system

Abstract

A system is provided.SOLUTION: A system includes a means for receiving and storing text data such as a manual, a guideline, and a FAQ related to each task, a means for vectorizing the received text data and calculating a cosine similarity matrix, a means for receiving a question from a user and generating a most relevant answer to the question, a means for providing feedback to the answer by the user, and a means for collecting the feedback and retraining a learning model based on the information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] When employees adapt to new work, many questions arise, and they often need to ask their superiors or senior colleagues to resolve them. However, busy times, environments that make it difficult to ask questions, and the guilt of asking again about something that has already been explained to them can all become obstacles that hinder efficient work execution. Furthermore, the senior colleagues who answer the questions are burdened by being asked the same questions multiple times. This can lead to issues such as reduced work efficiency and increased employee stress. [Means for solving the problem]

[0005] To address this issue, the present invention provides a system that includes: means for receiving and storing text data such as manuals, guidelines, and FAQs related to each task; means for vectorizing the received text data and calculating a cosine similarity matrix; means for receiving questions from users and generating the most relevant answers to those questions; means for users to provide feedback on the answers; and means for collecting feedback and retraining a learning model based on that information. This system creates an environment where employees can ask questions 24 hours a day, 365 days a year, improving work efficiency. It also reduces the burden on senior employees and increases productivity and employee satisfaction throughout the organization.

[0006] A "manual" is a document that describes the procedures and rules for a specific task or procedure.

[0007] A "guideline" is a document that explains specific guidelines and methods for carrying out a particular purpose or task.

[0008] An "FAQ" is a document that compiles frequently asked questions and their answers, with the aim of providing quick responses to frequently asked questions.

[0009] "Text data" is information expressed in the form of a string of characters, and includes documents, sentences, words, and the like.

[0010] "Vectorization" is the process of converting text data into numerical arrays, making it easier for machine learning models to work with.

[0011] "Cosine similarity" is a method for numerically evaluating the similarity between two vectors, and its value falls within the range of -1 to 1.

[0012] A "question" is a question that a user enters into the Elder GPT system to gain knowledge or information.

[0013] An "answer" is a provision of information or knowledge generated by the Elder GPT system in response to a question.

[0014] "Feedback" refers to the evaluation or suggestions for improvement that a user provides in response to a received answer.

[0015] A "learning model" is a system that learns patterns and relationships based on given data and makes predictions and classifications for new data.

[0016] "Retraining" is the process of retraining an existing model by adding new data or feedback information to improve the model's performance. [Brief explanation of the drawings]

[0017] [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

[0018] 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.

[0019] First, the terms used in the following description will be explained.

[0020] 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).

[0021] 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.

[0022] 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.

[0023] 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.

[0024] 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."

[0025] [First embodiment]

[0026] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0027] 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.

[0028] 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).

[0029] 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.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0034] 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.

[0035] 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.

[0036] 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.

[0037] 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."

[0038] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention provides business support using text data such as manuals, guidelines, and FAQs relating to each business.

[0039] 1. Uploading and saving data

[0040] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are stored in a specific directory on the server. This allows necessary work-related information to be managed centrally and easily accessible.

[0041] 2. Loading and Preprocessing Data

[0042] The server reads the saved file as text data and combines multiple sentences into a single list, allowing for consistent handling of the data in subsequent processing.

[0043] 3. Vectorization and Model Generation

[0044] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers. A cosine similarity matrix is ​​calculated based on the vectorized data, and a learning model is generated using this. This makes it possible to evaluate the similarity between text data.

[0045] 4. Question and Answer Processing

[0046] A user inputs a question in natural language to the Elder GPT system using a client device. For example, a question like, "Please tell me the format of the financial report." This question is sent from the device to the server via an API.

[0047] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained, the server searches for the most relevant document and generates its content as the answer.

[0048] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0049] For example, if a user asks, "What is the format of a financial report?", the server will search for the most relevant guidelines or manuals related to financial reports from text data that has been vectorized and learned in advance, and provide the content to the user, allowing the user to obtain accurate information quickly.

[0050] 5. Providing feedback and retraining

[0051] The user inputs feedback such as satisfaction with the answer provided, points for improvement, etc. For example, feedback such as "satisfied" or "insufficient information" is applicable.

[0052] The terminal transmits the input feedback information to the server.

[0053] The server collects the submitted feedback and stores it for use in improving the learning model in the future. The model is periodically retrained with data that reflects the feedback information, further improving the accuracy and responsiveness of the system.

[0054] Through the above process, the Elder GPT system provides a support environment where employees can easily ask questions at any time, improving work efficiency and reducing employee stress.

[0055] The processing flow will be explained below.

[0056] Step 1:

[0057] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a web interface or a dedicated application.

[0058] Step 2:

[0059] The server receives the uploaded files and stores them in the appropriate directory, allowing necessary business-related information to be managed centrally.

[0060] Step 3:

[0061] The server reads the saved files, treats them as text data, gets the contents of each file, and compiles them into a list.

[0062] Step 4:

[0063] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This vectorization makes it easier for machine learning models to handle text data.

[0064] Step 5:

[0065] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which makes it possible to evaluate the similarity between text data.

[0066] Step 6:

[0067] A user inputs a question in natural language to the Elder GPT system using a client terminal, and the input question is sent from the terminal to the server via an API.

[0068] Step 7:

[0069] The server vectorizes the received question text and performs cosine similarity calculations. Based on the similarity score, it searches for the most relevant documents.

[0070] Step 8:

[0071] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[0072] Step 9:

[0073] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0074] Step 10:

[0075] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[0076] Step 11:

[0077] The server collects the submitted feedback and stores it for future use in improving the learning model.

[0078] Step 12:

[0079] The server periodically retrains the model based on feedback information, improving the accuracy and responsiveness of the system and providing better support.

[0080] Example 1

[0081] 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."

[0082] Conventional systems have difficulty in centrally managing and effectively utilizing business document data. Furthermore, they lack mechanisms for providing quick and accurate answers to user questions, and lack the means to continuously improve the quality of answers. This makes it difficult to maximize business efficiency.

[0083] 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.

[0084] In this invention, the server includes a means for receiving and storing document data related to each business, a means for reading the received document data, listing it, and performing preprocessing, and a means for vectorizing the preprocessed document data and calculating a similarity matrix, thereby enabling unified management and effective use of document data.

[0085] The server further includes means for receiving question data, vectorizing the data, and generating the most relevant answer based on similarity, means for users to provide feedback on the answers, and means for collecting the feedback and retraining the generative model based on the feedback, thereby enabling the server to provide fast and accurate answers to user questions and continuously improve the quality of the answers by incorporating the feedback.

[0086] "Document data related to each business" is a general term for text data such as manuals, guidelines, FAQs, etc. related to the business operations of a company or organization.

[0087] "Means of receiving and storing" refers to the process by which the server receives document data uploaded by a user and stores the data in a specific directory or the like.

[0088] "Means of creating a list and pre-processing" refers to the process of reading the received data in list format and performing pre-processing such as cleaning the data and removing unnecessary information.

[0089] "Vectorization" refers to the process of converting document data into an array of numbers, or vectors, which makes it easier for computers to process.

[0090] The "means for calculating a similarity matrix" refers to a process for calculating the cosine similarity between vectorized data and expressing the result in a matrix format.

[0091] "Question data" refers to text data of the question that the user inputs to the system.

[0092] The "means for generating the most relevant answer based on similarity" refers to a process of vectorizing question data, identifying the most relevant document data using a pre-calculated similarity matrix, and generating an answer.

[0093] "Means for providing feedback" refers to the process by which users input their ratings and opinions on the answers provided into the system.

[0094] "Means for collecting feedback" refers to the process of storing feedback provided by users in a database or the like and aggregating it for use in subsequent processing.

[0095] A "generative model" is a model generated from document data using a specific learning algorithm, and is used to generate answers to questions.

[0096] "Retraining" refers to the process of updating an existing generative model based on collected feedback information to improve its accuracy and responsiveness.

[0097] This invention is a system that efficiently manages document data related to each task in a business support system and can provide quick and accurate answers to user questions. The system is implemented by clearly defining the roles of the server, terminal, and user. The hardware used includes a standard web server and client PC. The software used includes a natural language processing library (e.g., spaCy, Transformers), a data management library (e.g., pandas), and a library for similarity calculation (e.g., SciKit-Learn).

[0098] Data upload and management

[0099] Users upload document data related to their work (e.g., manuals, guidelines, FAQs, etc.) from their own devices to the server. Uploading is done through a web interface, and the selected file is saved in a specific directory on the server. For example, a user opens a browser, clicks the upload button, selects the desired document file, and executes the upload. At this time, the file's metadata (e.g., file name, upload date and time) is also saved.

[0100] Data loading and preprocessing

[0101] The server periodically scans the stored document data and reads newly uploaded files in text format. The read data is converted into a list format and then text cleaning (removal of whitespace and unnecessary characters) is performed. This process is performed using the Python pandas library. For example, the pandas read_csv method is used to read a CSV file and convert it to DataFrame format.

[0102] Vectorization and Similarity Calculation

[0103] After the text data has been read and preprocessed, it is converted into a numerical vector using a natural language processing library. This conversion is performed using libraries such as spaCy and Transformers. Next, a cosine similarity matrix is ​​calculated from the vectorized data to evaluate the relevance between the text data. The calculated similarity matrix is ​​then processed using the SciKit-Learn library.

[0104] Question and answer processing

[0105] A user inputs a question in natural language from their own device to the system. For example, a question might be, "What is the format of a financial report?" The question is entered through a web form and sent to the server. The server vectorizes the question and identifies the most relevant document data using a pre-calculated similarity matrix. An answer is generated based on the content and sent back to the device. The user can check the answer on their browser. As a specific example, if a user asks, "Please tell me about the marketing strategy for a new product," the server will search for relevant documents and generate an answer.

[0106] Feedback and retraining

[0107] Users provide feedback on the answers they provide, such as their satisfaction level or additional comments, through a web form. This feedback is sent from the device to the server and stored in a database. The server periodically retrains the model based on the feedback information, allowing the system to continuously improve its answer accuracy and user satisfaction.

[0108] Prompt Sentence Examples

[0109] Please tell me the format of the financial statement.

[0110] "Please tell us about your marketing strategy for the new product."

[0111] "How do I troubleshoot my product?"

[0112] This allows the system to not only centrally manage business-related information, but also provide quick and accurate answers to user questions. Furthermore, by improving the accuracy of the system through the feedback function, the quality of business support can be improved.

[0113] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0114] Step 1: Upload and save your data

[0115] A user uploads document data related to work by, for example, clicking the upload button on the browser and selecting the specified file.

[0116] Input: A document file selected by the user

[0117] The server receives the uploaded file and stores it in a specific directory (e.g., / data / uploads), along with metadata (file name, upload date and time).

[0118] Output: The saved file and its metadata

[0119] Step 2: Loading and preprocessing data

[0120] The server periodically scans a specific directory and reads newly uploaded document files, converts them as text files into a DataFrame, and then organizes them into a list.

[0121] Input: Saved document file

[0122] The server cleans the data, removing unnecessary whitespace and special characters, using the read_csv method from the pandas library.

[0123] Output: A list of preprocessed text data

[0124] Step 3: Vectorization and similarity matrix calculation

[0125] The server vectorizes the preprocessed text data and converts it into a numerical vector using a natural language processing library (e.g., spaCy, Transformers).

[0126] Input: A list of preprocessed text data

[0127] The server calculates the cosine similarity matrix based on the vectorized data. It uses the SciKit-Learn library to calculate the similarity between documents.

[0128] Output: Cosine similarity matrix

[0129] Step 4: Receiving questions and generating answers

[0130] The user inputs a question to the system from the client terminal, for example, a prompt such as "Please tell me the format of the financial statement."

[0131] Input: User question data

[0132] The device sends the question to the server via an API.

[0133] The server vectorizes the question, uses a pre-computed similarity matrix to identify the most relevant documents, and generates an answer.

[0134] Output: Generated response data

[0135] Step 5: View your answers

[0136] The server transmits the generated response data to the terminal.

[0137] Input: Generated response data

[0138] The device displays the received response to the user, which appears as text or a link in the browser.

[0139] Output: The answer that is displayed to the user

[0140] Step 6: Provide and collect feedback

[0141] Users provide feedback on the answers provided, including their satisfaction and any additional comments, by filling out a form.

[0142] Input: User feedback data

[0143] The device sends feedback to the server via an API.

[0144] The server collects the feedback and stores it in a database.

[0145] Output: Feedback data collected and stored

[0146] Step 7: Retrain with feedback

[0147] The server periodically retrains the generative model based on the collected feedback data, adding new data and recalculating the cosine similarity matrix to update the model.

[0148] Input: Collected feedback data

[0149] The server stores the updated generative model.

[0150] Output: Updated generative model

[0151] (Application example 1)

[0152] 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."

[0153] Picking work at logistics centers requires accuracy and efficiency, so it is important to have an environment where staff can quickly obtain the information they need. However, traditional methods often require staff to refer to paper-based manuals and guidelines, which is time-consuming and laborious. In addition, staff have limited means of obtaining immediate answers to their questions, which can lead to reduced work efficiency. Furthermore, there is a lack of mechanisms for effectively utilizing feedback to improve the quality of the answers provided. These issues need to be resolved to improve the efficiency and accuracy of picking work.

[0154] 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.

[0155] In this invention, the server includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer, thereby enabling pickers to quickly obtain necessary information and improve system performance based on the feedback.

[0156] A "manual" is a document that describes the procedures and methods for performing a specific task or work.

[0157] "Guidelines" are documents that provide basic guidelines and standards to be followed in business operations and procedures.

[0158] A "FAQ" is a document that compiles frequently asked questions and their answers.

[0159] "Vectorization" is the process of converting text data into an array of numbers.

[0160] A "cosine similarity matrix" is a matrix used to calculate the similarity between different vectors.

[0161] The "means for receiving a question" is a mechanism by which the system receives a question from a user.

[0162] The "means for generating an answer" is a process for generating the most appropriate answer to a question received from a user.

[0163] The "means for providing feedback" is a mechanism for users to input their opinions, such as ratings and suggestions for improvement, regarding the answers provided.

[0164] "Means for collecting feedback" refers to the process by which the system collects feedback information provided by users.

[0165] The "retraining method" is the process of re-learning the learning model to improve its performance based on the collected feedback information.

[0166] "Smart glasses" are a wearable eyeglass-type device equipped with a display that can display information and accept voice input.

[0167] A "means for asking questions in natural language" is a mechanism that allows users to ask questions using natural language, either spoken or written.

[0168] The "display means" is a method for showing the generated answers and related information to the user in real time.

[0169] The present invention comprises a system for supporting picking operations in a logistics center. The system includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer.

[0170] First, the distribution center manager uploads and stores text data such as manuals, guidelines, and FAQs for each operation on the server. The server then reads the stored text data and manages it centrally in list format.

[0171] The server then vectorizes the text data and calculates a cosine similarity matrix, which can improve the accuracy of generating answers to questions.

[0172] When a user wears smart glasses and speaks a question in natural language, the smart glasses send the question to the server via API. The server vectorizes the received question, performs cosine similarity calculations, and generates an answer based on the most relevant documents. The answer is displayed in real time on the smart glasses' display, allowing the user to immediately check the information they need.

[0173] Users can also provide feedback on the answers they provide. The feedback information is sent to the server and stored. The server uses this feedback information to retrain the learning model and improve the system's performance.

[0174] The hardware required to realize this system includes smart glasses (e.g., Google® Glass®, Vuzix Blade) and a server. The software used includes Python, scikit-learn, and spaCy for vectorizing text data, Flask and FastAPI for the server-side API, AWS® S3 and Google Cloud Storage for cloud storage, and OpenAI® GPT and Hugging Face Transformers for the question-answering model.

[0175] For example, when a picker at a logistics center asks through smart glasses, "What's the next item to pick?", the server vectorizes the question and generates an answer based on relevant documents from the stored data. This answer is displayed on the smart glasses in real time, allowing the picker to quickly proceed with the next task.

[0176] An example of a prompt sentence is, "What is the next item to pick?" In this way, the present invention provides a system that effectively supports picking operations in logistics centers and significantly improves work efficiency and accuracy.

[0177] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0178] Step 1:

[0179] The server receives and stores text data such as manuals, guidelines, and FAQs for each operation uploaded by the logistics center administrator. The input is the text data uploaded by the administrator, and the output is a file stored on the server. This process allows necessary operation-related information to be centrally managed on the server and easily accessible.

[0180] Step 2:

[0181] The server reads the saved text data and compiles multiple sentences into a single list. The input is the text data saved in the server, and the output is the text data compiled into a single list. This process allows the data to be handled consistently in subsequent processes.

[0182] Step 3:

[0183] The server vectorizes the text data compiled in list format and calculates the cosine similarity matrix. The input is text data in list format, and the output is the vectorized data and the cosine similarity matrix. This process makes it possible to evaluate the similarity between text data.

[0184] Step 4:

[0185] The user uses the smart glasses to input a question in natural language by voice. The input is the user's natural language question, and the output is a question recorded by the smart glasses as voice data. This process allows the user to immediately query the system for the information they need.

[0186] Step 5:

[0187] The device converts the question recorded as voice data into text data and sends it to the server via API. The input is the user's question recorded as voice data, and the output is the question as text data sent to the server. Through this process, the server receives the user's question.

[0188] Step 6:

[0189] The server vectorizes the received question and performs a cosine similarity calculation. The input is the question as text data, and the output is the most relevant documents based on the similarity score. Through this process, the server generates an appropriate answer.

[0190] Step 7:

[0191] The server sends the generated answers to the smart glasses via API and displays them to the user in real time using the display means. The input is the answer as the most relevant document, and the output is the answer displayed on the smart glasses. This process allows the user to quickly obtain information.

[0192] Step 8:

[0193] The user provides feedback on the displayed answers. The input is the displayed answer and the user's evaluation of it, and the output is feedback information sent to the server. This process allows the user's opinion to be reflected in the system.

[0194] Step 9:

[0195] The server retrains the learning model based on the collected feedback information. The input is the feedback information, and the output is an improved learning model. This process improves the system's performance and increases the accuracy of the next question answer.

[0196] 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.

[0197] The following is a detailed description of an embodiment of the present invention. The present invention is a system that provides business support using text data such as manuals, guidelines, and FAQs related to each business, and also combines it with an emotion engine that recognizes user emotions.

[0198] 1. Uploading and saving data

[0199] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are uploaded via a web interface or a dedicated application and are saved in a specific directory on the server. This allows work-related information to be managed centrally and be accessible at any time.

[0200] 2. Loading and Preprocessing Data

[0201] The server reads the saved file as text data and organizes multiple sentences into a list, allowing the data to be handled consistently in subsequent processing.

[0202] 3. Vectorization and Model Generation

[0203] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers, which makes it easier for machine learning models to handle text data. A cosine similarity matrix is ​​then calculated based on the vectorized data to generate a learning model. This learning model makes it possible to evaluate the similarity between text data.

[0204] 4. Question and Answer Processing

[0205] A user inputs a question in natural language to the ElderGPT system using a client device. For example, a question like, "How do I create a project plan?" The input question is sent from the device to the server via an API.

[0206] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, it searches for the most relevant document and generates an answer from that document. The generated answer is then sent back to the device.

[0207] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0208] 5. Emotion Recognition with Emotion Engine

[0209] The server uses an emotion engine to analyze the user's emotions regarding questions and answers. The emotion engine uses text analysis technology to recognize emotions from the user's input. For example, it can classify emotions as positive, negative, neutral, etc.

[0210] 6. Adjusting responses based on emotions

[0211] The server generates a response appropriate to the user based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a specific and detailed response.

[0212] 7. Providing and Saving Feedback

[0213] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[0214] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to periodically retrain the model.

[0215] 8. Retraining the model

[0216] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[0217] For example, if a user asks, "How do I create a project plan?", the server will search for relevant documents and provide them. At the same time, it will recognize emotions from the user's input and determine whether a positive response is required. In this way, a system can be realized that quickly and appropriately provides the information the user is looking for.

[0218] Through the above process, the Elder GPT system provides a support environment that can respond to users while recognizing their emotions 24 hours a day, 365 days a year, thereby improving work efficiency and reducing employee stress.

[0219] The processing flow will be explained below.

[0220] Step 1:

[0221] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a dedicated web interface or application.

[0222] Step 2:

[0223] The server receives the uploaded files and stores them in the specified directory, allowing business-related information to be managed centrally.

[0224] Step 3:

[0225] The server reads the saved files, identifies them as text data, and combines the contents of multiple files into a single list.

[0226] Step 4:

[0227] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This conversion makes it easier for machine learning models to handle text data.

[0228] Step 5:

[0229] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which is used to evaluate the similarity between text data.

[0230] Step 6:

[0231] A user inputs a question in natural language into the Elder GPT system using a client device. For example, a question like, "How do I create a project plan?" The question is sent from the device to the server via an API.

[0232] Step 7:

[0233] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, the server searches for the most relevant document. For example, the server searches for the most relevant "Guidelines for creating project plans."

[0234] Step 8:

[0235] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[0236] Step 9:

[0237] The terminal displays the answer received from the server to the user, allowing the user to quickly resolve their question.

[0238] Step 10:

[0239] The server uses an emotion engine to analyze the user's sentiment towards the question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input, which can be classified as negative, positive, neutral, etc.

[0240] Step 11:

[0241] The server further tailors the answer based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a more detailed and helpful answer.

[0242] Step 12:

[0243] The user provides feedback on the displayed answer. The feedback includes satisfaction with the answer and areas for improvement. The device sends this feedback information to the server.

[0244] Step 13:

[0245] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to retrain the model.

[0246] Step 14:

[0247] The server periodically retrains the learning model based on feedback and emotional information, improving the accuracy and responsiveness of the system and providing better support to users.

[0248] Example 2

[0249] 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."

[0250] While there is a demand for efficient information provision using business-related text data and improved user satisfaction, conventional systems have had problems with not being able to properly search text data or generate answers, making it difficult to quickly and accurately resolve user questions.In addition, because they are unable to generate answers that take user emotions into consideration, it has been difficult to improve the system to fully reflect user feedback.

[0251] 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.

[0252] In this invention, the server includes means for receiving and storing text data related to each task, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing the user's emotion using emotion analysis technology, means for adjusting the answer based on the recognized emotion information, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the information. This enables the system to provide quick and accurate answers to user questions, realizes detailed responses that take user emotions into consideration, and makes it possible to improve the system by utilizing feedback.

[0253] "Text data" is a data format that represents document information such as business manuals, guidelines, and FAQs.

[0254] "Vectorization" is the process of converting text data into an array of numbers, making it easier for machine learning models to handle text data.

[0255] A "cosine similarity matrix" is a matrix obtained by calculating cosine similarity in order to evaluate the similarity between vectorized text data.

[0256] "Receiving a question" refers to the system receiving and processing an inquiry from a user.

[0257] "Generating an answer" refers to the process of providing highly relevant information to a user's question and answering the user's doubt.

[0258] "Emotion analysis technology" is a technology for recognizing and classifying user emotions from text data.

[0259] "Feedback" refers to evaluation information such as satisfaction and areas for improvement provided by users to the system.

[0260] "Retraining" is the process of retraining an existing learning model using newly collected data in order to improve it.

[0261] The following is a specific description of an embodiment of the present invention. This system provides business support using text data related to each business, and is also combined with an emotion engine that recognizes the user's emotions.

[0262] First, users upload text data such as work-related manuals, guidelines, and FAQs to the server. Users can upload text data using a web interface or a dedicated application. The uploaded data is saved in a specific directory on the server. This allows work-related information to be managed centrally and can be accessed by users at any time.

[0263] The server then reads the stored text data and performs preprocessing: it opens the files, analyzes the text content, and organizes multiple documents into a list. During this process, the text is cleaned and standardized, ensuring consistency for subsequent processing.

[0264] The server then vectorizes the preprocessed text data using a machine learning model. This process uses natural language processing models such as Word2Vec and BERT. Based on the vectorized data, the server calculates a cosine similarity matrix and generates a learning model. This learning model allows the similarity between text data to be evaluated.

[0265] When a user enters a question into the ElderGPT system using a client device, the question is sent from the client device to the server via API. For example, a question like "How do I create a project plan?" is a good example. The server vectorizes the received question and performs a cosine similarity calculation. Based on the calculation results, the server searches for the most relevant document and generates its content as an answer. This generated answer is sent back to the client device and displayed to the user.

[0266] The server also uses an emotion engine to recognize the emotion of the user's question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input and classify them as positive, negative, or neutral. Based on the recognized emotion information, the server adjusts the answer. For example, if the server recognizes that the user is in trouble, it will provide a more specific and detailed answer.

[0267] Feedback can also be provided. Users can provide feedback on the displayed answers, such as their satisfaction level and areas for improvement. Feedback information is sent from the client device to the server, which collects and stores it to help improve the learning model in the future. The collected feedback information is used to periodically retrain the model.

[0268] Finally, the server periodically retrains the model based on feedback and emotion information. This improves the system's accuracy and responsiveness, enabling it to provide better support to users. For example, if a user asks, "How do I create a project plan?", the server searches for relevant documents and provides them. At the same time, it recognizes emotions from the user's input and determines whether a positive response is required. In this way, a system is realized that quickly and appropriately provides the information users are looking for.

[0269] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0270] Step 1:

[0271] A user uses a client terminal to upload text data related to business (such as manuals, guidelines, and FAQs). Specifically, the user accesses a web interface or a dedicated application, selects the text data using the file selection button, and clicks the upload button.

[0272] Input: Text file about business

[0273] Output: Text file saved on the server (specific directory)

[0274] Step 2:

[0275] The server reads the uploaded text data and performs preprocessing: it opens the file, extracts the text content, removes unnecessary spaces and special characters, and standardizes the text data (for example, standardizing uppercase and lowercase letters).

[0276] Input: A text file stored on the server

[0277] Output: Preprocessed text data in list format

[0278] Step 3:

[0279] The server vectorizes the preprocessed text data. Specifically, the server uses a natural language processing model such as Word2Vec or BERT to convert the text data into an array of numbers (vectors). The server then uses the vectorized data to calculate a cosine similarity matrix and generate a learning model.

[0280] Input: Preprocessed text data in list format

[0281] Output: Vectorized text data, cosine similarity matrix, and training model

[0282] Step 4:

[0283] The user enters a question into the ElderGPT system using a client terminal and presses the send button. The question may include how to create a project plan.

[0284] Input: Question text (prompt text)

[0285] Output: Question text sent to the server

[0286] Step 5:

[0287] The server vectorizes the received query and performs cosine similarity calculations. Based on the results, the server searches for the most relevant documents and generates their contents as the answer.

[0288] Input: Question text sent to the server, learning model

[0289] Output: The generated answer

[0290] Step 6:

[0291] The server uses an emotion engine to recognize the sentiment of the user's questions and answers, and uses text analysis technology to analyze the sentiment of the user's input, such as positive, negative, or neutral.

[0292] Input: Question text, generated answer

[0293] Output: Recognized emotion information

[0294] Step 7:

[0295] The server adjusts the answer based on the recognized emotional information. Specifically, if the server recognizes that the user is in trouble, it provides a more specific and detailed answer. This adjusted answer is then sent back to the client device.

[0296] Input: Recognized emotion information, generated answer

[0297] Output: Adjusted answer

[0298] Step 8:

[0299] The terminal displays the adjusted answers received from the server to the user, who can review the answers on the screen and ask further questions if necessary.

[0300] Input: Adjusted answer

[0301] Output: The answer displayed to the user

[0302] Step 9:

[0303] The user provides feedback on the displayed answer, such as satisfaction and areas for improvement. The user enters their feedback using the rating button or comment form and clicks the submit button.

[0304] Input: Satisfaction and feedback on improvements

[0305] Output: Feedback information sent to the server

[0306] Step 10:

[0307] The server collects feedback information and stores it to help improve the learning model in the future. The collected feedback information is used for periodic model retraining. The server stores this information in a database and uses it for the next retraining.

[0308] Input: Feedback information

[0309] Output: Saved feedback information

[0310] Step 11:

[0311] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[0312] Input: Stored feedback information, emotion information

[0313] Output: Retrained learning model

[0314] (Application example 2)

[0315] 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."

[0316] Conventional business support systems only provide information such as manuals, guidelines, and FAQs for each task, and lack the ability to recognize users' emotions and adjust responses accordingly, meaning users cannot receive appropriate support tailored to their individual situations and emotions. Furthermore, mechanisms for retraining models based on feedback and improving system performance are extremely limited, making sustained improvement difficult.

[0317] 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 receiving and storing text data such as manuals, guidelines, and FAQs related to each business, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing emotions from the user's input, means for adjusting the tone and level of detail of the answer based on the recognized emotion, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the feedback. This makes it possible to continuously improve the performance of the system while providing appropriate support according to the user's emotions.

[0318] "Text data such as manuals, guidelines, and FAQs for each task" refers to data in written form that contains information about the execution and procedures of tasks, and is used by employees and workers to refer to in order to carry out their tasks smoothly.

[0319] "Means for receiving and storing" refers to the mechanism for importing and storing data into storage such as a server.

[0320] "Vectorization" is the process of converting text data into an array of numbers (vectors), making it easier for machine learning models to process the text data.

[0321] A "cosine similarity matrix" is a matrix for calculating the similarity between different vectors, and the larger the value, the more similar the two vectors are.

[0322] "Means for receiving a question and generating the most relevant answer to that question" refers to technology for receiving a question posed by a user and providing the most relevant information or answer to that question.

[0323] "Means for recognizing emotions" refers to a mechanism for analyzing emotions from user input or voice and classifying them as, for example, positive, negative, neutral, etc.

[0324] "Means of adjusting the tone and level of detail of responses based on recognized emotions" refers to technology that appropriately changes the content of responses and how they are conveyed based on the results of emotion analysis.

[0325] "Means for providing feedback" refers to an interface that allows users to input and submit their impressions and opinions regarding answers, such as satisfaction and areas for improvement.

[0326] "Means for collecting feedback and retraining the learning model based on that information" refers to technology for collecting feedback information from users, systematically analyzing it, and reconstructing the machine learning model to improve the system's performance.

[0327] This invention is a system that supports business operations using text data such as manuals, guidelines, and FAQs related to each operation, and combines it with an emotion engine that recognizes user emotions. By installing this system on factory robots, it can efficiently support business operations within the factory.

[0328] The server has a means of receiving and storing text data such as manuals, guidelines, and FAQs for each task. This is done, for example, through a web interface or a dedicated application. The stored data is stored in built-in storage or cloud storage.

[0329] The server has a means to vectorize the received text data and calculate a cosine similarity matrix. For vectorization, Python's scikit-learn library and TfidfVectorizer are used. This converts the text data into an array of numerical values ​​(vectors), making it easier for machine learning models to process. The calculated cosine similarity matrix is ​​used to evaluate the similarity between different vectors.

[0330] A user inputs a question using a client device (e.g., a voice input device or a tablet). The question is sent to the server and processed by a means to generate the most relevant answer to the question. The server vectorizes the question, performs a cosine similarity calculation, searches for the most relevant text data, and generates the content as the answer.

[0331] The server has a means to recognize emotions from user input. It uses the TextBlob library as an emotion engine to analyze emotions from the user's input text and voice. The analysis results are classified as positive, negative, neutral, etc.

[0332] Based on the results of emotion recognition, the server has the means to adjust the tone and level of detail of the response, for example, providing gentler language and more detailed explanations if a negative emotion is recognized.

[0333] The system also provides a means for users to provide feedback on their answers. Feedback is collected by users entering their level of satisfaction and areas for improvement using a client terminal and sending the feedback to the server.

[0334] The server has the means to retrain the learning model based on the collected feedback. By periodically retraining the model, the performance of the system can be continuously improved.

[0335] Specific examples

[0336] For example, if a worker asks, "How do I deal with machine error message 1234?", the robot will search for relevant manual information and provide the most appropriate information. Furthermore, it will analyze the worker's emotions from their tone of voice and explain things more carefully if they are confused.

[0337] Prompt Sentence Examples

[0338] "For factory robots to assist workers in their work, please create a program that provides relevant manual information in response to worker questions and recognizes emotions to provide appropriate responses. For example, in response to the question, 'How do I deal with machine error message 1234?', the program should provide the most appropriate information from the machine's troubleshooting guide, adjusting the tone of the response depending on the worker's emotions (positive, negative, neutral)."

[0339] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0340] Step 1:

[0341] Users upload text data such as manuals, guidelines, and FAQs for each business to the server. The uploaded files are saved in a specific directory on the server. The input here is the text data uploaded by the user, and the output is the text data saved on the server.

[0342] Step 2:

[0343] The server reads the saved file and captures the text data, which is extracted and compiled into a list. The input is the text file saved on the server, and the output is the text data in list format.

[0344] Step 3:

[0345] The server vectorizes the text data and calculates the cosine similarity matrix. Here, the text data is converted into a numerical array using Python's TfidfVectorizer. The input is text data in list format, and the output is the vectorized text data and the cosine similarity matrix.

[0346] Step 4:

[0347] A user inputs a question using a client terminal. The input question is sent to the server via API. The input is the user's question text, and the output is the question sent to the server.

[0348] Step 5:

[0349] The server vectorizes the question text and performs cosine similarity calculations. Based on the results of the calculations, it searches for the most relevant documents and generates their contents as an answer. The input is the user's question text and existing text data, and the output is the generated answer.

[0350] Step 6:

[0351] The server uses an emotion engine to recognize the sentiment of the user's question and answer. It uses the TextBlob library to perform sentiment analysis. The input is the user's question text, and the output is the sentiment classification result (positive, negative, neutral).

[0352] Step 7:

[0353] The server adjusts the tone and level of detail of the response based on the recognized emotion. For example, if a negative emotion is recognized, it provides a more detailed and polite explanation. The input is the emotion classification result and the generated response, and the output is the adjusted response.

[0354] Step 8:

[0355] The user provides feedback on the answer. The feedback is input using a client terminal and sent to the server. The input is the user's feedback, and the output is the feedback information stored on the server.

[0356] Step 9:

[0357] The server retrains the learning model based on the collected feedback information. It periodically analyzes the feedback information and updates the model. The input is the feedback information, and the output is an improved learning model.

[0358] Step 10:

[0359] The retrained model is used to improve future question answering and emotion recognition accuracy. The inputs are new questions and existing datasets, and the output is improved system responses.

[0360] 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.

[0361] 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.

[0362] 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.

[0363] [Second embodiment]

[0364] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0365] 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.

[0366] 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).

[0367] 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.

[0368] 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.

[0369] 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).

[0370] 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.

[0371] 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.

[0372] 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.

[0373] 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.

[0374] 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.

[0375] 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."

[0376] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention provides business support using text data such as manuals, guidelines, and FAQs relating to each business.

[0377] 1. Uploading and saving data

[0378] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are stored in a specific directory on the server. This allows necessary work-related information to be managed centrally and easily accessible.

[0379] 2. Loading and Preprocessing Data

[0380] The server reads the saved file as text data and combines multiple sentences into a single list, allowing for consistent handling of the data in subsequent processing.

[0381] 3. Vectorization and Model Generation

[0382] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers. A cosine similarity matrix is ​​calculated based on the vectorized data, and a learning model is generated using this. This makes it possible to evaluate the similarity between text data.

[0383] 4. Question and Answer Processing

[0384] A user inputs a question in natural language to the Elder GPT system using a client device. For example, a question like, "Please tell me the format of the financial report." This question is sent from the device to the server via an API.

[0385] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained, the server searches for the most relevant document and generates its content as the answer.

[0386] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0387] For example, if a user asks, "What is the format of a financial report?", the server will search for the most relevant guidelines or manuals related to financial reports from text data that has been vectorized and learned in advance, and provide the content to the user, allowing the user to obtain accurate information quickly.

[0388] 5. Providing feedback and retraining

[0389] The user inputs feedback such as satisfaction with the answer provided, points for improvement, etc. For example, feedback such as "satisfied" or "insufficient information" is applicable.

[0390] The terminal transmits the input feedback information to the server.

[0391] The server collects the submitted feedback and stores it for use in improving the learning model in the future. The model is periodically retrained with data that reflects the feedback information, further improving the accuracy and responsiveness of the system.

[0392] Through the above process, the Elder GPT system provides a support environment where employees can easily ask questions at any time, improving work efficiency and reducing employee stress.

[0393] The processing flow will be explained below.

[0394] Step 1:

[0395] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a web interface or a dedicated application.

[0396] Step 2:

[0397] The server receives the uploaded files and stores them in the appropriate directory, allowing necessary business-related information to be managed centrally.

[0398] Step 3:

[0399] The server reads the saved files, treats them as text data, gets the contents of each file, and compiles them into a list.

[0400] Step 4:

[0401] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This vectorization makes it easier for machine learning models to handle text data.

[0402] Step 5:

[0403] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which makes it possible to evaluate the similarity between text data.

[0404] Step 6:

[0405] A user inputs a question in natural language to the Elder GPT system using a client terminal, and the input question is sent from the terminal to the server via an API.

[0406] Step 7:

[0407] The server vectorizes the received question text and performs cosine similarity calculations. Based on the similarity score, it searches for the most relevant documents.

[0408] Step 8:

[0409] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[0410] Step 9:

[0411] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0412] Step 10:

[0413] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[0414] Step 11:

[0415] The server collects the submitted feedback and stores it for future use in improving the learning model.

[0416] Step 12:

[0417] The server periodically retrains the model based on feedback information, improving the accuracy and responsiveness of the system and providing better support.

[0418] Example 1

[0419] 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."

[0420] Conventional systems have difficulty in centrally managing and effectively utilizing business document data. Furthermore, they lack mechanisms for providing quick and accurate answers to user questions, and lack the means to continuously improve the quality of answers. This makes it difficult to maximize business efficiency.

[0421] 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.

[0422] In this invention, the server includes a means for receiving and storing document data related to each business, a means for reading the received document data, listing it, and performing preprocessing, and a means for vectorizing the preprocessed document data and calculating a similarity matrix, thereby enabling unified management and effective use of document data.

[0423] The server further includes means for receiving question data, vectorizing the data, and generating the most relevant answer based on similarity, means for users to provide feedback on the answers, and means for collecting the feedback and retraining the generative model based on the feedback, thereby enabling the server to provide fast and accurate answers to user questions and continuously improve the quality of the answers by incorporating the feedback.

[0424] "Document data related to each business" is a general term for text data such as manuals, guidelines, FAQs, etc. related to the business operations of a company or organization.

[0425] "Means of receiving and storing" refers to the process by which the server receives document data uploaded by a user and stores the data in a specific directory or the like.

[0426] "Means of creating a list and pre-processing" refers to the process of reading the received data in list format and performing pre-processing such as cleaning the data and removing unnecessary information.

[0427] "Vectorization" refers to the process of converting document data into an array of numbers, or vectors, which makes it easier for computers to process.

[0428] The "means for calculating a similarity matrix" refers to a process for calculating the cosine similarity between vectorized data and expressing the result in a matrix format.

[0429] "Question data" refers to text data of the question that the user inputs to the system.

[0430] The "means for generating the most relevant answer based on similarity" refers to a process of vectorizing question data, identifying the most relevant document data using a pre-calculated similarity matrix, and generating an answer.

[0431] "Means for providing feedback" refers to the process by which users input their ratings and opinions on the answers provided into the system.

[0432] "Means for collecting feedback" refers to the process of storing feedback provided by users in a database or the like and aggregating it for use in subsequent processing.

[0433] A "generative model" is a model generated from document data using a specific learning algorithm, and is used to generate answers to questions.

[0434] "Retraining" refers to the process of updating an existing generative model based on collected feedback information to improve its accuracy and responsiveness.

[0435] This invention is a system that efficiently manages document data related to each task in a business support system and can provide quick and accurate answers to user questions. The system is implemented by clearly defining the roles of the server, terminal, and user. The hardware used includes a standard web server and client PC. The software used includes a natural language processing library (e.g., spaCy, Transformers), a data management library (e.g., pandas), and a library for similarity calculation (e.g., SciKit-Learn).

[0436] Data upload and management

[0437] Users upload document data related to their work (e.g., manuals, guidelines, FAQs, etc.) from their own devices to the server. Uploading is done through a web interface, and the selected file is saved in a specific directory on the server. For example, a user opens a browser, clicks the upload button, selects the desired document file, and executes the upload. At this time, the file's metadata (e.g., file name, upload date and time) is also saved.

[0438] Data loading and preprocessing

[0439] The server periodically scans the stored document data and reads newly uploaded files in text format. The read data is converted into a list format and then text cleaning (removal of whitespace and unnecessary characters) is performed. This process is performed using the Python pandas library. For example, the pandas read_csv method is used to read a CSV file and convert it to DataFrame format.

[0440] Vectorization and Similarity Calculation

[0441] After the text data has been read and preprocessed, it is converted into a numerical vector using a natural language processing library. This conversion is performed using libraries such as spaCy and Transformers. Next, a cosine similarity matrix is ​​calculated from the vectorized data to evaluate the relevance between the text data. The calculated similarity matrix is ​​then processed using the SciKit-Learn library.

[0442] Question and answer processing

[0443] A user inputs a question in natural language from their own device to the system. For example, a question might be, "What is the format of a financial report?" The question is entered through a web form and sent to the server. The server vectorizes the question and identifies the most relevant document data using a pre-calculated similarity matrix. An answer is generated based on the content and sent back to the device. The user can check the answer on their browser. As a specific example, if a user asks, "Please tell me about the marketing strategy for a new product," the server will search for relevant documents and generate an answer.

[0444] Feedback and retraining

[0445] Users provide feedback on the answers they provide, such as their satisfaction level or additional comments, through a web form. This feedback is sent from the device to the server and stored in a database. The server periodically retrains the model based on the feedback information, allowing the system to continuously improve its answer accuracy and user satisfaction.

[0446] Prompt Sentence Examples

[0447] Please tell me the format of the financial statement.

[0448] "Please tell us about your marketing strategy for the new product."

[0449] "How do I troubleshoot my product?"

[0450] This allows the system to not only centrally manage business-related information, but also provide quick and accurate answers to user questions. Furthermore, by improving the accuracy of the system through the feedback function, the quality of business support can be improved.

[0451] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0452] Step 1: Upload and save your data

[0453] A user uploads document data related to work by, for example, clicking the upload button on the browser and selecting the specified file.

[0454] Input: A document file selected by the user

[0455] The server receives the uploaded file and stores it in a specific directory (e.g., / data / uploads), along with metadata (file name, upload date and time).

[0456] Output: The saved file and its metadata

[0457] Step 2: Loading and preprocessing data

[0458] The server periodically scans a specific directory and reads newly uploaded document files, converts them as text files into a DataFrame, and then organizes them into a list.

[0459] Input: Saved document file

[0460] The server cleans the data, removing unnecessary whitespace and special characters, using the read_csv method from the pandas library.

[0461] Output: A list of preprocessed text data

[0462] Step 3: Vectorization and similarity matrix calculation

[0463] The server vectorizes the preprocessed text data and converts it into a numerical vector using a natural language processing library (e.g., spaCy, Transformers).

[0464] Input: A list of preprocessed text data

[0465] The server calculates the cosine similarity matrix based on the vectorized data. It uses the SciKit-Learn library to calculate the similarity between documents.

[0466] Output: Cosine similarity matrix

[0467] Step 4: Receiving questions and generating answers

[0468] The user inputs a question to the system from the client terminal, for example, a prompt such as "Please tell me the format of the financial statement."

[0469] Input: User question data

[0470] The device sends the question to the server via an API.

[0471] The server vectorizes the question, uses a pre-computed similarity matrix to identify the most relevant documents, and generates an answer.

[0472] Output: Generated response data

[0473] Step 5: View your answers

[0474] The server transmits the generated response data to the terminal.

[0475] Input: Generated response data

[0476] The device displays the received response to the user, which appears as text or a link in the browser.

[0477] Output: The answer that is displayed to the user

[0478] Step 6: Provide and collect feedback

[0479] Users provide feedback on the answers provided, including their satisfaction and any additional comments, by filling out a form.

[0480] Input: User feedback data

[0481] The device sends feedback to the server via an API.

[0482] The server collects the feedback and stores it in a database.

[0483] Output: Feedback data collected and stored

[0484] Step 7: Retrain with feedback

[0485] The server periodically retrains the generative model based on the collected feedback data, adding new data and recalculating the cosine similarity matrix to update the model.

[0486] Input: Collected feedback data

[0487] The server stores the updated generative model.

[0488] Output: Updated generative model

[0489] (Application example 1)

[0490] 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."

[0491] Picking work at logistics centers requires accuracy and efficiency, so it is important to have an environment where staff can quickly obtain the information they need. However, traditional methods often require staff to refer to paper-based manuals and guidelines, which is time-consuming and laborious. In addition, staff have limited means of obtaining immediate answers to their questions, which can lead to reduced work efficiency. Furthermore, there is a lack of mechanisms for effectively utilizing feedback to improve the quality of the answers provided. These issues need to be resolved to improve the efficiency and accuracy of picking work.

[0492] 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.

[0493] In this invention, the server includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer, thereby enabling pickers to quickly obtain necessary information and improve system performance based on the feedback.

[0494] A "manual" is a document that describes the procedures and methods for performing a specific task or work.

[0495] "Guidelines" are documents that provide basic guidelines and standards to be followed in business operations and procedures.

[0496] A "FAQ" is a document that compiles frequently asked questions and their answers.

[0497] "Vectorization" is the process of converting text data into an array of numbers.

[0498] A "cosine similarity matrix" is a matrix used to calculate the similarity between different vectors.

[0499] The "means for receiving a question" is a mechanism by which the system receives a question from a user.

[0500] The "means for generating an answer" is a process for generating the most appropriate answer to a question received from a user.

[0501] The "means for providing feedback" is a mechanism for users to input their opinions, such as ratings and suggestions for improvement, regarding the answers provided.

[0502] "Means for collecting feedback" refers to the process by which the system collects feedback information provided by users.

[0503] The "retraining method" is the process of re-learning the learning model to improve its performance based on the collected feedback information.

[0504] "Smart glasses" are a wearable eyeglass-type device equipped with a display that can display information and accept voice input.

[0505] A "means for asking questions in natural language" is a mechanism that allows users to ask questions using natural language, either spoken or written.

[0506] The "display means" is a method for showing the generated answers and related information to the user in real time.

[0507] The present invention comprises a system for supporting picking operations in a logistics center. The system includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer.

[0508] First, the distribution center manager uploads and stores text data such as manuals, guidelines, and FAQs for each operation on the server. The server then reads the stored text data and manages it centrally in list format.

[0509] The server then vectorizes the text data and calculates a cosine similarity matrix, which can improve the accuracy of generating answers to questions.

[0510] When a user wears smart glasses and speaks a question in natural language, the smart glasses send the question to the server via API. The server vectorizes the received question, performs cosine similarity calculations, and generates an answer based on the most relevant documents. The answer is displayed in real time on the smart glasses' display, allowing the user to immediately check the information they need.

[0511] Users can also provide feedback on the answers they provide. The feedback information is sent to the server and stored. The server uses this feedback information to retrain the learning model and improve the system's performance.

[0512] The hardware required to realize this system is smart glasses (e.g., Google Glass, Vuzix Blade) and a server. The software used includes Python, scikit-learn, and spaCy for vectorizing text data, Flask and FastAPI for the server-side API, AWS S3 and Google Cloud Storage for cloud storage, and OpenAI GPT and Hugging Face Transformers for the question-answering model.

[0513] For example, when a picker at a logistics center asks through smart glasses, "What's the next item to pick?", the server vectorizes the question and generates an answer based on relevant documents from the stored data. This answer is displayed on the smart glasses in real time, allowing the picker to quickly proceed with the next task.

[0514] An example of a prompt sentence is, "What is the next item to pick?" In this way, the present invention provides a system that effectively supports picking operations in logistics centers and significantly improves work efficiency and accuracy.

[0515] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0516] Step 1:

[0517] The server receives and stores text data such as manuals, guidelines, and FAQs for each operation uploaded by the logistics center administrator. The input is the text data uploaded by the administrator, and the output is a file stored on the server. This process allows necessary operation-related information to be centrally managed on the server and easily accessible.

[0518] Step 2:

[0519] The server reads the saved text data and compiles multiple sentences into a single list. The input is the text data saved in the server, and the output is the text data compiled into a single list. This process allows the data to be handled consistently in subsequent processes.

[0520] Step 3:

[0521] The server vectorizes the text data compiled in list format and calculates the cosine similarity matrix. The input is text data in list format, and the output is the vectorized data and the cosine similarity matrix. This process makes it possible to evaluate the similarity between text data.

[0522] Step 4:

[0523] The user uses the smart glasses to input a question in natural language by voice. The input is the user's natural language question, and the output is a question recorded by the smart glasses as voice data. This process allows the user to immediately query the system for the information they need.

[0524] Step 5:

[0525] The device converts the question recorded as voice data into text data and sends it to the server via API. The input is the user's question recorded as voice data, and the output is the question as text data sent to the server. Through this process, the server receives the user's question.

[0526] Step 6:

[0527] The server vectorizes the received question and performs a cosine similarity calculation. The input is the question as text data, and the output is the most relevant documents based on the similarity score. Through this process, the server generates an appropriate answer.

[0528] Step 7:

[0529] The server sends the generated answers to the smart glasses via API and displays them to the user in real time using the display means. The input is the answer as the most relevant document, and the output is the answer displayed on the smart glasses. This process allows the user to quickly obtain information.

[0530] Step 8:

[0531] The user provides feedback on the displayed answers. The input is the displayed answer and the user's evaluation of it, and the output is feedback information sent to the server. This process allows the user's opinion to be reflected in the system.

[0532] Step 9:

[0533] The server retrains the learning model based on the collected feedback information. The input is the feedback information, and the output is an improved learning model. This process improves the system's performance and increases the accuracy of the next question answer.

[0534] 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.

[0535] The following is a detailed description of an embodiment of the present invention. The present invention is a system that provides business support using text data such as manuals, guidelines, and FAQs related to each business, and also combines it with an emotion engine that recognizes user emotions.

[0536] 1. Uploading and saving data

[0537] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are uploaded via a web interface or a dedicated application and are saved in a specific directory on the server. This allows work-related information to be managed centrally and be accessible at any time.

[0538] 2. Loading and Preprocessing Data

[0539] The server reads the saved file as text data and organizes multiple sentences into a list, allowing the data to be handled consistently in subsequent processing.

[0540] 3. Vectorization and Model Generation

[0541] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers, which makes it easier for machine learning models to handle text data. A cosine similarity matrix is ​​then calculated based on the vectorized data to generate a learning model. This learning model makes it possible to evaluate the similarity between text data.

[0542] 4. Question and Answer Processing

[0543] A user inputs a question in natural language to the ElderGPT system using a client device. For example, a question like, "How do I create a project plan?" The input question is sent from the device to the server via an API.

[0544] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, it searches for the most relevant document and generates an answer from that document. The generated answer is then sent back to the device.

[0545] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0546] 5. Emotion Recognition with Emotion Engine

[0547] The server uses an emotion engine to analyze the user's emotions regarding questions and answers. The emotion engine uses text analysis technology to recognize emotions from the user's input. For example, it can classify emotions as positive, negative, neutral, etc.

[0548] 6. Adjusting responses based on emotions

[0549] The server generates a response appropriate to the user based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a specific and detailed response.

[0550] 7. Providing and Saving Feedback

[0551] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[0552] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to periodically retrain the model.

[0553] 8. Retraining the model

[0554] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[0555] For example, if a user asks, "How do I create a project plan?", the server will search for relevant documents and provide them. At the same time, it will recognize emotions from the user's input and determine whether a positive response is required. In this way, a system can be realized that quickly and appropriately provides the information the user is looking for.

[0556] Through the above process, the Elder GPT system provides a support environment that can respond to users while recognizing their emotions 24 hours a day, 365 days a year, thereby improving work efficiency and reducing employee stress.

[0557] The processing flow will be explained below.

[0558] Step 1:

[0559] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a dedicated web interface or application.

[0560] Step 2:

[0561] The server receives the uploaded files and stores them in the specified directory, allowing business-related information to be managed centrally.

[0562] Step 3:

[0563] The server reads the saved files, identifies them as text data, and combines the contents of multiple files into a single list.

[0564] Step 4:

[0565] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This conversion makes it easier for machine learning models to handle text data.

[0566] Step 5:

[0567] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which is used to evaluate the similarity between text data.

[0568] Step 6:

[0569] A user inputs a question in natural language into the Elder GPT system using a client device. For example, a question like, "How do I create a project plan?" The question is sent from the device to the server via an API.

[0570] Step 7:

[0571] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, the server searches for the most relevant document. For example, the server searches for the most relevant "Guidelines for creating project plans."

[0572] Step 8:

[0573] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[0574] Step 9:

[0575] The terminal displays the answer received from the server to the user, allowing the user to quickly resolve their question.

[0576] Step 10:

[0577] The server uses an emotion engine to analyze the user's sentiment towards the question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input, which can be classified as negative, positive, neutral, etc.

[0578] Step 11:

[0579] The server further tailors the answer based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a more detailed and helpful answer.

[0580] Step 12:

[0581] The user provides feedback on the displayed answer. The feedback includes satisfaction with the answer and areas for improvement. The device sends this feedback information to the server.

[0582] Step 13:

[0583] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to retrain the model.

[0584] Step 14:

[0585] The server periodically retrains the learning model based on feedback and emotional information, improving the accuracy and responsiveness of the system and providing better support to users.

[0586] Example 2

[0587] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0588] While there is a demand for efficient information provision using business-related text data and improved user satisfaction, conventional systems have had problems with not being able to properly search text data or generate answers, making it difficult to quickly and accurately resolve user questions.In addition, because they are unable to generate answers that take user emotions into consideration, it has been difficult to improve the system to fully reflect user feedback.

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

[0590] In this invention, the server includes means for receiving and storing text data related to each task, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing the user's emotion using emotion analysis technology, means for adjusting the answer based on the recognized emotion information, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the information. This enables the system to provide quick and accurate answers to user questions, realizes detailed responses that take user emotions into consideration, and makes it possible to improve the system by utilizing feedback.

[0591] "Text data" is a data format that represents document information such as business manuals, guidelines, and FAQs.

[0592] "Vectorization" is the process of converting text data into an array of numbers, making it easier for machine learning models to handle text data.

[0593] A "cosine similarity matrix" is a matrix obtained by calculating cosine similarity in order to evaluate the similarity between vectorized text data.

[0594] "Receiving a question" refers to the system receiving and processing an inquiry from a user.

[0595] "Generating an answer" refers to the process of providing highly relevant information to a user's question and answering the user's doubt.

[0596] "Emotion analysis technology" is a technology for recognizing and classifying user emotions from text data.

[0597] "Feedback" refers to evaluation information such as satisfaction and areas for improvement provided by users to the system.

[0598] "Retraining" is the process of retraining an existing learning model using newly collected data in order to improve it.

[0599] The following is a specific description of an embodiment of the present invention. This system provides business support using text data related to each business, and is also combined with an emotion engine that recognizes the user's emotions.

[0600] First, users upload text data such as work-related manuals, guidelines, and FAQs to the server. Users can upload text data using a web interface or a dedicated application. The uploaded data is saved in a specific directory on the server. This allows work-related information to be managed centrally and can be accessed by users at any time.

[0601] The server then reads the stored text data and performs preprocessing: it opens the files, analyzes the text content, and organizes multiple documents into a list. During this process, the text is cleaned and standardized, ensuring consistency for subsequent processing.

[0602] The server then vectorizes the preprocessed text data using a machine learning model. This process uses natural language processing models such as Word2Vec and BERT. Based on the vectorized data, the server calculates a cosine similarity matrix and generates a learning model. This learning model allows the similarity between text data to be evaluated.

[0603] When a user enters a question into the ElderGPT system using a client device, the question is sent from the client device to the server via API. For example, a question like "How do I create a project plan?" is a good example. The server vectorizes the received question and performs a cosine similarity calculation. Based on the calculation results, the server searches for the most relevant document and generates its content as an answer. This generated answer is sent back to the client device and displayed to the user.

[0604] The server also uses an emotion engine to recognize the emotion of the user's question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input and classify them as positive, negative, or neutral. Based on the recognized emotion information, the server adjusts the answer. For example, if the server recognizes that the user is in trouble, it will provide a more specific and detailed answer.

[0605] Feedback can also be provided. Users can provide feedback on the displayed answers, such as their satisfaction level and areas for improvement. Feedback information is sent from the client device to the server, which collects and stores it to help improve the learning model in the future. The collected feedback information is used to periodically retrain the model.

[0606] Finally, the server periodically retrains the model based on feedback and emotion information. This improves the system's accuracy and responsiveness, enabling it to provide better support to users. For example, if a user asks, "How do I create a project plan?", the server searches for relevant documents and provides them. At the same time, it recognizes emotions from the user's input and determines whether a positive response is required. In this way, a system is realized that quickly and appropriately provides the information users are looking for.

[0607] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0608] Step 1:

[0609] A user uses a client terminal to upload text data related to business (such as manuals, guidelines, and FAQs). Specifically, the user accesses a web interface or a dedicated application, selects the text data using the file selection button, and clicks the upload button.

[0610] Input: Text file about business

[0611] Output: Text file saved on the server (specific directory)

[0612] Step 2:

[0613] The server reads the uploaded text data and performs preprocessing: it opens the file, extracts the text content, removes unnecessary spaces and special characters, and standardizes the text data (for example, standardizing uppercase and lowercase letters).

[0614] Input: A text file stored on the server

[0615] Output: Preprocessed text data in list format

[0616] Step 3:

[0617] The server vectorizes the preprocessed text data. Specifically, the server uses a natural language processing model such as Word2Vec or BERT to convert the text data into an array of numbers (vectors). The server then uses the vectorized data to calculate a cosine similarity matrix and generate a learning model.

[0618] Input: Preprocessed text data in list format

[0619] Output: Vectorized text data, cosine similarity matrix, and training model

[0620] Step 4:

[0621] The user enters a question into the ElderGPT system using a client terminal and presses the send button. The question may include how to create a project plan.

[0622] Input: Question text (prompt text)

[0623] Output: Question text sent to the server

[0624] Step 5:

[0625] The server vectorizes the received query and performs cosine similarity calculations. Based on the results, the server searches for the most relevant documents and generates their contents as the answer.

[0626] Input: Question text sent to the server, learning model

[0627] Output: The generated answer

[0628] Step 6:

[0629] The server uses an emotion engine to recognize the sentiment of the user's questions and answers, and uses text analysis technology to analyze the sentiment of the user's input, such as positive, negative, or neutral.

[0630] Input: Question text, generated answer

[0631] Output: Recognized emotion information

[0632] Step 7:

[0633] The server adjusts the answer based on the recognized emotional information. Specifically, if the server recognizes that the user is in trouble, it provides a more specific and detailed answer. This adjusted answer is then sent back to the client device.

[0634] Input: Recognized emotion information, generated answer

[0635] Output: Adjusted answer

[0636] Step 8:

[0637] The terminal displays the adjusted answers received from the server to the user, who can review the answers on the screen and ask further questions if necessary.

[0638] Input: Adjusted answer

[0639] Output: The answer displayed to the user

[0640] Step 9:

[0641] The user provides feedback on the displayed answer, such as satisfaction and areas for improvement. The user enters their feedback using the rating button or comment form and clicks the submit button.

[0642] Input: Satisfaction and feedback on improvements

[0643] Output: Feedback information sent to the server

[0644] Step 10:

[0645] The server collects feedback information and stores it to help improve the learning model in the future. The collected feedback information is used for periodic model retraining. The server stores this information in a database and uses it for the next retraining.

[0646] Input: Feedback information

[0647] Output: Saved feedback information

[0648] Step 11:

[0649] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[0650] Input: Stored feedback information, emotion information

[0651] Output: Retrained learning model

[0652] (Application example 2)

[0653] 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."

[0654] Conventional business support systems only provide information such as manuals, guidelines, and FAQs for each task, and lack the ability to recognize users' emotions and adjust responses accordingly, meaning users cannot receive appropriate support tailored to their individual situations and emotions. Furthermore, mechanisms for retraining models based on feedback and improving system performance are extremely limited, making sustained improvement difficult.

[0655] 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 receiving and storing text data such as manuals, guidelines, and FAQs related to each business, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing emotions from the user's input, means for adjusting the tone and level of detail of the answer based on the recognized emotion, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the feedback. This makes it possible to continuously improve the performance of the system while providing appropriate support according to the user's emotions.

[0656] "Text data such as manuals, guidelines, and FAQs for each task" refers to data in written form that contains information about the execution and procedures of tasks, and is used by employees and workers to refer to in order to carry out their tasks smoothly.

[0657] "Means for receiving and storing" refers to the mechanism for importing and storing data into storage such as a server.

[0658] "Vectorization" is the process of converting text data into an array of numbers (vectors), making it easier for machine learning models to process the text data.

[0659] A "cosine similarity matrix" is a matrix for calculating the similarity between different vectors, and the larger the value, the more similar the two vectors are.

[0660] "Means for receiving a question and generating the most relevant answer to that question" refers to technology for receiving a question posed by a user and providing the most relevant information or answer to that question.

[0661] "Means for recognizing emotions" refers to a mechanism for analyzing emotions from user input or voice and classifying them as, for example, positive, negative, neutral, etc.

[0662] "Means of adjusting the tone and level of detail of responses based on recognized emotions" refers to technology that appropriately changes the content of responses and how they are conveyed based on the results of emotion analysis.

[0663] "Means for providing feedback" refers to an interface that allows users to input and submit their impressions and opinions regarding answers, such as satisfaction and areas for improvement.

[0664] "Means for collecting feedback and retraining the learning model based on that information" refers to technology for collecting feedback information from users, systematically analyzing it, and reconstructing the machine learning model to improve the system's performance.

[0665] This invention is a system that supports business operations using text data such as manuals, guidelines, and FAQs related to each operation, and combines it with an emotion engine that recognizes user emotions. By installing this system on factory robots, it can efficiently support business operations within the factory.

[0666] The server has a means of receiving and storing text data such as manuals, guidelines, and FAQs for each task. This is done, for example, through a web interface or a dedicated application. The stored data is stored in built-in storage or cloud storage.

[0667] The server has a means to vectorize the received text data and calculate a cosine similarity matrix. For vectorization, Python's scikit-learn library and TfidfVectorizer are used. This converts the text data into an array of numerical values ​​(vectors), making it easier for machine learning models to process. The calculated cosine similarity matrix is ​​used to evaluate the similarity between different vectors.

[0668] A user inputs a question using a client device (e.g., a voice input device or a tablet). The question is sent to the server and processed by a means to generate the most relevant answer to the question. The server vectorizes the question, performs a cosine similarity calculation, searches for the most relevant text data, and generates the content as the answer.

[0669] The server has a means to recognize emotions from user input. It uses the TextBlob library as an emotion engine to analyze emotions from the user's input text and voice. The analysis results are classified as positive, negative, neutral, etc.

[0670] Based on the results of emotion recognition, the server has the means to adjust the tone and level of detail of the response, for example, providing gentler language and more detailed explanations if a negative emotion is recognized.

[0671] The system also provides a means for users to provide feedback on their answers. Feedback is collected by users entering their level of satisfaction and areas for improvement using a client terminal and sending the feedback to the server.

[0672] The server has the means to retrain the learning model based on the collected feedback. By periodically retraining the model, the performance of the system can be continuously improved.

[0673] Specific examples

[0674] For example, if a worker asks, "How do I deal with machine error message 1234?", the robot will search for relevant manual information and provide the most appropriate information. Furthermore, it will analyze the worker's emotions from their tone of voice and explain things more carefully if they are confused.

[0675] Prompt Sentence Examples

[0676] "For factory robots to assist workers in their work, please create a program that provides relevant manual information in response to worker questions and recognizes emotions to provide appropriate responses. For example, in response to the question, 'How do I deal with machine error message 1234?', the program should provide the most appropriate information from the machine's troubleshooting guide, adjusting the tone of the response depending on the worker's emotions (positive, negative, neutral)."

[0677] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0678] Step 1:

[0679] Users upload text data such as manuals, guidelines, and FAQs for each business to the server. The uploaded files are saved in a specific directory on the server. The input here is the text data uploaded by the user, and the output is the text data saved on the server.

[0680] Step 2:

[0681] The server reads the saved file and captures the text data, which is extracted and compiled into a list. The input is the text file saved on the server, and the output is the text data in list format.

[0682] Step 3:

[0683] The server vectorizes the text data and calculates the cosine similarity matrix. Here, the text data is converted into a numerical array using Python's TfidfVectorizer. The input is text data in list format, and the output is the vectorized text data and the cosine similarity matrix.

[0684] Step 4:

[0685] A user inputs a question using a client terminal. The input question is sent to the server via API. The input is the user's question text, and the output is the question sent to the server.

[0686] Step 5:

[0687] The server vectorizes the question text and performs cosine similarity calculations. Based on the results of the calculations, it searches for the most relevant documents and generates their contents as an answer. The input is the user's question text and existing text data, and the output is the generated answer.

[0688] Step 6:

[0689] The server uses an emotion engine to recognize the sentiment of the user's question and answer. It uses the TextBlob library to perform sentiment analysis. The input is the user's question text, and the output is the sentiment classification result (positive, negative, neutral).

[0690] Step 7:

[0691] The server adjusts the tone and level of detail of the response based on the recognized emotion. For example, if a negative emotion is recognized, it provides a more detailed and polite explanation. The input is the emotion classification result and the generated response, and the output is the adjusted response.

[0692] Step 8:

[0693] The user provides feedback on the answer. The feedback is input using a client terminal and sent to the server. The input is the user's feedback, and the output is the feedback information stored on the server.

[0694] Step 9:

[0695] The server retrains the learning model based on the collected feedback information. It periodically analyzes the feedback information and updates the model. The input is the feedback information, and the output is an improved learning model.

[0696] Step 10:

[0697] The retrained model is used to improve future question answering and emotion recognition accuracy. The inputs are new questions and existing datasets, and the output is improved system responses.

[0698] 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.

[0699] 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.

[0700] 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.

[0701] [Third embodiment]

[0702] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0703] 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.

[0704] 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).

[0705] 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.

[0706] 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.

[0707] 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).

[0708] 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.

[0709] 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.

[0710] 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.

[0711] 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.

[0712] 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.

[0713] 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."

[0714] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention provides business support using text data such as manuals, guidelines, and FAQs relating to each business.

[0715] 1. Uploading and saving data

[0716] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are stored in a specific directory on the server. This allows necessary work-related information to be managed centrally and easily accessible.

[0717] 2. Loading and Preprocessing Data

[0718] The server reads the saved file as text data and combines multiple sentences into a single list, allowing for consistent handling of the data in subsequent processing.

[0719] 3. Vectorization and Model Generation

[0720] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers. A cosine similarity matrix is ​​calculated based on the vectorized data, and a learning model is generated using this. This makes it possible to evaluate the similarity between text data.

[0721] 4. Question and Answer Processing

[0722] A user inputs a question in natural language to the Elder GPT system using a client device. For example, a question like, "Please tell me the format of the financial report." This question is sent from the device to the server via an API.

[0723] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained, the server searches for the most relevant document and generates its content as the answer.

[0724] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0725] For example, if a user asks, "What is the format of a financial report?", the server will search for the most relevant guidelines or manuals related to financial reports from text data that has been vectorized and learned in advance, and provide the content to the user, allowing the user to obtain accurate information quickly.

[0726] 5. Providing feedback and retraining

[0727] The user inputs feedback such as satisfaction with the answer provided, points for improvement, etc. For example, feedback such as "satisfied" or "insufficient information" is applicable.

[0728] The terminal transmits the input feedback information to the server.

[0729] The server collects the submitted feedback and stores it for use in improving the learning model in the future. The model is periodically retrained with data that reflects the feedback information, further improving the accuracy and responsiveness of the system.

[0730] Through the above process, the Elder GPT system provides a support environment where employees can easily ask questions at any time, improving work efficiency and reducing employee stress.

[0731] The processing flow will be explained below.

[0732] Step 1:

[0733] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a web interface or a dedicated application.

[0734] Step 2:

[0735] The server receives the uploaded files and stores them in the appropriate directory, allowing necessary business-related information to be managed centrally.

[0736] Step 3:

[0737] The server reads the saved files, treats them as text data, gets the contents of each file, and compiles them into a list.

[0738] Step 4:

[0739] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This vectorization makes it easier for machine learning models to handle text data.

[0740] Step 5:

[0741] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which makes it possible to evaluate the similarity between text data.

[0742] Step 6:

[0743] A user inputs a question in natural language to the Elder GPT system using a client terminal, and the input question is sent from the terminal to the server via an API.

[0744] Step 7:

[0745] The server vectorizes the received question text and performs cosine similarity calculations. Based on the similarity score, it searches for the most relevant documents.

[0746] Step 8:

[0747] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[0748] Step 9:

[0749] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0750] Step 10:

[0751] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[0752] Step 11:

[0753] The server collects the submitted feedback and stores it for future use in improving the learning model.

[0754] Step 12:

[0755] The server periodically retrains the model based on feedback information, improving the accuracy and responsiveness of the system and providing better support.

[0756] Example 1

[0757] 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."

[0758] Conventional systems have difficulty in centrally managing and effectively utilizing business document data. Furthermore, they lack mechanisms for providing quick and accurate answers to user questions, and lack the means to continuously improve the quality of answers. This makes it difficult to maximize business efficiency.

[0759] 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.

[0760] In this invention, the server includes a means for receiving and storing document data related to each business, a means for reading the received document data, listing it, and performing preprocessing, and a means for vectorizing the preprocessed document data and calculating a similarity matrix, thereby enabling unified management and effective use of document data.

[0761] The server further includes means for receiving question data, vectorizing the data, and generating the most relevant answer based on similarity, means for users to provide feedback on the answers, and means for collecting the feedback and retraining the generative model based on the feedback, thereby enabling the server to provide fast and accurate answers to user questions and continuously improve the quality of the answers by incorporating the feedback.

[0762] "Document data related to each business" is a general term for text data such as manuals, guidelines, FAQs, etc. related to the business operations of a company or organization.

[0763] "Means of receiving and storing" refers to the process by which the server receives document data uploaded by a user and stores the data in a specific directory or the like.

[0764] "Means of creating a list and pre-processing" refers to the process of reading the received data in list format and performing pre-processing such as cleaning the data and removing unnecessary information.

[0765] "Vectorization" refers to the process of converting document data into an array of numbers, or vectors, which makes it easier for computers to process.

[0766] The "means for calculating a similarity matrix" refers to a process for calculating the cosine similarity between vectorized data and expressing the result in a matrix format.

[0767] "Question data" refers to text data of the question that the user inputs to the system.

[0768] The "means for generating the most relevant answer based on similarity" refers to a process of vectorizing question data, identifying the most relevant document data using a pre-calculated similarity matrix, and generating an answer.

[0769] "Means for providing feedback" refers to the process by which users input their ratings and opinions on the answers provided into the system.

[0770] "Means for collecting feedback" refers to the process of storing feedback provided by users in a database or the like and aggregating it for use in subsequent processing.

[0771] A "generative model" is a model generated from document data using a specific learning algorithm, and is used to generate answers to questions.

[0772] "Retraining" refers to the process of updating an existing generative model based on collected feedback information to improve its accuracy and responsiveness.

[0773] This invention is a system that efficiently manages document data related to each task in a business support system and can provide quick and accurate answers to user questions. The system is implemented by clearly defining the roles of the server, terminal, and user. The hardware used includes a standard web server and client PC. The software used includes a natural language processing library (e.g., spaCy, Transformers), a data management library (e.g., pandas), and a library for similarity calculation (e.g., SciKit-Learn).

[0774] Data upload and management

[0775] Users upload document data related to their work (e.g., manuals, guidelines, FAQs, etc.) from their own devices to the server. Uploading is done through a web interface, and the selected file is saved in a specific directory on the server. For example, a user opens a browser, clicks the upload button, selects the desired document file, and executes the upload. At this time, the file's metadata (e.g., file name, upload date and time) is also saved.

[0776] Data loading and preprocessing

[0777] The server periodically scans the stored document data and reads newly uploaded files in text format. The read data is converted into a list format and then text cleaning (removal of whitespace and unnecessary characters) is performed. This process is performed using the Python pandas library. For example, the pandas read_csv method is used to read a CSV file and convert it to DataFrame format.

[0778] Vectorization and Similarity Calculation

[0779] After the text data has been read and preprocessed, it is converted into a numerical vector using a natural language processing library. This conversion is performed using libraries such as spaCy and Transformers. Next, a cosine similarity matrix is ​​calculated from the vectorized data to evaluate the relevance between the text data. The calculated similarity matrix is ​​then processed using the SciKit-Learn library.

[0780] Question and answer processing

[0781] A user inputs a question in natural language from their own device to the system. For example, a question might be, "What is the format of a financial report?" The question is entered through a web form and sent to the server. The server vectorizes the question and identifies the most relevant document data using a pre-calculated similarity matrix. An answer is generated based on the content and sent back to the device. The user can check the answer on their browser. As a specific example, if a user asks, "Please tell me about the marketing strategy for a new product," the server will search for relevant documents and generate an answer.

[0782] Feedback and retraining

[0783] Users provide feedback on the answers they provide, such as their satisfaction level or additional comments, through a web form. This feedback is sent from the device to the server and stored in a database. The server periodically retrains the model based on the feedback information, allowing the system to continuously improve its answer accuracy and user satisfaction.

[0784] Prompt Sentence Examples

[0785] Please tell me the format of the financial statement.

[0786] "Please tell us about your marketing strategy for the new product."

[0787] "How do I troubleshoot my product?"

[0788] This allows the system to not only centrally manage business-related information, but also provide quick and accurate answers to user questions. Furthermore, by improving the accuracy of the system through the feedback function, the quality of business support can be improved.

[0789] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0790] Step 1: Upload and save your data

[0791] A user uploads document data related to work by, for example, clicking the upload button on the browser and selecting the specified file.

[0792] Input: A document file selected by the user

[0793] The server receives the uploaded file and stores it in a specific directory (e.g., / data / uploads), along with metadata (file name, upload date and time).

[0794] Output: The saved file and its metadata

[0795] Step 2: Loading and preprocessing data

[0796] The server periodically scans a specific directory and reads newly uploaded document files, converts them as text files into a DataFrame, and then organizes them into a list.

[0797] Input: Saved document file

[0798] The server cleans the data, removing unnecessary whitespace and special characters, using the read_csv method from the pandas library.

[0799] Output: A list of preprocessed text data

[0800] Step 3: Vectorization and similarity matrix calculation

[0801] The server vectorizes the preprocessed text data and converts it into a numerical vector using a natural language processing library (e.g., spaCy, Transformers).

[0802] Input: A list of preprocessed text data

[0803] The server calculates the cosine similarity matrix based on the vectorized data. It uses the SciKit-Learn library to calculate the similarity between documents.

[0804] Output: Cosine similarity matrix

[0805] Step 4: Receiving questions and generating answers

[0806] The user inputs a question to the system from the client terminal, for example, a prompt such as "Please tell me the format of the financial statement."

[0807] Input: User question data

[0808] The device sends the question to the server via an API.

[0809] The server vectorizes the question, uses a pre-computed similarity matrix to identify the most relevant documents, and generates an answer.

[0810] Output: Generated response data

[0811] Step 5: View your answers

[0812] The server transmits the generated response data to the terminal.

[0813] Input: Generated response data

[0814] The device displays the received response to the user, which appears as text or a link in the browser.

[0815] Output: The answer that is displayed to the user

[0816] Step 6: Provide and collect feedback

[0817] Users provide feedback on the answers provided, including their satisfaction and any additional comments, by filling out a form.

[0818] Input: User feedback data

[0819] The device sends feedback to the server via an API.

[0820] The server collects the feedback and stores it in a database.

[0821] Output: Feedback data collected and stored

[0822] Step 7: Retrain with feedback

[0823] The server periodically retrains the generative model based on the collected feedback data, adding new data and recalculating the cosine similarity matrix to update the model.

[0824] Input: Collected feedback data

[0825] The server stores the updated generative model.

[0826] Output: Updated generative model

[0827] (Application example 1)

[0828] 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."

[0829] Picking work at logistics centers requires accuracy and efficiency, so it is important to have an environment where staff can quickly obtain the information they need. However, traditional methods often require staff to refer to paper-based manuals and guidelines, which is time-consuming and laborious. In addition, staff have limited means of obtaining immediate answers to their questions, which can lead to reduced work efficiency. Furthermore, there is a lack of mechanisms for effectively utilizing feedback to improve the quality of the answers provided. These issues need to be resolved to improve the efficiency and accuracy of picking work.

[0830] 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.

[0831] In this invention, the server includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer, thereby enabling pickers to quickly obtain necessary information and improve system performance based on the feedback.

[0832] A "manual" is a document that describes the procedures and methods for performing a specific task or work.

[0833] "Guidelines" are documents that provide basic guidelines and standards to be followed in business operations and procedures.

[0834] A "FAQ" is a document that compiles frequently asked questions and their answers.

[0835] "Vectorization" is the process of converting text data into an array of numbers.

[0836] A "cosine similarity matrix" is a matrix used to calculate the similarity between different vectors.

[0837] The "means for receiving a question" is a mechanism by which the system receives a question from a user.

[0838] The "means for generating an answer" is a process for generating the most appropriate answer to a question received from a user.

[0839] The "means for providing feedback" is a mechanism for users to input their opinions, such as ratings and suggestions for improvement, regarding the answers provided.

[0840] "Means for collecting feedback" refers to the process by which the system collects feedback information provided by users.

[0841] The "retraining method" is the process of re-learning the learning model to improve its performance based on the collected feedback information.

[0842] "Smart glasses" are a wearable eyeglass-type device equipped with a display that can display information and accept voice input.

[0843] A "means for asking questions in natural language" is a mechanism that allows users to ask questions using natural language, either spoken or written.

[0844] The "display means" is a method for showing the generated answers and related information to the user in real time.

[0845] The present invention comprises a system for supporting picking operations in a logistics center. The system includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer.

[0846] First, the distribution center manager uploads and stores text data such as manuals, guidelines, and FAQs for each operation on the server. The server then reads the stored text data and manages it centrally in list format.

[0847] The server then vectorizes the text data and calculates a cosine similarity matrix, which can improve the accuracy of generating answers to questions.

[0848] When a user wears smart glasses and speaks a question in natural language, the smart glasses send the question to the server via API. The server vectorizes the received question, performs cosine similarity calculations, and generates an answer based on the most relevant documents. The answer is displayed in real time on the smart glasses' display, allowing the user to immediately check the information they need.

[0849] Users can also provide feedback on the answers they provide. The feedback information is sent to the server and stored. The server uses this feedback information to retrain the learning model and improve the system's performance.

[0850] The hardware required to realize this system is smart glasses (e.g., Google Glass, Vuzix Blade) and a server. The software used includes Python, scikit-learn, and spaCy for vectorizing text data, Flask and FastAPI for the server-side API, AWS S3 and Google Cloud Storage for cloud storage, and OpenAI GPT and Hugging Face Transformers for the question-answering model.

[0851] For example, when a picker at a logistics center asks through smart glasses, "What's the next item to pick?", the server vectorizes the question and generates an answer based on relevant documents from the stored data. This answer is displayed on the smart glasses in real time, allowing the picker to quickly proceed with the next task.

[0852] An example of a prompt sentence is, "What is the next item to pick?" In this way, the present invention provides a system that effectively supports picking operations in logistics centers and significantly improves work efficiency and accuracy.

[0853] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0854] Step 1:

[0855] The server receives and stores text data such as manuals, guidelines, and FAQs for each operation uploaded by the logistics center administrator. The input is the text data uploaded by the administrator, and the output is a file stored on the server. This process allows necessary operation-related information to be centrally managed on the server and easily accessible.

[0856] Step 2:

[0857] The server reads the saved text data and compiles multiple sentences into a single list. The input is the text data saved in the server, and the output is the text data compiled into a single list. This process allows the data to be handled consistently in subsequent processes.

[0858] Step 3:

[0859] The server vectorizes the text data compiled in list format and calculates the cosine similarity matrix. The input is text data in list format, and the output is the vectorized data and the cosine similarity matrix. This process makes it possible to evaluate the similarity between text data.

[0860] Step 4:

[0861] The user uses the smart glasses to input a question in natural language by voice. The input is the user's natural language question, and the output is a question recorded by the smart glasses as voice data. This process allows the user to immediately query the system for the information they need.

[0862] Step 5:

[0863] The device converts the question recorded as voice data into text data and sends it to the server via API. The input is the user's question recorded as voice data, and the output is the question as text data sent to the server. Through this process, the server receives the user's question.

[0864] Step 6:

[0865] The server vectorizes the received question and performs a cosine similarity calculation. The input is the question as text data, and the output is the most relevant documents based on the similarity score. Through this process, the server generates an appropriate answer.

[0866] Step 7:

[0867] The server sends the generated answers to the smart glasses via API and displays them to the user in real time using the display means. The input is the answer as the most relevant document, and the output is the answer displayed on the smart glasses. This process allows the user to quickly obtain information.

[0868] Step 8:

[0869] The user provides feedback on the displayed answers. The input is the displayed answer and the user's evaluation of it, and the output is feedback information sent to the server. This process allows the user's opinion to be reflected in the system.

[0870] Step 9:

[0871] The server retrains the learning model based on the collected feedback information. The input is the feedback information, and the output is an improved learning model. This process improves the system's performance and increases the accuracy of the next question answer.

[0872] 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.

[0873] The following is a detailed description of an embodiment of the present invention. The present invention is a system that provides business support using text data such as manuals, guidelines, and FAQs related to each business, and also combines it with an emotion engine that recognizes user emotions.

[0874] 1. Uploading and saving data

[0875] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are uploaded via a web interface or a dedicated application and are saved in a specific directory on the server. This allows work-related information to be managed centrally and be accessible at any time.

[0876] 2. Loading and Preprocessing Data

[0877] The server reads the saved file as text data and organizes multiple sentences into a list, allowing the data to be handled consistently in subsequent processing.

[0878] 3. Vectorization and Model Generation

[0879] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers, which makes it easier for machine learning models to handle text data. A cosine similarity matrix is ​​then calculated based on the vectorized data to generate a learning model. This learning model makes it possible to evaluate the similarity between text data.

[0880] 4. Question and Answer Processing

[0881] A user inputs a question in natural language to the ElderGPT system using a client device. For example, a question like, "How do I create a project plan?" The input question is sent from the device to the server via an API.

[0882] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, it searches for the most relevant document and generates an answer from that document. The generated answer is then sent back to the device.

[0883] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[0884] 5. Emotion Recognition with Emotion Engine

[0885] The server uses an emotion engine to analyze the user's emotions regarding questions and answers. The emotion engine uses text analysis technology to recognize emotions from the user's input. For example, it can classify emotions as positive, negative, neutral, etc.

[0886] 6. Adjusting responses based on emotions

[0887] The server generates a response appropriate to the user based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a specific and detailed response.

[0888] 7. Providing and Saving Feedback

[0889] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[0890] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to periodically retrain the model.

[0891] 8. Retraining the model

[0892] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[0893] For example, if a user asks, "How do I create a project plan?", the server will search for relevant documents and provide them. At the same time, it will recognize emotions from the user's input and determine whether a positive response is required. In this way, a system can be realized that quickly and appropriately provides the information the user is looking for.

[0894] Through the above process, the Elder GPT system provides a support environment that can respond to users while recognizing their emotions 24 hours a day, 365 days a year, thereby improving work efficiency and reducing employee stress.

[0895] The processing flow will be explained below.

[0896] Step 1:

[0897] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a dedicated web interface or application.

[0898] Step 2:

[0899] The server receives the uploaded files and stores them in the specified directory, allowing business-related information to be managed centrally.

[0900] Step 3:

[0901] The server reads the saved files, identifies them as text data, and combines the contents of multiple files into a single list.

[0902] Step 4:

[0903] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This conversion makes it easier for machine learning models to handle text data.

[0904] Step 5:

[0905] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which is used to evaluate the similarity between text data.

[0906] Step 6:

[0907] A user inputs a question in natural language into the Elder GPT system using a client device. For example, a question like, "How do I create a project plan?" The question is sent from the device to the server via an API.

[0908] Step 7:

[0909] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, the server searches for the most relevant document. For example, the server searches for the most relevant "Guidelines for creating project plans."

[0910] Step 8:

[0911] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[0912] Step 9:

[0913] The terminal displays the answer received from the server to the user, allowing the user to quickly resolve their question.

[0914] Step 10:

[0915] The server uses an emotion engine to analyze the user's sentiment towards the question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input, which can be classified as negative, positive, neutral, etc.

[0916] Step 11:

[0917] The server further tailors the answer based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a more detailed and helpful answer.

[0918] Step 12:

[0919] The user provides feedback on the displayed answer. The feedback includes satisfaction with the answer and areas for improvement. The device sends this feedback information to the server.

[0920] Step 13:

[0921] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to retrain the model.

[0922] Step 14:

[0923] The server periodically retrains the learning model based on feedback and emotional information, improving the accuracy and responsiveness of the system and providing better support to users.

[0924] Example 2

[0925] 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."

[0926] While there is a demand for efficient information provision using business-related text data and improved user satisfaction, conventional systems have had problems with not being able to properly search text data or generate answers, making it difficult to quickly and accurately resolve user questions.In addition, because they are unable to generate answers that take user emotions into consideration, it has been difficult to improve the system to fully reflect user feedback.

[0927] 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.

[0928] In this invention, the server includes means for receiving and storing text data related to each task, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing the user's emotion using emotion analysis technology, means for adjusting the answer based on the recognized emotion information, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the information. This enables the system to provide quick and accurate answers to user questions, realizes detailed responses that take user emotions into consideration, and makes it possible to improve the system by utilizing feedback.

[0929] "Text data" is a data format that represents document information such as business manuals, guidelines, and FAQs.

[0930] "Vectorization" is the process of converting text data into an array of numbers, making it easier for machine learning models to handle text data.

[0931] A "cosine similarity matrix" is a matrix obtained by calculating cosine similarity in order to evaluate the similarity between vectorized text data.

[0932] "Receiving a question" refers to the system receiving and processing an inquiry from a user.

[0933] "Generating an answer" refers to the process of providing highly relevant information to a user's question and answering the user's doubt.

[0934] "Emotion analysis technology" is a technology for recognizing and classifying user emotions from text data.

[0935] "Feedback" refers to evaluation information such as satisfaction and areas for improvement provided by users to the system.

[0936] "Retraining" is the process of retraining an existing learning model using newly collected data in order to improve it.

[0937] The following is a specific description of an embodiment of the present invention. This system provides business support using text data related to each business, and is also combined with an emotion engine that recognizes the user's emotions.

[0938] First, users upload text data such as work-related manuals, guidelines, and FAQs to the server. Users can upload text data using a web interface or a dedicated application. The uploaded data is saved in a specific directory on the server. This allows work-related information to be managed centrally and can be accessed by users at any time.

[0939] The server then reads the stored text data and performs preprocessing: it opens the files, analyzes the text content, and organizes multiple documents into a list. During this process, the text is cleaned and standardized, ensuring consistency for subsequent processing.

[0940] The server then vectorizes the preprocessed text data using a machine learning model. This process uses natural language processing models such as Word2Vec and BERT. Based on the vectorized data, the server calculates a cosine similarity matrix and generates a learning model. This learning model allows the similarity between text data to be evaluated.

[0941] When a user enters a question into the ElderGPT system using a client device, the question is sent from the client device to the server via API. For example, a question like "How do I create a project plan?" is a good example. The server vectorizes the received question and performs a cosine similarity calculation. Based on the calculation results, the server searches for the most relevant document and generates its content as an answer. This generated answer is sent back to the client device and displayed to the user.

[0942] The server also uses an emotion engine to recognize the emotion of the user's question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input and classify them as positive, negative, or neutral. Based on the recognized emotion information, the server adjusts the answer. For example, if the server recognizes that the user is in trouble, it will provide a more specific and detailed answer.

[0943] Feedback can also be provided. Users can provide feedback on the displayed answers, such as their satisfaction level and areas for improvement. Feedback information is sent from the client device to the server, which collects and stores it to help improve the learning model in the future. The collected feedback information is used to periodically retrain the model.

[0944] Finally, the server periodically retrains the model based on feedback and emotion information. This improves the system's accuracy and responsiveness, enabling it to provide better support to users. For example, if a user asks, "How do I create a project plan?", the server searches for relevant documents and provides them. At the same time, it recognizes emotions from the user's input and determines whether a positive response is required. In this way, a system is realized that quickly and appropriately provides the information users are looking for.

[0945] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0946] Step 1:

[0947] A user uses a client terminal to upload text data related to business (such as manuals, guidelines, and FAQs). Specifically, the user accesses a web interface or a dedicated application, selects the text data using the file selection button, and clicks the upload button.

[0948] Input: Text file about business

[0949] Output: Text file saved on the server (specific directory)

[0950] Step 2:

[0951] The server reads the uploaded text data and performs preprocessing: it opens the file, extracts the text content, removes unnecessary spaces and special characters, and standardizes the text data (for example, standardizing uppercase and lowercase letters).

[0952] Input: A text file stored on the server

[0953] Output: Preprocessed text data in list format

[0954] Step 3:

[0955] The server vectorizes the preprocessed text data. Specifically, the server uses a natural language processing model such as Word2Vec or BERT to convert the text data into an array of numbers (vectors). The server then uses the vectorized data to calculate a cosine similarity matrix and generate a learning model.

[0956] Input: Preprocessed text data in list format

[0957] Output: Vectorized text data, cosine similarity matrix, and training model

[0958] Step 4:

[0959] The user enters a question into the ElderGPT system using a client terminal and presses the send button. The question may include how to create a project plan.

[0960] Input: Question text (prompt text)

[0961] Output: Question text sent to the server

[0962] Step 5:

[0963] The server vectorizes the received query and performs cosine similarity calculations. Based on the results, the server searches for the most relevant documents and generates their contents as the answer.

[0964] Input: Question text sent to the server, learning model

[0965] Output: The generated answer

[0966] Step 6:

[0967] The server uses an emotion engine to recognize the sentiment of the user's questions and answers, and uses text analysis technology to analyze the sentiment of the user's input, such as positive, negative, or neutral.

[0968] Input: Question text, generated answer

[0969] Output: Recognized emotion information

[0970] Step 7:

[0971] The server adjusts the answer based on the recognized emotional information. Specifically, if the server recognizes that the user is in trouble, it provides a more specific and detailed answer. This adjusted answer is then sent back to the client device.

[0972] Input: Recognized emotion information, generated answer

[0973] Output: Adjusted answer

[0974] Step 8:

[0975] The terminal displays the adjusted answers received from the server to the user, who can review the answers on the screen and ask further questions if necessary.

[0976] Input: Adjusted answer

[0977] Output: The answer displayed to the user

[0978] Step 9:

[0979] The user provides feedback on the displayed answer, such as satisfaction and areas for improvement. The user enters their feedback using the rating button or comment form and clicks the submit button.

[0980] Input: Satisfaction and feedback on improvements

[0981] Output: Feedback information sent to the server

[0982] Step 10:

[0983] The server collects feedback information and stores it to help improve the learning model in the future. The collected feedback information is used for periodic model retraining. The server stores this information in a database and uses it for the next retraining.

[0984] Input: Feedback information

[0985] Output: Saved feedback information

[0986] Step 11:

[0987] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[0988] Input: Stored feedback information, emotion information

[0989] Output: Retrained learning model

[0990] (Application example 2)

[0991] 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."

[0992] Conventional business support systems only provide information such as manuals, guidelines, and FAQs for each task, and lack the ability to recognize users' emotions and adjust responses accordingly, meaning users cannot receive appropriate support tailored to their individual situations and emotions. Furthermore, mechanisms for retraining models based on feedback and improving system performance are extremely limited, making sustained improvement difficult.

[0993] 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 receiving and storing text data such as manuals, guidelines, and FAQs related to each business, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing emotions from the user's input, means for adjusting the tone and level of detail of the answer based on the recognized emotion, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the feedback. This makes it possible to continuously improve the performance of the system while providing appropriate support according to the user's emotions.

[0994] "Text data such as manuals, guidelines, and FAQs for each task" refers to data in written form that contains information about the execution and procedures of tasks, and is used by employees and workers to refer to in order to carry out their tasks smoothly.

[0995] "Means for receiving and storing" refers to the mechanism for importing and storing data into storage such as a server.

[0996] "Vectorization" is the process of converting text data into an array of numbers (vectors), making it easier for machine learning models to process the text data.

[0997] A "cosine similarity matrix" is a matrix for calculating the similarity between different vectors, and the larger the value, the more similar the two vectors are.

[0998] "Means for receiving a question and generating the most relevant answer to that question" refers to technology for receiving a question posed by a user and providing the most relevant information or answer to that question.

[0999] "Means for recognizing emotions" refers to a mechanism for analyzing emotions from user input or voice and classifying them as, for example, positive, negative, neutral, etc.

[1000] "Means of adjusting the tone and level of detail of responses based on recognized emotions" refers to technology that appropriately changes the content of responses and how they are conveyed based on the results of emotion analysis.

[1001] "Means for providing feedback" refers to an interface that allows users to input and submit their impressions and opinions regarding answers, such as satisfaction and areas for improvement.

[1002] "Means for collecting feedback and retraining the learning model based on that information" refers to technology for collecting feedback information from users, systematically analyzing it, and reconstructing the machine learning model to improve the system's performance.

[1003] This invention is a system that supports business operations using text data such as manuals, guidelines, and FAQs related to each operation, and combines it with an emotion engine that recognizes user emotions. By installing this system on factory robots, it can efficiently support business operations within the factory.

[1004] The server has a means of receiving and storing text data such as manuals, guidelines, and FAQs for each task. This is done, for example, through a web interface or a dedicated application. The stored data is stored in built-in storage or cloud storage.

[1005] The server has a means to vectorize the received text data and calculate a cosine similarity matrix. For vectorization, Python's scikit-learn library and TfidfVectorizer are used. This converts the text data into an array of numerical values ​​(vectors), making it easier for machine learning models to process. The calculated cosine similarity matrix is ​​used to evaluate the similarity between different vectors.

[1006] A user inputs a question using a client device (e.g., a voice input device or a tablet). The question is sent to the server and processed by a means to generate the most relevant answer to the question. The server vectorizes the question, performs a cosine similarity calculation, searches for the most relevant text data, and generates the content as the answer.

[1007] The server has a means to recognize emotions from user input. It uses the TextBlob library as an emotion engine to analyze emotions from the user's input text and voice. The analysis results are classified as positive, negative, neutral, etc.

[1008] Based on the results of emotion recognition, the server has the means to adjust the tone and level of detail of the response, for example, providing gentler language and more detailed explanations if a negative emotion is recognized.

[1009] The system also provides a means for users to provide feedback on their answers. Feedback is collected by users entering their level of satisfaction and areas for improvement using a client terminal and sending the feedback to the server.

[1010] The server has the means to retrain the learning model based on the collected feedback. By periodically retraining the model, the performance of the system can be continuously improved.

[1011] Specific examples

[1012] For example, if a worker asks, "How do I deal with machine error message 1234?", the robot will search for relevant manual information and provide the most appropriate information. Furthermore, it will analyze the worker's emotions from their tone of voice and explain things more carefully if they are confused.

[1013] Prompt Sentence Examples

[1014] "For factory robots to assist workers in their work, please create a program that provides relevant manual information in response to worker questions and recognizes emotions to provide appropriate responses. For example, in response to the question, 'How do I deal with machine error message 1234?', the program should provide the most appropriate information from the machine's troubleshooting guide, adjusting the tone of the response depending on the worker's emotions (positive, negative, neutral)."

[1015] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1016] Step 1:

[1017] Users upload text data such as manuals, guidelines, and FAQs for each business to the server. The uploaded files are saved in a specific directory on the server. The input here is the text data uploaded by the user, and the output is the text data saved on the server.

[1018] Step 2:

[1019] The server reads the saved file and captures the text data, which is extracted and compiled into a list. The input is the text file saved on the server, and the output is the text data in list format.

[1020] Step 3:

[1021] The server vectorizes the text data and calculates the cosine similarity matrix. Here, the text data is converted into a numerical array using Python's TfidfVectorizer. The input is text data in list format, and the output is the vectorized text data and the cosine similarity matrix.

[1022] Step 4:

[1023] A user inputs a question using a client terminal. The input question is sent to the server via API. The input is the user's question text, and the output is the question sent to the server.

[1024] Step 5:

[1025] The server vectorizes the question text and performs cosine similarity calculations. Based on the results of the calculations, it searches for the most relevant documents and generates their contents as an answer. The input is the user's question text and existing text data, and the output is the generated answer.

[1026] Step 6:

[1027] The server uses an emotion engine to recognize the sentiment of the user's question and answer. It uses the TextBlob library to perform sentiment analysis. The input is the user's question text, and the output is the sentiment classification result (positive, negative, neutral).

[1028] Step 7:

[1029] The server adjusts the tone and level of detail of the response based on the recognized emotion. For example, if a negative emotion is recognized, it provides a more detailed and polite explanation. The input is the emotion classification result and the generated response, and the output is the adjusted response.

[1030] Step 8:

[1031] The user provides feedback on the answer. The feedback is input using a client terminal and sent to the server. The input is the user's feedback, and the output is the feedback information stored on the server.

[1032] Step 9:

[1033] The server retrains the learning model based on the collected feedback information. It periodically analyzes the feedback information and updates the model. The input is the feedback information, and the output is an improved learning model.

[1034] Step 10:

[1035] The retrained model is used to improve future question answering and emotion recognition accuracy. The inputs are new questions and existing datasets, and the output is improved system responses.

[1036] 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.

[1037] 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.

[1038] 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.

[1039] [Fourth embodiment]

[1040] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1041] 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.

[1042] 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).

[1043] 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.

[1044] 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.

[1045] 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).

[1046] 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.

[1047] 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.

[1048] 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.

[1049] 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.

[1050] 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.

[1051] 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.

[1052] 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."

[1053] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention provides business support using text data such as manuals, guidelines, and FAQs relating to each business.

[1054] 1. Uploading and saving data

[1055] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are stored in a specific directory on the server. This allows necessary work-related information to be managed centrally and easily accessible.

[1056] 2. Loading and Preprocessing Data

[1057] The server reads the saved file as text data and combines multiple sentences into a single list, allowing for consistent handling of the data in subsequent processing.

[1058] 3. Vectorization and Model Generation

[1059] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers. A cosine similarity matrix is ​​calculated based on the vectorized data, and a learning model is generated using this. This makes it possible to evaluate the similarity between text data.

[1060] 4. Question and Answer Processing

[1061] A user inputs a question in natural language to the Elder GPT system using a client device. For example, a question like, "Please tell me the format of the financial report." This question is sent from the device to the server via an API.

[1062] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained, the server searches for the most relevant document and generates its content as the answer.

[1063] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[1064] For example, if a user asks, "What is the format of a financial report?", the server will search for the most relevant guidelines or manuals related to financial reports from text data that has been vectorized and learned in advance, and provide the content to the user, allowing the user to obtain accurate information quickly.

[1065] 5. Providing feedback and retraining

[1066] The user inputs feedback such as satisfaction with the answer provided, points for improvement, etc. For example, feedback such as "satisfied" or "insufficient information" is applicable.

[1067] The terminal transmits the input feedback information to the server.

[1068] The server collects the submitted feedback and stores it for use in improving the learning model in the future. The model is periodically retrained with data that reflects the feedback information, further improving the accuracy and responsiveness of the system.

[1069] Through the above process, the Elder GPT system provides a support environment where employees can easily ask questions at any time, improving work efficiency and reducing employee stress.

[1070] The processing flow will be explained below.

[1071] Step 1:

[1072] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a web interface or a dedicated application.

[1073] Step 2:

[1074] The server receives the uploaded files and stores them in the appropriate directory, allowing necessary business-related information to be managed centrally.

[1075] Step 3:

[1076] The server reads the saved files, treats them as text data, gets the contents of each file, and compiles them into a list.

[1077] Step 4:

[1078] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This vectorization makes it easier for machine learning models to handle text data.

[1079] Step 5:

[1080] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which makes it possible to evaluate the similarity between text data.

[1081] Step 6:

[1082] A user inputs a question in natural language to the Elder GPT system using a client terminal, and the input question is sent from the terminal to the server via an API.

[1083] Step 7:

[1084] The server vectorizes the received question text and performs cosine similarity calculations. Based on the similarity score, it searches for the most relevant documents.

[1085] Step 8:

[1086] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[1087] Step 9:

[1088] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[1089] Step 10:

[1090] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[1091] Step 11:

[1092] The server collects the submitted feedback and stores it for future use in improving the learning model.

[1093] Step 12:

[1094] The server periodically retrains the model based on feedback information, improving the accuracy and responsiveness of the system and providing better support.

[1095] Example 1

[1096] 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."

[1097] Conventional systems have difficulty in centrally managing and effectively utilizing business document data. Furthermore, they lack mechanisms for providing quick and accurate answers to user questions, and lack the means to continuously improve the quality of answers. This makes it difficult to maximize business efficiency.

[1098] 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.

[1099] In this invention, the server includes a means for receiving and storing document data related to each business, a means for reading the received document data, listing it, and performing preprocessing, and a means for vectorizing the preprocessed document data and calculating a similarity matrix, thereby enabling unified management and effective use of document data.

[1100] The server further includes means for receiving question data, vectorizing the data, and generating the most relevant answer based on similarity, means for users to provide feedback on the answers, and means for collecting the feedback and retraining the generative model based on the feedback, thereby enabling the server to provide fast and accurate answers to user questions and continuously improve the quality of the answers by incorporating the feedback.

[1101] "Document data related to each business" is a general term for text data such as manuals, guidelines, FAQs, etc. related to the business operations of a company or organization.

[1102] "Means of receiving and storing" refers to the process by which the server receives document data uploaded by a user and stores the data in a specific directory or the like.

[1103] "Means of creating a list and pre-processing" refers to the process of reading the received data in list format and performing pre-processing such as cleaning the data and removing unnecessary information.

[1104] "Vectorization" refers to the process of converting document data into an array of numbers, or vectors, which makes it easier for computers to process.

[1105] The "means for calculating a similarity matrix" refers to a process for calculating the cosine similarity between vectorized data and expressing the result in a matrix format.

[1106] "Question data" refers to text data of the question that the user inputs to the system.

[1107] The "means for generating the most relevant answer based on similarity" refers to a process of vectorizing question data, identifying the most relevant document data using a pre-calculated similarity matrix, and generating an answer.

[1108] "Means for providing feedback" refers to the process by which users input their ratings and opinions on the answers provided into the system.

[1109] "Means for collecting feedback" refers to the process of storing feedback provided by users in a database or the like and aggregating it for use in subsequent processing.

[1110] A "generative model" is a model generated from document data using a specific learning algorithm, and is used to generate answers to questions.

[1111] "Retraining" refers to the process of updating an existing generative model based on collected feedback information to improve its accuracy and responsiveness.

[1112] This invention is a system that efficiently manages document data related to each task in a business support system and can provide quick and accurate answers to user questions. The system is implemented by clearly defining the roles of the server, terminal, and user. The hardware used includes a standard web server and client PC. The software used includes a natural language processing library (e.g., spaCy, Transformers), a data management library (e.g., pandas), and a library for similarity calculation (e.g., SciKit-Learn).

[1113] Data upload and management

[1114] Users upload document data related to their work (e.g., manuals, guidelines, FAQs, etc.) from their own devices to the server. Uploading is done through a web interface, and the selected file is saved in a specific directory on the server. For example, a user opens a browser, clicks the upload button, selects the desired document file, and executes the upload. At this time, the file's metadata (e.g., file name, upload date and time) is also saved.

[1115] Data loading and preprocessing

[1116] The server periodically scans the stored document data and reads newly uploaded files in text format. The read data is converted into a list format and then text cleaning (removal of whitespace and unnecessary characters) is performed. This process is performed using the Python pandas library. For example, the pandas read_csv method is used to read a CSV file and convert it to DataFrame format.

[1117] Vectorization and Similarity Calculation

[1118] After the text data has been read and preprocessed, it is converted into a numerical vector using a natural language processing library. This conversion is performed using libraries such as spaCy and Transformers. Next, a cosine similarity matrix is ​​calculated from the vectorized data to evaluate the relevance between the text data. The calculated similarity matrix is ​​then processed using the SciKit-Learn library.

[1119] Question and answer processing

[1120] A user inputs a question in natural language from their own device to the system. For example, a question might be, "What is the format of a financial report?" The question is entered through a web form and sent to the server. The server vectorizes the question and identifies the most relevant document data using a pre-calculated similarity matrix. An answer is generated based on the content and sent back to the device. The user can check the answer on their browser. As a specific example, if a user asks, "Please tell me about the marketing strategy for a new product," the server will search for relevant documents and generate an answer.

[1121] Feedback and retraining

[1122] Users provide feedback on the answers they provide, such as their satisfaction level or additional comments, through a web form. This feedback is sent from the device to the server and stored in a database. The server periodically retrains the model based on the feedback information, allowing the system to continuously improve its answer accuracy and user satisfaction.

[1123] Prompt Sentence Examples

[1124] Please tell me the format of the financial statement.

[1125] "Please tell us about your marketing strategy for the new product."

[1126] "How do I troubleshoot my product?"

[1127] This allows the system to not only centrally manage business-related information, but also provide quick and accurate answers to user questions. Furthermore, by improving the accuracy of the system through the feedback function, the quality of business support can be improved.

[1128] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1129] Step 1: Upload and save your data

[1130] A user uploads document data related to work by, for example, clicking the upload button on the browser and selecting the specified file.

[1131] Input: A document file selected by the user

[1132] The server receives the uploaded file and stores it in a specific directory (e.g., / data / uploads), along with metadata (file name, upload date and time).

[1133] Output: The saved file and its metadata

[1134] Step 2: Loading and preprocessing data

[1135] The server periodically scans a specific directory and reads newly uploaded document files, converts them as text files into a DataFrame, and then organizes them into a list.

[1136] Input: Saved document file

[1137] The server cleans the data, removing unnecessary whitespace and special characters, using the read_csv method from the pandas library.

[1138] Output: A list of preprocessed text data

[1139] Step 3: Vectorization and similarity matrix calculation

[1140] The server vectorizes the preprocessed text data and converts it into a numerical vector using a natural language processing library (e.g., spaCy, Transformers).

[1141] Input: A list of preprocessed text data

[1142] The server calculates the cosine similarity matrix based on the vectorized data. It uses the SciKit-Learn library to calculate the similarity between documents.

[1143] Output: Cosine similarity matrix

[1144] Step 4: Receiving questions and generating answers

[1145] The user inputs a question to the system from the client terminal, for example, a prompt such as "Please tell me the format of the financial statement."

[1146] Input: User question data

[1147] The device sends the question to the server via an API.

[1148] The server vectorizes the question, uses a pre-computed similarity matrix to identify the most relevant documents, and generates an answer.

[1149] Output: Generated response data

[1150] Step 5: View your answers

[1151] The server transmits the generated response data to the terminal.

[1152] Input: Generated response data

[1153] The device displays the received response to the user, which appears as text or a link in the browser.

[1154] Output: The answer that is displayed to the user

[1155] Step 6: Provide and collect feedback

[1156] Users provide feedback on the answers provided, including their satisfaction and any additional comments, by filling out a form.

[1157] Input: User feedback data

[1158] The device sends feedback to the server via an API.

[1159] The server collects the feedback and stores it in a database.

[1160] Output: Feedback data collected and stored

[1161] Step 7: Retrain with feedback

[1162] The server periodically retrains the generative model based on the collected feedback data, adding new data and recalculating the cosine similarity matrix to update the model.

[1163] Input: Collected feedback data

[1164] The server stores the updated generative model.

[1165] Output: Updated generative model

[1166] (Application example 1)

[1167] 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."

[1168] Picking work at logistics centers requires accuracy and efficiency, so it is important to have an environment where staff can quickly obtain the information they need. However, traditional methods often require staff to refer to paper-based manuals and guidelines, which is time-consuming and laborious. In addition, staff have limited means of obtaining immediate answers to their questions, which can lead to reduced work efficiency. Furthermore, there is a lack of mechanisms for effectively utilizing feedback to improve the quality of the answers provided. These issues need to be resolved to improve the efficiency and accuracy of picking work.

[1169] 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.

[1170] In this invention, the server includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer, thereby enabling pickers to quickly obtain necessary information and improve system performance based on the feedback.

[1171] A "manual" is a document that describes the procedures and methods for performing a specific task or work.

[1172] "Guidelines" are documents that provide basic guidelines and standards to be followed in business operations and procedures.

[1173] A "FAQ" is a document that compiles frequently asked questions and their answers.

[1174] "Vectorization" is the process of converting text data into an array of numbers.

[1175] A "cosine similarity matrix" is a matrix used to calculate the similarity between different vectors.

[1176] The "means for receiving a question" is a mechanism by which the system receives a question from a user.

[1177] The "means for generating an answer" is a process for generating the most appropriate answer to a question received from a user.

[1178] The "means for providing feedback" is a mechanism for users to input their opinions, such as ratings and suggestions for improvement, regarding the answers provided.

[1179] "Means for collecting feedback" refers to the process by which the system collects feedback information provided by users.

[1180] The "retraining method" is the process of re-learning the learning model to improve its performance based on the collected feedback information.

[1181] "Smart glasses" are a wearable eyeglass-type device equipped with a display that can display information and accept voice input.

[1182] A "means for asking questions in natural language" is a mechanism that allows users to ask questions using natural language, either spoken or written.

[1183] The "display means" is a method for showing the generated answers and related information to the user in real time.

[1184] The present invention comprises a system for supporting picking operations in a logistics center. The system includes a means for a user to ask a question in natural language using smart glasses, a means for displaying an answer on the smart glasses in real time using a display means, and a means for the user to provide feedback on the answer.

[1185] First, the distribution center manager uploads and stores text data such as manuals, guidelines, and FAQs for each operation on the server. The server then reads the stored text data and manages it centrally in list format.

[1186] The server then vectorizes the text data and calculates a cosine similarity matrix, which can improve the accuracy of generating answers to questions.

[1187] When a user wears smart glasses and speaks a question in natural language, the smart glasses send the question to the server via API. The server vectorizes the received question, performs cosine similarity calculations, and generates an answer based on the most relevant documents. The answer is displayed in real time on the smart glasses' display, allowing the user to immediately check the information they need.

[1188] Users can also provide feedback on the answers they provide. The feedback information is sent to the server and stored. The server uses this feedback information to retrain the learning model and improve the system's performance.

[1189] The hardware required to realize this system is smart glasses (e.g., Google Glass, Vuzix Blade) and a server. The software used includes Python, scikit-learn, and spaCy for vectorizing text data, Flask and FastAPI for the server-side API, AWS S3 and Google Cloud Storage for cloud storage, and OpenAI GPT and Hugging Face Transformers for the question-answering model.

[1190] For example, when a picker at a logistics center asks through smart glasses, "What's the next item to pick?", the server vectorizes the question and generates an answer based on relevant documents from the stored data. This answer is displayed on the smart glasses in real time, allowing the picker to quickly proceed with the next task.

[1191] An example of a prompt sentence is, "What is the next item to pick?" In this way, the present invention provides a system that effectively supports picking operations in logistics centers and significantly improves work efficiency and accuracy.

[1192] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1193] Step 1:

[1194] The server receives and stores text data such as manuals, guidelines, and FAQs for each operation uploaded by the logistics center administrator. The input is the text data uploaded by the administrator, and the output is a file stored on the server. This process allows necessary operation-related information to be centrally managed on the server and easily accessible.

[1195] Step 2:

[1196] The server reads the saved text data and compiles multiple sentences into a single list. The input is the text data saved in the server, and the output is the text data compiled into a single list. This process allows the data to be handled consistently in subsequent processes.

[1197] Step 3:

[1198] The server vectorizes the text data compiled in list format and calculates the cosine similarity matrix. The input is text data in list format, and the output is the vectorized data and the cosine similarity matrix. This process makes it possible to evaluate the similarity between text data.

[1199] Step 4:

[1200] The user uses the smart glasses to input a question in natural language by voice. The input is the user's natural language question, and the output is a question recorded by the smart glasses as voice data. This process allows the user to immediately query the system for the information they need.

[1201] Step 5:

[1202] The device converts the question recorded as voice data into text data and sends it to the server via API. The input is the user's question recorded as voice data, and the output is the question as text data sent to the server. Through this process, the server receives the user's question.

[1203] Step 6:

[1204] The server vectorizes the received question and performs a cosine similarity calculation. The input is the question as text data, and the output is the most relevant documents based on the similarity score. Through this process, the server generates an appropriate answer.

[1205] Step 7:

[1206] The server sends the generated answers to the smart glasses via API and displays them to the user in real time using the display means. The input is the answer as the most relevant document, and the output is the answer displayed on the smart glasses. This process allows the user to quickly obtain information.

[1207] Step 8:

[1208] The user provides feedback on the displayed answers. The input is the displayed answer and the user's evaluation of it, and the output is feedback information sent to the server. This process allows the user's opinion to be reflected in the system.

[1209] Step 9:

[1210] The server retrains the learning model based on the collected feedback information. The input is the feedback information, and the output is an improved learning model. This process improves the system's performance and increases the accuracy of the next question answer.

[1211] 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.

[1212] The following is a detailed description of an embodiment of the present invention. The present invention is a system that provides business support using text data such as manuals, guidelines, and FAQs related to each business, and also combines it with an emotion engine that recognizes user emotions.

[1213] 1. Uploading and saving data

[1214] Users upload text data such as manuals, guidelines, and FAQs related to their work to the server. Uploaded files are uploaded via a web interface or a dedicated application and are saved in a specific directory on the server. This allows work-related information to be managed centrally and be accessible at any time.

[1215] 2. Loading and Preprocessing Data

[1216] The server reads the saved file as text data and organizes multiple sentences into a list, allowing the data to be handled consistently in subsequent processing.

[1217] 3. Vectorization and Model Generation

[1218] The text data loaded by the server is vectorized. Vectorization is the process of converting text data into an array of numbers, which makes it easier for machine learning models to handle text data. A cosine similarity matrix is ​​then calculated based on the vectorized data to generate a learning model. This learning model makes it possible to evaluate the similarity between text data.

[1219] 4. Question and Answer Processing

[1220] A user inputs a question in natural language to the ElderGPT system using a client device. For example, a question like, "How do I create a project plan?" The input question is sent from the device to the server via an API.

[1221] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, it searches for the most relevant document and generates an answer from that document. The generated answer is then sent back to the device.

[1222] The terminal displays the answer received from the server to the user, allowing the user to instantly resolve their question.

[1223] 5. Emotion Recognition with Emotion Engine

[1224] The server uses an emotion engine to analyze the user's emotions regarding questions and answers. The emotion engine uses text analysis technology to recognize emotions from the user's input. For example, it can classify emotions as positive, negative, neutral, etc.

[1225] 6. Adjusting responses based on emotions

[1226] The server generates a response appropriate to the user based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a specific and detailed response.

[1227] 7. Providing and Saving Feedback

[1228] The user provides feedback on the displayed answers, such as satisfaction and areas for improvement. The feedback information is sent from the device to the server.

[1229] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to periodically retrain the model.

[1230] 8. Retraining the model

[1231] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[1232] For example, if a user asks, "How do I create a project plan?", the server will search for relevant documents and provide them. At the same time, it will recognize emotions from the user's input and determine whether a positive response is required. In this way, a system can be realized that quickly and appropriately provides the information the user is looking for.

[1233] Through the above process, the Elder GPT system provides a support environment that can respond to users while recognizing their emotions 24 hours a day, 365 days a year, thereby improving work efficiency and reducing employee stress.

[1234] The processing flow will be explained below.

[1235] Step 1:

[1236] Users upload text data such as manuals, guidelines, and FAQs for each task to the server via a dedicated web interface or application.

[1237] Step 2:

[1238] The server receives the uploaded files and stores them in the specified directory, allowing business-related information to be managed centrally.

[1239] Step 3:

[1240] The server reads the saved files, identifies them as text data, and combines the contents of multiple files into a single list.

[1241] Step 4:

[1242] The server vectorizes the text data it reads. Vectorization is the process of converting text data into an array of numbers. This conversion makes it easier for machine learning models to handle text data.

[1243] Step 5:

[1244] The server calculates a cosine similarity matrix based on the vectorized data and generates a learning model, which is used to evaluate the similarity between text data.

[1245] Step 6:

[1246] A user inputs a question in natural language into the Elder GPT system using a client device. For example, a question like, "How do I create a project plan?" The question is sent from the device to the server via an API.

[1247] Step 7:

[1248] The server vectorizes the received question text and performs a cosine similarity calculation. Based on the similarity score obtained through the calculation, the server searches for the most relevant document. For example, the server searches for the most relevant "Guidelines for creating project plans."

[1249] Step 8:

[1250] The server generates an answer based on the content of relevant documents, and sends the answer back to the device.

[1251] Step 9:

[1252] The terminal displays the answer received from the server to the user, allowing the user to quickly resolve their question.

[1253] Step 10:

[1254] The server uses an emotion engine to analyze the user's sentiment towards the question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input, which can be classified as negative, positive, neutral, etc.

[1255] Step 11:

[1256] The server further tailors the answer based on the emotional information recognized by the emotion engine. For example, if the server recognizes that the user is in trouble, it will provide a more detailed and helpful answer.

[1257] Step 12:

[1258] The user provides feedback on the displayed answer. The feedback includes satisfaction with the answer and areas for improvement. The device sends this feedback information to the server.

[1259] Step 13:

[1260] The server collects the submitted feedback and stores it for future use in improving the learning model. The collected feedback information is used to retrain the model.

[1261] Step 14:

[1262] The server periodically retrains the learning model based on feedback and emotional information, improving the accuracy and responsiveness of the system and providing better support to users.

[1263] Example 2

[1264] 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."

[1265] While there is a demand for efficient information provision using business-related text data and improved user satisfaction, conventional systems have had problems with not being able to properly search text data or generate answers, making it difficult to quickly and accurately resolve user questions.In addition, because they are unable to generate answers that take user emotions into consideration, it has been difficult to improve the system to fully reflect user feedback.

[1266] 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.

[1267] In this invention, the server includes means for receiving and storing text data related to each task, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing the user's emotion using emotion analysis technology, means for adjusting the answer based on the recognized emotion information, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the information. This enables the system to provide quick and accurate answers to user questions, realizes detailed responses that take user emotions into consideration, and makes it possible to improve the system by utilizing feedback.

[1268] "Text data" is a data format that represents document information such as business manuals, guidelines, and FAQs.

[1269] "Vectorization" is the process of converting text data into an array of numbers, making it easier for machine learning models to handle text data.

[1270] A "cosine similarity matrix" is a matrix obtained by calculating cosine similarity in order to evaluate the similarity between vectorized text data.

[1271] "Receiving a question" refers to the system receiving and processing an inquiry from a user.

[1272] "Generating an answer" refers to the process of providing highly relevant information to a user's question and answering the user's doubt.

[1273] "Emotion analysis technology" is a technology for recognizing and classifying user emotions from text data.

[1274] "Feedback" refers to evaluation information such as satisfaction and areas for improvement provided by users to the system.

[1275] "Retraining" is the process of retraining an existing learning model using newly collected data in order to improve it.

[1276] The following is a specific description of an embodiment of the present invention. This system provides business support using text data related to each business, and is also combined with an emotion engine that recognizes the user's emotions.

[1277] First, users upload text data such as work-related manuals, guidelines, and FAQs to the server. Users can upload text data using a web interface or a dedicated application. The uploaded data is saved in a specific directory on the server. This allows work-related information to be managed centrally and can be accessed by users at any time.

[1278] The server then reads the stored text data and performs preprocessing: it opens the files, analyzes the text content, and organizes multiple documents into a list. During this process, the text is cleaned and standardized, ensuring consistency for subsequent processing.

[1279] The server then vectorizes the preprocessed text data using a machine learning model. This process uses natural language processing models such as Word2Vec and BERT. Based on the vectorized data, the server calculates a cosine similarity matrix and generates a learning model. This learning model allows the similarity between text data to be evaluated.

[1280] When a user enters a question into the ElderGPT system using a client device, the question is sent from the client device to the server via API. For example, a question like "How do I create a project plan?" is a good example. The server vectorizes the received question and performs a cosine similarity calculation. Based on the calculation results, the server searches for the most relevant document and generates its content as an answer. This generated answer is sent back to the client device and displayed to the user.

[1281] The server also uses an emotion engine to recognize the emotion of the user's question and answer. The emotion engine uses text analysis technology to recognize emotions from the user's input and classify them as positive, negative, or neutral. Based on the recognized emotion information, the server adjusts the answer. For example, if the server recognizes that the user is in trouble, it will provide a more specific and detailed answer.

[1282] Feedback can also be provided. Users can provide feedback on the displayed answers, such as their satisfaction level and areas for improvement. Feedback information is sent from the client device to the server, which collects and stores it to help improve the learning model in the future. The collected feedback information is used to periodically retrain the model.

[1283] Finally, the server periodically retrains the model based on feedback and emotion information. This improves the system's accuracy and responsiveness, enabling it to provide better support to users. For example, if a user asks, "How do I create a project plan?", the server searches for relevant documents and provides them. At the same time, it recognizes emotions from the user's input and determines whether a positive response is required. In this way, a system is realized that quickly and appropriately provides the information users are looking for.

[1284] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1285] Step 1:

[1286] A user uses a client terminal to upload text data related to business (such as manuals, guidelines, and FAQs). Specifically, the user accesses a web interface or a dedicated application, selects the text data using the file selection button, and clicks the upload button.

[1287] Input: Text file about business

[1288] Output: Text file saved on the server (specific directory)

[1289] Step 2:

[1290] The server reads the uploaded text data and performs preprocessing: it opens the file, extracts the text content, removes unnecessary spaces and special characters, and standardizes the text data (for example, standardizing uppercase and lowercase letters).

[1291] Input: A text file stored on the server

[1292] Output: Preprocessed text data in list format

[1293] Step 3:

[1294] The server vectorizes the preprocessed text data. Specifically, the server uses a natural language processing model such as Word2Vec or BERT to convert the text data into an array of numbers (vectors). The server then uses the vectorized data to calculate a cosine similarity matrix and generate a learning model.

[1295] Input: Preprocessed text data in list format

[1296] Output: Vectorized text data, cosine similarity matrix, and training model

[1297] Step 4:

[1298] The user enters a question into the ElderGPT system using a client terminal and presses the send button. The question may include how to create a project plan.

[1299] Input: Question text (prompt text)

[1300] Output: Question text sent to the server

[1301] Step 5:

[1302] The server vectorizes the received query and performs cosine similarity calculations. Based on the results, the server searches for the most relevant documents and generates their contents as the answer.

[1303] Input: Question text sent to the server, learning model

[1304] Output: The generated answer

[1305] Step 6:

[1306] The server uses an emotion engine to recognize the sentiment of the user's questions and answers, and uses text analysis technology to analyze the sentiment of the user's input, such as positive, negative, or neutral.

[1307] Input: Question text, generated answer

[1308] Output: Recognized emotion information

[1309] Step 7:

[1310] The server adjusts the answer based on the recognized emotional information. Specifically, if the server recognizes that the user is in trouble, it provides a more specific and detailed answer. This adjusted answer is then sent back to the client device.

[1311] Input: Recognized emotion information, generated answer

[1312] Output: Adjusted answer

[1313] Step 8:

[1314] The terminal displays the adjusted answers received from the server to the user, who can review the answers on the screen and ask further questions if necessary.

[1315] Input: Adjusted answer

[1316] Output: The answer displayed to the user

[1317] Step 9:

[1318] The user provides feedback on the displayed answer, such as satisfaction and areas for improvement. The user enters their feedback using the rating button or comment form and clicks the submit button.

[1319] Input: Satisfaction and feedback on improvements

[1320] Output: Feedback information sent to the server

[1321] Step 10:

[1322] The server collects feedback information and stores it to help improve the learning model in the future. The collected feedback information is used for periodic model retraining. The server stores this information in a database and uses it for the next retraining.

[1323] Input: Feedback information

[1324] Output: Saved feedback information

[1325] Step 11:

[1326] The server periodically retrains the model based on feedback and sentiment information, improving the accuracy and responsiveness of the system and providing better support.

[1327] Input: Stored feedback information, emotion information

[1328] Output: Retrained learning model

[1329] (Application example 2)

[1330] 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."

[1331] Conventional business support systems only provide information such as manuals, guidelines, and FAQs for each task, and lack the ability to recognize users' emotions and adjust responses accordingly, meaning users cannot receive appropriate support tailored to their individual situations and emotions. Furthermore, mechanisms for retraining models based on feedback and improving system performance are extremely limited, making sustained improvement difficult.

[1332] 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 receiving and storing text data such as manuals, guidelines, and FAQs related to each business, means for vectorizing the received text data and calculating a cosine similarity matrix, means for receiving a question from a user and generating the most relevant answer to the question, means for recognizing emotions from the user's input, means for adjusting the tone and level of detail of the answer based on the recognized emotion, means for the user to provide feedback on the answer, and means for collecting feedback and retraining the learning model based on the feedback. This makes it possible to continuously improve the performance of the system while providing appropriate support according to the user's emotions.

[1333] "Text data such as manuals, guidelines, and FAQs for each task" refers to data in written form that contains information about the execution and procedures of tasks, and is used by employees and workers to refer to in order to carry out their tasks smoothly.

[1334] "Means for receiving and storing" refers to the mechanism for importing and storing data into storage such as a server.

[1335] "Vectorization" is the process of converting text data into an array of numbers (vectors), making it easier for machine learning models to process the text data.

[1336] A "cosine similarity matrix" is a matrix for calculating the similarity between different vectors, and the larger the value, the more similar the two vectors are.

[1337] "Means for receiving a question and generating the most relevant answer to that question" refers to technology for receiving a question posed by a user and providing the most relevant information or answer to that question.

[1338] "Means for recognizing emotions" refers to a mechanism for analyzing emotions from user input or voice and classifying them as, for example, positive, negative, neutral, etc.

[1339] "Means of adjusting the tone and level of detail of responses based on recognized emotions" refers to technology that appropriately changes the content of responses and how they are conveyed based on the results of emotion analysis.

[1340] "Means for providing feedback" refers to an interface that allows users to input and submit their impressions and opinions regarding answers, such as satisfaction and areas for improvement.

[1341] "Means for collecting feedback and retraining the learning model based on that information" refers to technology for collecting feedback information from users, systematically analyzing it, and reconstructing the machine learning model to improve the system's performance.

[1342] This invention is a system that supports business operations using text data such as manuals, guidelines, and FAQs related to each operation, and combines it with an emotion engine that recognizes user emotions. By installing this system on factory robots, it can efficiently support business operations within the factory.

[1343] The server has a means of receiving and storing text data such as manuals, guidelines, and FAQs for each task. This is done, for example, through a web interface or a dedicated application. The stored data is stored in built-in storage or cloud storage.

[1344] The server has a means to vectorize the received text data and calculate a cosine similarity matrix. For vectorization, Python's scikit-learn library and TfidfVectorizer are used. This converts the text data into an array of numerical values ​​(vectors), making it easier for machine learning models to process. The calculated cosine similarity matrix is ​​used to evaluate the similarity between different vectors.

[1345] A user inputs a question using a client device (e.g., a voice input device or a tablet). The question is sent to the server and processed by a means to generate the most relevant answer to the question. The server vectorizes the question, performs a cosine similarity calculation, searches for the most relevant text data, and generates the content as the answer.

[1346] The server has a means to recognize emotions from user input. It uses the TextBlob library as an emotion engine to analyze emotions from the user's input text and voice. The analysis results are classified as positive, negative, neutral, etc.

[1347] Based on the results of emotion recognition, the server has the means to adjust the tone and level of detail of the response, for example, providing gentler language and more detailed explanations if a negative emotion is recognized.

[1348] The system also provides a means for users to provide feedback on their answers. Feedback is collected by users entering their level of satisfaction and areas for improvement using a client terminal and sending the feedback to the server.

[1349] The server has the means to retrain the learning model based on the collected feedback. By periodically retraining the model, the performance of the system can be continuously improved.

[1350] Specific examples

[1351] For example, if a worker asks, "How do I deal with machine error message 1234?", the robot will search for relevant manual information and provide the most appropriate information. Furthermore, it will analyze the worker's emotions from their tone of voice and explain things more carefully if they are confused.

[1352] Prompt Sentence Examples

[1353] "For factory robots to assist workers in their work, please create a program that provides relevant manual information in response to worker questions and recognizes emotions to provide appropriate responses. For example, in response to the question, 'How do I deal with machine error message 1234?', the program should provide the most appropriate information from the machine's troubleshooting guide, adjusting the tone of the response depending on the worker's emotions (positive, negative, neutral)."

[1354] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1355] Step 1:

[1356] Users upload text data such as manuals, guidelines, and FAQs for each business to the server. The uploaded files are saved in a specific directory on the server. The input here is the text data uploaded by the user, and the output is the text data saved on the server.

[1357] Step 2:

[1358] The server reads the saved file and captures the text data, which is extracted and compiled into a list. The input is the text file saved on the server, and the output is the text data in list format.

[1359] Step 3:

[1360] The server vectorizes the text data and calculates the cosine similarity matrix. Here, the text data is converted into a numerical array using Python's TfidfVectorizer. The input is text data in list format, and the output is the vectorized text data and the cosine similarity matrix.

[1361] Step 4:

[1362] A user inputs a question using a client terminal. The input question is sent to the server via API. The input is the user's question text, and the output is the question sent to the server.

[1363] Step 5:

[1364] The server vectorizes the question text and performs cosine similarity calculations. Based on the results of the calculations, it searches for the most relevant documents and generates their contents as an answer. The input is the user's question text and existing text data, and the output is the generated answer.

[1365] Step 6:

[1366] The server uses an emotion engine to recognize the sentiment of the user's question and answer. It uses the TextBlob library to perform sentiment analysis. The input is the user's question text, and the output is the sentiment classification result (positive, negative, neutral).

[1367] Step 7:

[1368] The server adjusts the tone and level of detail of the response based on the recognized emotion. For example, if a negative emotion is recognized, it provides a more detailed and polite explanation. The input is the emotion classification result and the generated response, and the output is the adjusted response.

[1369] Step 8:

[1370] The user provides feedback on the answer. The feedback is input using a client terminal and sent to the server. The input is the user's feedback, and the output is the feedback information stored on the server.

[1371] Step 9:

[1372] The server retrains the learning model based on the collected feedback information. It periodically analyzes the feedback information and updates the model. The input is the feedback information, and the output is an improved learning model.

[1373] Step 10:

[1374] The retrained model is used to improve future question answering and emotion recognition accuracy. The inputs are new questions and existing datasets, and the output is improved system responses.

[1375] 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.

[1376] 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.

[1377] 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.

[1378] 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.

[1379] 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.

[1380] 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.

[1381] 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).

[1382] 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.

[1383] 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."

[1384] 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.

[1385] 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).

[1386] 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.

[1387] 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.

[1388] 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.

[1389] 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.

[1390] 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.

[1391] 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.

[1392] 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.

[1393] 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.

[1394] 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.

[1395] 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.

[1396] The following is further disclosed regarding the above embodiment.

[1397] (Claim 1)

[1398] A means to receive and store text data such as manuals, guidelines, and FAQs for each business,

[1399] means for vectorizing the received text data and calculating a cosine similarity matrix;

[1400] means for receiving a question from a user and generating a most relevant answer to the question;

[1401] a means for the user to provide feedback on the answers;

[1402] A means of collecting feedback and using that information to retrain the learning model; and

[1403] A system including:

[1404] (Claim 2)

[1405] 10. The system of claim 1, further comprising means for using vectorization and cosine similarity calculations in generating answers to user questions.

[1406] (Claim 3)

[1407] 10. The system of claim 1, further comprising means for periodically retraining the learning model to improve its performance based on the feedback information.

[1408] "Example 1"

[1409] (Claim 1)

[1410] A means for receiving and storing document data related to each business;

[1411] A means for reading received document data, listing it, and performing preprocessing;

[1412] means for vectorizing the preprocessed document data and calculating a similarity matrix;

[1413] means for receiving question data and vectorizing it to generate the most relevant answers based on similarity;

[1414] a means for the user to provide feedback on the answers;

[1415] A means of collecting feedback and using that information to retrain the generative model;

[1416] A system including:

[1417] (Claim 2)

[1418] 10. The system of claim 1, further comprising means for using vectorization and similarity calculations when generating answers to user questions based on similarity calculations.

[1419] (Claim 3)

[1420] 10. The system of claim 1, further comprising means for periodically retraining the generative model to improve its performance based on the feedback information.

[1421] "Application Example 1"

[1422] (Claim 1)

[1423] A means to receive and store text data such as manuals, guidelines, and FAQs for each business,

[1424] means for vectorizing the received text data and calculating a cosine similarity matrix;

[1425] means for receiving a question from a user and generating a most relevant answer to the question;

[1426] a means for a user to ask questions in natural language using the smart glasses;

[1427] means for displaying the answers in real time on the smart glasses using a display means;

[1428] a means for the user to provide feedback on the answers;

[1429] A means of collecting feedback and using that information to retrain the learning model; and

[1430] A system including:

[1431] (Claim 2)

[1432] 10. The system of claim 1, further comprising means for using vectorization and cosine similarity calculations in generating answers to user questions.

[1433] (Claim 3)

[1434] 10. The system of claim 1, further comprising means for periodically retraining the learning model to improve its performance based on the feedback information.

[1435] "Example 2: Combining Emotion Engines"

[1436] (Claim 1)

[1437] A means for receiving and storing text data related to each task;

[1438] means for vectorizing the received text data and calculating a cosine similarity matrix;

[1439] means for receiving a question from a user and generating a most relevant answer to the question;

[1440] a means for the user to provide feedback on the answers;

[1441] A means of collecting feedback and using that information to retrain the learning model; and

[1442] means for using emotion analysis techniques to recognize the emotion of a user;

[1443] a means for adjusting responses based on the perceived emotional information;

[1444] A system including:

[1445] (Claim 2)

[1446] 10. The system of claim 1, further comprising means for using vectorization and cosine similarity calculations in generating answers to user questions.

[1447] (Claim 3)

[1448] 10. The system of claim 1, further comprising means for periodically retraining the learning model to improve its performance based on the feedback information.

[1449] "Application example 2 when combining emotion engines"

[1450] (Claim 1)

[1451] A means to receive and store text data such as manuals, guidelines, and FAQs for each business,

[1452] means for vectorizing the received text data and calculating a cosine similarity matrix;

[1453] means for receiving a question from a user and generating a most relevant answer to the question;

[1454] means for recognizing emotions from user input;

[1455] A means to adjust the tone and level of detail of responses based on perceived sentiment;

[1456] a means for the user to provide feedback on the answers;

[1457] A means of collecting feedback and using that information to retrain the learning model; and

[1458] A system including:

[1459] (Claim 2)

[1460] 10. The system of claim 1, further comprising means for using vectorization and cosine similarity calculations in generating answers to user questions.

[1461] (Claim 3)

[1462] 10. The system of claim 1, further comprising means for periodically retraining the learning model to improve its performance based on the feedback information. [Explanation of symbols]

[1463] 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 to receive and store text data such as manuals, guidelines, and FAQs for each business, means for vectorizing the received text data and calculating a cosine similarity matrix; means for receiving a question from a user and generating a most relevant answer to the question; a means for the user to provide feedback on the answers; A means of collecting feedback and using that information to retrain the learning model; and A system including:

2. 10. The system of claim 1, further comprising means for using vectorization and cosine similarity calculations in generating answers to user questions.

3. 10. The system of claim 1, further comprising means for periodically retraining the learning model to improve its performance based on the feedback information.

Images