as a chatbot.

[0072] The "server" is the core technology of the system that receives instructions from users and devices, collects training data, performs initial filtering and processing using cleaning AI, and then supplies the data to the re-training AI to re-train the model.

[0073] A "data collection instruction" is an instruction to command the server to start collecting learning data based on information input by the user into the terminal.

[0074] "Training data" refers to text data and information collected to train an AI chatbot model.

[0075] "Initial filtering" is a process to remove HTML tags and unnecessary line breaks and spaces from the data collected by the server and unify the format.

[0076] "Cleaning AI" is an artificial intelligence that uses natural language processing technology to correct input errors and inconsistencies in initially filtered data and mask personal information.

[0077] "Cleaning" is the process by which cleaning AI corrects errors in data and masks personal information.

[0078] "Retraining AI" is artificial intelligence that uses data reviewed and corrected by users to retrain new models to improve chatbot performance.

[0079] "Retraining" is the process of updating an AI model based on corrected data and repeating the training process.

[0080] An "artificial intelligence model" is a mathematical model generated by a retraining AI based on collected training data and used by the chatbot to respond to questions.

[0081] A "chatbot" is a computer program that implements a retrained artificial intelligence model to automatically interact with users.

[0082] "Monitoring" is the process of monitoring the quality of responses of a chatbot during operation, collecting feedback from users, and reflecting it in the next re-training.

[0083] "Feedback" refers to the user's evaluation and opinion of the chatbot's responses, and is information that can be used to further improve the chatbot's performance when it is retrained.

[0084] The present invention aims to automate the development and operation process of an AI chatbot system. A specific method for implementing the present invention will be described in detail below.

[0085] First, we will explain the process of collecting training data. The user provides the information to be collected to the device. For example, the user might enter the URL of a company's FAQ page or product review into the device. Based on this information, the device sends a data collection command to the server. The server accesses the specified website or database and obtains the required text data by scraping it. This data collection can be performed using, for example, the Python libraries "BeautifulSoup" or "Scrapy."

[0086] Next, the server performs initial filtering of the collected data. The server removes HTML tags, unnecessary line breaks, and spaces from the collected data to unify the format. Specifically, this filtering process is performed using regular expressions (Regex) and the Python "BeautifulSoup" library.

[0087] Next, the data is cleaned. The cleaning AI uses natural language processing technology to correct typos and inconsistencies in the filtered data. For example, it uses natural language processing libraries such as Python's "spaCy" and "NLTK." The cleaning AI also automatically recognizes and masks personal and confidential information. This process uses custom rules and machine learning models for personal information detection.

[0088] The server then presents the cleaned data to the user, who can then review the data and make any necessary corrections. This review and correction can be performed via a client-side web interface. For example, a user-friendly interface can be built using a front-end framework such as React or Angular.

[0089] The data that the user has confirmed and corrected is then sent back to the server. The server then provides this data to the re-learning AI and retrains the model. Deep learning frameworks such as TENSORFLOW (registered trademark) and PyTorch are suitable for re-learning.

[0090] The new retrained model is sent from the server to the device. The device then installs the new model in the chatbot and begins operation. Chatbots can be built using, for example, Dialogflow or the Microsoft® Bot Framework.

[0091] During operation, the server monitors the quality of the chatbot's responses and collects user feedback, which is used during the next retraining phase to continuously improve the chatbot's accuracy.

[0092] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. The generated new model is installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[0093] Examples of prompts include:

[0094] "Gather training data from a company's FAQ page and past support emails, remove unnecessary information, mask personal information, and generate a new AI model."

[0095] The above is a specific form for implementing the present invention, which enables the efficient construction and operation of a highly accurate AI chatbot system.

[0096] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0097] Step 1: Instructions for collecting training data

[0098] The user provides the information to be collected to the device. Specifically, they input the URL of a company's website's FAQ page or a product review into the device. Based on this input data, the device sends a data collection instruction to the server. The input data may be a URL or a database path, and the device organizes it and sends it to the server. The specific operation of the device is to collect the information entered by the user and generate instructions for the server in an appropriate format.

[0099] Input: User-specified URL or database path

[0100] Output: Organized data collection instructions

[0101] Step 2: Collect training data

[0102] The server accesses the specified website or database and scrapes the required text data. The server connects to the specified URL, analyzes the HTML content of the page, and extracts the text data. For example, scraping libraries such as "BeautifulSoup" and "Scrapy" are used to collect the data.

[0103] Input: Organized data collection instructions

[0104] Output: Collected text data

[0105] Step 3: Initial filtering

[0106] The server performs initial filtering of the collected data, specifically removing HTML tags and formatting line breaks and whitespace. It uses regular expressions (Regex) and libraries such as the BeautifulSoup library to remove unnecessary parts of the data and standardize the format.

[0107] Input: Collected text data

[0108] Output: filtered text data

[0109] Step 4: Cleaning the data

[0110] The cleaning AI cleans the filtered data. Specifically, it uses natural language processing technology to correct typos and inconsistencies and mask personal information. It uses Python libraries such as "spaCy" and "NLTK." This improves data accuracy and protects personal information.

[0111] Input: filtered text data

[0112] Output: Cleaned text data

[0113] Step 5: Check and correct data

[0114] The server presents the cleaned data to the user, who then checks it and makes corrections as necessary. The server receives the corrected data and records it in a log. Since the user can check and correct the data through a web interface, it is effective to use a front-end framework such as React or Angular.

[0115] Input: Cleaned text data

[0116] Output: Data reviewed and corrected by the user

[0117] Step 6: Retrain

[0118] The server provides the verified and corrected data to the retraining AI, retraining the model, and builds a new generative AI model using a deep learning framework such as TensorFlow or PyTorch.

[0119] Input: Data confirmed and corrected by the user

[0120] Output: Retrained generative AI model

[0121] Step 7: Update and install the model

[0122] The server sends the generated new model to the device, which then installs it in the chatbot. The device then updates the chatbot with the new model and starts using it. For example, using Dialogflow or the Microsoft Bot Framework.

[0123] Input: Retrained generative AI model

[0124] Output: The new model installed in the chatbot

[0125] Step 8: Operate and monitor your chatbot

[0126] Users use the chatbot equipped with the new model to answer their questions. The server monitors the quality of the chatbot's responses and collects feedback. This feedback is used for the next retraining. The collected feedback is stored in a database and used for continuous improvement.

[0127] Input: User questions and feedback

[0128] Output: Monitoring results and collected feedback

[0129] Through these steps, this system will enable the development and operation of highly accurate and efficient AI chatbot systems.

[0130] (Application example 1)

[0131] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0132] In today's brick-and-mortar stores, the knowledge and skills of store clerks are important for smoothly responding to a wide range of customer questions. However, because not every store clerk has complete knowledge of all products and services, customer service is often inconsistent. Furthermore, there are limited means to search and provide vast amounts of information in real time to accurately answer customer questions. In such situations, there is a risk of customer satisfaction decreasing, so a smart system to assist store clerks is needed.

[0133] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0134] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, and means for implementing the re-trained artificial intelligence model and operating it as a chatbot to assist store clerks in customer service in real time using smart glasses. This enables store clerks to receive questions from customers through the smart glasses and provide appropriate answers in real time based on a vast amount of information, thereby increasing customer satisfaction.

[0135] "Training data" is the data set used to train an AI model.

[0136] "Means of collection" refers to the method or device used to obtain the required information from the website or database specified by the user.

[0137] "Initial filtering" is a process of removing unnecessary information from collected data and standardizing the data format.

[0138] "Cleaning AI" is an AI technology that cleans collected data to eliminate errors and inconsistencies, and properly processes personal and confidential information.

[0139] The "means for accepting user confirmation and correction" is an interface that allows the user to confirm the cleaned data and correct it if necessary.

[0140] "Retraining" is a method of retraining an existing AI model with the latest information using data that has been confirmed and corrected by the user.

[0141] "Means of implementation and operation as a chatbot" refers to the method or device for incorporating the retrained AI model into a chatbot and using it in an actual operational environment.

[0142] "Smart glasses" are wearable devices equipped with displays and sensors that display information within the store clerk's field of vision.

[0143] The "means of assisting customers in real time" is a system that allows store clerks to instantly provide appropriate answers to customers' questions through smart glasses.

[0144] As a concrete example of implementing this invention, we will explain a store clerk support system for brick-and-mortar stores using smart glasses. This system collects learning data, performs initial filtering, cleaning, checking and correcting, re-learning, and operates as a chatbot. The details are described below.

[0145] First, the device collects training data from websites and databases specified by the user. At this stage, Scrapy and BeautifulSoup are used as information collection tools. These tools allow the device to efficiently extract the necessary text data.

[0146] Next, the server performs initial filtering on the collected data, which involves removing HTML tags, line breaks, and whitespace formatting, using the standard Python library and regular expressions.

[0147] The collected and initially filtered data is then further processed by the Cleaning AI, which uses natural language processing techniques to correct data typos and inconsistencies, and masks personal and confidential information. This process uses natural language processing libraries such as SpaCy and NLTK.

[0148] Once the data has been cleaned, it is presented to the user for review and correction. This review process is carried out using a web-based interface, allowing the user to correct the data in their browser. This stage ensures the accuracy and consistency of the data.

[0149] Based on the revised data, the server updates the AI ​​model for retraining. This retraining process uses deep learning frameworks such as TensorFlow or PyTorch. The retrained model improves based on the new information and can generate more accurate responses.

[0150] The retrained AI model will be implemented as a chatbot and made available to store associates in real time through smart glasses, a wearable device equipped with a camera and a display, allowing store associates to instantly view the information they need within their field of vision.

[0151] As a specific use case, when a customer asks, "What features does this product have?", the chatbot in the smart glasses will instantly respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[0152] An example prompt is:

[0153] Customer: What features does this product have?

[0154] AI: This product has the latest Bluetooth functionality, long battery life, and is waterproof.

[0155] This system will enable store staff to respond to customers quickly and accurately, which is expected to improve customer satisfaction.

[0156] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0157] Step 1:

[0158] The device collects training data from websites and databases specified by the user. The user provides the device with the URLs and database information to be collected. The device uses Scrapy or BeautifulSoup to extract the necessary text data from the websites and databases and send it to the server. The input for this step is the URL of the website or database, and the output is the collected text data.

[0159] Step 2:

[0160] The server performs initial filtering of the collected data. It uses the Python standard library and regular expressions to remove HTML tags and format lines and spaces. In this process, the data format is unified and unnecessary information is removed. The input of this step is the collected text data, and the output is the initially filtered text data.

[0161] Step 3:

[0162] The server cleans the initially filtered data. This process is performed by a cleaning AI, which uses natural language processing libraries such as SpaCy and NLTK to correct typos and inconsistencies in the data, as well as masking personal and confidential information. The input of this step is the initially filtered text data, and the output is the cleaned text data.

[0163] Step 4:

[0164] The server presents the cleaned data to the user for review and correction. The user reviews the cleaned data through a web-based interface and makes corrections as needed. A browser is used for this review and correction work. The input of the step is the cleaned text data, and the output is the text data reviewed and corrected by the user.

[0165] Step 5:

[0166] The server retrains the AI ​​model based on the confirmed and corrected data. It uses a deep learning framework such as TensorFlow or PyTorch to update the retrained AI model. The input for this step is the text data confirmed and corrected by the user, and the output is the retrained AI model.

[0167] Step 6:

[0168] The server uses the retrained AI model to operate as a chatbot. The smart glasses use this chatbot to assist store clerks in dealing with customers in real time. Through the display and camera installed in the smart glasses, store clerks can receive customer questions and instantly display appropriate answers. The input for this step is the retrained AI model, and the output is customer service information provided in real time.

[0169] For example, if a customer asks, "What features does this product have?" the smart glasses' display will respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[0170] 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.

[0171] The purpose of this invention is to provide responses that correspond to the user's emotions and improve their accuracy by combining an emotion engine that recognizes the user's emotions with an AI chatbot system. The specific operation of the system's program is explained below.

[0172] First, the user inputs the website or database information to be collected into the device. The device receives this information and sends a collection instruction to the server. The server then accesses the specified website or database and collects the necessary text data. For example, it can extract question and answer pairs from an FAQ page.

[0173] Next, the server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. Once the initial filtering is complete, the data moves on to the next step.

[0174] Cleaning AI then takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0175] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This process ensures the accuracy of the data.

[0176] The server then passes the revised data to a retraining AI, which uses the data to retrain the chatbot's model, generating a new model in the process.

[0177] The generated new model is sent from the server to the device. The device installs this model in the chatbot, and the chatbot begins operation. The user asks a question using the chatbot, and the chatbot generates an appropriate response based on the new model.

[0178] In the present invention, the system is further improved by adding an emotion engine. When a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. The analyzed emotion data is sent to the server and reflected in the response. For example, if the user is feeling angry or stressed, the chatbot will provide a calm and gentle response accordingly.

[0179] Furthermore, emotional data is collected and used for the next retraining. This allows the chatbot to continually improve and respond better to user emotions. For example, in a customer support chatbot, when a user expresses dissatisfaction, the dissatisfaction can be handled quickly and appropriately, improving customer satisfaction.

[0180] In this way, by combining the collection, cleaning, and relearning of learning data with an emotion engine, the present invention enables the construction and operation of a highly accurate chatbot system that takes emotions into consideration.

[0181] The processing flow will be explained below.

[0182] Step 1:

[0183] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[0184] Step 2:

[0185] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[0186] Step 3:

[0187] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[0188] Step 4:

[0189] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0190] Step 5:

[0191] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[0192] Step 6:

[0193] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[0194] Step 7:

[0195] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[0196] Step 8:

[0197] The user uses the chatbot to ask a question, and the chatbot generates an appropriate response based on the new model. During this process, the emotion engine recognizes the user's emotions and analyzes the user's emotional state from the text. For example, if the user is angry, the chatbot will provide a calm and gentle response accordingly.

[0198] Step 9:

[0199] The server monitors the chatbot's responses and records the emotional data collected by the emotion engine. The server collects this data as feedback and uses it for the next re-training.

[0200] Step 10:

[0201] During retraining, the server passes feedback, including emotional data, to the retraining AI to train it to improve its emotional responses, allowing the chatbot to better respond to the user's emotions.

[0202] Example 2

[0203] 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."

[0204] Existing AI chatbot systems often lack the ability to recognize user emotions and respond inappropriately. Furthermore, there is a need for efficient and accurate methods for the entire process, from data collection to model retraining and data cleansing. In particular, challenges remain in automatically masking personal and confidential information and removing duplicate data. Furthermore, the accuracy of chatbot responses has not been sufficiently improved due to a lack of continuous training using user feedback and sentiment analysis data.

[0205] 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.

[0206] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, and artificial intelligence means for cleaning the initially filtered data. It also includes means for presenting the cleaned data to a user and accepting user confirmation and correction, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing emotions from text input by the user, means for generating chatbot responses based on this emotional data, and means for collecting emotional data and using it for the next relearning. This enables the generation of appropriate responses based on the user's emotions, thereby improving the accuracy of chatbot responses while protecting personal and confidential information.

[0207] "Training data" refers to information or data sets used to train an artificial intelligence model.

[0208] "Collection means" refers to the method or device used to obtain the necessary information from websites, databases, etc.

[0209] "Initial filtering" is a process for removing unnecessary information from collected data and standardizing the data format.

[0210] "Artificial intelligence means" refers to techniques that use machine learning algorithms and deep learning models to analyze and process data.

[0211] "Cleaning" is an operation to remove noise and errors from collected and filtered data and to arrange the data.

[0212] A "presentation means" is an interface that displays the cleaned data to the user and allows for review and modification.

[0213] "Retraining" is the process of retraining an artificial intelligence model using updated data.

[0214] A "model" refers to an artificial intelligence structure or algorithm that has been trained for a specific purpose.

[0215] A "chatbot" is a computer program that automatically handles interactions with users.

[0216] "Sentiment analysis" is a technique for detecting and classifying user emotions from text data.

[0217] "Response generation" is the process by which a chatbot creates a response to a user's input.

[0218] "Feedback" refers to usage evaluations and opinions provided by users, and is information that is used to improve the system.

[0219] "Auto-masking" is the process of automatically identifying personal or sensitive information from collected data and masking it to protect it.

[0220] This invention is an AI chatbot system that recognizes a user's emotions and provides responses according to those emotions. The purpose is to improve the accuracy of the data collected from the user while also increasing the accuracy of the chatbot. The implementation method of this system is specifically described below, specifying the hardware and software used.

[0221] First, the user inputs the website or database information to be collected into the device. For example, the user can input the URL of a customer support FAQ page. The device receives this input information and sends a collection instruction to the server using an HTTP request.

[0222] The server then accesses the specified websites and databases to collect the required text data, using a scraping tool to extract the text from FAQ pages and store it in a temporary database.

[0223] The collected data is initially filtered by the server, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. For example, the data is cleaned by using regular expressions to remove unnecessary HTML tags.

[0224] The server then uses cleaning AI to further clean the initially filtered data. Cleaning includes removing duplicate data, correcting typos, and automatically detecting and masking personal and confidential information. For example, if the same question appears multiple times, the cleaning AI will remove the duplicates.

[0225] The cleaned data is presented to the user by the server. The cleaned FAQ data is displayed through a user interface, and the user checks it. For example, the user checks whether the correct answer matches the question correctly. After the user makes the necessary corrections, the results are sent to the server.

[0226] The server then passes the corrected data to a retraining AI to retrain the chatbot's model. This process uses frameworks such as TensorFlow or PyTorch. Once the new model is generated, the model data is sent to the device using a secure protocol (e.g., HTTPS).

[0227] The terminal installs the new model into the chatbot. The chatbot begins operation by replacing the existing model with the new one. The user uses the chatbot based on the new model and asks questions. For example, if the user asks, "How do I return a product?", the chatbot responds, "You can return a product from your My Page on the website."

[0228] Furthermore, when a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. For example, natural language processing technology is used to classify emotions such as "anger," "joy," and "sadness." The server reflects the analyzed emotional data in the response; for example, if the server determines that the user is feeling "anger," it will provide a calm and gentle response accordingly.

[0229] The emotional data is stored on the server and used for the next re-learning, allowing the chatbot to continue generating appropriate responses according to the user's emotions.

[0230] As a concrete example, the following prompt sentence can be used:

[0231] "How do I return a product?"

[0232] In this way, the system of the present invention can provide responses that correspond to the user's emotions, making it possible to realize a highly accurate chatbot system.

[0233] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0234] Step 1:

[0235] The user inputs the information from the website or database to be collected into the terminal.

[0236] Input: URLs and queries for websites and databases to be collected

[0237] Specific operation: The user enters information into an input field in a browser or application, and a collection instruction is sent to the terminal.

[0238] Output: The terminal generates data along with collection instructions to convert the input information into a format to be sent to the server.

[0239] Step 2:

[0240] The terminal sends a collection instruction to the server using an HTTP request.

[0241] Input: Collected instruction data from the user

[0242] Specific operation: The terminal packages the input information into an HTTP request and sends it to the server.

[0243] Output: The server receives the HTTP request and interprets the collection instructions data.

[0244] Step 3:

[0245] The server accesses the specified website or database and collects the required text data.

[0246] Input: Collection instruction data

[0247] What happens: The server uses a scraping tool to extract text data from the specified URL, for example, extracting question and answer pairs from an FAQ page.

[0248] Output: Save the extracted text data in a temporary database.

[0249] Step 4:

[0250] The server performs an initial filtering of the collected data.

[0251] Input: Collected text data

[0252] Specific operation: The server removes unnecessary HTML tags from the data, formats line breaks and whitespace, unifies date and time formats, etc. It uses regular expressions and text processing libraries.

[0253] Output: Generates the initial filtered data.

[0254] Step 5:

[0255] The server further cleans the initially filtered data using cleaning AI.

[0256] Input: Initial filtered data

[0257] What it does: Removes duplicate data, corrects typos, automatically detects and masks personal and confidential information, and uses natural language processing techniques to identify specific patterns and data formats.

[0258] Output: Produces the cleaned data.

[0259] Step 6:

[0260] The server presents the cleaned data to the user and accepts confirmation and correction by the user.

[0261] Input: Cleaned data

[0262] Specific operation: The server displays the data through a user interface, allowing the user to review and modify it. It also provides an interface for receiving modified data.

[0263] Output: Data reviewed and corrected by the user

[0264] Step 7:

[0265] The server retrains the artificial intelligence model based on the data confirmed and corrected by the user.

[0266] Input: Verified and corrected data

[0267] What it does: Retrain the model using a framework such as TensorFlow or PyTorch, using the corrected data as training data to update the neural network weights.

[0268] Output: A new retrained AI model

[0269] Step 8:

[0270] The server transmits the retrained artificial intelligence model to the terminal.

[0271] Input: Retrained AI model

[0272] Specific operation: The server sends the new model to the device using a secure protocol (e.g., HTTPS).

[0273] Output: The new model data sent to the terminal.

[0274] Step 9:

[0275] The device installs the retrained artificial intelligence model into the chatbot and begins operating the chatbot.

[0276] Input: Retrained AI model

[0277] Specific behavior: Replace the existing model with the new one. Restart the chatbot software and load the new model.

[0278] Output: A running chatbot with the new model installed

[0279] Step 10:

[0280] Users use chatbots based on the new model to ask questions.

[0281] Input: User question text

[0282] Specific behavior: For example, ask, "How do I return an item?" The chatbot uses the new model to generate an appropriate response.

[0283] Output: Chatbot response text

[0284] Step 11:

[0285] When a user types a question into the chatbot, the emotion engine analyzes the user's emotions.

[0286] Input: User question text

[0287] Specific operation: Using natural language processing technology, emotions such as "anger," "joy," and "sadness" are classified.

[0288] Output: Parsed emotion data

[0289] Step 12:

[0290] The server reflects the analyzed emotional data in its response and takes appropriate action.

[0291] Input: Parsed emotion data, user question text

[0292] Specific behavior: If the user is analyzed as feeling "anger," the chatbot will provide a calm and gentle response.

[0293] Output: Chatbot response depending on the emotion

[0294] Step 13:

[0295] The emotion data is stored on the server and used for the next re-learning.

[0296] Input: Parsed emotion data

[0297] Specific operation: Emotion data is saved in the database and used the next time the model is retrained.

[0298] Output: Saved emotion data

[0299] (Application example 2)

[0300] 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."

[0301] Current customer support systems in brick-and-mortar stores face the problem of being unable to analyze customers' emotions, such as facial expressions and tone of voice, in real time and provide appropriate support based on that analysis. Furthermore, the inability to respond appropriately to customer emotions makes it difficult to improve customer satisfaction. To solve this problem, a system is needed that analyzes customers' emotions and provides instructions on how to respond in real time based on the results.

[0302] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing user emotions using an emotion analysis engine, and means for reflecting the user's emotion data in the response content. This makes it possible to analyze customer emotions in real time and instruct store staff on appropriate responses based on the analysis results.

[0303] "Means for collecting training data" refers to the mechanism for obtaining the necessary text data from designated websites and databases.

[0304] "Initial filtering" refers to the process of removing unnecessary information from collected data and shaping the data.

[0305] "Artificial intelligence means" refers to a system that uses deep learning and natural language processing techniques to clean and analyze data.

[0306] The "means for presenting to the user and accepting confirmation and correction by the user" refers to an interface that shows the cleaned data to the user and accepts confirmation and correction.

[0307] The "means for retraining an artificial intelligence model" is a function for retraining a model based on data that has been checked and corrected by the user.

[0308] "Means of operating as a chatbot" refers to the process of installing and operating the retrained artificial intelligence model in an actual chatbot system.

[0309] "Means for analyzing user emotions using an emotion analysis engine" refers to a process that uses an engine to analyze an emotional state from text or voice data entered by a user.

[0310] The "means for reflecting the user's emotional data in the response content" is a function for adjusting the response content to the user based on the analyzed emotional data.

[0311] To implement this invention, store staff wear smart glasses and use the glasses' camera and microphone to capture customer facial expressions and voices. The data is then sent to a server in real time, where an emotion analysis engine is used to analyze the customer's emotional state. Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[0312] This system uses the following hardware and software:

[0313] Hardware

[0314] Smart glasses (with built-in camera and microphone)

[0315] Server (high-performance cloud server, e.g., Amazon EC2, Microsoft Azure (registered trademark))

[0316] software

[0317] Sentiment analysis engine (e.g., Microsoft Azure Cognitive Services Emotion API)

[0318] Data processing and communication (Python, TensorFlow, Flask)

[0319] Overview of data processing and calculation

[0320] 1. Capture customer facial and voice data

[0321] The smart glasses' camera and microphone capture the customer's facial and voice data in real time.

[0322] 2. Sending data to the server

[0323] The captured data is sent from the smart glasses to a server.

[0324] 3. Pretreatment

[0325] The received data is preprocessed on the server side (noise removal, data formatting, etc.) and then passed to the sentiment analysis engine.

[0326] 4. Emotion analysis

[0327] The sentiment analysis engine analyzes the customer's emotional state (e.g., "anger," "sadness," "joy," etc.).

[0328] 5. Feedback of response instructions

[0329] Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[0330] Specific examples

[0331] For example, if a customer looks dissatisfied while asking about a product at a counter, their facial expression and voice data will be captured, and the server will perform emotion analysis and determine that they are dissatisfied. As a result, the smart glasses will display a message saying, "The customer seems dissatisfied. Please be courteous and considerate."

[0332] Prompt Sentence Examples

[0333] "Please analyze the customer's emotions from the following facial and voice data:

[0334] Facial expression data: {image_data}

[0335] Audio data: {audio_data}"

[0336] In this way, it becomes possible to provide more effective customer service in physical stores, which contributes to improving customer satisfaction.

[0337] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0338] Step 1:

[0339] The smart glasses capture the customer's facial expression and voice data. They use the customer's real-time video and audio as input, which are captured by the smart glasses' built-in camera and microphone. The captured data is output in byte stream format.

[0340] Step 2:

[0341] The smart glasses send the captured data to the server. The captured byte stream data is the input and is sent to the server through the communication module. The output is the byte stream data received on the server side.

[0342] Step 3:

[0343] The server preprocesses the received data. The input data is the byte stream sent from the smart glasses, and it undergoes noise removal and data shaping (e.g., frame extraction and audio cleaning). The preprocessed shaped data is output as image data and audio data.

[0344] Step 4:

[0345] The server passes the preprocessed data to the sentiment analysis engine. The input data is the preprocessed image data and audio data, which are then input to the sentiment analysis engine. The sentiment analysis engine uses a deep learning model to identify the emotional state (e.g., "anger," "sadness," "joy," etc.). The output is text data that indicates the emotional state.

[0346] Step 5:

[0347] The server generates response instructions based on the analysis results. The output text data from the sentiment analysis engine is used as input, and response instructions for staff are generated based on that. For example, a message such as "The customer seems dissatisfied. Please try to respond politely and kindly" is generated. The output is text data showing response instructions.

[0348] Step 6:

[0349] The server sends the generated response instruction to the smart glasses. The input is text data indicating the response instruction, which is sent to the smart glasses through the communication module. The output is a text message displayed on the smart glasses.

[0350] Step 7:

[0351] The smart glasses display instructions to the staff. Text data indicating instructions sent from the server is used as input and displayed on the smart glasses' display. The staff responds to the customer based on this displayed information. The output is an instruction message that is displayed to the staff.

[0352] The above is a flow of specific processing steps for carrying out the invention, which is expected to improve the quality of customer service in brick-and-mortar stores and increase customer satisfaction.

[0353] 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.

[0354] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0355] 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.

[0356] [Second embodiment]

[0357] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0358] 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.

[0359] 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).

[0360] 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.

[0361] 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.

[0362] 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).

[0363] 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.

[0364] 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.

[0365] 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.

[0366] 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.

[0367] 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.

[0368] 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."

[0369] The purpose of this invention is to automate the development and operation process of an AI chatbot system. The specific operation of the system's program is explained below.

[0370] First, the device collects learning data. The user provides the device with information about the websites and databases to be collected. The device uses the specified information to send a data collection command to the server. The server then accesses the specified websites and databases and retrieves the necessary text data. For example, the data needed for learning can be collected from FAQ pages or user comments.

[0371] Next, the server performs initial filtering on the collected data to standardize the data format and remove unnecessary information, such as removing HTML tags and formatting line breaks and whitespace.

[0372] Next, the cleaning AI cleans the filtered data. It uses natural language processing technology to correct typos and inconsistencies in the data. It also automatically recognizes and masks personal and confidential information. For example, it appropriately conceals personal information such as email addresses and credit card information.

[0373] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This checking and correction is an important step in maintaining the accuracy of the data.

[0374] With the corrected data, the server passes it to the retraining AI, which uses the new data to retrain the chatbot's model. This process ensures that the AI ​​model is updated with the latest information.

[0375] The new model generated through retraining is sent from the server to the device. The device then installs this model into the chatbot and begins operation. Users ask questions using the chatbot, and the chatbot responds based on the data it has learned.

[0376] While the chatbot is running, the server monitors the quality of responses, collects user feedback, and uses it for the next retraining phase, allowing the chatbot to continually improve its accuracy.

[0377] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. A new model is generated and installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[0378] In this way, the present invention effectively automates the collection, cleaning, relearning, and operation of learning data, enabling the construction and operation of a highly accurate chatbot system.

[0379] The processing flow will be explained below.

[0380] Step 1:

[0381] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[0382] Step 2:

[0383] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[0384] Step 3:

[0385] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[0386] Step 4:

[0387] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0388] Step 5:

[0389] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[0390] Step 6:

[0391] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[0392] Step 7:

[0393] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[0394] Step 8:

[0395] A user asks a question using the chatbot, the chatbot generates an appropriate response based on the new model, and the server monitors and quality-controls this response.

[0396] Step 9:

[0397] The server collects user feedback and evaluates the quality of responses. This feedback is used during the next retraining phase. This continuous feedback loop helps improve the chatbot's performance.

[0398] Example 1

[0399] 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."

[0400] Conventional AI chatbot systems required a great deal of time and effort, from collecting and filtering training data, cleaning the data, to retraining the model, and then operating it. Furthermore, manual verification work was required to protect personal information and maintain data accuracy, and automation was insufficient, resulting in low efficiency. Furthermore, there was an insufficient mechanism for continuously improving the quality of chatbot responses during operation, making it difficult to appropriately incorporate customer feedback.

[0401] 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.

[0402] In this invention, the server includes: means for a user to provide information to be collected and a terminal to send a data collection instruction to the server based on that information; means for the server to access specified websites and databases to collect training data; means for the server to initially filter the training data and remove unnecessary information; means for a cleaning AI to correct input errors and inconsistencies in the filtered data using natural language processing technology and mask personal information; means for presenting the cleaned data to the user and accepting confirmation and correction by the user; means for a re-training AI to re-train an AI model based on the data confirmed and corrected by the user; means for sending the AI ​​model generated by the re-training to the terminal and operating it as a chatbot; and means for monitoring the quality of the chatbot's responses, collecting user feedback, and using it for the next re-training. This automates the process from collecting training data to re-training the model, operation, and continuous improvement of response quality, enabling the construction and operation of an efficient and highly accurate AI chatbot system.

[0403] "User" means a person or organization that uses the system to specify the targets for collecting learning data and to review and correct the data.

[0404] A "terminal" is a hardware or software component that allows a user to input information, send data collection instructions to a server, install a retrained model, and operate it as a chatbot.

[0405] The "server" is the core technology of the system that receives instructions from users and devices, collects training data, performs initial filtering and processing using cleaning AI, and then supplies the data to the re-training AI to re-train the model.

[0406] A "data collection instruction" is an instruction to command the server to start collecting learning data based on information input by the user into the terminal.

[0407] "Training data" refers to text data and information collected to train an AI chatbot model.

[0408] "Initial filtering" is a process to remove HTML tags and unnecessary line breaks and spaces from the data collected by the server and unify the format.

[0409] "Cleaning AI" is an artificial intelligence that uses natural language processing technology to correct input errors and inconsistencies in initially filtered data and mask personal information.

[0410] "Cleaning" is the process by which cleaning AI corrects errors in data and masks personal information.

[0411] "Retraining AI" is artificial intelligence that uses data reviewed and corrected by users to retrain new models to improve chatbot performance.

[0412] "Retraining" is the process of updating an AI model based on corrected data and repeating the training process.

[0413] An "artificial intelligence model" is a mathematical model generated by a retraining AI based on collected training data and used by the chatbot to respond to questions.

[0414] A "chatbot" is a computer program that implements a retrained artificial intelligence model to automatically interact with users.

[0415] "Monitoring" is the process of monitoring the quality of responses of a chatbot during operation, collecting feedback from users, and reflecting it in the next re-training.

[0416] "Feedback" refers to the user's evaluation and opinion of the chatbot's responses, and is information that can be used to further improve the chatbot's performance when it is retrained.

[0417] The present invention aims to automate the development and operation process of an AI chatbot system. A specific method for implementing the present invention will be described in detail below.

[0418] First, we will explain the process of collecting training data. The user provides the information to be collected to the device. For example, the user might enter the URL of a company's FAQ page or product review into the device. Based on this information, the device sends a data collection command to the server. The server accesses the specified website or database and obtains the required text data by scraping it. This data collection can be performed using, for example, the Python libraries "BeautifulSoup" or "Scrapy."

[0419] Next, the server performs initial filtering of the collected data. The server removes HTML tags, unnecessary line breaks, and spaces from the collected data to unify the format. Specifically, this filtering process is performed using regular expressions (Regex) and the Python "BeautifulSoup" library.

[0420] Next, the data is cleaned. The cleaning AI uses natural language processing technology to correct typos and inconsistencies in the filtered data. For example, it uses natural language processing libraries such as Python's "spaCy" and "NLTK." The cleaning AI also automatically recognizes and masks personal and confidential information. This process uses custom rules and machine learning models for personal information detection.

[0421] The server then presents the cleaned data to the user, who can then review the data and make any necessary corrections. This review and correction can be performed via a client-side web interface. For example, a user-friendly interface can be built using a front-end framework such as React or Angular.

[0422] The data that the user has confirmed and corrected is then sent back to the server, which then provides this data to the retraining AI and retrains the model. Deep learning frameworks such as TensorFlow and PyTorch are suitable for retraining.

[0423] The new retrained model is sent from the server to the device. The device then installs the new model in the chatbot and begins operation. Chatbots can be built using, for example, Dialogflow or the Microsoft Bot Framework.

[0424] During operation, the server monitors the quality of the chatbot's responses and collects user feedback, which is used during the next retraining phase to continuously improve the chatbot's accuracy.

[0425] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. The generated new model is installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[0426] Examples of prompts include:

[0427] "Gather training data from a company's FAQ page and past support emails, remove unnecessary information, mask personal information, and generate a new AI model."

[0428] The above is a specific form for implementing the present invention, which enables the efficient construction and operation of a highly accurate AI chatbot system.

[0429] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0430] Step 1: Instructions for collecting training data

[0431] The user provides the information to be collected to the device. Specifically, they input the URL of a company's website's FAQ page or a product review into the device. Based on this input data, the device sends a data collection instruction to the server. The input data may be a URL or a database path, and the device organizes it and sends it to the server. The specific operation of the device is to collect the information entered by the user and generate instructions for the server in an appropriate format.

[0432] Input: User-specified URL or database path

[0433] Output: Organized data collection instructions

[0434] Step 2: Collect training data

[0435] The server accesses the specified website or database and scrapes the required text data. The server connects to the specified URL, analyzes the HTML content of the page, and extracts the text data. For example, scraping libraries such as "BeautifulSoup" and "Scrapy" are used to collect the data.

[0436] Input: Organized data collection instructions

[0437] Output: Collected text data

[0438] Step 3: Initial filtering

[0439] The server performs initial filtering of the collected data, specifically removing HTML tags and formatting line breaks and whitespace. It uses regular expressions (Regex) and libraries such as the BeautifulSoup library to remove unnecessary parts of the data and standardize the format.

[0440] Input: Collected text data

[0441] Output: filtered text data

[0442] Step 4: Cleaning the data

[0443] The cleaning AI cleans the filtered data. Specifically, it uses natural language processing technology to correct typos and inconsistencies and mask personal information. It uses Python libraries such as "spaCy" and "NLTK." This improves data accuracy and protects personal information.

[0444] Input: filtered text data

[0445] Output: Cleaned text data

[0446] Step 5: Check and correct data

[0447] The server presents the cleaned data to the user, who then checks it and makes corrections as necessary. The server receives the corrected data and records it in a log. Since the user can check and correct the data through a web interface, it is effective to use a front-end framework such as React or Angular.

[0448] Input: Cleaned text data

[0449] Output: Data reviewed and corrected by the user

[0450] Step 6: Retrain

[0451] The server provides the verified and corrected data to the retraining AI, retraining the model, and builds a new generative AI model using a deep learning framework such as TensorFlow or PyTorch.

[0452] Input: Data confirmed and corrected by the user

[0453] Output: Retrained generative AI model

[0454] Step 7: Update and install the model

[0455] The server sends the generated new model to the device, which then installs it in the chatbot. The device then updates the chatbot with the new model and starts using it. For example, using Dialogflow or the Microsoft Bot Framework.

[0456] Input: Retrained generative AI model

[0457] Output: The new model installed in the chatbot

[0458] Step 8: Operate and monitor your chatbot

[0459] Users use the chatbot equipped with the new model to answer their questions. The server monitors the quality of the chatbot's responses and collects feedback. This feedback is used for the next retraining. The collected feedback is stored in a database and used for continuous improvement.

[0460] Input: User questions and feedback

[0461] Output: Monitoring results and collected feedback

[0462] Through these steps, this system will enable the development and operation of highly accurate and efficient AI chatbot systems.

[0463] (Application example 1)

[0464] 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."

[0465] In today's brick-and-mortar stores, the knowledge and skills of store clerks are important for smoothly responding to a wide range of customer questions. However, because not every store clerk has complete knowledge of all products and services, customer service is often inconsistent. Furthermore, there are limited means to search and provide vast amounts of information in real time to accurately answer customer questions. In such situations, there is a risk of customer satisfaction decreasing, so a smart system to assist store clerks is needed.

[0466] 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.

[0467] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, and means for implementing the re-trained artificial intelligence model and operating it as a chatbot to assist store clerks in customer service in real time using smart glasses. This enables store clerks to receive questions from customers through the smart glasses and provide appropriate answers in real time based on a vast amount of information, thereby increasing customer satisfaction.

[0468] "Training data" is the data set used to train an AI model.

[0469] "Means of collection" refers to the method or device used to obtain the required information from the website or database specified by the user.

[0470] "Initial filtering" is a process of removing unnecessary information from collected data and standardizing the data format.

[0471] "Cleaning AI" is an AI technology that cleans collected data to eliminate errors and inconsistencies, and properly processes personal and confidential information.

[0472] The "means for accepting user confirmation and correction" is an interface that allows the user to confirm the cleaned data and correct it if necessary.

[0473] "Retraining" is a method of retraining an existing AI model with the latest information using data that has been confirmed and corrected by the user.

[0474] "Means of implementation and operation as a chatbot" refers to the method or device for incorporating the retrained AI model into a chatbot and using it in an actual operational environment.

[0475] "Smart glasses" are wearable devices equipped with displays and sensors that display information within the store clerk's field of vision.

[0476] The "means of assisting customers in real time" is a system that allows store clerks to instantly provide appropriate answers to customers' questions through smart glasses.

[0477] As a concrete example of implementing this invention, we will explain a store clerk support system for brick-and-mortar stores using smart glasses. This system collects learning data, performs initial filtering, cleaning, checking and correcting, re-learning, and operates as a chatbot. The details are described below.

[0478] First, the device collects training data from websites and databases specified by the user. At this stage, Scrapy and BeautifulSoup are used as information collection tools. These tools allow the device to efficiently extract the necessary text data.

[0479] Next, the server performs initial filtering on the collected data, which involves removing HTML tags, line breaks, and whitespace formatting, using the standard Python library and regular expressions.

[0480] The collected and initially filtered data is then further processed by the Cleaning AI, which uses natural language processing techniques to correct data typos and inconsistencies, and masks personal and confidential information. This process uses natural language processing libraries such as SpaCy and NLTK.

[0481] Once the data has been cleaned, it is presented to the user for review and correction. This review process is carried out using a web-based interface, allowing the user to correct the data in their browser. This stage ensures the accuracy and consistency of the data.

[0482] Based on the revised data, the server updates the AI ​​model for retraining. This retraining process uses deep learning frameworks such as TensorFlow or PyTorch. The retrained model improves based on the new information and can generate more accurate responses.

[0483] The retrained AI model will be implemented as a chatbot and made available to store associates in real time through smart glasses, a wearable device equipped with a camera and a display, allowing store associates to instantly view the information they need within their field of vision.

[0484] As a specific use case, when a customer asks, "What features does this product have?", the chatbot in the smart glasses will instantly respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[0485] An example prompt is:

[0486] Customer: What features does this product have?

[0487] AI: This product has the latest Bluetooth functionality, long battery life, and is waterproof.

[0488] This system will enable store staff to respond to customers quickly and accurately, which is expected to improve customer satisfaction.

[0489] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0490] Step 1:

[0491] The device collects training data from websites and databases specified by the user. The user provides the device with the URLs and database information to be collected. The device uses Scrapy or BeautifulSoup to extract the necessary text data from the websites and databases and send it to the server. The input for this step is the URL of the website or database, and the output is the collected text data.

[0492] Step 2:

[0493] The server performs initial filtering of the collected data. It uses the Python standard library and regular expressions to remove HTML tags and format lines and spaces. In this process, the data format is unified and unnecessary information is removed. The input of this step is the collected text data, and the output is the initially filtered text data.

[0494] Step 3:

[0495] The server cleans the initially filtered data. This process is performed by a cleaning AI, which uses natural language processing libraries such as SpaCy and NLTK to correct typos and inconsistencies in the data, as well as masking personal and confidential information. The input of this step is the initially filtered text data, and the output is the cleaned text data.

[0496] Step 4:

[0497] The server presents the cleaned data to the user for review and correction. The user reviews the cleaned data through a web-based interface and makes corrections as needed. A browser is used for this review and correction work. The input of the step is the cleaned text data, and the output is the text data reviewed and corrected by the user.

[0498] Step 5:

[0499] The server retrains the AI ​​model based on the confirmed and corrected data. It uses a deep learning framework such as TensorFlow or PyTorch to update the retrained AI model. The input for this step is the text data confirmed and corrected by the user, and the output is the retrained AI model.

[0500] Step 6:

[0501] The server uses the retrained AI model to operate as a chatbot. The smart glasses use this chatbot to assist store clerks in dealing with customers in real time. Through the display and camera installed in the smart glasses, store clerks can receive customer questions and instantly display appropriate answers. The input for this step is the retrained AI model, and the output is customer service information provided in real time.

[0502] For example, if a customer asks, "What features does this product have?" the smart glasses' display will respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[0503] 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.

[0504] The purpose of this invention is to provide responses that correspond to the user's emotions and improve their accuracy by combining an emotion engine that recognizes the user's emotions with an AI chatbot system. The specific operation of the system's program is explained below.

[0505] First, the user inputs the website or database information to be collected into the device. The device receives this information and sends a collection instruction to the server. The server then accesses the specified website or database and collects the necessary text data. For example, it can extract question and answer pairs from an FAQ page.

[0506] Next, the server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. Once the initial filtering is complete, the data moves on to the next step.

[0507] Cleaning AI then takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0508] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This process ensures the accuracy of the data.

[0509] The server then passes the revised data to a retraining AI, which uses the data to retrain the chatbot's model, generating a new model in the process.

[0510] The generated new model is sent from the server to the device. The device installs this model in the chatbot, and the chatbot begins operation. The user asks a question using the chatbot, and the chatbot generates an appropriate response based on the new model.

[0511] In the present invention, the system is further improved by adding an emotion engine. When a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. The analyzed emotion data is sent to the server and reflected in the response. For example, if the user is feeling angry or stressed, the chatbot will provide a calm and gentle response accordingly.

[0512] Furthermore, emotional data is collected and used for the next retraining. This allows the chatbot to continually improve and respond better to user emotions. For example, in a customer support chatbot, when a user expresses dissatisfaction, the dissatisfaction can be handled quickly and appropriately, improving customer satisfaction.

[0513] In this way, by combining the collection, cleaning, and relearning of learning data with an emotion engine, the present invention enables the construction and operation of a highly accurate chatbot system that takes emotions into consideration.

[0514] The processing flow will be explained below.

[0515] Step 1:

[0516] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[0517] Step 2:

[0518] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[0519] Step 3:

[0520] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[0521] Step 4:

[0522] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0523] Step 5:

[0524] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[0525] Step 6:

[0526] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[0527] Step 7:

[0528] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[0529] Step 8:

[0530] The user uses the chatbot to ask a question, and the chatbot generates an appropriate response based on the new model. During this process, the emotion engine recognizes the user's emotions and analyzes the user's emotional state from the text. For example, if the user is angry, the chatbot will provide a calm and gentle response accordingly.

[0531] Step 9:

[0532] The server monitors the chatbot's responses and records the emotional data collected by the emotion engine. The server collects this data as feedback and uses it for the next re-training.

[0533] Step 10:

[0534] During retraining, the server passes feedback, including emotional data, to the retraining AI to train it to improve its emotional responses, allowing the chatbot to better respond to the user's emotions.

[0535] Example 2

[0536] 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."

[0537] Existing AI chatbot systems often lack the ability to recognize user emotions and respond inappropriately. Furthermore, there is a need for efficient and accurate methods for the entire process, from data collection to model retraining and data cleansing. In particular, challenges remain in automatically masking personal and confidential information and removing duplicate data. Furthermore, the accuracy of chatbot responses has not been sufficiently improved due to a lack of continuous training using user feedback and sentiment analysis data.

[0538] 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.

[0539] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, and artificial intelligence means for cleaning the initially filtered data. It also includes means for presenting the cleaned data to a user and accepting user confirmation and correction, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing emotions from text input by the user, means for generating chatbot responses based on this emotional data, and means for collecting emotional data and using it for the next relearning. This enables the generation of appropriate responses based on the user's emotions, thereby improving the accuracy of chatbot responses while protecting personal and confidential information.

[0540] "Training data" refers to information or data sets used to train an artificial intelligence model.

[0541] "Collection means" refers to the method or device used to obtain the necessary information from websites, databases, etc.

[0542] "Initial filtering" is a process for removing unnecessary information from collected data and standardizing the data format.

[0543] "Artificial intelligence means" refers to techniques that use machine learning algorithms and deep learning models to analyze and process data.

[0544] "Cleaning" is an operation to remove noise and errors from collected and filtered data and to arrange the data.

[0545] A "presentation means" is an interface that displays the cleaned data to the user and allows for review and modification.

[0546] "Retraining" is the process of retraining an artificial intelligence model using updated data.

[0547] A "model" refers to an artificial intelligence structure or algorithm that has been trained for a specific purpose.

[0548] A "chatbot" is a computer program that automatically handles interactions with users.

[0549] "Sentiment analysis" is a technique for detecting and classifying user emotions from text data.

[0550] "Response generation" is the process by which a chatbot creates a response to a user's input.

[0551] "Feedback" refers to usage evaluations and opinions provided by users, and is information that is used to improve the system.

[0552] "Auto-masking" is the process of automatically identifying personal or sensitive information from collected data and masking it to protect it.

[0553] This invention is an AI chatbot system that recognizes a user's emotions and provides responses according to those emotions. The purpose is to improve the accuracy of the data collected from the user while also increasing the accuracy of the chatbot. The implementation method of this system is specifically described below, specifying the hardware and software used.

[0554] First, the user inputs the website or database information to be collected into the device. For example, the user can input the URL of a customer support FAQ page. The device receives this input information and sends a collection instruction to the server using an HTTP request.

[0555] The server then accesses the specified websites and databases to collect the required text data, using a scraping tool to extract the text from FAQ pages and store it in a temporary database.

[0556] The collected data is initially filtered by the server, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. For example, the data is cleaned by using regular expressions to remove unnecessary HTML tags.

[0557] The server then uses cleaning AI to further clean the initially filtered data. Cleaning includes removing duplicate data, correcting typos, and automatically detecting and masking personal and confidential information. For example, if the same question appears multiple times, the cleaning AI will remove the duplicates.

[0558] The cleaned data is presented to the user by the server. The cleaned FAQ data is displayed through a user interface, and the user checks it. For example, the user checks whether the correct answer matches the question correctly. After the user makes the necessary corrections, the results are sent to the server.

[0559] The server then passes the corrected data to a retraining AI to retrain the chatbot's model. This process uses frameworks such as TensorFlow or PyTorch. Once the new model is generated, the model data is sent to the device using a secure protocol (e.g., HTTPS).

[0560] The terminal installs the new model into the chatbot. The chatbot begins operation by replacing the existing model with the new one. The user uses the chatbot based on the new model and asks questions. For example, if the user asks, "How do I return a product?", the chatbot responds, "You can return a product from your My Page on the website."

[0561] Furthermore, when a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. For example, natural language processing technology is used to classify emotions such as "anger," "joy," and "sadness." The server reflects the analyzed emotional data in the response; for example, if the server determines that the user is feeling "anger," it will provide a calm and gentle response accordingly.

[0562] The emotional data is stored on the server and used for the next re-learning, allowing the chatbot to continue generating appropriate responses according to the user's emotions.

[0563] As a concrete example, the following prompt sentence can be used:

[0564] "How do I return a product?"

[0565] In this way, the system of the present invention can provide responses that correspond to the user's emotions, making it possible to realize a highly accurate chatbot system.

[0566] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0567] Step 1:

[0568] The user inputs the information from the website or database to be collected into the terminal.

[0569] Input: URLs and queries for websites and databases to be collected

[0570] Specific operation: The user enters information into an input field in a browser or application, and a collection instruction is sent to the terminal.

[0571] Output: The terminal generates data along with collection instructions to convert the input information into a format to be sent to the server.

[0572] Step 2:

[0573] The terminal sends a collection instruction to the server using an HTTP request.

[0574] Input: Collected instruction data from the user

[0575] Specific operation: The terminal packages the input information into an HTTP request and sends it to the server.

[0576] Output: The server receives the HTTP request and interprets the collection instructions data.

[0577] Step 3:

[0578] The server accesses the specified website or database and collects the required text data.

[0579] Input: Collection instruction data

[0580] What happens: The server uses a scraping tool to extract text data from the specified URL, for example, extracting question and answer pairs from an FAQ page.

[0581] Output: Save the extracted text data in a temporary database.

[0582] Step 4:

[0583] The server performs an initial filtering of the collected data.

[0584] Input: Collected text data

[0585] Specific operation: The server removes unnecessary HTML tags from the data, formats line breaks and whitespace, unifies date and time formats, etc. It uses regular expressions and text processing libraries.

[0586] Output: Generates the initial filtered data.

[0587] Step 5:

[0588] The server further cleans the initially filtered data using cleaning AI.

[0589] Input: Initial filtered data

[0590] What it does: Removes duplicate data, corrects typos, automatically detects and masks personal and confidential information, and uses natural language processing techniques to identify specific patterns and data formats.

[0591] Output: Produces the cleaned data.

[0592] Step 6:

[0593] The server presents the cleaned data to the user and accepts confirmation and correction by the user.

[0594] Input: Cleaned data

[0595] Specific operation: The server displays the data through a user interface, allowing the user to review and modify it. It also provides an interface for receiving modified data.

[0596] Output: Data reviewed and corrected by the user

[0597] Step 7:

[0598] The server retrains the artificial intelligence model based on the data confirmed and corrected by the user.

[0599] Input: Verified and corrected data

[0600] What it does: Retrain the model using a framework such as TensorFlow or PyTorch, using the corrected data as training data to update the neural network weights.

[0601] Output: A new retrained AI model

[0602] Step 8:

[0603] The server transmits the retrained artificial intelligence model to the terminal.

[0604] Input: Retrained AI model

[0605] Specific operation: The server sends the new model to the device using a secure protocol (e.g., HTTPS).

[0606] Output: The new model data sent to the terminal.

[0607] Step 9:

[0608] The device installs the retrained artificial intelligence model into the chatbot and begins operating the chatbot.

[0609] Input: Retrained AI model

[0610] Specific behavior: Replace the existing model with the new one. Restart the chatbot software and load the new model.

[0611] Output: A running chatbot with the new model installed

[0612] Step 10:

[0613] Users use chatbots based on the new model to ask questions.

[0614] Input: User question text

[0615] Specific behavior: For example, ask, "How do I return an item?" The chatbot uses the new model to generate an appropriate response.

[0616] Output: Chatbot response text

[0617] Step 11:

[0618] When a user types a question into the chatbot, the emotion engine analyzes the user's emotions.

[0619] Input: User question text

[0620] Specific operation: Using natural language processing technology, emotions such as "anger," "joy," and "sadness" are classified.

[0621] Output: Parsed emotion data

[0622] Step 12:

[0623] The server reflects the analyzed emotional data in its response and takes appropriate action.

[0624] Input: Parsed emotion data, user question text

[0625] Specific behavior: If the user is analyzed as feeling "anger," the chatbot will provide a calm and gentle response.

[0626] Output: Chatbot response depending on the emotion

[0627] Step 13:

[0628] The emotion data is stored on the server and used for the next re-learning.

[0629] Input: Parsed emotion data

[0630] Specific operation: Emotion data is saved in the database and used the next time the model is retrained.

[0631] Output: Saved emotion data

[0632] (Application example 2)

[0633] 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."

[0634] Current customer support systems in brick-and-mortar stores face the problem of being unable to analyze customers' emotions, such as facial expressions and tone of voice, in real time and provide appropriate support based on that analysis. Furthermore, the inability to respond appropriately to customer emotions makes it difficult to improve customer satisfaction. To solve this problem, a system is needed that analyzes customers' emotions and provides instructions on how to respond in real time based on the results.

[0635] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing user emotions using an emotion analysis engine, and means for reflecting the user's emotion data in the response content. This makes it possible to analyze customer emotions in real time and instruct store staff on appropriate responses based on the analysis results.

[0636] "Means for collecting training data" refers to the mechanism for obtaining the necessary text data from designated websites and databases.

[0637] "Initial filtering" refers to the process of removing unnecessary information from collected data and shaping the data.

[0638] "Artificial intelligence means" refers to a system that uses deep learning and natural language processing techniques to clean and analyze data.

[0639] The "means for presenting to the user and accepting confirmation and correction by the user" refers to an interface that shows the cleaned data to the user and accepts confirmation and correction.

[0640] The "means for retraining an artificial intelligence model" is a function for retraining a model based on data that has been checked and corrected by the user.

[0641] "Means of operating as a chatbot" refers to the process of installing and operating the retrained artificial intelligence model in an actual chatbot system.

[0642] "Means for analyzing user emotions using an emotion analysis engine" refers to a process that uses an engine to analyze an emotional state from text or voice data entered by a user.

[0643] The "means for reflecting the user's emotional data in the response content" is a function for adjusting the response content to the user based on the analyzed emotional data.

[0644] To implement this invention, store staff wear smart glasses and use the glasses' camera and microphone to capture customer facial expressions and voices. The data is then sent to a server in real time, where an emotion analysis engine is used to analyze the customer's emotional state. Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[0645] This system uses the following hardware and software:

[0646] Hardware

[0647] Smart glasses (with built-in camera and microphone)

[0648] Server (high-performance cloud server, e.g., Amazon EC2, Microsoft Azure)

[0649] software

[0650] Sentiment analysis engine (e.g., Microsoft Azure Cognitive Services Emotion API)

[0651] Data processing and communication (Python, TensorFlow, Flask)

[0652] Overview of data processing and calculation

[0653] 1. Capture customer facial and voice data

[0654] The smart glasses' camera and microphone capture the customer's facial and voice data in real time.

[0655] 2. Sending data to the server

[0656] The captured data is sent from the smart glasses to a server.

[0657] 3. Pretreatment

[0658] The received data is preprocessed on the server side (noise removal, data formatting, etc.) and then passed to the sentiment analysis engine.

[0659] 4. Emotion analysis

[0660] The sentiment analysis engine analyzes the customer's emotional state (e.g., "anger," "sadness," "joy," etc.).

[0661] 5. Feedback of response instructions

[0662] Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[0663] Specific examples

[0664] For example, if a customer looks dissatisfied while asking about a product at a counter, their facial expression and voice data will be captured, and the server will perform emotion analysis and determine that they are dissatisfied. As a result, the smart glasses will display a message saying, "The customer seems dissatisfied. Please be courteous and considerate."

[0665] Prompt Sentence Examples

[0666] "Please analyze the customer's emotions from the following facial and voice data:

[0667] Facial expression data: {image_data}

[0668] Audio data: {audio_data}"

[0669] In this way, it becomes possible to provide more effective customer service in physical stores, which contributes to improving customer satisfaction.

[0670] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0671] Step 1:

[0672] The smart glasses capture the customer's facial expression and voice data. They use the customer's real-time video and audio as input, which are captured by the smart glasses' built-in camera and microphone. The captured data is output in byte stream format.

[0673] Step 2:

[0674] The smart glasses send the captured data to the server. The captured byte stream data is the input and is sent to the server through the communication module. The output is the byte stream data received on the server side.

[0675] Step 3:

[0676] The server preprocesses the received data. The input data is the byte stream sent from the smart glasses, and it undergoes noise removal and data shaping (e.g., frame extraction and audio cleaning). The preprocessed shaped data is output as image data and audio data.

[0677] Step 4:

[0678] The server passes the preprocessed data to the sentiment analysis engine. The input data is the preprocessed image data and audio data, which are then input to the sentiment analysis engine. The sentiment analysis engine uses a deep learning model to identify the emotional state (e.g., "anger," "sadness," "joy," etc.). The output is text data that indicates the emotional state.

[0679] Step 5:

[0680] The server generates response instructions based on the analysis results. The output text data from the sentiment analysis engine is used as input, and response instructions for staff are generated based on that. For example, a message such as "The customer seems dissatisfied. Please try to respond politely and kindly" is generated. The output is text data showing response instructions.

[0681] Step 6:

[0682] The server sends the generated response instruction to the smart glasses. The input is text data indicating the response instruction, which is sent to the smart glasses through the communication module. The output is a text message displayed on the smart glasses.

[0683] Step 7:

[0684] The smart glasses display instructions to the staff. Text data indicating instructions sent from the server is used as input and displayed on the smart glasses' display. The staff responds to the customer based on this displayed information. The output is an instruction message that is displayed to the staff.

[0685] The above is a flow of specific processing steps for carrying out the invention, which is expected to improve the quality of customer service in brick-and-mortar stores and increase customer satisfaction.

[0686] 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.

[0687] 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.

[0688] 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.

[0689] [Third embodiment]

[0690] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0691] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0692] 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).

[0693] 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.

[0694] 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.

[0695] 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).

[0696] 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.

[0697] 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.

[0698] 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.

[0699] 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.

[0700] 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.

[0701] 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."

[0702] The purpose of this invention is to automate the development and operation process of an AI chatbot system. The specific operation of the system's program is explained below.

[0703] First, the device collects learning data. The user provides the device with information about the websites and databases to be collected. The device uses the specified information to send a data collection command to the server. The server then accesses the specified websites and databases and retrieves the necessary text data. For example, the data needed for learning can be collected from FAQ pages or user comments.

[0704] Next, the server performs initial filtering on the collected data to standardize the data format and remove unnecessary information, such as removing HTML tags and formatting line breaks and whitespace.

[0705] Next, the cleaning AI cleans the filtered data. It uses natural language processing technology to correct typos and inconsistencies in the data. It also automatically recognizes and masks personal and confidential information. For example, it appropriately conceals personal information such as email addresses and credit card information.

[0706] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This checking and correction is an important step in maintaining the accuracy of the data.

[0707] With the corrected data, the server passes it to the retraining AI, which uses the new data to retrain the chatbot's model. This process ensures that the AI ​​model is updated with the latest information.

[0708] The new model generated through retraining is sent from the server to the device. The device then installs this model into the chatbot and begins operation. Users ask questions using the chatbot, and the chatbot responds based on the data it has learned.

[0709] While the chatbot is running, the server monitors the quality of responses, collects user feedback, and uses it for the next retraining phase, allowing the chatbot to continually improve its accuracy.

[0710] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. A new model is generated and installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[0711] In this way, the present invention effectively automates the collection, cleaning, relearning, and operation of learning data, enabling the construction and operation of a highly accurate chatbot system.

[0712] The processing flow will be explained below.

[0713] Step 1:

[0714] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[0715] Step 2:

[0716] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[0717] Step 3:

[0718] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[0719] Step 4:

[0720] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0721] Step 5:

[0722] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[0723] Step 6:

[0724] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[0725] Step 7:

[0726] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[0727] Step 8:

[0728] A user asks a question using the chatbot, the chatbot generates an appropriate response based on the new model, and the server monitors and quality-controls this response.

[0729] Step 9:

[0730] The server collects user feedback and evaluates the quality of responses. This feedback is used during the next retraining phase. This continuous feedback loop helps improve the chatbot's performance.

[0731] Example 1

[0732] 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."

[0733] Conventional AI chatbot systems required a great deal of time and effort, from collecting and filtering training data, cleaning the data, to retraining the model, and then operating it. Furthermore, manual verification work was required to protect personal information and maintain data accuracy, and automation was insufficient, resulting in low efficiency. Furthermore, there was an insufficient mechanism for continuously improving the quality of chatbot responses during operation, making it difficult to appropriately incorporate customer feedback.

[0734] 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.

[0735] In this invention, the server includes: means for a user to provide information to be collected and a terminal to send a data collection instruction to the server based on that information; means for the server to access specified websites and databases to collect training data; means for the server to initially filter the training data and remove unnecessary information; means for a cleaning AI to correct input errors and inconsistencies in the filtered data using natural language processing technology and mask personal information; means for presenting the cleaned data to the user and accepting confirmation and correction by the user; means for a re-training AI to re-train an AI model based on the data confirmed and corrected by the user; means for sending the AI ​​model generated by the re-training to the terminal and operating it as a chatbot; and means for monitoring the quality of the chatbot's responses, collecting user feedback, and using it for the next re-training. This automates the process from collecting training data to re-training the model, operation, and continuous improvement of response quality, enabling the construction and operation of an efficient and highly accurate AI chatbot system.

[0736] "User" means a person or organization that uses the system to specify the targets for collecting learning data and to review and correct the data.

[0737] A "terminal" is a hardware or software component that allows a user to input information, send data collection instructions to a server, install a retrained model, and operate it as a chatbot.

[0738] The "server" is the core technology of the system that receives instructions from users and devices, collects training data, performs initial filtering and processing using cleaning AI, and then supplies the data to the re-training AI to re-train the model.

[0739] A "data collection instruction" is an instruction to command the server to start collecting learning data based on information input by the user into the terminal.

[0740] "Training data" refers to text data and information collected to train an AI chatbot model.

[0741] "Initial filtering" is a process to remove HTML tags and unnecessary line breaks and spaces from the data collected by the server and unify the format.

[0742] "Cleaning AI" is an artificial intelligence that uses natural language processing technology to correct input errors and inconsistencies in initially filtered data and mask personal information.

[0743] "Cleaning" is the process by which cleaning AI corrects errors in data and masks personal information.

[0744] "Retraining AI" is artificial intelligence that uses data reviewed and corrected by users to retrain new models to improve chatbot performance.

[0745] "Retraining" is the process of updating an AI model based on corrected data and repeating the training process.

[0746] An "artificial intelligence model" is a mathematical model generated by a retraining AI based on collected training data and used by the chatbot to respond to questions.

[0747] A "chatbot" is a computer program that implements a retrained artificial intelligence model to automatically interact with users.

[0748] "Monitoring" is the process of monitoring the quality of responses of a chatbot during operation, collecting feedback from users, and reflecting it in the next re-training.

[0749] "Feedback" refers to the user's evaluation and opinion of the chatbot's responses, and is information that can be used to further improve the chatbot's performance when it is retrained.

[0750] The present invention aims to automate the development and operation process of an AI chatbot system. A specific method for implementing the present invention will be described in detail below.

[0751] First, we will explain the process of collecting training data. The user provides the information to be collected to the device. For example, the user might enter the URL of a company's FAQ page or product review into the device. Based on this information, the device sends a data collection command to the server. The server accesses the specified website or database and obtains the required text data by scraping it. This data collection can be performed using, for example, the Python libraries "BeautifulSoup" or "Scrapy."

[0752] Next, the server performs initial filtering of the collected data. The server removes HTML tags, unnecessary line breaks, and spaces from the collected data to unify the format. Specifically, this filtering process is performed using regular expressions (Regex) and the Python "BeautifulSoup" library.

[0753] Next, the data is cleaned. The cleaning AI uses natural language processing technology to correct typos and inconsistencies in the filtered data. For example, it uses natural language processing libraries such as Python's "spaCy" and "NLTK." The cleaning AI also automatically recognizes and masks personal and confidential information. This process uses custom rules and machine learning models for personal information detection.

[0754] The server then presents the cleaned data to the user, who can then review the data and make any necessary corrections. This review and correction can be performed via a client-side web interface. For example, a user-friendly interface can be built using a front-end framework such as React or Angular.

[0755] The data that the user has confirmed and corrected is then sent back to the server, which then provides this data to the retraining AI and retrains the model. Deep learning frameworks such as TensorFlow and PyTorch are suitable for retraining.

[0756] The new retrained model is sent from the server to the device. The device then installs the new model in the chatbot and begins operation. Chatbots can be built using, for example, Dialogflow or the Microsoft Bot Framework.

[0757] During operation, the server monitors the quality of the chatbot's responses and collects user feedback, which is used during the next retraining phase to continuously improve the chatbot's accuracy.

[0758] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. The generated new model is installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[0759] Examples of prompts include:

[0760] "Gather training data from a company's FAQ page and past support emails, remove unnecessary information, mask personal information, and generate a new AI model."

[0761] The above is a specific form for implementing the present invention, which enables the efficient construction and operation of a highly accurate AI chatbot system.

[0762] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0763] Step 1: Instructions for collecting training data

[0764] The user provides the information to be collected to the device. Specifically, they input the URL of a company's website's FAQ page or a product review into the device. Based on this input data, the device sends a data collection instruction to the server. The input data may be a URL or a database path, and the device organizes it and sends it to the server. The specific operation of the device is to collect the information entered by the user and generate instructions for the server in an appropriate format.

[0765] Input: User-specified URL or database path

[0766] Output: Organized data collection instructions

[0767] Step 2: Collect training data

[0768] The server accesses the specified website or database and scrapes the required text data. The server connects to the specified URL, analyzes the HTML content of the page, and extracts the text data. For example, scraping libraries such as "BeautifulSoup" and "Scrapy" are used to collect the data.

[0769] Input: Organized data collection instructions

[0770] Output: Collected text data

[0771] Step 3: Initial filtering

[0772] The server performs initial filtering of the collected data, specifically removing HTML tags and formatting line breaks and whitespace. It uses regular expressions (Regex) and libraries such as the BeautifulSoup library to remove unnecessary parts of the data and standardize the format.

[0773] Input: Collected text data

[0774] Output: filtered text data

[0775] Step 4: Cleaning the data

[0776] The cleaning AI cleans the filtered data. Specifically, it uses natural language processing technology to correct typos and inconsistencies and mask personal information. It uses Python libraries such as "spaCy" and "NLTK." This improves data accuracy and protects personal information.

[0777] Input: filtered text data

[0778] Output: Cleaned text data

[0779] Step 5: Check and correct data

[0780] The server presents the cleaned data to the user, who then checks it and makes corrections as necessary. The server receives the corrected data and records it in a log. Since the user can check and correct the data through a web interface, it is effective to use a front-end framework such as React or Angular.

[0781] Input: Cleaned text data

[0782] Output: Data reviewed and corrected by the user

[0783] Step 6: Retrain

[0784] The server provides the verified and corrected data to the retraining AI, retraining the model, and builds a new generative AI model using a deep learning framework such as TensorFlow or PyTorch.

[0785] Input: Data confirmed and corrected by the user

[0786] Output: Retrained generative AI model

[0787] Step 7: Update and install the model

[0788] The server sends the generated new model to the device, which then installs it in the chatbot. The device then updates the chatbot with the new model and starts using it. For example, using Dialogflow or the Microsoft Bot Framework.

[0789] Input: Retrained generative AI model

[0790] Output: The new model installed in the chatbot

[0791] Step 8: Operate and monitor your chatbot

[0792] Users use the chatbot equipped with the new model to answer their questions. The server monitors the quality of the chatbot's responses and collects feedback. This feedback is used for the next retraining. The collected feedback is stored in a database and used for continuous improvement.

[0793] Input: User questions and feedback

[0794] Output: Monitoring results and collected feedback

[0795] Through these steps, this system will enable the development and operation of highly accurate and efficient AI chatbot systems.

[0796] (Application example 1)

[0797] 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."

[0798] In today's brick-and-mortar stores, the knowledge and skills of store clerks are important for smoothly responding to a wide range of customer questions. However, because not every store clerk has complete knowledge of all products and services, customer service is often inconsistent. Furthermore, there are limited means to search and provide vast amounts of information in real time to accurately answer customer questions. In such situations, there is a risk of customer satisfaction decreasing, so a smart system to assist store clerks is needed.

[0799] 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.

[0800] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, and means for implementing the re-trained artificial intelligence model and operating it as a chatbot to assist store clerks in customer service in real time using smart glasses. This enables store clerks to receive questions from customers through the smart glasses and provide appropriate answers in real time based on a vast amount of information, thereby increasing customer satisfaction.

[0801] "Training data" is the data set used to train an AI model.

[0802] "Means of collection" refers to the method or device used to obtain the required information from the website or database specified by the user.

[0803] "Initial filtering" is a process of removing unnecessary information from collected data and standardizing the data format.

[0804] "Cleaning AI" is an AI technology that cleans collected data to eliminate errors and inconsistencies, and properly processes personal and confidential information.

[0805] The "means for accepting user confirmation and correction" is an interface that allows the user to confirm the cleaned data and correct it if necessary.

[0806] "Retraining" is a method of retraining an existing AI model with the latest information using data that has been confirmed and corrected by the user.

[0807] "Means of implementation and operation as a chatbot" refers to the method or device for incorporating the retrained AI model into a chatbot and using it in an actual operational environment.

[0808] "Smart glasses" are wearable devices equipped with displays and sensors that display information within the store clerk's field of vision.

[0809] The "means of assisting customers in real time" is a system that allows store clerks to instantly provide appropriate answers to customers' questions through smart glasses.

[0810] As a concrete example of implementing this invention, we will explain a store clerk support system for brick-and-mortar stores using smart glasses. This system collects learning data, performs initial filtering, cleaning, checking and correcting, re-learning, and operates as a chatbot. The details are described below.

[0811] First, the device collects training data from websites and databases specified by the user. At this stage, Scrapy and BeautifulSoup are used as information collection tools. These tools allow the device to efficiently extract the necessary text data.

[0812] Next, the server performs initial filtering on the collected data, which involves removing HTML tags, line breaks, and whitespace formatting, using the standard Python library and regular expressions.

[0813] The collected and initially filtered data is then further processed by the Cleaning AI, which uses natural language processing techniques to correct data typos and inconsistencies, and masks personal and confidential information. This process uses natural language processing libraries such as SpaCy and NLTK.

[0814] Once the data has been cleaned, it is presented to the user for review and correction. This review process is carried out using a web-based interface, allowing the user to correct the data in their browser. This stage ensures the accuracy and consistency of the data.

[0815] Based on the revised data, the server updates the AI ​​model for retraining. This retraining process uses deep learning frameworks such as TensorFlow or PyTorch. The retrained model improves based on the new information and can generate more accurate responses.

[0816] The retrained AI model will be implemented as a chatbot and made available to store associates in real time through smart glasses, a wearable device equipped with a camera and a display, allowing store associates to instantly view the information they need within their field of vision.

[0817] As a specific use case, when a customer asks, "What features does this product have?", the chatbot in the smart glasses will instantly respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[0818] An example prompt is:

[0819] Customer: What features does this product have?

[0820] AI: This product has the latest Bluetooth functionality, long battery life, and is waterproof.

[0821] This system will enable store staff to respond to customers quickly and accurately, which is expected to improve customer satisfaction.

[0822] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0823] Step 1:

[0824] The device collects training data from websites and databases specified by the user. The user provides the device with the URLs and database information to be collected. The device uses Scrapy or BeautifulSoup to extract the necessary text data from the websites and databases and send it to the server. The input for this step is the URL of the website or database, and the output is the collected text data.

[0825] Step 2:

[0826] The server performs initial filtering of the collected data. It uses the Python standard library and regular expressions to remove HTML tags and format lines and spaces. In this process, the data format is unified and unnecessary information is removed. The input of this step is the collected text data, and the output is the initially filtered text data.

[0827] Step 3:

[0828] The server cleans the initially filtered data. This process is performed by a cleaning AI, which uses natural language processing libraries such as SpaCy and NLTK to correct typos and inconsistencies in the data, as well as masking personal and confidential information. The input of this step is the initially filtered text data, and the output is the cleaned text data.

[0829] Step 4:

[0830] The server presents the cleaned data to the user for review and correction. The user reviews the cleaned data through a web-based interface and makes corrections as needed. A browser is used for this review and correction work. The input of the step is the cleaned text data, and the output is the text data reviewed and corrected by the user.

[0831] Step 5:

[0832] The server retrains the AI ​​model based on the confirmed and corrected data. It uses a deep learning framework such as TensorFlow or PyTorch to update the retrained AI model. The input for this step is the text data confirmed and corrected by the user, and the output is the retrained AI model.

[0833] Step 6:

[0834] The server uses the retrained AI model to operate as a chatbot. The smart glasses use this chatbot to assist store clerks in dealing with customers in real time. Through the display and camera installed in the smart glasses, store clerks can receive customer questions and instantly display appropriate answers. The input for this step is the retrained AI model, and the output is customer service information provided in real time.

[0835] For example, if a customer asks, "What features does this product have?" the smart glasses' display will respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[0836] 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.

[0837] The purpose of this invention is to provide responses that correspond to the user's emotions and improve their accuracy by combining an emotion engine that recognizes the user's emotions with an AI chatbot system. The specific operation of the system's program is explained below.

[0838] First, the user inputs the website or database information to be collected into the device. The device receives this information and sends a collection instruction to the server. The server then accesses the specified website or database and collects the necessary text data. For example, it can extract question and answer pairs from an FAQ page.

[0839] Next, the server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. Once the initial filtering is complete, the data moves on to the next step.

[0840] Cleaning AI then takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0841] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This process ensures the accuracy of the data.

[0842] The server then passes the revised data to a retraining AI, which uses the data to retrain the chatbot's model, generating a new model in the process.

[0843] The generated new model is sent from the server to the device. The device installs this model in the chatbot, and the chatbot begins operation. The user asks a question using the chatbot, and the chatbot generates an appropriate response based on the new model.

[0844] In the present invention, the system is further improved by adding an emotion engine. When a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. The analyzed emotion data is sent to the server and reflected in the response. For example, if the user is feeling angry or stressed, the chatbot will provide a calm and gentle response accordingly.

[0845] Furthermore, emotional data is collected and used for the next retraining. This allows the chatbot to continually improve and respond better to user emotions. For example, in a customer support chatbot, when a user expresses dissatisfaction, the dissatisfaction can be handled quickly and appropriately, improving customer satisfaction.

[0846] In this way, by combining the collection, cleaning, and relearning of learning data with an emotion engine, the present invention enables the construction and operation of a highly accurate chatbot system that takes emotions into consideration.

[0847] The processing flow will be explained below.

[0848] Step 1:

[0849] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[0850] Step 2:

[0851] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[0852] Step 3:

[0853] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[0854] Step 4:

[0855] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[0856] Step 5:

[0857] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[0858] Step 6:

[0859] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[0860] Step 7:

[0861] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[0862] Step 8:

[0863] The user uses the chatbot to ask a question, and the chatbot generates an appropriate response based on the new model. During this process, the emotion engine recognizes the user's emotions and analyzes the user's emotional state from the text. For example, if the user is angry, the chatbot will provide a calm and gentle response accordingly.

[0864] Step 9:

[0865] The server monitors the chatbot's responses and records the emotional data collected by the emotion engine. The server collects this data as feedback and uses it for the next re-training.

[0866] Step 10:

[0867] During retraining, the server passes feedback, including emotional data, to the retraining AI to train it to improve its emotional responses, allowing the chatbot to better respond to the user's emotions.

[0868] Example 2

[0869] 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."

[0870] Existing AI chatbot systems often lack the ability to recognize user emotions and respond inappropriately. Furthermore, there is a need for efficient and accurate methods for the entire process, from data collection to model retraining and data cleansing. In particular, challenges remain in automatically masking personal and confidential information and removing duplicate data. Furthermore, the accuracy of chatbot responses has not been sufficiently improved due to a lack of continuous training using user feedback and sentiment analysis data.

[0871] 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.

[0872] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, and artificial intelligence means for cleaning the initially filtered data. It also includes means for presenting the cleaned data to a user and accepting user confirmation and correction, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing emotions from text input by the user, means for generating chatbot responses based on this emotional data, and means for collecting emotional data and using it for the next relearning. This enables the generation of appropriate responses based on the user's emotions, thereby improving the accuracy of chatbot responses while protecting personal and confidential information.

[0873] "Training data" refers to information or data sets used to train an artificial intelligence model.

[0874] "Collection means" refers to the method or device used to obtain the necessary information from websites, databases, etc.

[0875] "Initial filtering" is a process for removing unnecessary information from collected data and standardizing the data format.

[0876] "Artificial intelligence means" refers to techniques that use machine learning algorithms and deep learning models to analyze and process data.

[0877] "Cleaning" is an operation to remove noise and errors from collected and filtered data and to arrange the data.

[0878] A "presentation means" is an interface that displays the cleaned data to the user and allows for review and modification.

[0879] "Retraining" is the process of retraining an artificial intelligence model using updated data.

[0880] A "model" refers to an artificial intelligence structure or algorithm that has been trained for a specific purpose.

[0881] A "chatbot" is a computer program that automatically handles interactions with users.

[0882] "Sentiment analysis" is a technique for detecting and classifying user emotions from text data.

[0883] "Response generation" is the process by which a chatbot creates a response to a user's input.

[0884] "Feedback" refers to usage evaluations and opinions provided by users, and is information that is used to improve the system.

[0885] "Auto-masking" is the process of automatically identifying personal or sensitive information from collected data and masking it to protect it.

[0886] This invention is an AI chatbot system that recognizes a user's emotions and provides responses according to those emotions. The purpose is to improve the accuracy of the data collected from the user while also increasing the accuracy of the chatbot. The implementation method of this system is specifically described below, specifying the hardware and software used.

[0887] First, the user inputs the website or database information to be collected into the device. For example, the user can input the URL of a customer support FAQ page. The device receives this input information and sends a collection instruction to the server using an HTTP request.

[0888] The server then accesses the specified websites and databases to collect the required text data, using a scraping tool to extract the text from FAQ pages and store it in a temporary database.

[0889] The collected data is initially filtered by the server, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. For example, the data is cleaned by using regular expressions to remove unnecessary HTML tags.

[0890] The server then uses cleaning AI to further clean the initially filtered data. Cleaning includes removing duplicate data, correcting typos, and automatically detecting and masking personal and confidential information. For example, if the same question appears multiple times, the cleaning AI will remove the duplicates.

[0891] The cleaned data is presented to the user by the server. The cleaned FAQ data is displayed through a user interface, and the user checks it. For example, the user checks whether the correct answer matches the question correctly. After the user makes the necessary corrections, the results are sent to the server.

[0892] The server then passes the corrected data to a retraining AI to retrain the chatbot's model. This process uses frameworks such as TensorFlow or PyTorch. Once the new model is generated, the model data is sent to the device using a secure protocol (e.g., HTTPS).

[0893] The terminal installs the new model into the chatbot. The chatbot begins operation by replacing the existing model with the new one. The user uses the chatbot based on the new model and asks questions. For example, if the user asks, "How do I return a product?", the chatbot responds, "You can return a product from your My Page on the website."

[0894] Furthermore, when a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. For example, natural language processing technology is used to classify emotions such as "anger," "joy," and "sadness." The server reflects the analyzed emotional data in the response; for example, if the server determines that the user is feeling "anger," it will provide a calm and gentle response accordingly.

[0895] The emotional data is stored on the server and used for the next re-learning, allowing the chatbot to continue generating appropriate responses according to the user's emotions.

[0896] As a concrete example, the following prompt sentence can be used:

[0897] "How do I return a product?"

[0898] In this way, the system of the present invention can provide responses that correspond to the user's emotions, making it possible to realize a highly accurate chatbot system.

[0899] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0900] Step 1:

[0901] The user inputs the information from the website or database to be collected into the terminal.

[0902] Input: URLs and queries for websites and databases to be collected

[0903] Specific operation: The user enters information into an input field in a browser or application, and a collection instruction is sent to the terminal.

[0904] Output: The terminal generates data along with collection instructions to convert the input information into a format to be sent to the server.

[0905] Step 2:

[0906] The terminal sends a collection instruction to the server using an HTTP request.

[0907] Input: Collected instruction data from the user

[0908] Specific operation: The terminal packages the input information into an HTTP request and sends it to the server.

[0909] Output: The server receives the HTTP request and interprets the collection instructions data.

[0910] Step 3:

[0911] The server accesses the specified website or database and collects the required text data.

[0912] Input: Collection instruction data

[0913] What happens: The server uses a scraping tool to extract text data from the specified URL, for example, extracting question and answer pairs from an FAQ page.

[0914] Output: Save the extracted text data in a temporary database.

[0915] Step 4:

[0916] The server performs an initial filtering of the collected data.

[0917] Input: Collected text data

[0918] Specific operation: The server removes unnecessary HTML tags from the data, formats line breaks and whitespace, unifies date and time formats, etc. It uses regular expressions and text processing libraries.

[0919] Output: Generates the initial filtered data.

[0920] Step 5:

[0921] The server further cleans the initially filtered data using cleaning AI.

[0922] Input: Initial filtered data

[0923] What it does: Removes duplicate data, corrects typos, automatically detects and masks personal and confidential information, and uses natural language processing techniques to identify specific patterns and data formats.

[0924] Output: Produces the cleaned data.

[0925] Step 6:

[0926] The server presents the cleaned data to the user and accepts confirmation and correction by the user.

[0927] Input: Cleaned data

[0928] Specific operation: The server displays the data through a user interface, allowing the user to review and modify it. It also provides an interface for receiving modified data.

[0929] Output: Data reviewed and corrected by the user

[0930] Step 7:

[0931] The server retrains the artificial intelligence model based on the data confirmed and corrected by the user.

[0932] Input: Verified and corrected data

[0933] What it does: Retrain the model using a framework such as TensorFlow or PyTorch, using the corrected data as training data to update the neural network weights.

[0934] Output: A new retrained AI model

[0935] Step 8:

[0936] The server transmits the retrained artificial intelligence model to the terminal.

[0937] Input: Retrained AI model

[0938] Specific operation: The server sends the new model to the device using a secure protocol (e.g., HTTPS).

[0939] Output: The new model data sent to the terminal.

[0940] Step 9:

[0941] The device installs the retrained artificial intelligence model into the chatbot and begins operating the chatbot.

[0942] Input: Retrained AI model

[0943] Specific behavior: Replace the existing model with the new one. Restart the chatbot software and load the new model.

[0944] Output: A running chatbot with the new model installed

[0945] Step 10:

[0946] Users use chatbots based on the new model to ask questions.

[0947] Input: User question text

[0948] Specific behavior: For example, ask, "How do I return an item?" The chatbot uses the new model to generate an appropriate response.

[0949] Output: Chatbot response text

[0950] Step 11:

[0951] When a user types a question into the chatbot, the emotion engine analyzes the user's emotions.

[0952] Input: User question text

[0953] Specific operation: Using natural language processing technology, emotions such as "anger," "joy," and "sadness" are classified.

[0954] Output: Parsed emotion data

[0955] Step 12:

[0956] The server reflects the analyzed emotional data in its response and takes appropriate action.

[0957] Input: Parsed emotion data, user question text

[0958] Specific behavior: If the user is analyzed as feeling "anger," the chatbot will provide a calm and gentle response.

[0959] Output: Chatbot response depending on the emotion

[0960] Step 13:

[0961] The emotion data is stored on the server and used for the next re-learning.

[0962] Input: Parsed emotion data

[0963] Specific operation: Emotion data is saved in the database and used the next time the model is retrained.

[0964] Output: Saved emotion data

[0965] (Application example 2)

[0966] 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."

[0967] Current customer support systems in brick-and-mortar stores face the problem of being unable to analyze customers' emotions, such as facial expressions and tone of voice, in real time and provide appropriate support based on that analysis. Furthermore, the inability to respond appropriately to customer emotions makes it difficult to improve customer satisfaction. To solve this problem, a system is needed that analyzes customers' emotions and provides instructions on how to respond in real time based on the results.

[0968] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing user emotions using an emotion analysis engine, and means for reflecting the user's emotion data in the response content. This makes it possible to analyze customer emotions in real time and instruct store staff on appropriate responses based on the analysis results.

[0969] "Means for collecting training data" refers to the mechanism for obtaining the necessary text data from designated websites and databases.

[0970] "Initial filtering" refers to the process of removing unnecessary information from collected data and shaping the data.

[0971] "Artificial intelligence means" refers to a system that uses deep learning and natural language processing techniques to clean and analyze data.

[0972] The "means for presenting to the user and accepting confirmation and correction by the user" refers to an interface that shows the cleaned data to the user and accepts confirmation and correction.

[0973] The "means for retraining an artificial intelligence model" is a function for retraining a model based on data that has been checked and corrected by the user.

[0974] "Means of operating as a chatbot" refers to the process of installing and operating the retrained artificial intelligence model in an actual chatbot system.

[0975] "Means for analyzing user emotions using an emotion analysis engine" refers to a process that uses an engine to analyze an emotional state from text or voice data entered by a user.

[0976] The "means for reflecting the user's emotional data in the response content" is a function for adjusting the response content to the user based on the analyzed emotional data.

[0977] To implement this invention, store staff wear smart glasses and use the glasses' camera and microphone to capture customer facial expressions and voices. The data is then sent to a server in real time, where an emotion analysis engine is used to analyze the customer's emotional state. Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[0978] This system uses the following hardware and software:

[0979] Hardware

[0980] Smart glasses (with built-in camera and microphone)

[0981] Server (high-performance cloud server, e.g., Amazon EC2, Microsoft Azure)

[0982] software

[0983] Sentiment analysis engine (e.g., Microsoft Azure Cognitive Services Emotion API)

[0984] Data processing and communication (Python, TensorFlow, Flask)

[0985] Overview of data processing and calculation

[0986] 1. Capture customer facial and voice data

[0987] The smart glasses' camera and microphone capture the customer's facial and voice data in real time.

[0988] 2. Sending data to the server

[0989] The captured data is sent from the smart glasses to a server.

[0990] 3. Pretreatment

[0991] The received data is preprocessed on the server side (noise removal, data formatting, etc.) and then passed to the sentiment analysis engine.

[0992] 4. Emotion analysis

[0993] The sentiment analysis engine analyzes the customer's emotional state (e.g., "anger," "sadness," "joy," etc.).

[0994] 5. Feedback of response instructions

[0995] Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[0996] Specific examples

[0997] For example, if a customer looks dissatisfied while asking about a product at a counter, their facial expression and voice data will be captured, and the server will perform emotion analysis and determine that they are dissatisfied. As a result, the smart glasses will display a message saying, "The customer seems dissatisfied. Please be courteous and considerate."

[0998] Prompt Sentence Examples

[0999] "Please analyze the customer's emotions from the following facial and voice data:

[1000] Facial expression data: {image_data}

[1001] Audio data: {audio_data}"

[1002] In this way, it becomes possible to provide more effective customer service in physical stores, which contributes to improving customer satisfaction.

[1003] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1004] Step 1:

[1005] The smart glasses capture the customer's facial expression and voice data. They use the customer's real-time video and audio as input, which are captured by the smart glasses' built-in camera and microphone. The captured data is output in byte stream format.

[1006] Step 2:

[1007] The smart glasses send the captured data to the server. The captured byte stream data is the input and is sent to the server through the communication module. The output is the byte stream data received on the server side.

[1008] Step 3:

[1009] The server preprocesses the received data. The input data is the byte stream sent from the smart glasses, and it undergoes noise removal and data shaping (e.g., frame extraction and audio cleaning). The preprocessed shaped data is output as image data and audio data.

[1010] Step 4:

[1011] The server passes the preprocessed data to the sentiment analysis engine. The input data is the preprocessed image data and audio data, which are then input to the sentiment analysis engine. The sentiment analysis engine uses a deep learning model to identify the emotional state (e.g., "anger," "sadness," "joy," etc.). The output is text data that indicates the emotional state.

[1012] Step 5:

[1013] The server generates response instructions based on the analysis results. The output text data from the sentiment analysis engine is used as input, and response instructions for staff are generated based on that. For example, a message such as "The customer seems dissatisfied. Please try to respond politely and kindly" is generated. The output is text data showing response instructions.

[1014] Step 6:

[1015] The server sends the generated response instruction to the smart glasses. The input is text data indicating the response instruction, which is sent to the smart glasses through the communication module. The output is a text message displayed on the smart glasses.

[1016] Step 7:

[1017] The smart glasses display instructions to the staff. Text data indicating instructions sent from the server is used as input and displayed on the smart glasses' display. The staff responds to the customer based on this displayed information. The output is an instruction message that is displayed to the staff.

[1018] The above is a flow of specific processing steps for carrying out the invention, which is expected to improve the quality of customer service in brick-and-mortar stores and increase customer satisfaction.

[1019] 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.

[1020] 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.

[1021] 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.

[1022] [Fourth embodiment]

[1023] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1024] 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.

[1025] 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).

[1026] 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.

[1027] 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.

[1028] 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).

[1029] 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.

[1030] 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.

[1031] 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.

[1032] 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.

[1033] 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.

[1034] 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.

[1035] 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."

[1036] The purpose of this invention is to automate the development and operation process of an AI chatbot system. The specific operation of the system's program is explained below.

[1037] First, the device collects learning data. The user provides the device with information about the websites and databases to be collected. The device uses the specified information to send a data collection command to the server. The server then accesses the specified websites and databases and retrieves the necessary text data. For example, the data needed for learning can be collected from FAQ pages or user comments.

[1038] Next, the server performs initial filtering on the collected data to standardize the data format and remove unnecessary information, such as removing HTML tags and formatting line breaks and whitespace.

[1039] Next, the cleaning AI cleans the filtered data. It uses natural language processing technology to correct typos and inconsistencies in the data. It also automatically recognizes and masks personal and confidential information. For example, it appropriately conceals personal information such as email addresses and credit card information.

[1040] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This checking and correction is an important step in maintaining the accuracy of the data.

[1041] With the corrected data, the server passes it to the retraining AI, which uses the new data to retrain the chatbot's model. This process ensures that the AI ​​model is updated with the latest information.

[1042] The new model generated through retraining is sent from the server to the device. The device then installs this model into the chatbot and begins operation. Users ask questions using the chatbot, and the chatbot responds based on the data it has learned.

[1043] While the chatbot is running, the server monitors the quality of responses, collects user feedback, and uses it for the next retraining phase, allowing the chatbot to continually improve its accuracy.

[1044] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. A new model is generated and installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[1045] In this way, the present invention effectively automates the collection, cleaning, relearning, and operation of learning data, enabling the construction and operation of a highly accurate chatbot system.

[1046] The processing flow will be explained below.

[1047] Step 1:

[1048] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[1049] Step 2:

[1050] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[1051] Step 3:

[1052] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[1053] Step 4:

[1054] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[1055] Step 5:

[1056] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[1057] Step 6:

[1058] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[1059] Step 7:

[1060] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[1061] Step 8:

[1062] A user asks a question using the chatbot, the chatbot generates an appropriate response based on the new model, and the server monitors and quality-controls this response.

[1063] Step 9:

[1064] The server collects user feedback and evaluates the quality of responses. This feedback is used during the next retraining phase. This continuous feedback loop helps improve the chatbot's performance.

[1065] Example 1

[1066] 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."

[1067] Conventional AI chatbot systems required a great deal of time and effort, from collecting and filtering training data, cleaning the data, to retraining the model, and then operating it. Furthermore, manual verification work was required to protect personal information and maintain data accuracy, and automation was insufficient, resulting in low efficiency. Furthermore, there was an insufficient mechanism for continuously improving the quality of chatbot responses during operation, making it difficult to appropriately incorporate customer feedback.

[1068] 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.

[1069] In this invention, the server includes: means for a user to provide information to be collected and a terminal to send a data collection instruction to the server based on that information; means for the server to access specified websites and databases to collect training data; means for the server to initially filter the training data and remove unnecessary information; means for a cleaning AI to correct input errors and inconsistencies in the filtered data using natural language processing technology and mask personal information; means for presenting the cleaned data to the user and accepting confirmation and correction by the user; means for a re-training AI to re-train an AI model based on the data confirmed and corrected by the user; means for sending the AI ​​model generated by the re-training to the terminal and operating it as a chatbot; and means for monitoring the quality of the chatbot's responses, collecting user feedback, and using it for the next re-training. This automates the process from collecting training data to re-training the model, operation, and continuous improvement of response quality, enabling the construction and operation of an efficient and highly accurate AI chatbot system.

[1070] "User" means a person or organization that uses the system to specify the targets for collecting learning data and to review and correct the data.

[1071] A "terminal" is a hardware or software component that allows a user to input information, send data collection instructions to a server, install a retrained model, and operate it as a chatbot.

[1072] The "server" is the core technology of the system that receives instructions from users and devices, collects training data, performs initial filtering and processing using cleaning AI, and then supplies the data to the re-training AI to re-train the model.

[1073] A "data collection instruction" is an instruction to command the server to start collecting learning data based on information input by the user into the terminal.

[1074] "Training data" refers to text data and information collected to train an AI chatbot model.

[1075] "Initial filtering" is a process to remove HTML tags and unnecessary line breaks and spaces from the data collected by the server and unify the format.

[1076] "Cleaning AI" is an artificial intelligence that uses natural language processing technology to correct input errors and inconsistencies in initially filtered data and mask personal information.

[1077] "Cleaning" is the process by which cleaning AI corrects errors in data and masks personal information.

[1078] "Retraining AI" is artificial intelligence that uses data reviewed and corrected by users to retrain new models to improve chatbot performance.

[1079] "Retraining" is the process of updating an AI model based on corrected data and repeating the training process.

[1080] An "artificial intelligence model" is a mathematical model generated by a retraining AI based on collected training data and used by the chatbot to respond to questions.

[1081] A "chatbot" is a computer program that implements a retrained artificial intelligence model to automatically interact with users.

[1082] "Monitoring" is the process of monitoring the quality of responses of a chatbot during operation, collecting feedback from users, and reflecting it in the next re-training.

[1083] "Feedback" refers to the user's evaluation and opinion of the chatbot's responses, and is information that can be used to further improve the chatbot's performance when it is retrained.

[1084] The present invention aims to automate the development and operation process of an AI chatbot system. A specific method for implementing the present invention will be described in detail below.

[1085] First, we will explain the process of collecting training data. The user provides the information to be collected to the device. For example, the user might enter the URL of a company's FAQ page or product review into the device. Based on this information, the device sends a data collection command to the server. The server accesses the specified website or database and obtains the required text data by scraping it. This data collection can be performed using, for example, the Python libraries "BeautifulSoup" or "Scrapy."

[1086] Next, the server performs initial filtering of the collected data. The server removes HTML tags, unnecessary line breaks, and spaces from the collected data to unify the format. Specifically, this filtering process is performed using regular expressions (Regex) and the Python "BeautifulSoup" library.

[1087] Next, the data is cleaned. The cleaning AI uses natural language processing technology to correct typos and inconsistencies in the filtered data. For example, it uses natural language processing libraries such as Python's "spaCy" and "NLTK." The cleaning AI also automatically recognizes and masks personal and confidential information. This process uses custom rules and machine learning models for personal information detection.

[1088] The server then presents the cleaned data to the user, who can then review the data and make any necessary corrections. This review and correction can be performed via a client-side web interface. For example, a user-friendly interface can be built using a front-end framework such as React or Angular.

[1089] The data that the user has confirmed and corrected is then sent back to the server, which then provides this data to the retraining AI and retrains the model. Deep learning frameworks such as TensorFlow and PyTorch are suitable for retraining.

[1090] The new retrained model is sent from the server to the device. The device then installs the new model in the chatbot and begins operation. Chatbots can be built using, for example, Dialogflow or the Microsoft Bot Framework.

[1091] During operation, the server monitors the quality of the chatbot's responses and collects user feedback, which is used during the next retraining phase to continuously improve the chatbot's accuracy.

[1092] As a concrete example, consider the case where a company introduces a chatbot for customer support. Learning data is collected from the company's FAQ page and past support emails, cleaned, and then provided to a retraining AI. The generated new model is installed in the chatbot, allowing customers to receive more accurate and useful answers. Automating this process significantly reduces the amount of human effort required and makes it possible to maintain the quality of the chatbot.

[1093] Examples of prompts include:

[1094] "Gather training data from a company's FAQ page and past support emails, remove unnecessary information, mask personal information, and generate a new AI model."

[1095] The above is a specific form for implementing the present invention, which enables the efficient construction and operation of a highly accurate AI chatbot system.

[1096] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1097] Step 1: Instructions for collecting training data

[1098] The user provides the information to be collected to the device. Specifically, they input the URL of a company's website's FAQ page or a product review into the device. Based on this input data, the device sends a data collection instruction to the server. The input data may be a URL or a database path, and the device organizes it and sends it to the server. The specific operation of the device is to collect the information entered by the user and generate instructions for the server in an appropriate format.

[1099] Input: User-specified URL or database path

[1100] Output: Organized data collection instructions

[1101] Step 2: Collect training data

[1102] The server accesses the specified website or database and scrapes the required text data. The server connects to the specified URL, analyzes the HTML content of the page, and extracts the text data. For example, scraping libraries such as "BeautifulSoup" and "Scrapy" are used to collect the data.

[1103] Input: Organized data collection instructions

[1104] Output: Collected text data

[1105] Step 3: Initial filtering

[1106] The server performs initial filtering of the collected data, specifically removing HTML tags and formatting line breaks and whitespace. It uses regular expressions (Regex) and libraries such as the BeautifulSoup library to remove unnecessary parts of the data and standardize the format.

[1107] Input: Collected text data

[1108] Output: filtered text data

[1109] Step 4: Cleaning the data

[1110] The cleaning AI cleans the filtered data. Specifically, it uses natural language processing technology to correct typos and inconsistencies and mask personal information. It uses Python libraries such as "spaCy" and "NLTK." This improves data accuracy and protects personal information.

[1111] Input: filtered text data

[1112] Output: Cleaned text data

[1113] Step 5: Check and correct data

[1114] The server presents the cleaned data to the user, who then checks it and makes corrections as necessary. The server receives the corrected data and records it in a log. Since the user can check and correct the data through a web interface, it is effective to use a front-end framework such as React or Angular.

[1115] Input: Cleaned text data

[1116] Output: Data reviewed and corrected by the user

[1117] Step 6: Retrain

[1118] The server provides the verified and corrected data to the retraining AI, retraining the model, and builds a new generative AI model using a deep learning framework such as TensorFlow or PyTorch.

[1119] Input: Data confirmed and corrected by the user

[1120] Output: Retrained generative AI model

[1121] Step 7: Update and install the model

[1122] The server sends the generated new model to the device, which then installs it in the chatbot. The device then updates the chatbot with the new model and starts using it. For example, using Dialogflow or the Microsoft Bot Framework.

[1123] Input: Retrained generative AI model

[1124] Output: The new model installed in the chatbot

[1125] Step 8: Operate and monitor your chatbot

[1126] Users use the chatbot equipped with the new model to answer their questions. The server monitors the quality of the chatbot's responses and collects feedback. This feedback is used for the next retraining. The collected feedback is stored in a database and used for continuous improvement.

[1127] Input: User questions and feedback

[1128] Output: Monitoring results and collected feedback

[1129] Through these steps, this system will enable the development and operation of highly accurate and efficient AI chatbot systems.

[1130] (Application example 1)

[1131] 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."

[1132] In today's brick-and-mortar stores, the knowledge and skills of store clerks are important for smoothly responding to a wide range of customer questions. However, because not every store clerk has complete knowledge of all products and services, customer service is often inconsistent. Furthermore, there are limited means to search and provide vast amounts of information in real time to accurately answer customer questions. In such situations, there is a risk of customer satisfaction decreasing, so a smart system to assist store clerks is needed.

[1133] 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.

[1134] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, and means for implementing the re-trained artificial intelligence model and operating it as a chatbot to assist store clerks in customer service in real time using smart glasses. This enables store clerks to receive questions from customers through the smart glasses and provide appropriate answers in real time based on a vast amount of information, thereby increasing customer satisfaction.

[1135] "Training data" is the data set used to train an AI model.

[1136] "Means of collection" refers to the method or device used to obtain the required information from the website or database specified by the user.

[1137] "Initial filtering" is a process of removing unnecessary information from collected data and standardizing the data format.

[1138] "Cleaning AI" is an AI technology that cleans collected data to eliminate errors and inconsistencies, and properly processes personal and confidential information.

[1139] The "means for accepting user confirmation and correction" is an interface that allows the user to confirm the cleaned data and correct it if necessary.

[1140] "Retraining" is a method of retraining an existing AI model with the latest information using data that has been confirmed and corrected by the user.

[1141] "Means of implementation and operation as a chatbot" refers to the method or device for incorporating the retrained AI model into a chatbot and using it in an actual operational environment.

[1142] "Smart glasses" are wearable devices equipped with displays and sensors that display information within the store clerk's field of vision.

[1143] The "means of assisting customers in real time" is a system that allows store clerks to instantly provide appropriate answers to customers' questions through smart glasses.

[1144] As a concrete example of implementing this invention, we will explain a store clerk support system for brick-and-mortar stores using smart glasses. This system collects learning data, performs initial filtering, cleaning, checking and correcting, re-learning, and operates as a chatbot. The details are described below.

[1145] First, the device collects training data from websites and databases specified by the user. At this stage, Scrapy and BeautifulSoup are used as information collection tools. These tools allow the device to efficiently extract the necessary text data.

[1146] Next, the server performs initial filtering on the collected data, which involves removing HTML tags, line breaks, and whitespace formatting, using the standard Python library and regular expressions.

[1147] The collected and initially filtered data is then further processed by the Cleaning AI, which uses natural language processing techniques to correct data typos and inconsistencies, and masks personal and confidential information. This process uses natural language processing libraries such as SpaCy and NLTK.

[1148] Once the data has been cleaned, it is presented to the user for review and correction. This review process is carried out using a web-based interface, allowing the user to correct the data in their browser. This stage ensures the accuracy and consistency of the data.

[1149] Based on the revised data, the server updates the AI ​​model for retraining. This retraining process uses deep learning frameworks such as TensorFlow or PyTorch. The retrained model improves based on the new information and can generate more accurate responses.

[1150] The retrained AI model will be implemented as a chatbot and made available to store associates in real time through smart glasses, a wearable device equipped with a camera and a display, allowing store associates to instantly view the information they need within their field of vision.

[1151] As a specific use case, when a customer asks, "What features does this product have?", the chatbot in the smart glasses will instantly respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[1152] An example prompt is:

[1153] Customer: What features does this product have?

[1154] AI: This product has the latest Bluetooth functionality, long battery life, and is waterproof.

[1155] This system will enable store staff to respond to customers quickly and accurately, which is expected to improve customer satisfaction.

[1156] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1157] Step 1:

[1158] The device collects training data from websites and databases specified by the user. The user provides the device with the URLs and database information to be collected. The device uses Scrapy or BeautifulSoup to extract the necessary text data from the websites and databases and send it to the server. The input for this step is the URL of the website or database, and the output is the collected text data.

[1159] Step 2:

[1160] The server performs initial filtering of the collected data. It uses the Python standard library and regular expressions to remove HTML tags and format lines and spaces. In this process, the data format is unified and unnecessary information is removed. The input of this step is the collected text data, and the output is the initially filtered text data.

[1161] Step 3:

[1162] The server cleans the initially filtered data. This process is performed by a cleaning AI, which uses natural language processing libraries such as SpaCy and NLTK to correct typos and inconsistencies in the data, as well as masking personal and confidential information. The input of this step is the initially filtered text data, and the output is the cleaned text data.

[1163] Step 4:

[1164] The server presents the cleaned data to the user for review and correction. The user reviews the cleaned data through a web-based interface and makes corrections as needed. A browser is used for this review and correction work. The input of the step is the cleaned text data, and the output is the text data reviewed and corrected by the user.

[1165] Step 5:

[1166] The server retrains the AI ​​model based on the confirmed and corrected data. It uses a deep learning framework such as TensorFlow or PyTorch to update the retrained AI model. The input for this step is the text data confirmed and corrected by the user, and the output is the retrained AI model.

[1167] Step 6:

[1168] The server uses the retrained AI model to operate as a chatbot. The smart glasses use this chatbot to assist store clerks in dealing with customers in real time. Through the display and camera installed in the smart glasses, store clerks can receive customer questions and instantly display appropriate answers. The input for this step is the retrained AI model, and the output is customer service information provided in real time.

[1169] For example, if a customer asks, "What features does this product have?" the smart glasses' display will respond, "This product has the latest Bluetooth functionality, a long battery life, and is waterproof."

[1170] 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.

[1171] The purpose of this invention is to provide responses that correspond to the user's emotions and improve their accuracy by combining an emotion engine that recognizes the user's emotions with an AI chatbot system. The specific operation of the system's program is explained below.

[1172] First, the user inputs the website or database information to be collected into the device. The device receives this information and sends a collection instruction to the server. The server then accesses the specified website or database and collects the necessary text data. For example, it can extract question and answer pairs from an FAQ page.

[1173] Next, the server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. Once the initial filtering is complete, the data moves on to the next step.

[1174] Cleaning AI then takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[1175] Once the cleaning is complete, the server presents the data to the user. The user then checks the data and makes any necessary corrections. The data that the user has checked and corrected is then sent back to the server. This process ensures the accuracy of the data.

[1176] The server then passes the revised data to a retraining AI, which uses the data to retrain the chatbot's model, generating a new model in the process.

[1177] The generated new model is sent from the server to the device. The device installs this model in the chatbot, and the chatbot begins operation. The user asks a question using the chatbot, and the chatbot generates an appropriate response based on the new model.

[1178] In the present invention, the system is further improved by adding an emotion engine. When a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. The analyzed emotion data is sent to the server and reflected in the response. For example, if the user is feeling angry or stressed, the chatbot will provide a calm and gentle response accordingly.

[1179] Furthermore, emotional data is collected and used for the next retraining. This allows the chatbot to continually improve and respond better to user emotions. For example, in a customer support chatbot, when a user expresses dissatisfaction, the dissatisfaction can be handled quickly and appropriately, improving customer satisfaction.

[1180] In this way, by combining the collection, cleaning, and relearning of learning data with an emotion engine, the present invention enables the construction and operation of a highly accurate chatbot system that takes emotions into consideration.

[1181] The processing flow will be explained below.

[1182] Step 1:

[1183] The user inputs information about the websites and databases to be collected into the terminal. The terminal receives this information and sends a collection instruction to the server.

[1184] Step 2:

[1185] The server accesses the specified website or database and collects the required text data (for example, extracting question and answer pairs from an FAQ page). The collected data is temporarily stored on the server.

[1186] Step 3:

[1187] The server performs initial filtering on the collected data, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. The filtered data is then passed on to the next step.

[1188] Step 4:

[1189] Cleaning AI takes the filtered data and uses deep learning and natural language processing techniques to clean it, including removing duplicates, correcting typos, and automatically detecting and masking personal and sensitive information.

[1190] Step 5:

[1191] The server presents the cleaned data to the user, who then checks the data through their device and makes any necessary corrections. The corrected data is then sent back to the server.

[1192] Step 6:

[1193] The server prepares the data reflecting the user's corrections and passes it to the retraining AI, which then retrains the chatbot model based on this data. This process generates a new model.

[1194] Step 7:

[1195] The server sends the newly trained model to the terminal, which then installs it into the chatbot, and the chatbot begins operating.

[1196] Step 8:

[1197] The user uses the chatbot to ask a question, and the chatbot generates an appropriate response based on the new model. During this process, the emotion engine recognizes the user's emotions and analyzes the user's emotional state from the text. For example, if the user is angry, the chatbot will provide a calm and gentle response accordingly.

[1198] Step 9:

[1199] The server monitors the chatbot's responses and records the emotional data collected by the emotion engine. The server collects this data as feedback and uses it for the next re-training.

[1200] Step 10:

[1201] During retraining, the server passes feedback, including emotional data, to the retraining AI to train it to improve its emotional responses, allowing the chatbot to better respond to the user's emotions.

[1202] Example 2

[1203] 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."

[1204] Existing AI chatbot systems often lack the ability to recognize user emotions and respond inappropriately. Furthermore, there is a need for efficient and accurate methods for the entire process, from data collection to model retraining and data cleansing. In particular, challenges remain in automatically masking personal and confidential information and removing duplicate data. Furthermore, the accuracy of chatbot responses has not been sufficiently improved due to a lack of continuous training using user feedback and sentiment analysis data.

[1205] 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.

[1206] In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, and artificial intelligence means for cleaning the initially filtered data. It also includes means for presenting the cleaned data to a user and accepting user confirmation and correction, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing emotions from text input by the user, means for generating chatbot responses based on this emotional data, and means for collecting emotional data and using it for the next relearning. This enables the generation of appropriate responses based on the user's emotions, thereby improving the accuracy of chatbot responses while protecting personal and confidential information.

[1207] "Training data" refers to information or data sets used to train an artificial intelligence model.

[1208] "Collection means" refers to the method or device used to obtain the necessary information from websites, databases, etc.

[1209] "Initial filtering" is a process for removing unnecessary information from collected data and standardizing the data format.

[1210] "Artificial intelligence means" refers to techniques that use machine learning algorithms and deep learning models to analyze and process data.

[1211] "Cleaning" is an operation to remove noise and errors from collected and filtered data and to arrange the data.

[1212] A "presentation means" is an interface that displays the cleaned data to the user and allows for review and modification.

[1213] "Retraining" is the process of retraining an artificial intelligence model using updated data.

[1214] A "model" refers to an artificial intelligence structure or algorithm that has been trained for a specific purpose.

[1215] A "chatbot" is a computer program that automatically handles interactions with users.

[1216] "Sentiment analysis" is a technique for detecting and classifying user emotions from text data.

[1217] "Response generation" is the process by which a chatbot creates a response to a user's input.

[1218] "Feedback" refers to usage evaluations and opinions provided by users, and is information that is used to improve the system.

[1219] "Auto-masking" is the process of automatically identifying personal or sensitive information from collected data and masking it to protect it.

[1220] This invention is an AI chatbot system that recognizes a user's emotions and provides responses according to those emotions. The purpose is to improve the accuracy of the data collected from the user while also increasing the accuracy of the chatbot. The implementation method of this system is specifically described below, specifying the hardware and software used.

[1221] First, the user inputs the website or database information to be collected into the device. For example, the user can input the URL of a customer support FAQ page. The device receives this input information and sends a collection instruction to the server using an HTTP request.

[1222] The server then accesses the specified websites and databases to collect the required text data, using a scraping tool to extract the text from FAQ pages and store it in a temporary database.

[1223] The collected data is initially filtered by the server, which includes removing HTML tags, formatting lines and spaces, standardizing date and time formats, etc. For example, the data is cleaned by using regular expressions to remove unnecessary HTML tags.

[1224] The server then uses cleaning AI to further clean the initially filtered data. Cleaning includes removing duplicate data, correcting typos, and automatically detecting and masking personal and confidential information. For example, if the same question appears multiple times, the cleaning AI will remove the duplicates.

[1225] The cleaned data is presented to the user by the server. The cleaned FAQ data is displayed through a user interface, and the user checks it. For example, the user checks whether the correct answer matches the question correctly. After the user makes the necessary corrections, the results are sent to the server.

[1226] The server then passes the corrected data to a retraining AI to retrain the chatbot's model. This process uses frameworks such as TensorFlow or PyTorch. Once the new model is generated, the model data is sent to the device using a secure protocol (e.g., HTTPS).

[1227] The terminal installs the new model into the chatbot. The chatbot begins operation by replacing the existing model with the new one. The user uses the chatbot based on the new model and asks questions. For example, if the user asks, "How do I return a product?", the chatbot responds, "You can return a product from your My Page on the website."

[1228] Furthermore, when a user inputs a question into the chatbot, the emotion engine analyzes the user's emotions from the text. For example, natural language processing technology is used to classify emotions such as "anger," "joy," and "sadness." The server reflects the analyzed emotional data in the response; for example, if the server determines that the user is feeling "anger," it will provide a calm and gentle response accordingly.

[1229] The emotional data is stored on the server and used for the next re-learning, allowing the chatbot to continue generating appropriate responses according to the user's emotions.

[1230] As a concrete example, the following prompt sentence can be used:

[1231] "How do I return a product?"

[1232] In this way, the system of the present invention can provide responses that correspond to the user's emotions, making it possible to realize a highly accurate chatbot system.

[1233] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1234] Step 1:

[1235] The user inputs the information from the website or database to be collected into the terminal.

[1236] Input: URLs and queries for websites and databases to be collected

[1237] Specific operation: The user enters information into an input field in a browser or application, and a collection instruction is sent to the terminal.

[1238] Output: The terminal generates data along with collection instructions to convert the input information into a format to be sent to the server.

[1239] Step 2:

[1240] The terminal sends a collection instruction to the server using an HTTP request.

[1241] Input: Collected instruction data from the user

[1242] Specific operation: The terminal packages the input information into an HTTP request and sends it to the server.

[1243] Output: The server receives the HTTP request and interprets the collection instructions data.

[1244] Step 3:

[1245] The server accesses the specified website or database and collects the required text data.

[1246] Input: Collection instruction data

[1247] What happens: The server uses a scraping tool to extract text data from the specified URL, for example, extracting question and answer pairs from an FAQ page.

[1248] Output: Save the extracted text data in a temporary database.

[1249] Step 4:

[1250] The server performs an initial filtering of the collected data.

[1251] Input: Collected text data

[1252] Specific operation: The server removes unnecessary HTML tags from the data, formats line breaks and whitespace, unifies date and time formats, etc. It uses regular expressions and text processing libraries.

[1253] Output: Generates the initial filtered data.

[1254] Step 5:

[1255] The server further cleans the initially filtered data using cleaning AI.

[1256] Input: Initial filtered data

[1257] What it does: Removes duplicate data, corrects typos, automatically detects and masks personal and confidential information, and uses natural language processing techniques to identify specific patterns and data formats.

[1258] Output: Produces the cleaned data.

[1259] Step 6:

[1260] The server presents the cleaned data to the user and accepts confirmation and correction by the user.

[1261] Input: Cleaned data

[1262] Specific operation: The server displays the data through a user interface, allowing the user to review and modify it. It also provides an interface for receiving modified data.

[1263] Output: Data reviewed and corrected by the user

[1264] Step 7:

[1265] The server retrains the artificial intelligence model based on the data confirmed and corrected by the user.

[1266] Input: Verified and corrected data

[1267] What it does: Retrain the model using a framework such as TensorFlow or PyTorch, using the corrected data as training data to update the neural network weights.

[1268] Output: A new retrained AI model

[1269] Step 8:

[1270] The server transmits the retrained artificial intelligence model to the terminal.

[1271] Input: Retrained AI model

[1272] Specific operation: The server sends the new model to the device using a secure protocol (e.g., HTTPS).

[1273] Output: The new model data sent to the terminal.

[1274] Step 9:

[1275] The device installs the retrained artificial intelligence model into the chatbot and begins operating the chatbot.

[1276] Input: Retrained AI model

[1277] Specific behavior: Replace the existing model with the new one. Restart the chatbot software and load the new model.

[1278] Output: A running chatbot with the new model installed

[1279] Step 10:

[1280] Users use chatbots based on the new model to ask questions.

[1281] Input: User question text

[1282] Specific behavior: For example, ask, "How do I return an item?" The chatbot uses the new model to generate an appropriate response.

[1283] Output: Chatbot response text

[1284] Step 11:

[1285] When a user types a question into the chatbot, the emotion engine analyzes the user's emotions.

[1286] Input: User question text

[1287] Specific operation: Using natural language processing technology, emotions such as "anger," "joy," and "sadness" are classified.

[1288] Output: Parsed emotion data

[1289] Step 12:

[1290] The server reflects the analyzed emotional data in its response and takes appropriate action.

[1291] Input: Parsed emotion data, user question text

[1292] Specific behavior: If the user is analyzed as feeling "anger," the chatbot will provide a calm and gentle response.

[1293] Output: Chatbot response depending on the emotion

[1294] Step 13:

[1295] The emotion data is stored on the server and used for the next re-learning.

[1296] Input: Parsed emotion data

[1297] Specific operation: Emotion data is saved in the database and used the next time the model is retrained.

[1298] Output: Saved emotion data

[1299] (Application example 2)

[1300] 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."

[1301] Current customer support systems in brick-and-mortar stores face the problem of being unable to analyze customers' emotions, such as facial expressions and tone of voice, in real time and provide appropriate support based on that analysis. Furthermore, the inability to respond appropriately to customer emotions makes it difficult to improve customer satisfaction. To solve this problem, a system is needed that analyzes customers' emotions and provides instructions on how to respond in real time based on the results.

[1302] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting training data, means for initially filtering the collected training data, artificial intelligence means for cleaning the initially filtered data, means for presenting the cleaned data to a user and accepting confirmation and correction by the user, means for relearning an artificial intelligence model based on the data confirmed and corrected by the user, means for implementing the re-trained artificial intelligence model and operating it as a chatbot, means for analyzing user emotions using an emotion analysis engine, and means for reflecting the user's emotion data in the response content. This makes it possible to analyze customer emotions in real time and instruct store staff on appropriate responses based on the analysis results.

[1303] "Means for collecting training data" refers to the mechanism for obtaining the necessary text data from designated websites and databases.

[1304] "Initial filtering" refers to the process of removing unnecessary information from collected data and shaping the data.

[1305] "Artificial intelligence means" refers to a system that uses deep learning and natural language processing techniques to clean and analyze data.

[1306] The "means for presenting to the user and accepting confirmation and correction by the user" refers to an interface that shows the cleaned data to the user and accepts confirmation and correction.

[1307] The "means for retraining an artificial intelligence model" is a function for retraining a model based on data that has been checked and corrected by the user.

[1308] "Means of operating as a chatbot" refers to the process of installing and operating the retrained artificial intelligence model in an actual chatbot system.

[1309] "Means for analyzing user emotions using an emotion analysis engine" refers to a process that uses an engine to analyze an emotional state from text or voice data entered by a user.

[1310] The "means for reflecting the user's emotional data in the response content" is a function for adjusting the response content to the user based on the analyzed emotional data.

[1311] To implement this invention, store staff wear smart glasses and use the glasses' camera and microphone to capture customer facial expressions and voices. The data is then sent to a server in real time, where an emotion analysis engine is used to analyze the customer's emotional state. Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[1312] This system uses the following hardware and software:

[1313] Hardware

[1314] Smart glasses (with built-in camera and microphone)

[1315] Server (high-performance cloud server, e.g., Amazon EC2, Microsoft Azure)

[1316] software

[1317] Sentiment analysis engine (e.g., Microsoft Azure Cognitive Services Emotion API)

[1318] Data processing and communication (Python, TensorFlow, Flask)

[1319] Overview of data processing and calculation

[1320] 1. Capture customer facial and voice data

[1321] The smart glasses' camera and microphone capture the customer's facial and voice data in real time.

[1322] 2. Sending data to the server

[1323] The captured data is sent from the smart glasses to a server.

[1324] 3. Pretreatment

[1325] The received data is preprocessed on the server side (noise removal, data formatting, etc.) and then passed to the sentiment analysis engine.

[1326] 4. Emotion analysis

[1327] The sentiment analysis engine analyzes the customer's emotional state (e.g., "anger," "sadness," "joy," etc.).

[1328] 5. Feedback of response instructions

[1329] Based on the analysis results, appropriate instructions are displayed on the smart glasses' display.

[1330] Specific examples

[1331] For example, if a customer looks dissatisfied while asking about a product at a counter, their facial expression and voice data will be captured, and the server will perform emotion analysis and determine that they are dissatisfied. As a result, the smart glasses will display a message saying, "The customer seems dissatisfied. Please be courteous and considerate."

[1332] Prompt Sentence Examples

[1333] "Please analyze the customer's emotions from the following facial and voice data:

[1334] Facial expression data: {image_data}

[1335] Audio data: {audio_data}"

[1336] In this way, it becomes possible to provide more effective customer service in physical stores, which contributes to improving customer satisfaction.

[1337] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1338] Step 1:

[1339] The smart glasses capture the customer's facial expression and voice data. They use the customer's real-time video and audio as input, which are captured by the smart glasses' built-in camera and microphone. The captured data is output in byte stream format.

[1340] Step 2:

[1341] The smart glasses send the captured data to the server. The captured byte stream data is the input and is sent to the server through the communication module. The output is the byte stream data received on the server side.

[1342] Step 3:

[1343] The server preprocesses the received data. The input data is the byte stream sent from the smart glasses, and it undergoes noise removal and data shaping (e.g., frame extraction and audio cleaning). The preprocessed shaped data is output as image data and audio data.

[1344] Step 4:

[1345] The server passes the preprocessed data to the sentiment analysis engine. The input data is the preprocessed image data and audio data, which are then input to the sentiment analysis engine. The sentiment analysis engine uses a deep learning model to identify the emotional state (e.g., "anger," "sadness," "joy," etc.). The output is text data that indicates the emotional state.

[1346] Step 5:

[1347] The server generates response instructions based on the analysis results. The output text data from the sentiment analysis engine is used as input, and response instructions for staff are generated based on that. For example, a message such as "The customer seems dissatisfied. Please try to respond politely and kindly" is generated. The output is text data showing response instructions.

[1348] Step 6:

[1349] The server sends the generated response instruction to the smart glasses. The input is text data indicating the response instruction, which is sent to the smart glasses through the communication module. The output is a text message displayed on the smart glasses.

[1350] Step 7:

[1351] The smart glasses display instructions to the staff. Text data indicating instructions sent from the server is used as input and displayed on the smart glasses' display. The staff responds to the customer based on this displayed information. The output is an instruction message that is displayed to the staff.

[1352] The above is a flow of specific processing steps for carrying out the invention, which is expected to improve the quality of customer service in brick-and-mortar stores and increase customer satisfaction.

[1353] 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.

[1354] 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.

[1355] 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.

[1356] 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.

[1357] 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.

[1358] 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.

[1359] 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).

[1360] 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.

[1361] 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."

[1362] 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.

[1363] 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).

[1364] 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.

[1365] 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.

[1366] 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.

[1367] 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.

[1368] 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.

[1369] 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.

[1370] 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.

[1371] 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.

[1372] 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.

[1373] 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.

[1374] The following is further disclosed regarding the above embodiment.

[1375] (Claim 1)

[1376] a means for collecting training data;

[1377] means for initially filtering the collected training data;

[1378] artificial intelligence means for cleaning the initially filtered data;

[1379] means for presenting the cleaned data to a user and accepting confirmation and correction by the user;

[1380] A means for retraining the artificial intelligence model based on the data confirmed and corrected by the user;

[1381] A means to implement the retrained AI model and operate it as a chatbot,

[1382] A system including:

[1383] (Claim 2)

[1384] A means to automatically recognize and mask personal and confidential information from training data, and

[1385] a means for continuously retraining the artificial intelligence model based on the cleaned data; and

[1386] The system of claim 1 further comprising:

[1387] (Claim 3)

[1388] A means of monitoring the cleaned data and collecting user feedback to reflect on the data;

[1389] A means of improving the chatbot's response accuracy by re-learning based on user feedback;

[1390] The system of claim 1 further comprising:

[1391] "Example 1"

[1392] (Claim 1)

[1393] A means for a user to provide information to be collected, and for a terminal to send a data collection instruction to a server based on the information;

[1394] A means for the server to access designated websites and databases to collect learning data;

[1395] A means for the server to initially filter the training data to remove unnecessary information;

[1396] Cleaning AI uses natural language processing technology to correct input errors and inconsistencies in the filtered data and mask personal information.

[1397] means for presenting the cleaned data to a user and accepting confirmation and correction by the user;

[1398] A means for the re-learning AI to re-train the artificial intelligence model based on the data confirmed and corrected by the user;

[1399] A means to send the artificial intelligence model generated by re-learning to a terminal and operate it as a chatbot;

[1400] A means to monitor the quality of the chatbot's responses and gather user feedback for use in the next refresher course; and

[1401] A system including:

[1402] (Claim 2)

[1403] A means to automatically recognize and mask personal and confidential information from training data, and

[1404] a means for continuously retraining the artificial intelligence model based on the cleaned data; and

[1405] The system of claim 1 further comprising:

[1406] (Claim 3)

[1407] A means of monitoring the cleaned data and collecting user feedback to reflect on the data;

[1408] A means of improving the chatbot's response accuracy by re-learning based on user feedback;

[1409] The system of claim 1 further comprising:

[1410] "Application Example 1"

[1411] (Claim 1)

[1412] a means for collecting training data;

[1413] means for initially filtering the collected training data;

[1414] artificial intelligence means for cleaning the initially filtered data;

[1415] means for presenting the cleaned data to a user and accepting confirmation and correction by the user;

[1416] A means for retraining the artificial intelligence model based on the data confirmed and corrected by the user;

[1417] The retrained AI model will be implemented and operated as a chatbot, and smart glasses will be used to assist store staff in dealing with customers in real time.

[1418] A system including:

[1419] (Claim 2)

[1420] A means to automatically recognize and mask personal and confidential information from training data, and

[1421] a means for continuously retraining the artificial intelligence model based on the cleaned data; and

[1422] A means of receiving customer questions through smart glasses and providing appropriate answers in real time;

[1423] The system of claim 1 further comprising:

[1424] (Claim 3)

[1425] A means of monitoring the cleaned data and collecting user feedback to reflect on the data;

[1426] A means of improving the chatbot's response accuracy by re-learning based on user feedback;

[1427] means for displaying the answer on a display of the smart glasses;

[1428] The system of claim 1 further comprising:

[1429] "Example 2: Combining Emotion Engines"

[1430] (Claim 1)

[1431] a means for collecting training data;

[1432] means for initially filtering the collected training data;

[1433] artificial intelligence means for cleaning the initially filtered data;

[1434] means for presenting the cleaned data to a user and accepting confirmation and correction by the user;

[1435] A means for retraining the artificial intelligence model based on the data confirmed and corrected by the user;

[1436] A means to implement the retrained AI model and operate it as a chatbot,

[1437] A means for analyzing emotions from a user's input text, and a means for generating a chatbot response based on the emotion data;

[1438] A means of collecting emotion data to be used for next retraining;

[1439] A system including:

[1440] (Claim 2)

[1441] A means to automatically recognize and mask personal and confidential information from training data, and

[1442] a means for continuously retraining the artificial intelligence model based on the cleaned data; and

[1443] A means for collecting and storing sentiment analysis data of user input text and using it for retraining;

[1444] The system of claim 1 further comprising:

[1445] (Claim 3)

[1446] A means of monitoring the cleaned data and collecting user feedback to reflect on the data;

[1447] A means of improving the chatbot's response accuracy by re-learning based on user feedback;

[1448] means for evaluating the accuracy of the responses generated using the sentiment analysis engine;

[1449] The system of claim 1 further comprising:

[1450] "Application example 2 when combining emotion engines"

[1451] (Claim 1)

[1452] a means for collecting training data;

[1453] means for initially filtering the collected training data;

[1454] artificial intelligence means for cleaning the initially filtered data;

[1455] means for presenting the cleaned data to a user and accepting confirmation and correction by the user;

[1456] A means for retraining the artificial intelligence model based on the data confirmed and corrected by the user;

[1457] A means to implement the retrained AI model and operate it as a chatbot,

[1458] means for analyzing user emotions using a sentiment analysis engine;

[1459] A means for reflecting user emotion data in response content;

[1460] A system including:

[1461] (Claim 2)

[1462] A means to automatically recognize and mask personal and confidential information from training data, and

[1463] a means for continuously retraining the artificial intelligence model based on the cleaned data; and

[1464] A means for analyzing the user's emotional state in real time using an emotion analysis engine and providing appropriate response instructions to staff;

[1465] The system of claim 1 further comprising:

[1466] (Claim 3)

[1467] A means of monitoring the cleaned data and collecting user feedback to reflect on the data;

[1468] A means of improving the chatbot's response accuracy by re-learning based on user feedback;

[1469] means for providing a response adapted to the user's emotions using emotion data analyzed by the emotion analysis engine;

[1470] The system of claim 1 further comprising: [Explanation of symbols]

[1471] 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 training data; means for initially filtering the collected training data; artificial intelligence means for cleaning the initially filtered data; means for presenting the cleaned data to a user and accepting confirmation and correction by the user; A means for retraining the artificial intelligence model based on the data confirmed and corrected by the user; A means to implement the retrained AI model and operate it as a chatbot, A system including:

2. A means to automatically recognize and mask personal and confidential information from training data, and a means for continuously retraining the artificial intelligence model based on the cleaned data; and The system of claim 1 further comprising:

3. A means of monitoring the cleaned data and collecting user feedback to reflect on the data; A means of improving the chatbot's response accuracy by re-learning based on user feedback; The system of claim 1 further comprising:

Images