system

Abstract

We provide the system. [Solution] A means for collecting internal company documents and storing them in an electronic format, A means for converting stored documents into vector data and saving the vector data in a searchable database, A means for receiving questions from users and retrieving information related to the received questions from the database, A means of generating answers to questions using a large-scale language model based on relevant information, A system that includes means for providing the generated response to the user.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 discloses a method for controlling a persona chatbot, which is performed by at least one processor, including steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a 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

Summary of the Invention

Problems to be Solved by the Invention

[0004] Internal company documents and manuals are in a huge amount, and it is difficult to quickly search for necessary information. In particular, when the manual is paper-based or has limited search functions even if it is digitized, it takes time and effort to obtain information. This problem may lead to a decrease in work efficiency and occurrence of mistakes. Also, there is a demand for a system that can quickly provide users with appropriate answers to specific questions.

Means for Solving the Problems

[0005] The present invention provides a system that includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and saving the vector data in a searchable database, means for receiving questions from users and retrieving information related to the received questions from the database, means for generating answers to the questions using a large-scale language model based on the related information, and means for providing the generated answers to users. This makes it possible to quickly and accurately retrieve necessary information from internal company documents and immediately provide appropriate answers to user questions.

[0006] "Internal documents" refer to documents such as operational manuals, reports, and guidelines created within a company or organization.

[0007] "Collection" refers to the act of gathering scattered documents together.

[0008] "Digitization" refers to the act of converting paper documents into a digital format.

[0009] "Format" refers to a specific form or method related to the storage and display of data or documents.

[0010] "Storage" refers to the act of saving data or documents in a specific location.

[0011] "Vector data" refers to a format that converts documents and information into numerical data, allowing for efficient processing on a computer.

[0012] "Conversion" refers to the act of replacing data or information with a different format or representation.

[0013] A "database" refers to a system that organizes and stores data so that it can be efficiently searched, added, and updated.

[0014] "Searchable" means that users can find data based on specific conditions or keywords.

[0015] "User" refers to an individual or a corporation that uses this system.

[0016] "Question" refers to the act of a user inputting the information or questions they want to obtain from the system.

[0017] "Relevant" refers to data or information related to a specific question.

[0018] "Information" refers to data or knowledge including the answers to questions.

[0019] "Large language model" refers to a machine learning model trained using a large amount of text data for natural language processing.

[0020] "Generate" refers to the act of creating new data or information.

[0021] "Provide" refers to the act of delivering information or answers to the user.

[0022] "System" refers to a mechanism in which multiple components or means operate in cooperation.

Brief Explanation of Drawings

[0023] [Figure 1] It is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] It is a conceptual diagram showing an example of the main functions of a data processing device and a smart device according to the first embodiment. [Figure 3] It is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] It is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] It is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] It is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] This shows an emotion map where multiple emotions are mapped. [Figure 10] This shows an emotion map where multiple emotions are mapped. [Figure 11] This is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12] This is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] This is a sequence diagram showing the processing flow of the data processing system in Example 2, which incorporates an emotion engine. [Figure 14] This is a sequence diagram showing the processing flow of the data processing system in Application Example 2, which combines an emotion engine. [Modes for carrying out the invention]

[0024] Hereinafter, an example of an embodiment of the system relating to the technology of this disclosure will be described with reference to the attached drawings.

[0025] First, let's explain the terminology used in the following explanation.

[0026] In the following embodiments, the signed processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Furthermore, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), and APU (Accelerated Processing Unit).

[0027] In the following embodiments, signed RAM (Random Access Memory) is a memory that temporarily stores information and is used as work memory by the processor.

[0028] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0029] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0030] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0031] [First Embodiment]

[0032] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0033] As shown in Figure 1, the 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.

[0034] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0035] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0036] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and 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.

[0037] 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 perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0038] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0039] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0040] As shown in Figure 2, in the data processing device 12, a specific processing 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" related to the technology of this 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 according to the specific processing program 56 executed on the RAM 30.

[0041] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0042] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0043] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0044] This invention provides a system for collecting internal company documents, storing them in an electronic format, converting them into vector data, and saving them in a database. The aim is to generate rapid and accurate answers to user inquiries using a large-scale language model.

[0045] System program

[0046] 1. Data Collection

[0047] The server first collects internal company documents. These documents include various documents used within the company, such as operational manuals, reports, and guidelines. Because the collected documents are difficult to search in their original form, they are converted into a digital format. For example, paper manuals are scanned and converted into text data using OCR (Optical Character Recognition). The converted data is temporarily stored in memory or on the file system.

[0048] 2. Generating vector data

[0049] Next, the server converts the collected text data into vector data. This is the process of converting text data into a numerical representation, which is necessary for efficiently searching the content of the text. The vector data is generated and stored using natural language processing techniques. Specifically, each document is quantified as a feature vector and stored in a searchable database.

[0050] 3. Loading and initializing the model

[0051] The server loads a Large-Scale Language Model (LLM). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a large amount of text data. Once loaded, the model is ready to operate.

[0052] 4. Question reception and answer generation

[0053] The user sends a question from their terminal to the server. For example, a question like, "How do I create an invoice?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the manual on creating invoices.

[0054] The server, having retrieved relevant information, then uses a large-scale language model to generate an answer to the question. The model takes the question and related information as input and generates a natural language response based on it. After generating the answer, the server returns the result to the user. The user can view and refer to this answer through their terminal.

[0055] Specific example

[0056] User: "How do I create an invoice?"

[0057] server:

[0058] Receive questions from users.

[0059] Search for information on invoice creation on the Vector Store.

[0060] Use a large-scale language model to generate specific answers to questions.

[0061] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount," is returned to the user.

[0062] User: Check the answer on the device and understand the steps to create the invoice.

[0063] In this way, the system of the present invention enables efficient management of internal company documents and prompt responses to user inquiries.

[0064] The following describes the processing flow.

[0065] Step 1:

[0066] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper copies.

[0067] Step 2:

[0068] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR technology. PDF and Word documents are also converted into text data. As a result, all documents are temporarily stored in text format.

[0069] Step 3:

[0070] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert the text data into numerical representations (vectors). The vector data captures the semantic characteristics of the text content.

[0071] Step 4:

[0072] The server stores the vector data in a searchable database. The vector data is stored in a specific database (e.g., ElasticSearch®) so that it can be searched quickly and efficiently.

[0073] Step 5:

[0074] The server loads a Large-Scale Language Model (LLM). It loads a pre-trained model and prepares it for operation. This model is responsible for natural language processing in response to the question.

[0075] Step 6:

[0076] The user enters a question from their device. The question might be something like, "How do I create an invoice?"

[0077] Step 7:

[0078] The server receives a question from the user. The received question is then formatted into an appropriate format for analysis.

[0079] Step 8:

[0080] The server searches a vector database to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[0081] Step 9:

[0082] The server uses a large-scale language model to generate answers to questions based on relevant information. The model takes the question and search results as input and outputs answers in natural language.

[0083] Step 10:

[0084] The server provides the user with the generated response. The response is sent to the user via the terminal.

[0085] Step 11:

[0086] Users view the answers on their devices and obtain specific information related to their questions. For example, they can receive immediately useful information such as, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the products and services, 4. Calculate and enter the total amount."

[0087] (Example 1)

[0088] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0089] In modern businesses, a large volume of internal documents exists, and efficiently collecting, digitizing, and managing these documents is crucial. However, converting paper documents into digital data and saving that data in a searchable format is time-consuming and difficult to do efficiently. Furthermore, generating quick and accurate answers to user inquiries also requires a lot of manual work, making it inefficient.

[0090] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0091] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting paper documents into text data using OCR technology; means for converting the stored documents into vector data using natural language processing technology and storing the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to questions using a generative AI model based on the related information; and means for providing the generated answers to users. This enables efficient management of internal company documents and the generation of quick and accurate answers to questions from users.

[0092] "Internal company documents" refer to various documents created and used within a company or organization, including operational manuals, reports, and guidelines.

[0093] "Means of collection" refers to the general term for methods and equipment used to collect and consolidate internal company documents in one place.

[0094] "Digitization methods" is a general term for technologies and devices used to convert paper documents into digital format, including scanners and OCR (optical character recognition) technology.

[0095] "OCR technology" is a technology that optically recognizes characters and converts paper documents into text data.

[0096] "Text data" refers to character information represented in digital format.

[0097] "Natural language processing technology" is a general term for technologies that process and understand human language using computers.

[0098] "Vector data" refers to text and other data that have been converted into numerical representations, making them easier to search and analyze.

[0099] A "searchable database" is a database system designed to allow for quick and efficient retrieval of stored data.

[0100] A "generative AI model" is a model that uses artificial intelligence (AI) to generate new information or answers based on input data.

[0101] "Question receiving means" refers to the general term for interfaces and functions used to receive questions submitted by users.

[0102] "Question analysis means" refers to technologies and methods for understanding the content of a received question and identifying the information necessary to generate an appropriate answer.

[0103] "Answer generation means" refers to the technologies and methods used to generate answers to questions, and often involves the use of AI models.

[0104] "User-provided means" refers to the collective technologies and functions used to provide generated answers to users.

[0105] This invention is a system that efficiently collects, digitizes, and manages a wide variety of internal company documents, and further generates quick and accurate answers to user inquiries. This system mainly consists of three elements: a server, terminals, and users.

[0106] The server first collects internal company documents. These documents include various types of documents used within the company, such as operational manuals, reports, and guidelines. The server uses a dedicated scanner to capture paper documents as digital images and converts these images into text data using OCR (Optical Character Recognition) software (e.g., Tesseract OCR). The converted text data is temporarily stored in memory or storage.

[0107] Next, the server converts the collected text data into vector data using natural language processing techniques (e.g., spaCy or the transformers library). This vector data is a numerical representation of the text data, enabling efficient searching. The converted vector data is stored in a searchable database (e.g., Elasticsearch).

[0108] The server also loads and initializes large-scale language models (e.g., the generative AI model OpenAI® GPT-3®). These models are highly effective for natural language understanding and response generation. Once loaded, the models become operational on the server.

[0109] When a user sends a question from their device, the server receives and analyzes it. Specifically, suppose a user sends the question, "How do I create an invoice?" The server analyzes this question and searches for relevant information in its vector database. For example, it might use Elasticsearch to search for vector data containing information about creating invoices.

[0110] Based on the searched relevant information, the server uses a generative AI model to generate specific answers to the questions. The generated answers are returned to the user in the format of: "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount." The user can review and refer to these answers through their device.

[0111] In this way, the system of the present invention enables the efficient digitization and management of internal company documents, as well as the rapid and accurate generation of answers to user inquiries. Users can obtain detailed answers simply by asking questions, thereby improving the efficiency of their work.

[0112] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0113] Step 1:

[0114] The server collects internal company documents. It receives paper documents such as operational manuals, reports, and guidelines as input. The server uses a scanner to capture these paper documents as digital images and uses OCR (Optical Character Recognition) software to extract text data from the images. The extracted text data is temporarily stored in memory or storage as output. Specifically, the server operates a dedicated scanner to generate digital images. Then, it launches OCR software to analyze the character information within the images and convert it into text data.

[0115] Step 2:

[0116] The server converts collected text data into vector data using natural language processing techniques. It receives text data as input and uses natural language processing libraries (e.g., spaCy, transformers) to convert the text into numerical vectors. The generated vector data is stored in a searchable database (e.g., Elasticsearch). Specifically, the server loads text data, applies natural language processing algorithms to generate feature vectors, and then inserts the generated vectors into the database.

[0117] Step 3:

[0118] The server loads and initializes large-scale language models. It receives model files and configuration data as input. It imports the generative AI model library (e.g., OpenAI GPT-3), loads the model into memory, and initializes it. The output is a generative AI model ready to run. Specifically, the server loads the model files from the specified path, performs the initialization procedure, and makes the model usable.

[0119] Step 4:

[0120] The user enters a question from their terminal and sends it to the server. The server receives the user's question as input. For example, the user might send a question like, "How do I create an invoice?" The server analyzes the received question and queries a vector database using the question as input. The server then retrieves relevant information as output from this query. Specifically, the server analyzes the question content and extracts relevant keywords. It then sends queries to a search engine (e.g., Elasticsearch) to retrieve relevant information.

[0121] Step 5:

[0122] The server generates an answer using a generative AI model based on the relevant information it has acquired. The relevant information and the question are input to the generative AI model. The generative AI model processes the information and generates an appropriate answer. The output is the answer to the user's question. Specifically, the server provides the generative AI model with the question and relevant information as input and executes the process of generating an answer.

[0123] Step 6:

[0124] The server provides the user with the generated response. It receives the generated response as input and sends it to the terminal. The output is the response displayed on the user's terminal. Specifically, the server processes the generated response data and sends it to the user's terminal, which then displays this response on its screen.

[0125] The above outlines the specific processing flow of the system. This system is designed to achieve efficient document management and rapid response generation.

[0126] (Application Example 1)

[0127] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0128] In logistics centers, many workers perform their tasks based on various documents and manuals on a daily basis, but it is difficult to quickly search for this information and obtain answers. Furthermore, in order to proceed with work efficiently, there is a need for a means to immediately resolve questions and problems that arise during work. In this situation, a system is needed that efficiently manages internal documents and provides quick and accurate answers to questions from workers.

[0129] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0130] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting the stored documents into vector data and saving the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to the questions using a large-scale language model based on the related information; means for providing the generated answers to users; and means for logistics center workers to input questions via smartphone and receive immediate answers to those questions. This enables workers to obtain quick and accurate answers to questions and problems that arise during their work.

[0131] "Internal company documents" refer to all types of documents used within a company, such as operational manuals, reports, and guidelines.

[0132] "Digitization" refers to the process of scanning paper documents and converting them into a digital format using optical character recognition (OCR) technology.

[0133] "Vector data" is a numerical representation of text data, and is used to efficiently search the contents of a document.

[0134] A "database" refers to an information system that stores and manages digital documents and vector data in a searchable format.

[0135] A "user" refers to a person or worker who uses the system to input questions.

[0136] A "question" refers to the information a user seeks from the system or the problem they want to solve.

[0137] A "large-scale language model" is a model that has been trained on a vast amount of text data based on natural language processing technology, and is used to generate answers to questions.

[0138] A "smartphone" is a type of portable mobile phone that allows for internet connectivity and advanced operations using applications.

[0139] A "logistics center" refers to a facility where logistics operations such as storing, organizing, and shipping goods are carried out.

[0140] This invention aims to improve operational efficiency in logistics centers, and to that end, provides a system that efficiently manages internal documents and responds quickly to questions from workers.

[0141] System program

[0142] 1. Data collection and digitization

[0143] The server first collects internal company documents within the logistics center. These include operational manuals, reports, and guidelines. If the collected documents are in paper format, they are converted to digital format by scanning and then converted into text data using optical character recognition (OCR) technology. The converted data is temporarily stored in memory or on the file system.

[0144] 2. Generating vector data

[0145] Next, the server converts the collected text data into vector data. This is a process that utilizes natural language processing techniques to convert text data into numerical representations. This vector data is then stored in a database in a searchable format. Specifically, each document is quantified as a feature vector and stored in the database.

[0146] 3. Loading and initializing the model

[0147] The server loads a large-scale language model (e.g., Hugging Face's Transformers library). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a vast amount of text data. Once loaded, the model is ready to generate answers.

[0148] 4. Question reception and answer generation

[0149] An employee sends a question to the server via their smartphone. For example, they might send a question like, "How do I update the inventory list?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the inventory management manual.

[0150] Based on the relevant information obtained, the server uses a large-scale language model (e.g., Hugging Face's Transformers) to generate an answer to the question. The generated answer might be something like, "The procedure for updating the inventory list is as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes." The server sends this answer back to the worker's smartphone, and the worker can use it as a reference to proceed with the task.

[0151] Hardware and software

[0152] Hardware: Primarily uses a standard smartphone and a cloud server (e.g., AWS® EC2).

[0153] Software: We use Hugging Face Transformers for natural language processing, PostgreSQL as the database management system, and Django as the web framework.

[0154] Specific example

[0155] 1. The worker enters the question "How do I update the inventory list?" into the smartphone app.

[0156] 2. The app sends a question to the server.

[0157] 3. The server searches the database for relevant documents and generates answers using Hugging Face Transformers.

[0158] 4. The server sends the generated response back to the worker's smartphone.

[0159] Example of a prompt

[0160] "How do I update the inventory list?"

[0161] This allows logistics center workers to receive quick and accurate answers to questions that arise during their work, improving overall work efficiency.

[0162] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0163] Step 1:

[0164] The server collects and digitizes internal company documents within the logistics center. Specifically, it scans paper documents such as operational manuals, reports, and guidelines, and converts them into text data using optical character recognition (OCR) technology. The input is paper documents, and the output is digital text data. This data is temporarily stored in memory or the file system.

[0165] Step 2:

[0166] The server converts the collected text data into vector data. Specifically, it uses natural language processing techniques to convert the text data into numerical representations (feature vectors). The input is digital text data, and the output is vector data. This vector data is stored in a searchable database.

[0167] Step 3:

[0168] The server loads and initializes a large-scale language model. Here, the Hugging Face Transformers library is used. The input is pre-trained model data, and the output is the usable language model. This model is used for natural language understanding and generation.

[0169] Step 4:

[0170] The user (worker) inputs the question into the terminal via their smartphone and sends it to the server. The input is the text data of the question, and the output received by the server is the question content.

[0171] Step 5:

[0172] The server analyzes the received question and retrieves relevant information from the vector database based on that analysis. The input is the question's text data and the vector database, and the output is the relevant information (a specific section of the document).

[0173] Step 6:

[0174] The server generates answers to questions using a large-scale language model based on relevant information. The input consists of text data of the relevant information and the question, and the output generated by the model is a natural language answer.

[0175] Step 7:

[0176] The server sends the generated response back to the user's smartphone. The input is the text data of the generated response, and the output is the response displayed to the user.

[0177] Specific examples of operation:

[0178] 1. The worker types the question "How do I update the inventory list?" into their smartphone and sends it.

[0179] 2. The server receives the question and searches the database for relevant information on the corresponding internal document.

[0180] 3. The server uses a large-scale language model to generate an answer to the question, producing the following response: "The steps to update the inventory list are as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes."

[0181] 4. The server sends the generated response back to the worker's smartphone, allowing the worker to use it as a reference to proceed with their work.

[0182] Example prompt: "How do I update the inventory list?"

[0183] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0184] This invention provides a system that generates appropriate and emotion-responsive answers to user questions by combining a system that collects internal company documents, stores them in an electronic format, converts them into vector data, and stores them in a database with an emotion engine that recognizes user emotions. Specific embodiments are shown below.

[0185] System program

[0186] 1. Data collection (collection and digitization of internal company documents)

[0187] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, and other documents from various departments and systems using manual or automated scripts. The collected documents are stored in various formats, including PDF, Word, and paper copies. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[0188] 2. Generating vector data

[0189] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is converted into numerical representations (vectors). The vector data captures the semantic features of the text content and is stored in a searchable database.

[0190] 3. Loading and initializing the model

[0191] The server loads a Large-Scale Language Model (LLM) and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[0192] 4. Introduction of an emotional engine

[0193] The server will also incorporate an emotion engine that recognizes user emotions. This emotion engine analyzes the word choices and sentence structure in the user's questions to identify their emotions. In addition, by integrating voice input and facial recognition technology, it will also be able to analyze voice and facial expression data.

[0194] 5. Question reception and sentiment analysis

[0195] The user enters a question from their device. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion engine to identify the user's emotion (e.g., confusion, excitement, anger).

[0196] 6. Searching for related information

[0197] The server searches a vector database to identify documents and information related to the query. Using vector data allows for effective retrieval of semantically relevant information.

[0198] 7. Emotion-based response generation

[0199] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the sentiment engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and helpful explanations to confused users.

[0200] 8. Providing responses and user feedback

[0201] The server provides the user with the generated response. The response is sent to the user via their device. The user views the response on their device and obtains specific information about the question. In addition, the sentiment engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[0202] Specific example

[0203] User: Typed "How do I create an invoice?"

[0204] server:

[0205] Receive questions from users.

[0206] The emotion engine analyzes the question content and identifies when the user is confused.

[0207] Search the vector database for information related to invoice creation.

[0208] We use large-scale language models to generate detailed and thorough answers.

[0209] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[0210] User: Check the answer on your device to help solve the problem.

[0211] In this way, the system of the present invention enables efficient management of internal company documents and prompt, emotionally responsive responses to user inquiries.

[0212] The following describes the processing flow.

[0213] Step 1:

[0214] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are in various formats, such as PDF, Word, and paper copies.

[0215] Step 2:

[0216] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[0217] Step 3:

[0218] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert the text data into numerical representations (vectors). This allows for the capture of semantic features of the text content.

[0219] Step 4:

[0220] The server stores the vector data in a searchable database. The vector data is stored in a specific database so that it can be searched quickly and efficiently.

[0221] Step 5:

[0222] The server loads the Large-Scale Language Model (LLM) and prepares to operate. This is a crucial component for generating answers to user questions.

[0223] Step 6:

[0224] The server loads an emotion engine that recognizes the user's emotions. The emotion engine analyzes the words and sentence structure entered by the user to identify their emotions.

[0225] Step 7:

[0226] The user enters a question from their device. For example, they might send a question like "How do I create an invoice?" to the server.

[0227] Step 8:

[0228] The server receives a question from the user. The received question is analyzed by the sentiment engine, and the user's emotions are identified. For example, the wording and tone of the question can be used to determine that the user is confused.

[0229] Step 9:

[0230] The server searches the vector database to identify information related to the question. This extracts documents and information that are semantically relevant to the question.

[0231] Step 10:

[0232] The server generates answers to questions using a large-scale language model based on relevant information. The response is then adjusted to an appropriate tone and level of detail based on the user's emotions, as identified by the sentiment engine.

[0233] Step 11:

[0234] The server provides the user with a generated response. The response is sent to the user via their device. For example, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions."

[0235] Step 12:

[0236] The user views the answers on their device and obtains specific information regarding the question. Simultaneously, the sentiment engine monitors the user's emotional changes in real time and adjusts the answers as needed. For example, if the user remains confused, it provides further details.

[0237] Through the above series of steps, the system of the present invention enables efficient management of internal company documents and provides prompt and emotionally responsive answers to user inquiries.

[0238] (Example 2)

[0239] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0240] Traditional information management systems struggled to provide appropriate and timely answers to user inquiries. Furthermore, they lacked the ability to consider user emotions, resulting in a poor user experience. This often led to users being unable to quickly access the information they needed, ultimately reducing operational efficiency.

[0241] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users and searching the database for information related to the received questions, means for generating answers to questions using a large-scale language model based on the related information, means for providing the generated answers to users, means for analyzing the content of the questions and the emotional state of the users using an emotion recognition engine that recognizes the emotions of the users, and means for generating appropriate answers according to the emotions of the users based on the analysis results of the emotion recognition engine. This makes it possible to provide quick and appropriate answers according to the emotions of users to their questions.

[0242] "Internal documents" refer to documents created and used by a company or organization in connection with its operations, and include work manuals, guidelines, reports, etc.

[0243] "Digitization" refers to the process of converting physical paper documents into a digital format, which is achieved using scanners and OCR technology.

[0244] "Vector data" refers to data obtained by converting text data into numerical representations, and is generated using natural language processing techniques and machine learning algorithms.

[0245] A "searchable database" is a database system designed to store vector data and other types of data, and to allow for efficient searching.

[0246] A "large-scale language model" is a machine learning model that excels at understanding and generating natural language, and is used to generate answers to user questions.

[0247] An "emotion recognition engine" is a software module that analyzes user input text, voice, facial expressions, etc., to identify their emotional state.

[0248] "Information related to the question" refers to the documents and data necessary to generate an answer to what the user asked.

[0249] An "appropriate answer" is one that provides accurate and useful information in response to a question, and is also tailored to facilitate communication that takes into account the user's emotional state.

[0250] This invention provides a system that digitizes internal company documents, converts them into vector data for storage, and generates answers to user questions using a large-scale language model. By further combining this with an emotion recognition engine, the system provides appropriate answers that correspond to the user's emotions.

[0251] 1. Data collection (collection and digitization of internal company documents)

[0252] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, etc., from each department and system using manual or automated scripts (e.g., "DocFetch"). The collected documents are saved in various formats, including PDF, Word, and paper. Paper documents are scanned using a scanner (e.g., "ScanMaster") and converted into text data using OCR (Optical Character Recognition) technology (e.g., "TextExtractor"). PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[0253] 2. Generating vector data

[0254] The server converts digitized text data into vector data. Using natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec), it generates vector data that captures the semantic features of the text content. This vector data is then stored in a searchable database (e.g., "VectorDB").

[0255] 3. Loading and initializing the model

[0256] The server loads a large-scale language model (e.g., "GPT-3") and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[0257] 4. Introduction of an emotion recognition engine

[0258] The server also incorporates an emotion recognition engine (e.g., "EmotionDetect") that recognizes the user's emotions. This emotion recognition engine analyzes the word choices and sentence structure included in the user's questions to identify emotions. It also integrates voice input and facial recognition technology, enabling the analysis of voice and facial expression data.

[0259] 5. Question reception and sentiment analysis

[0260] The user enters a question using a terminal. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion recognition engine to identify the user's emotion (e.g., confusion, excitement, anger).

[0261] 6. Searching for related information

[0262] The server searches a searchable database (e.g., "VectorDB") to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[0263] 7. Emotion-based response generation

[0264] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the emotion recognition engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and helpful explanations to confused users.

[0265] 8. Providing responses and user feedback

[0266] The server provides the user with the generated response. The user reviews the response on their device and uses it to solve the problem. Furthermore, the emotion recognition engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[0267] Specific example

[0268] Example prompt:

[0269] User: "Please tell me how to create an invoice."

[0270] server:

[0271] Receive questions from users.

[0272] The emotion recognition engine analyzes the question and identifies that the user is confused.

[0273] Search for information related to invoice creation in the searchable database ("VectorDB").

[0274] Using a large-scale language model ("GPT-3"), detailed and thorough answers are generated.

[0275] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[0276] User: Check the answer on your device to help solve the problem.

[0277] In this way, the system of the present invention can achieve efficient management of internal company documents and prompt, emotionally responsive answers to user inquiries.

[0278] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0279] Step 1:

[0280] This system collects and digitizes internal company documents. The server uses manual or automated scripts (DocFetch) to collect documents such as work manuals, guidelines, and reports from various departments and systems. Paper documents are digitized using a scanner (ScanMaster) and converted into text data using OCR technology (TextExtractor). Input consists of various internal company documents, and output is in text format.

[0281] Specific operation: The server runs the "DocFetch" script to collect documents from shared folders on the network. It then uses the "ScanMaster" application to scan paper documents and converts them to text using the "TextExtractor" OCR system.

[0282] Step 2:

[0283] Convert the digitized text data into vector data. The server uses natural language processing technologies and machine learning algorithms (e.g., BERT and Word2Vec) to generate vector data that captures the semantic features of the text content. The input is text data, and the output is vector data.

[0284] Specific operation: The server uses the "VectorTransform" module to convert text data into numerical vectors and inserts them into "VectorDB".

[0285] Step 3:

[0286] Load and initialize a large language model. The server loads and initializes a pre-trained large language model (GPT-3). This prepares for generating answers to the user's questions. The input is the configuration information of the language model, and the output is the initialized model.

[0287] Specific operation: The server executes the "ModelLoader" script to load the "GPT-3" model into memory and initialize it.

[0288] Step 4:

[0289] Introduce an emotion recognition engine. The server operates an emotion recognition engine (EmotionDetect) that recognizes the user's emotions and adds the function of analyzing the selection of words and sentence structures in the user's questions to identify emotions. Voice input and face recognition technologies are also integrated to enable the analysis of voice and facial expression data. The input is the user's question and voice / facial expression data, and the output is the data with emotion tags attached.

[0290] Specific operation: The server runs the "EmotionDetect" module and works in conjunction with the "FaceReader" and "VoiceAnalyzer" modules to analyze text, voice, and facial expression data in real time.

[0291] Step 5:

[0292] The system receives questions from users and performs sentiment analysis. Users input questions using a terminal, which the server receives and analyzes using a sentiment recognition engine. The input is the user's question, and the output is question data with analyzed sentiment tags.

[0293] Specific operation: The user enters a question using the "HelpDesk" application, and the server analyzes it using the "EmotionDetect" module and assigns an emotion tag.

[0294] Step 6:

[0295] This function searches for related information. The server searches "VectorDB" based on question data tagged with sentiment, and identifies information related to the question. The input is question data tagged with sentiment, and the output is the extracted related information.

[0296] Specific operation: The server uses the "SearchEngine" module to search "VectorDB" and retrieve relevant information.

[0297] Step 7:

[0298] It generates responses that respond to emotions. The server uses a large-scale language model based on relevant information to generate appropriate responses that reflect the user's emotions. The input is question data with relevant information and emotion tags, and the output is adjusted response data.

[0299] Specific operation: The server generates responses using the "GPT-3" model and adjusts the responses according to emotions using the "EmotionFilter" module.

[0300] Step 8:

[0301] The system provides answers and user feedback. The server delivers the generated answers to the user via the terminal. The user reviews the answers on the terminal and provides feedback if necessary. The emotion recognition engine monitors the user's emotional changes in real time and adjusts the answers as needed. The input is the generated answer data and user feedback, and the output is the answer, if possible, further adjusted.

[0302] Specific operation: The server sends the response to the terminal using the "ResponseSender" module, and the user views the response in the "HelpDesk" application and enters feedback as needed. The "EmotionTracker" module analyzes the feedback in real time and readjusts the response as necessary.

[0303] (Application Example 2)

[0304] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0305] Traditional information delivery systems lacked the ability to generate appropriate responses that reflected user emotions, and the technology to make the user experience personal and effective was absent. This resulted in users consistently receiving unsatisfactory responses. Furthermore, even content recommendation systems struggled to address individual user needs based on their emotions.

[0306] The specific processing by the specific processing unit 290 of the data processing apparatus 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting in-house documents and storing them in an electronic form, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving a question from a user, analyzing the received question with a sentiment analysis engine to identify the user's sentiment, means for generating an answer to the question using a large language model based on relevant information, means for adjusting and providing the generated answer according to the user's sentiment, means for monitoring the change in the user's sentiment in real time and readjusting the answer as necessary, and means for recommending content according to the sentiment. This enables appropriate answers and content recommendations that take into account the sentiment for each user.

[0307] "In-house documents" refer to documents such as documents, reports, operation manuals, and guidelines created and managed within an organization.

[0308] "Digitization" refers to a technology that converts paper-based documents and other analog-formatted information into digital data.

[0309] "Vector data" is data obtained by converting text data into a numerical representation, which is a data format that makes it easier for a computer to perform processing and analysis.

[0310] "Database" is a system that stores text data and other information in a structured manner for efficient management and retrieval.

[0311] "User's question" refers to input data representing the information a user who uses the system wants to know or the problem they want to solve.

[0312] "Receiving a question" means receiving a question sent by a user and taking it in for processing.

[0313] An "emotion analysis engine" is an analytical tool that identifies emotions from user input data (text, voice, facial expressions).

[0314] A "large-scale language model" is a natural language processing model trained on vast amounts of text data, and is an algorithm capable of generating and understanding language.

[0315] "Answer generation" is the process by which a system creates an appropriate answer to a user's question based on relevant information.

[0316] "Adjusting to emotions" means changing the content and tone of the generated response according to the user's emotional state.

[0317] "Real-time monitoring" refers to constantly monitoring the user's emotional state and taking action as needed.

[0318] "Content recommendation" is a process that suggests entertainment and information, such as movies, music, and articles, based on the user's emotional state.

[0319] This invention is a system that provides appropriate responses and recommends content based on the user's emotions. Specific embodiments are shown below.

[0320] System program

[0321] 1. Collection and digitization of internal company documents

[0322] First, the server collects internal documents from each department and system. This is done manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[0323] 2. Generating vector data

[0324] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is transformed into numerical representations (vectors). This vector data captures the semantic features of the text content and is stored in a searchable database.

[0325] 3. Question reception and sentiment analysis

[0326] The user enters a question from their device. For example, they might send a question like, "I'm tired today, but I want to watch a relaxing movie." The server analyzes the received question using an emotion analysis engine to identify the user's emotions (e.g., fatigue, confusion, excitement, anger, etc.). The OpenAI API is used for this emotion analysis.

[0327] 4. Searching for related information and generating answers

[0328] The server searches a vector database to identify documents and information related to the question. Next, it uses a large-scale language model (GPT-3) based on this relevant information to generate an answer to the question. This large-scale language model may include OpenAI's GPT-3. The generated answer is then adjusted according to the user's sentiment.

[0329] 5. Facial Expression and Voice Analysis

[0330] Furthermore, the server utilizes cameras and microphones to analyze the user's facial expressions and voice. By using the OpenCV library to analyze facial expressions and the SpeechRecognition library to convert speech to text, the system captures the user's emotions from multiple perspectives.

[0331] 6. Content Recommendation and Delivery

[0332] Based on the results of sentiment analysis, the server recommends content. If the user is fatigued, relaxing movies or music articles will be recommended. For example, specific content might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[0333] Specific example

[0334] Prompt example:

[0335] 1. Sentiment Analysis: "Analyze the sentiment of the following sentence and label it: 'I'm tired today, but I want to watch a relaxing movie.'"

[0336] 2. Content Recommendation: "Please recommend three relaxing movies that you would suggest to a user who is feeling tired."

[0337] By following these specific steps, the server can provide appropriate responses and content recommendations that take each user's emotions into consideration.

[0338] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0339] Step 1:

[0340] The server collects internal company documents and stores them in electronic format. Specifically, documents are collected manually or using automated scripts from each department and system. If the collected documents are in PDF or Word format, they are directly converted into text data. Paper documents are scanned and converted into text data using OCR technology. As a result, all documents are temporarily stored in text format.

[0341] input:

[0342] PDF, Word, and paper-based internal documents

[0343] Data processing:

[0344] Digitization of paper documents using scanners

[0345] Text conversion using OCR technology

[0346] output:

[0347] Text-format internal documents

[0348] Step 2:

[0349] The server converts digitized text data into vector data. This uses natural language processing techniques and machine learning algorithms. Specifically, it uses document vectorizers such as the BERT model to convert text data into numerical representations. The converted vector data captures its semantic features and is stored in a searchable database.

[0350] input:

[0351] Text-format internal documents

[0352] Data calculation:

[0353] Vectorization using natural language processing techniques (such as the BERT model)

[0354] output:

[0355] Numerical representation data stored in a vector database

[0356] Step 3:

[0357] The user enters a question from their device. For example, they might enter a question like, "I'm tired today, so I'd like to watch a relaxing movie." The entered question is then sent to the server.

[0358] input:

[0359] User questions (text data)

[0360] Data calculation:

[0361] Receiving questions and preparing for analysis

[0362] output:

[0363] Text data on the server

[0364] Step 4:

[0365] The server analyzes the received question using an emotion analysis engine. It uses the OpenAI API and other tools to identify the user's emotions from the question. Specifically, it sends the question text to the API and retrieves an emotion label.

[0366] input:

[0367] User questions (text data)

[0368] Data calculation:

[0369] Sentiment analysis using OpenAI API, etc.

[0370] output:

[0371] Emotional labels (fatigue, confusion, excitement, anger, etc.)

[0372] Step 5:

[0373] The server searches a vector database to identify documents and information related to the question. Specifically, it uses vector data to effectively search for semantically relevant information. For example, it can find manuals and guidelines related to the user's question.

[0374] input:

[0375] Vector database

[0376] User Questions

[0377] Data calculation:

[0378] Search using vector data

[0379] output:

[0380] Information (documents) related to the question

[0381] Step 6:

[0382] The server generates answers to questions using large-scale language models based on relevant information. This may include OpenAI's GPT-3. Furthermore, it adjusts the answers according to the user's emotions based on sentiment analysis results.

[0383] input:

[0384] Related information (documents)

[0385] Emotional labels

[0386] Data calculation:

[0387] Response generation and adjustment using GPT-3, etc.

[0388] output:

[0389] Emotionally appropriate response

[0390] Step 7:

[0391] The server sends the generated response to the device and provides it to the user. Furthermore, the server monitors the user's emotional changes in real time and readjusts the response as needed. To do this, the device uses its camera and microphone to send the user's facial expressions and voice to the server.

[0392] input:

[0393] Adjusted answer

[0394] User facial expressions and voice data

[0395] Data calculation:

[0396] Real-time monitoring of emotions and response adjustment

[0397] output:

[0398] Answers provided to users

[0399] Step 8:

[0400] The server recommends content to the user based on the results of sentiment analysis. For example, if the user is tired, it will recommend relaxing movies or music. A specific example might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[0401] input:

[0402] Emotional labels

[0403] Data calculation:

[0404] Content recommendation using large-scale language models

[0405] output:

[0406] Recommended content provided to users

[0407] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating 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.

[0408] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0409] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0410] [Second Embodiment]

[0411] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0412] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0413] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0414] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0415] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0416] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0417] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0418] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0419] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0420] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0421] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0422] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. 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".

[0423] This invention provides a system for collecting internal company documents, storing them in an electronic format, converting them into vector data, and saving them in a database. The aim is to generate rapid and accurate answers to user inquiries using a large-scale language model.

[0424] System program

[0425] 1. Data Collection

[0426] The server first collects internal company documents. These documents include various documents used within the company, such as operational manuals, reports, and guidelines. Because the collected documents are difficult to search in their original form, they are converted into a digital format. For example, paper manuals are scanned and converted into text data using OCR (Optical Character Recognition). The converted data is temporarily stored in memory or on the file system.

[0427] 2. Generating vector data

[0428] Next, the server converts the collected text data into vector data. This is the process of converting text data into a numerical representation, which is necessary for efficiently searching the content of the text. The vector data is generated and stored using natural language processing techniques. Specifically, each document is quantified as a feature vector and stored in a searchable database.

[0429] 3. Loading and initializing the model

[0430] The server loads a Large-Scale Language Model (LLM). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a large amount of text data. Once loaded, the model is ready to operate.

[0431] 4. Question reception and answer generation

[0432] The user sends a question from their terminal to the server. For example, a question like, "How do I create an invoice?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the manual on creating invoices.

[0433] The server, having retrieved relevant information, then uses a large-scale language model to generate an answer to the question. The model takes the question and related information as input and generates a natural language response based on it. After generating the answer, the server returns the result to the user. The user can view and refer to this answer through their terminal.

[0434] Specific example

[0435] User: "How do I create an invoice?"

[0436] server:

[0437] Receive questions from users.

[0438] Search for information on invoice creation on the Vector Store.

[0439] Use a large-scale language model to generate specific answers to questions.

[0440] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount," is returned to the user.

[0441] User: Check the answer on the device and understand the steps to create the invoice.

[0442] In this way, the system of the present invention enables efficient management of internal company documents and prompt responses to user inquiries.

[0443] The following describes the processing flow.

[0444] Step 1:

[0445] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper copies.

[0446] Step 2:

[0447] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR technology. PDF and Word documents are also converted into text data. As a result, all documents are temporarily stored in text format.

[0448] Step 3:

[0449] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert the text data into numerical representations (vectors). The vector data captures the semantic characteristics of the text content.

[0450] Step 4:

[0451] The server stores the vector data in a searchable database. The vector data is stored in a specific database (e.g., Elasticsearch) to allow for quick and efficient searching.

[0452] Step 5:

[0453] The server loads a Large-Scale Language Model (LLM). It loads a pre-trained model and prepares it for operation. This model is responsible for natural language processing in response to the question.

[0454] Step 6:

[0455] The user enters a question from their device. The question might be something like, "How do I create an invoice?"

[0456] Step 7:

[0457] The server receives a question from the user. The received question is then formatted into an appropriate format for analysis.

[0458] Step 8:

[0459] The server searches a vector database to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[0460] Step 9:

[0461] The server uses a large-scale language model to generate answers to questions based on relevant information. The model takes the question and search results as input and outputs answers in natural language.

[0462] Step 10:

[0463] The server provides the user with the generated response. The response is sent to the user via the terminal.

[0464] Step 11:

[0465] Users view the answers on their devices and obtain specific information related to their questions. For example, they can receive immediately useful information such as, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the products and services, 4. Calculate and enter the total amount."

[0466] (Example 1)

[0467] Next, we will describe Example 1. 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."

[0468] In modern businesses, a large volume of internal documents exists, and efficiently collecting, digitizing, and managing these documents is crucial. However, converting paper documents into digital data and saving that data in a searchable format is time-consuming and difficult to do efficiently. Furthermore, generating quick and accurate answers to user inquiries also requires a lot of manual work, making it inefficient.

[0469] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0470] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting paper documents into text data using OCR technology; means for converting the stored documents into vector data using natural language processing technology and storing the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to questions using a generative AI model based on the related information; and means for providing the generated answers to users. This enables efficient management of internal company documents and the generation of quick and accurate answers to questions from users.

[0471] "Internal company documents" refer to various documents created and used within a company or organization, including operational manuals, reports, and guidelines.

[0472] "Means of collection" refers to the general term for methods and equipment used to collect and consolidate internal company documents in one place.

[0473] "Digitization methods" is a general term for technologies and devices used to convert paper documents into digital format, including scanners and OCR (optical character recognition) technology.

[0474] "OCR technology" is a technology that optically recognizes characters and converts paper documents into text data.

[0475] "Text data" refers to character information represented in digital format.

[0476] "Natural language processing technology" is a general term for technologies that process and understand human language using computers.

[0477] "Vector data" refers to text and other data that have been converted into numerical representations, making them easier to search and analyze.

[0478] A "searchable database" is a database system designed to allow for quick and efficient retrieval of stored data.

[0479] A "generative AI model" is a model that uses artificial intelligence (AI) to generate new information or answers based on input data.

[0480] "Question receiving means" refers to the general term for interfaces and functions used to receive questions submitted by users.

[0481] "Question analysis means" refers to technologies and methods for understanding the content of a received question and identifying the information necessary to generate an appropriate answer.

[0482] "Answer generation means" refers to the technologies and methods used to generate answers to questions, and often involves the use of AI models.

[0483] "User-provided means" refers to the collective technologies and functions used to provide generated answers to users.

[0484] This invention is a system that efficiently collects, digitizes, and manages a wide variety of internal company documents, and further generates quick and accurate answers to user inquiries. This system mainly consists of three elements: a server, terminals, and users.

[0485] The server first collects internal company documents. These documents include various types of documents used within the company, such as operational manuals, reports, and guidelines. The server uses a dedicated scanner to capture paper documents as digital images and converts these images into text data using OCR (Optical Character Recognition) software (e.g., Tesseract OCR). The converted text data is temporarily stored in memory or storage.

[0486] Next, the server converts the collected text data into vector data using natural language processing techniques (e.g., spaCy or the transformers library). This vector data is a numerical representation of the text data, enabling efficient searching. The converted vector data is stored in a searchable database (e.g., Elasticsearch).

[0487] The server also loads and initializes large-scale language models (e.g., the generative AI model OpenAI GPT-3), which are excellent models for natural language understanding and response generation. Once loaded, the models become operational on the server.

[0488] When a user sends a question from their device, the server receives and analyzes it. Specifically, suppose a user sends the question, "How do I create an invoice?" The server analyzes this question and searches for relevant information in its vector database. For example, it might use Elasticsearch to search for vector data containing information about creating invoices.

[0489] Based on the searched relevant information, the server uses a generative AI model to generate specific answers to the questions. The generated answers are returned to the user in the format of: "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount." The user can review and refer to these answers through their device.

[0490] In this way, the system of the present invention enables the efficient digitization and management of internal company documents, as well as the rapid and accurate generation of answers to user inquiries. Users can obtain detailed answers simply by asking questions, thereby improving the efficiency of their work.

[0491] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0492] Step 1:

[0493] The server collects internal company documents. It receives paper documents such as operational manuals, reports, and guidelines as input. The server uses a scanner to capture these paper documents as digital images and uses OCR (Optical Character Recognition) software to extract text data from the images. The extracted text data is temporarily stored in memory or storage as output. Specifically, the server operates a dedicated scanner to generate digital images. Then, it launches OCR software to analyze the character information within the images and convert it into text data.

[0494] Step 2:

[0495] The server converts collected text data into vector data using natural language processing techniques. It receives text data as input and uses natural language processing libraries (e.g., spaCy, transformers) to convert the text into numerical vectors. The generated vector data is stored in a searchable database (e.g., Elasticsearch). Specifically, the server loads text data, applies natural language processing algorithms to generate feature vectors, and then inserts the generated vectors into the database.

[0496] Step 3:

[0497] The server loads and initializes large-scale language models. It receives model files and configuration data as input. It imports the generative AI model library (e.g., OpenAI GPT-3), loads the model into memory, and initializes it. The output is a generative AI model ready to run. Specifically, the server loads the model files from the specified path, performs the initialization procedure, and makes the model usable.

[0498] Step 4:

[0499] The user enters a question from their terminal and sends it to the server. The server receives the user's question as input. For example, the user might send a question like, "How do I create an invoice?" The server analyzes the received question and queries a vector database using the question as input. The server then retrieves relevant information as output from this query. Specifically, the server analyzes the question content and extracts relevant keywords. It then sends queries to a search engine (e.g., Elasticsearch) to retrieve relevant information.

[0500] Step 5:

[0501] The server generates an answer using a generative AI model based on the relevant information it has acquired. The relevant information and the question are input to the generative AI model. The generative AI model processes the information and generates an appropriate answer. The output is the answer to the user's question. Specifically, the server provides the generative AI model with the question and relevant information as input and executes the process of generating an answer.

[0502] Step 6:

[0503] The server provides the user with the generated response. It receives the generated response as input and sends it to the terminal. The output is the response displayed on the user's terminal. Specifically, the server processes the generated response data and sends it to the user's terminal, which then displays this response on its screen.

[0504] The above outlines the specific processing flow of the system. This system is designed to achieve efficient document management and rapid response generation.

[0505] (Application Example 1)

[0506] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0507] In logistics centers, many workers perform their tasks based on various documents and manuals on a daily basis, but it is difficult to quickly search for this information and obtain answers. Furthermore, in order to proceed with work efficiently, there is a need for a means to immediately resolve questions and problems that arise during work. In this situation, a system is needed that efficiently manages internal documents and provides quick and accurate answers to questions from workers.

[0508] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0509] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting the stored documents into vector data and saving the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to the questions using a large-scale language model based on the related information; means for providing the generated answers to users; and means for logistics center workers to input questions via smartphone and receive immediate answers to those questions. This enables workers to obtain quick and accurate answers to questions and problems that arise during their work.

[0510] "Internal company documents" refer to all types of documents used within a company, such as operational manuals, reports, and guidelines.

[0511] "Digitization" refers to the process of scanning paper documents and converting them into a digital format using optical character recognition (OCR) technology.

[0512] "Vector data" is a numerical representation of text data, and is used to efficiently search the contents of a document.

[0513] A "database" refers to an information system that stores and manages digital documents and vector data in a searchable format.

[0514] A "user" refers to a person or worker who uses the system to input questions.

[0515] A "question" refers to the information a user seeks from the system or the problem they want to solve.

[0516] A "large-scale language model" is a model that has been trained on a vast amount of text data based on natural language processing technology, and is used to generate answers to questions.

[0517] A "smartphone" is a type of portable mobile phone that allows for internet connectivity and advanced operations using applications.

[0518] A "logistics center" refers to a facility where logistics operations such as storing, organizing, and shipping goods are carried out.

[0519] This invention aims to improve operational efficiency in logistics centers, and to that end, provides a system that efficiently manages internal documents and responds quickly to questions from workers.

[0520] System program

[0521] 1. Data collection and digitization

[0522] The server first collects internal company documents within the logistics center. These include operational manuals, reports, and guidelines. If the collected documents are in paper format, they are converted to digital format by scanning and then converted into text data using optical character recognition (OCR) technology. The converted data is temporarily stored in memory or on the file system.

[0523] 2. Generating vector data

[0524] Next, the server converts the collected text data into vector data. This is a process that utilizes natural language processing techniques to convert text data into numerical representations. This vector data is then stored in a database in a searchable format. Specifically, each document is quantified as a feature vector and stored in the database.

[0525] 3. Loading and initializing the model

[0526] The server loads a large-scale language model (e.g., Hugging Face's Transformers library). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a vast amount of text data. Once loaded, the model is ready to generate answers.

[0527] 4. Question reception and answer generation

[0528] An employee sends a question to the server via their smartphone. For example, they might send a question like, "How do I update the inventory list?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the inventory management manual.

[0529] Based on the relevant information obtained, the server uses a large-scale language model (e.g., Hugging Face's Transformers) to generate an answer to the question. The generated answer might be something like, "The procedure for updating the inventory list is as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes." The server sends this answer back to the worker's smartphone, and the worker can use it as a reference to proceed with the task.

[0530] Hardware and software

[0531] Hardware: Primarily uses a standard smartphone and a cloud server (e.g., AWS EC2).

[0532] Software: We use Hugging Face Transformers for natural language processing, PostgreSQL as the database management system, and Django as the web framework.

[0533] Specific example

[0534] 1. The worker enters the question "How do I update the inventory list?" into the smartphone app.

[0535] 2. The app sends a question to the server.

[0536] 3. The server searches the database for relevant documents and generates answers using Hugging Face Transformers.

[0537] 4. The server sends the generated response back to the worker's smartphone.

[0538] Example of a prompt

[0539] "How do I update the inventory list?"

[0540] This allows logistics center workers to receive quick and accurate answers to questions that arise during their work, improving overall work efficiency.

[0541] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0542] Step 1:

[0543] The server collects and digitizes internal company documents within the logistics center. Specifically, it scans paper documents such as operational manuals, reports, and guidelines, and converts them into text data using optical character recognition (OCR) technology. The input is paper documents, and the output is digital text data. This data is temporarily stored in memory or the file system.

[0544] Step 2:

[0545] The server converts the collected text data into vector data. Specifically, it uses natural language processing techniques to convert the text data into numerical representations (feature vectors). The input is digital text data, and the output is vector data. This vector data is stored in a searchable database.

[0546] Step 3:

[0547] The server loads and initializes a large-scale language model. Here, the Hugging Face Transformers library is used. The input is pre-trained model data, and the output is the usable language model. This model is used for natural language understanding and generation.

[0548] Step 4:

[0549] The user (worker) inputs the question into the terminal via their smartphone and sends it to the server. The input is the text data of the question, and the output received by the server is the question content.

[0550] Step 5:

[0551] The server analyzes the received question and retrieves relevant information from the vector database based on that analysis. The input is the question's text data and the vector database, and the output is the relevant information (a specific section of the document).

[0552] Step 6:

[0553] The server generates answers to questions using a large-scale language model based on relevant information. The input consists of text data of the relevant information and the question, and the output generated by the model is a natural language answer.

[0554] Step 7:

[0555] The server sends the generated response back to the user's smartphone. The input is the text data of the generated response, and the output is the response displayed to the user.

[0556] Specific examples of operation:

[0557] 1. The worker types the question "How do I update the inventory list?" into their smartphone and sends it.

[0558] 2. The server receives the question and searches the database for relevant information on the corresponding internal document.

[0559] 3. The server uses a large-scale language model to generate an answer to the question, producing the following response: "The steps to update the inventory list are as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes."

[0560] 4. The server sends the generated response back to the worker's smartphone, allowing the worker to use it as a reference to proceed with their work.

[0561] Example prompt: "How do I update the inventory list?"

[0562] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0563] This invention provides a system that generates appropriate and emotion-responsive answers to user questions by combining a system that collects internal company documents, stores them in an electronic format, converts them into vector data, and stores them in a database with an emotion engine that recognizes user emotions. Specific embodiments are shown below.

[0564] System program

[0565] 1. Data collection (collection and digitization of internal company documents)

[0566] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, and other documents from various departments and systems using manual or automated scripts. The collected documents are stored in various formats, including PDF, Word, and paper copies. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[0567] 2. Generating vector data

[0568] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is converted into numerical representations (vectors). The vector data captures the semantic features of the text content and is stored in a searchable database.

[0569] 3. Loading and initializing the model

[0570] The server loads a Large-Scale Language Model (LLM) and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[0571] 4. Introduction of an emotional engine

[0572] The server will also incorporate an emotion engine that recognizes user emotions. This emotion engine analyzes the word choices and sentence structure in the user's questions to identify their emotions. In addition, by integrating voice input and facial recognition technology, it will also be able to analyze voice and facial expression data.

[0573] 5. Question reception and sentiment analysis

[0574] The user enters a question from their device. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion engine to identify the user's emotion (e.g., confusion, excitement, anger).

[0575] 6. Searching for related information

[0576] The server searches a vector database to identify documents and information related to the query. Using vector data allows for effective retrieval of semantically relevant information.

[0577] 7. Emotion-based response generation

[0578] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the sentiment engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and polite explanations to confused users.

[0579] 8. Providing responses and user feedback

[0580] The server provides the user with the generated response. The response is sent to the user via their device. The user views the response on their device and obtains specific information about the question. In addition, the sentiment engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[0581] Specific example

[0582] User: Typed "How do I create an invoice?"

[0583] server:

[0584] Receive questions from users.

[0585] The emotion engine analyzes the question content and identifies when the user is confused.

[0586] Search the vector database for information related to invoice creation.

[0587] We use large-scale language models to generate detailed and thorough answers.

[0588] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[0589] User: Check the answer on your device to help solve the problem.

[0590] In this way, the system of the present invention enables efficient management of internal company documents and prompt, emotionally responsive responses to user inquiries.

[0591] The following describes the processing flow.

[0592] Step 1:

[0593] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are in various formats, such as PDF, Word, and paper copies.

[0594] Step 2:

[0595] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[0596] Step 3:

[0597] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert text data into numerical representations (vectors). This allows for the capture of semantic features of the text content.

[0598] Step 4:

[0599] The server stores the vector data in a searchable database. The vector data is stored in a specific database so that it can be searched quickly and efficiently.

[0600] Step 5:

[0601] The server loads the Large-Scale Language Model (LLM) and prepares to operate. This is a crucial component for generating answers to user questions.

[0602] Step 6:

[0603] The server loads an emotion engine that recognizes the user's emotions. The emotion engine analyzes the words and sentence structure entered by the user to identify their emotions.

[0604] Step 7:

[0605] The user enters a question from their device. For example, they might send a question like "How do I create an invoice?" to the server.

[0606] Step 8:

[0607] The server receives a question from the user. The received question is analyzed by the sentiment engine, and the user's emotions are identified. For example, the wording and tone of the question can be used to determine that the user is confused.

[0608] Step 9:

[0609] The server searches the vector database to identify information related to the question. This extracts documents and information that are semantically relevant to the question.

[0610] Step 10:

[0611] The server generates answers to questions using a large-scale language model based on relevant information. The response is then adjusted to an appropriate tone and level of detail based on the user's emotions, as identified by the sentiment engine.

[0612] Step 11:

[0613] The server provides the user with a generated response. The response is sent to the user via their device. For example, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions."

[0614] Step 12:

[0615] The user views the answers on their device and obtains specific information regarding the question. Simultaneously, the sentiment engine monitors the user's emotional changes in real time and adjusts the answers as needed. For example, if the user remains confused, it provides further details.

[0616] Through the above series of steps, the system of the present invention enables efficient management of internal company documents and provides prompt and emotionally responsive answers to user inquiries.

[0617] (Example 2)

[0618] Next, we will describe Example 2. 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".

[0619] Traditional information management systems struggled to provide appropriate and timely answers to user inquiries. Furthermore, they lacked the ability to consider user emotions, resulting in a poor user experience. This often led to users being unable to quickly access the information they needed, ultimately reducing operational efficiency.

[0620] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users and searching the database for information related to the received questions, means for generating answers to questions using a large-scale language model based on the related information, means for providing the generated answers to users, means for analyzing the content of the questions and the emotional state of the users using an emotion recognition engine that recognizes the emotions of the users, and means for generating appropriate answers according to the emotions of the users based on the analysis results of the emotion recognition engine. This makes it possible to provide quick and appropriate answers according to the emotions of users to their questions.

[0621] "Internal documents" refer to documents created and used by a company or organization in connection with its operations, and include work manuals, guidelines, reports, etc.

[0622] "Digitization" refers to the process of converting physical paper documents into a digital format, which is achieved using scanners and OCR technology.

[0623] "Vector data" refers to data obtained by converting text data into numerical representations, and is generated using natural language processing techniques and machine learning algorithms.

[0624] A "searchable database" is a database system designed to store vector data and other types of data, and to enable efficient searching.

[0625] A "large-scale language model" is a machine learning model that excels at understanding and generating natural language, and is used to generate answers to user questions.

[0626] An "emotion recognition engine" is a software module that analyzes user input text, voice, facial expressions, etc., to identify their emotional state.

[0627] "Information related to the question" refers to the documents and data necessary to generate an answer to what the user asked.

[0628] An "appropriate answer" is one that provides accurate and useful information in response to a question, and is also tailored to facilitate communication that takes into account the user's emotional state.

[0629] This invention provides a system that digitizes internal company documents, converts them into vector data for storage, and generates answers to user questions using a large-scale language model. By further combining this with an emotion recognition engine, the system provides appropriate answers that correspond to the user's emotions.

[0630] 1. Data collection (collection and digitization of internal company documents)

[0631] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, etc., from each department and system using manual or automated scripts (e.g., "DocFetch"). The collected documents are saved in various formats, including PDF, Word, and paper. Paper documents are scanned using a scanner (e.g., "ScanMaster") and converted into text data using OCR (Optical Character Recognition) technology (e.g., "TextExtractor"). PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[0632] 2. Generating vector data

[0633] The server converts digitized text data into vector data. Using natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec), it generates vector data that captures the semantic features of the text content. This vector data is then stored in a searchable database (e.g., "VectorDB").

[0634] 3. Loading and initializing the model

[0635] The server loads a large-scale language model (e.g., "GPT-3") and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[0636] 4. Introduction of an emotion recognition engine

[0637] The server also incorporates an emotion recognition engine (e.g., "EmotionDetect") that recognizes the user's emotions. This emotion recognition engine analyzes the word choices and sentence structure included in the user's questions to identify emotions. It also integrates voice input and facial recognition technology, enabling the analysis of voice and facial expression data.

[0638] 5. Question reception and sentiment analysis

[0639] The user enters a question using a terminal. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion recognition engine to identify the user's emotion (e.g., confusion, excitement, anger).

[0640] 6. Searching for related information

[0641] The server searches a searchable database (e.g., "VectorDB") to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[0642] 7. Emotion-based response generation

[0643] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the emotion recognition engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and helpful explanations to confused users.

[0644] 8. Providing responses and user feedback

[0645] The server provides the user with the generated response. The user reviews the response on their device and uses it to solve the problem. Furthermore, the emotion recognition engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[0646] Specific example

[0647] Example prompt:

[0648] User: "Please tell me how to create an invoice."

[0649] server:

[0650] Receive questions from users.

[0651] The emotion recognition engine analyzes the question and identifies that the user is confused.

[0652] Search for information related to invoice creation in the searchable database ("VectorDB").

[0653] Using a large-scale language model ("GPT-3"), detailed and thorough answers are generated.

[0654] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[0655] User: Check the answer on your device to help solve the problem.

[0656] In this way, the system of the present invention can achieve efficient management of internal company documents and prompt, emotionally responsive answers to user inquiries.

[0657] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0658] Step 1:

[0659] This system collects and digitizes internal company documents. The server uses manual or automated scripts (DocFetch) to collect documents such as work manuals, guidelines, and reports from various departments and systems. Paper documents are digitized using a scanner (ScanMaster) and converted into text data using OCR technology (TextExtractor). Input consists of various internal company documents, and output is in text format.

[0660] Specific operation: The server runs the "DocFetch" script to collect documents from shared folders on the network. It then uses the "ScanMaster" application to scan paper documents and converts them to text using the "TextExtractor" OCR system.

[0661] Step 2:

[0662] This process converts digitized text data into vector data. The server uses natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec) to generate vector data that captures the semantic features of the text content. The input is text data, and the output is vector data.

[0663] Specific operation: The server uses the "VectorTransform" module to convert text data into numerical vectors and insert them into "VectorDB".

[0664] Step 3:

[0665] This process loads and initializes a large-scale language model. The server loads and initializes a pre-trained large-scale language model (GPT-3), preparing it to generate answers to user questions. The input is the language model configuration information, and the output is the initialized model.

[0666] Specific operation: The server executes the "ModelLoader" script, loads the "GPT-3" model into memory, and performs initialization.

[0667] Step 4:

[0668] An emotion recognition engine will be implemented. The server will run an emotion recognition engine (EmotionDetect) that recognizes user emotions, adding a function to identify emotions by analyzing the word choices and sentence structure included in the user's questions. Voice input and facial recognition technology will also be integrated, enabling the analysis of voice and facial expression data. Input will consist of user questions and voice / facial expression data, and output will be data with emotion tags attached.

[0669] Specific operation: The server runs the "EmotionDetect" module and works in conjunction with the "FaceReader" and "VoiceAnalyzer" modules to analyze text, voice, and facial expression data in real time.

[0670] Step 5:

[0671] The system receives questions from users and performs sentiment analysis. Users input questions using a terminal, which the server receives and analyzes using a sentiment recognition engine. The input is the user's question, and the output is question data with analyzed sentiment tags.

[0672] Specific operation: The user enters a question using the "HelpDesk" application, and the server analyzes it using the "EmotionDetect" module and assigns an emotion tag.

[0673] Step 6:

[0674] This function searches for related information. The server searches "VectorDB" based on question data tagged with sentiment, and identifies information related to the question. The input is question data tagged with sentiment, and the output is the extracted related information.

[0675] Specific operation: The server uses the "SearchEngine" module to search "VectorDB" and retrieve relevant information.

[0676] Step 7:

[0677] It generates responses that respond to emotions. The server uses a large-scale language model based on relevant information to generate appropriate responses that reflect the user's emotions. The input is question data with relevant information and emotion tags, and the output is adjusted response data.

[0678] Specific operation: The server generates responses using the "GPT-3" model and adjusts the responses according to emotions using the "EmotionFilter" module.

[0679] Step 8:

[0680] The system provides answers and user feedback. The server delivers the generated answers to the user via the terminal. The user reviews the answers on the terminal and provides feedback if necessary. The emotion recognition engine monitors the user's emotional changes in real time and adjusts the answers as needed. The input is the generated answer data and user feedback, and the output is the answer, if possible, further adjusted.

[0681] Specific operation: The server uses the "ResponseSender" module to send the response to the terminal, and the user views the response in the "HelpDesk" application and enters feedback as needed. The "EmotionTracker" module analyzes the feedback in real time and readjusts the response as necessary.

[0682] (Application Example 2)

[0683] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0684] Traditional information delivery systems lacked the ability to generate appropriate responses that reflected user emotions, and the technology to make the user experience personal and effective was absent. This resulted in users consistently receiving unsatisfactory responses. Furthermore, even content recommendation systems struggled to address individual user needs based on their emotions.

[0685] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users, analyzing the received questions with an emotion analysis engine to identify the user's emotions, means for generating answers to the questions using a large-scale language model based on relevant information, means for adjusting the generated answers to match the user's emotions and providing them, means for monitoring changes in the user's emotions in real time and readjusting the answers as necessary, and means for recommending content that matches the emotions. This makes it possible to provide appropriate answers and content recommendations that take emotions into consideration for each user.

[0686] "Internal documents" refer to documents such as reports, operational manuals, and guidelines that are created and managed within an organization.

[0687] "Digitization" refers to the technology of converting paper documents and other analog information into digital data.

[0688] "Vector data" is a data format that converts text data into a numerical representation, making it easier for computers to process and analyze.

[0689] A "database" is a system that structures and stores text data and other information so that it can be efficiently managed and searched.

[0690] "User questions" are input data that represent the information users want to know or the problems they want to solve when using the system.

[0691] "Receiving a question" means receiving a question submitted by a user and taking it in for processing.

[0692] An "emotion analysis engine" is an analytical tool that identifies emotions from user input data (text, voice, facial expressions).

[0693] A "large-scale language model" is a natural language processing model trained on vast amounts of text data, and is an algorithm capable of generating and understanding language.

[0694] "Answer generation" is the process by which a system creates an appropriate answer to a user's question based on relevant information.

[0695] "Adjusting to emotions" means changing the content and tone of the generated response according to the user's emotional state.

[0696] "Real-time monitoring" refers to constantly monitoring the user's emotional state and taking action as needed.

[0697] "Content recommendation" is a process that suggests entertainment and information, such as movies, music, and articles, based on the user's emotional state.

[0698] This invention is a system that provides appropriate responses and recommends content based on the user's emotions. Specific embodiments are shown below.

[0699] System program

[0700] 1. Collection and digitization of internal company documents

[0701] First, the server collects internal documents from each department and system. This is done manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[0702] 2. Generating vector data

[0703] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is transformed into numerical representations (vectors). This vector data captures the semantic features of the text content and is stored in a searchable database.

[0704] 3. Question reception and sentiment analysis

[0705] The user enters a question from their device. For example, they might send a question like, "I'm tired today, but I want to watch a relaxing movie." The server analyzes the received question using an emotion analysis engine to identify the user's emotions (e.g., fatigue, confusion, excitement, anger, etc.). The OpenAI API is used for this emotion analysis.

[0706] 4. Searching for related information and generating answers

[0707] The server searches a vector database to identify documents and information related to the question. Next, it uses a large-scale language model (GPT-3) based on this relevant information to generate an answer to the question. This large-scale language model may include OpenAI's GPT-3. The generated answer is then adjusted according to the user's sentiment.

[0708] 5. Facial Expression and Voice Analysis

[0709] Furthermore, the server utilizes cameras and microphones to analyze the user's facial expressions and voice. By using the OpenCV library to analyze facial expressions and the SpeechRecognition library to convert speech to text, the system captures the user's emotions from multiple perspectives.

[0710] 6. Content Recommendation and Delivery

[0711] Based on the results of sentiment analysis, the server recommends content. If the user is fatigued, relaxing movies or music articles will be recommended. For example, specific content might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[0712] Specific example

[0713] Prompt example:

[0714] 1. Sentiment Analysis: "Analyze the sentiment of the following sentence and label it: 'I'm tired today, but I want to watch a relaxing movie.'"

[0715] 2. Content Recommendation: "Please recommend three relaxing movies that you would suggest to a user who is feeling tired."

[0716] By following these specific steps, the server can provide appropriate responses and content recommendations that take each user's emotions into consideration.

[0717] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0718] Step 1:

[0719] The server collects internal company documents and stores them in electronic format. Specifically, documents are collected manually or using automated scripts from each department and system. If the collected documents are in PDF or Word format, they are directly converted into text data. Paper documents are scanned and converted into text data using OCR technology. As a result, all documents are temporarily stored in text format.

[0720] input:

[0721] PDF, Word, and paper-based internal documents

[0722] Data processing:

[0723] Digitization of paper documents using scanners

[0724] Text conversion using OCR technology

[0725] output:

[0726] Text-format internal documents

[0727] Step 2:

[0728] The server converts digitized text data into vector data. This uses natural language processing techniques and machine learning algorithms. Specifically, it uses document vectorizers such as the BERT model to convert text data into numerical representations. The converted vector data captures its semantic features and is stored in a searchable database.

[0729] input:

[0730] Text-format internal documents

[0731] Data calculation:

[0732] Vectorization using natural language processing techniques (such as the BERT model)

[0733] output:

[0734] Numerical representation data stored in a vector database

[0735] Step 3:

[0736] The user enters a question from their device. For example, they might enter a question like, "I'm tired today, so I'd like to watch a relaxing movie." The entered question is then sent to the server.

[0737] input:

[0738] User questions (text data)

[0739] Data calculation:

[0740] Receiving questions and preparing for analysis

[0741] output:

[0742] Text data on the server

[0743] Step 4:

[0744] The server analyzes the received question using an emotion analysis engine. It uses the OpenAI API and other tools to identify the user's emotions from the question. Specifically, it sends the question text to the API and retrieves an emotion label.

[0745] input:

[0746] User questions (text data)

[0747] Data calculation:

[0748] Sentiment analysis using OpenAI API, etc.

[0749] output:

[0750] Emotional labels (fatigue, confusion, excitement, anger, etc.)

[0751] Step 5:

[0752] The server searches a vector database to identify documents and information related to the question. Specifically, it uses vector data to effectively search for semantically relevant information. For example, it can find manuals and guidelines related to the user's question.

[0753] input:

[0754] Vector database

[0755] User Questions

[0756] Data calculation:

[0757] Search using vector data

[0758] output:

[0759] Information (documents) related to the question

[0760] Step 6:

[0761] The server generates answers to questions using large-scale language models based on relevant information. This may include OpenAI's GPT-3. Furthermore, it adjusts the answers according to the user's emotions based on sentiment analysis results.

[0762] input:

[0763] Related information (documents)

[0764] Emotional labels

[0765] Data calculation:

[0766] Response generation and adjustment using GPT-3, etc.

[0767] output:

[0768] Emotionally appropriate response

[0769] Step 7:

[0770] The server sends the generated response to the device and provides it to the user. Furthermore, the server monitors the user's emotional changes in real time and readjusts the response as needed. To do this, the device uses its camera and microphone to send the user's facial expressions and voice to the server.

[0771] input:

[0772] Adjusted answer

[0773] User facial expressions and voice data

[0774] Data calculation:

[0775] Real-time monitoring of emotions and response adjustment

[0776] output:

[0777] Answers provided to users

[0778] Step 8:

[0779] The server recommends content to the user based on the results of sentiment analysis. For example, if the user is tired, it will recommend relaxing movies or music. A specific example might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[0780] input:

[0781] Emotional labels

[0782] Data calculation:

[0783] Content recommendation using large-scale language models

[0784] output:

[0785] Recommended content provided to users

[0786] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0787] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0788] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0789] [Third Embodiment]

[0790] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0791] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0792] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0793] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0794] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0795] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0796] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0797] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0798] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0799] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0800] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0801] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0802] This invention provides a system for collecting internal company documents, storing them in an electronic format, converting them into vector data, and saving them in a database. The aim is to generate rapid and accurate answers to user inquiries using a large-scale language model.

[0803] System program

[0804] 1. Data Collection

[0805] The server first collects internal company documents. These documents include various documents used within the company, such as operational manuals, reports, and guidelines. Because the collected documents are difficult to search in their original form, they are converted into a digital format. For example, paper manuals are scanned and converted into text data using OCR (Optical Character Recognition). The converted data is temporarily stored in memory or on the file system.

[0806] 2. Generating vector data

[0807] Next, the server converts the collected text data into vector data. This is the process of converting text data into a numerical representation, which is necessary for efficiently searching the content of the text. The vector data is generated and stored using natural language processing techniques. Specifically, each document is quantified as a feature vector and stored in a searchable database.

[0808] 3. Loading and initializing the model

[0809] The server loads a Large-Scale Language Model (LLM). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a large amount of text data. Once loaded, the model is ready to operate.

[0810] 4. Question reception and answer generation

[0811] The user sends a question from their terminal to the server. For example, a question like, "How do I create an invoice?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the manual on creating invoices.

[0812] The server, having retrieved relevant information, then uses a large-scale language model to generate an answer to the question. The model takes the question and related information as input and generates a natural language response based on it. After generating the answer, the server returns the result to the user. The user can view and refer to this answer through their terminal.

[0813] Specific example

[0814] User: "How do I create an invoice?"

[0815] server:

[0816] Receive questions from users.

[0817] Search for information on invoice creation on the Vector Store.

[0818] Use a large-scale language model to generate specific answers to questions.

[0819] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount," is returned to the user.

[0820] User: Check the answer on the device and understand the steps to create the invoice.

[0821] In this way, the system of the present invention enables efficient management of internal company documents and prompt responses to user inquiries.

[0822] The following describes the processing flow.

[0823] Step 1:

[0824] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper copies.

[0825] Step 2:

[0826] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR technology. PDF and Word documents are also converted into text data. As a result, all documents are temporarily stored in text format.

[0827] Step 3:

[0828] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert the text data into numerical representations (vectors). The vector data captures the semantic characteristics of the text content.

[0829] Step 4:

[0830] The server stores the vector data in a searchable database. The vector data is stored in a specific database (e.g., Elasticsearch) to allow for quick and efficient searching.

[0831] Step 5:

[0832] The server loads a Large-Scale Language Model (LLM). It loads a pre-trained model and prepares it for operation. This model is responsible for natural language processing in response to the question.

[0833] Step 6:

[0834] The user enters a question from their device. The question might be something like, "How do I create an invoice?"

[0835] Step 7:

[0836] The server receives a question from the user. The received question is then formatted into an appropriate format for analysis.

[0837] Step 8:

[0838] The server searches a vector database to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[0839] Step 9:

[0840] The server uses a large-scale language model to generate answers to questions based on relevant information. The model takes the question and search results as input and outputs answers in natural language.

[0841] Step 10:

[0842] The server provides the user with the generated response. The response is sent to the user via the terminal.

[0843] Step 11:

[0844] Users view the answers on their devices and obtain specific information related to their questions. For example, they can receive immediately useful information such as, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the products and services, 4. Calculate and enter the total amount."

[0845] (Example 1)

[0846] Next, we will describe Example 1. 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."

[0847] In modern businesses, a large volume of internal documents exists, and efficiently collecting, digitizing, and managing these documents is crucial. However, converting paper documents into digital data and saving that data in a searchable format is time-consuming and difficult to do efficiently. Furthermore, generating quick and accurate answers to user inquiries also requires a lot of manual work, making it inefficient.

[0848] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0849] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting paper documents into text data using OCR technology; means for converting the stored documents into vector data using natural language processing technology and storing the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to questions using a generative AI model based on the related information; and means for providing the generated answers to users. This enables efficient management of internal company documents and the generation of quick and accurate answers to questions from users.

[0850] "Internal company documents" refer to various documents created and used within a company or organization, including operational manuals, reports, and guidelines.

[0851] "Means of collection" refers to the general term for methods and equipment used to collect and consolidate internal company documents in one place.

[0852] "Digitization methods" is a general term for technologies and devices used to convert paper documents into digital format, including scanners and OCR (optical character recognition) technology.

[0853] "OCR technology" is a technology that optically recognizes characters and converts paper documents into text data.

[0854] "Text data" refers to character information represented in digital format.

[0855] "Natural language processing technology" is a general term for technologies that process and understand human language using computers.

[0856] "Vector data" refers to text and other data that have been converted into numerical representations, making them easier to search and analyze.

[0857] A "searchable database" is a database system designed to allow for quick and efficient retrieval of stored data.

[0858] A "generative AI model" is a model that uses artificial intelligence (AI) to generate new information or answers based on input data.

[0859] "Question receiving means" refers to the general term for interfaces and functions used to receive questions submitted by users.

[0860] "Question analysis means" refers to technologies and methods for understanding the content of a received question and identifying the information necessary to generate an appropriate answer.

[0861] "Answer generation means" refers to the technologies and methods used to generate answers to questions, and often involves the use of AI models.

[0862] "User-provided means" refers to the collective technologies and functions used to provide generated answers to users.

[0863] This invention is a system that efficiently collects, digitizes, and manages a wide variety of internal company documents, and further generates quick and accurate answers to user inquiries. This system mainly consists of three elements: a server, terminals, and users.

[0864] The server first collects internal company documents. These documents include various types of documents used within the company, such as operational manuals, reports, and guidelines. The server uses a dedicated scanner to capture paper documents as digital images and converts these images into text data using OCR (Optical Character Recognition) software (e.g., Tesseract OCR). The converted text data is temporarily stored in memory or storage.

[0865] Next, the server converts the collected text data into vector data using natural language processing techniques (e.g., spaCy or the transformers library). This vector data is a numerical representation of the text data, enabling efficient searching. The converted vector data is stored in a searchable database (e.g., Elasticsearch).

[0866] The server also loads and initializes large-scale language models (e.g., the generative AI model OpenAI GPT-3), which are excellent models for natural language understanding and response generation. Once loaded, the models become operational on the server.

[0867] When a user sends a question from their device, the server receives and analyzes it. Specifically, suppose a user sends the question, "How do I create an invoice?" The server analyzes this question and searches for relevant information in its vector database. For example, it might use Elasticsearch to search for vector data containing information about creating invoices.

[0868] Based on the searched relevant information, the server uses a generative AI model to generate specific answers to the questions. The generated answers are returned to the user in the format of: "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount." The user can review and refer to these answers through their device.

[0869] In this way, the system of the present invention enables the efficient digitization and management of internal company documents, as well as the rapid and accurate generation of answers to user inquiries. Users can obtain detailed answers simply by asking questions, thereby improving the efficiency of their work.

[0870] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0871] Step 1:

[0872] The server collects internal company documents. It receives paper documents such as operational manuals, reports, and guidelines as input. The server uses a scanner to capture these paper documents as digital images and uses OCR (Optical Character Recognition) software to extract text data from the images. The extracted text data is temporarily stored in memory or storage as output. Specifically, the server operates a dedicated scanner to generate digital images. Then, it launches OCR software to analyze the character information within the images and convert it into text data.

[0873] Step 2:

[0874] The server converts collected text data into vector data using natural language processing techniques. It receives text data as input and uses natural language processing libraries (e.g., spaCy, transformers) to convert the text into numerical vectors. The generated vector data is stored in a searchable database (e.g., Elasticsearch). Specifically, the server loads text data, applies natural language processing algorithms to generate feature vectors, and then inserts the generated vectors into the database.

[0875] Step 3:

[0876] The server loads and initializes large-scale language models. It receives model files and configuration data as input. It imports the generative AI model library (e.g., OpenAI GPT-3), loads the model into memory, and initializes it. The output is a generative AI model ready to run. Specifically, the server loads the model files from the specified path, performs the initialization procedure, and makes the model usable.

[0877] Step 4:

[0878] The user enters a question from their terminal and sends it to the server. The server receives the user's question as input. For example, the user might send a question like, "How do I create an invoice?" The server analyzes the received question and queries a vector database using the question as input. The server then retrieves relevant information as output from this query. Specifically, the server analyzes the question content and extracts relevant keywords. It then sends queries to a search engine (e.g., Elasticsearch) to retrieve relevant information.

[0879] Step 5:

[0880] The server generates an answer using a generative AI model based on the relevant information it has acquired. The relevant information and the question are input to the generative AI model. The generative AI model processes the information and generates an appropriate answer. The output is the answer to the user's question. Specifically, the server provides the generative AI model with the question and relevant information as input and executes the process of generating an answer.

[0881] Step 6:

[0882] The server provides the user with the generated response. It receives the generated response as input and sends it to the terminal. The output is the response displayed on the user's terminal. Specifically, the server processes the generated response data and sends it to the user's terminal, which then displays this response on its screen.

[0883] The above outlines the specific processing flow of the system. This system is designed to achieve efficient document management and rapid response generation.

[0884] (Application Example 1)

[0885] Next, we will explain Application Example 1. In the following explanation, 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."

[0886] In logistics centers, many workers perform their tasks based on various documents and manuals on a daily basis, but it is difficult to quickly search for this information and obtain answers. Furthermore, in order to proceed with work efficiently, there is a need for a means to immediately resolve questions and problems that arise during work. In this situation, a system is needed that efficiently manages internal documents and provides quick and accurate answers to questions from workers.

[0887] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0888] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting the stored documents into vector data and saving the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to the questions using a large-scale language model based on the related information; means for providing the generated answers to users; and means for logistics center workers to input questions via smartphone and receive immediate answers to those questions. This enables workers to obtain quick and accurate answers to questions and problems that arise during their work.

[0889] "Internal company documents" refer to all types of documents used within a company, such as operational manuals, reports, and guidelines.

[0890] "Digitization" refers to the process of scanning paper documents and converting them into a digital format using optical character recognition (OCR) technology.

[0891] "Vector data" is a numerical representation of text data, and is used to efficiently search the contents of a document.

[0892] A "database" refers to an information system that stores and manages digital documents and vector data in a searchable format.

[0893] A "user" refers to a person or worker who uses the system to input questions.

[0894] A "question" refers to the information a user seeks from the system or the problem they want to solve.

[0895] A "large-scale language model" is a model that has been trained on a vast amount of text data based on natural language processing technology, and is used to generate answers to questions.

[0896] A "smartphone" is a type of portable mobile phone that allows for internet connectivity and advanced operations using applications.

[0897] A "logistics center" refers to a facility where logistics operations such as storing, organizing, and shipping goods are carried out.

[0898] This invention aims to improve operational efficiency in logistics centers, and to that end, provides a system that efficiently manages internal documents and responds quickly to questions from workers.

[0899] System program

[0900] 1. Data collection and digitization

[0901] The server first collects internal company documents within the logistics center. These include operational manuals, reports, and guidelines. If the collected documents are in paper format, they are converted to digital format by scanning and then converted into text data using optical character recognition (OCR) technology. The converted data is temporarily stored in memory or on the file system.

[0902] 2. Generating vector data

[0903] Next, the server converts the collected text data into vector data. This is a process that utilizes natural language processing techniques to convert text data into numerical representations. This vector data is then stored in a database in a searchable format. Specifically, each document is quantified as a feature vector and stored in the database.

[0904] 3. Loading and initializing the model

[0905] The server loads a large-scale language model (e.g., Hugging Face's Transformers library). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a vast amount of text data. Once loaded, the model is ready to generate answers.

[0906] 4. Question reception and answer generation

[0907] An employee sends a question to the server via their smartphone. For example, they might send a question like, "How do I update the inventory list?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the inventory management manual.

[0908] Based on the relevant information obtained, the server uses a large-scale language model (e.g., Hugging Face's Transformers) to generate an answer to the question. The generated answer might be something like, "The procedure for updating the inventory list is as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes." The server sends this answer back to the worker's smartphone, and the worker can use it as a reference to proceed with the task.

[0909] Hardware and software

[0910] Hardware: Primarily uses a standard smartphone and a cloud server (e.g., AWS EC2).

[0911] Software: We use Hugging Face Transformers for natural language processing, PostgreSQL as the database management system, and Django as the web framework.

[0912] Specific example

[0913] 1. The worker enters the question "How do I update the inventory list?" into the smartphone app.

[0914] 2. The app sends a question to the server.

[0915] 3. The server searches the database for relevant documents and generates answers using Hugging Face Transformers.

[0916] 4. The server sends the generated response back to the worker's smartphone.

[0917] Example of a prompt

[0918] "How do I update the inventory list?"

[0919] This allows logistics center workers to receive quick and accurate answers to questions that arise during their work, improving overall work efficiency.

[0920] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0921] Step 1:

[0922] The server collects and digitizes internal company documents within the logistics center. Specifically, it scans paper documents such as operational manuals, reports, and guidelines, and converts them into text data using optical character recognition (OCR) technology. The input is paper documents, and the output is digital text data. This data is temporarily stored in memory or the file system.

[0923] Step 2:

[0924] The server converts the collected text data into vector data. Specifically, it uses natural language processing techniques to convert the text data into numerical representations (feature vectors). The input is digital text data, and the output is vector data. This vector data is stored in a searchable database.

[0925] Step 3:

[0926] The server loads and initializes a large-scale language model. Here, the Hugging Face Transformers library is used. The input is pre-trained model data, and the output is the usable language model. This model is used for natural language understanding and generation.

[0927] Step 4:

[0928] The user (worker) inputs the question into the terminal via their smartphone and sends it to the server. The input is the text data of the question, and the output received by the server is the question content.

[0929] Step 5:

[0930] The server analyzes the received question and retrieves relevant information from the vector database based on that analysis. The input is the question's text data and the vector database, and the output is the relevant information (a specific section of the document).

[0931] Step 6:

[0932] The server generates answers to questions using a large-scale language model based on relevant information. The input consists of text data of the relevant information and the question, and the output generated by the model is a natural language answer.

[0933] Step 7:

[0934] The server sends the generated response back to the user's smartphone. The input is the text data of the generated response, and the output is the response displayed to the user.

[0935] Specific examples of operation:

[0936] 1. The worker types the question "How do I update the inventory list?" into their smartphone and sends it.

[0937] 2. The server receives the question and searches the database for relevant information on the corresponding internal document.

[0938] 3. The server uses a large-scale language model to generate an answer to the question, producing the following response: "The steps to update the inventory list are as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes."

[0939] 4. The server sends the generated response back to the worker's smartphone, allowing the worker to use it as a reference to proceed with their work.

[0940] Example prompt: "How do I update the inventory list?"

[0941] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0942] This invention provides a system that generates appropriate and emotion-responsive answers to user questions by combining a system that collects internal company documents, stores them in an electronic format, converts them into vector data, and stores them in a database with an emotion engine that recognizes user emotions. Specific embodiments are shown below.

[0943] System program

[0944] 1. Data collection (collection and digitization of internal company documents)

[0945] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, and other documents from various departments and systems using manual or automated scripts. The collected documents are stored in various formats, including PDF, Word, and paper copies. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[0946] 2. Generating vector data

[0947] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is converted into numerical representations (vectors). The vector data captures the semantic features of the text content and is stored in a searchable database.

[0948] 3. Loading and initializing the model

[0949] The server loads a Large-Scale Language Model (LLM) and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[0950] 4. Introduction of an emotional engine

[0951] The server will also incorporate an emotion engine that recognizes user emotions. This emotion engine analyzes the word choices and sentence structure in the user's questions to identify their emotions. In addition, by integrating voice input and facial recognition technology, it will also be able to analyze voice and facial expression data.

[0952] 5. Question reception and sentiment analysis

[0953] The user enters a question from their device. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion engine to identify the user's emotion (e.g., confusion, excitement, anger).

[0954] 6. Searching for related information

[0955] The server searches a vector database to identify documents and information related to the query. Using vector data allows for effective retrieval of semantically relevant information.

[0956] 7. Emotion-based response generation

[0957] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the sentiment engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and polite explanations to confused users.

[0958] 8. Providing responses and user feedback

[0959] The server provides the user with the generated response. The response is sent to the user via their device. The user views the response on their device and obtains specific information about the question. In addition, the sentiment engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[0960] Specific example

[0961] User: Typed "How do I create an invoice?"

[0962] server:

[0963] Receive questions from users.

[0964] The emotion engine analyzes the question content and identifies when the user is confused.

[0965] Search the vector database for information related to invoice creation.

[0966] We use large-scale language models to generate detailed and thorough answers.

[0967] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[0968] User: Check the answer on your device to help solve the problem.

[0969] In this way, the system of the present invention enables efficient management of internal company documents and prompt, emotionally responsive responses to user inquiries.

[0970] The following describes the processing flow.

[0971] Step 1:

[0972] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are in various formats, such as PDF, Word, and paper copies.

[0973] Step 2:

[0974] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[0975] Step 3:

[0976] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert text data into numerical representations (vectors). This allows for the capture of semantic features of the text content.

[0977] Step 4:

[0978] The server stores the vector data in a searchable database. The vector data is stored in a specific database so that it can be searched quickly and efficiently.

[0979] Step 5:

[0980] The server loads the Large-Scale Language Model (LLM) and prepares to operate. This is a crucial component for generating answers to user questions.

[0981] Step 6:

[0982] The server loads an emotion engine that recognizes the user's emotions. The emotion engine analyzes the words and sentence structure entered by the user to identify their emotions.

[0983] Step 7:

[0984] The user enters a question from their device. For example, they might send a question like "How do I create an invoice?" to the server.

[0985] Step 8:

[0986] The server receives a question from the user. The received question is analyzed by the sentiment engine, and the user's emotions are identified. For example, the wording and tone of the question can be used to determine that the user is confused.

[0987] Step 9:

[0988] The server searches the vector database to identify information related to the question. This extracts documents and information that are semantically relevant to the question.

[0989] Step 10:

[0990] The server generates answers to questions using a large-scale language model based on relevant information. The response is then adjusted to an appropriate tone and level of detail based on the user's emotions, as identified by the sentiment engine.

[0991] Step 11:

[0992] The server provides the user with a generated response. The response is sent to the user via their device. For example, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions."

[0993] Step 12:

[0994] The user views the answers on their device and obtains specific information regarding the question. Simultaneously, the sentiment engine monitors the user's emotional changes in real time and adjusts the answers as needed. For example, if the user remains confused, it provides further details.

[0995] Through the above series of steps, the system of the present invention enables efficient management of internal company documents and provides prompt and emotionally responsive answers to user inquiries.

[0996] (Example 2)

[0997] Next, we will describe Example 2. 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."

[0998] Traditional information management systems struggled to provide appropriate and timely answers to user inquiries. Furthermore, they lacked the ability to consider user emotions, resulting in a poor user experience. This often led to users being unable to quickly access the information they needed, ultimately reducing operational efficiency.

[0999] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users and searching the database for information related to the received questions, means for generating answers to questions using a large-scale language model based on the related information, means for providing the generated answers to users, means for analyzing the content of the questions and the emotional state of the users using an emotion recognition engine that recognizes the emotions of the users, and means for generating appropriate answers according to the emotions of the users based on the analysis results of the emotion recognition engine. This makes it possible to provide quick and appropriate answers according to the emotions of users to their questions.

[1000] "Internal documents" refer to documents created and used by a company or organization in connection with its operations, and include work manuals, guidelines, reports, etc.

[1001] "Digitization" refers to the process of converting physical paper documents into a digital format, which is achieved using scanners and OCR technology.

[1002] "Vector data" refers to data obtained by converting text data into numerical representations, and is generated using natural language processing techniques and machine learning algorithms.

[1003] A "searchable database" is a database system designed to store vector data and other types of data, and to enable efficient searching.

[1004] A "large-scale language model" is a machine learning model that excels at understanding and generating natural language, and is used to generate answers to user questions.

[1005] An "emotion recognition engine" is a software module that analyzes user input text, voice, facial expressions, etc., to identify their emotional state.

[1006] "Information related to the question" refers to the documents and data necessary to generate an answer to what the user asked.

[1007] An "appropriate answer" is one that provides accurate and useful information in response to a question, and is also tailored to facilitate communication that takes into account the user's emotional state.

[1008] This invention provides a system that digitizes internal company documents, converts them into vector data for storage, and generates answers to user questions using a large-scale language model. By further combining this with an emotion recognition engine, the system provides appropriate answers that correspond to the user's emotions.

[1009] 1. Data collection (collection and digitization of internal company documents)

[1010] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, etc., from each department and system using manual or automated scripts (e.g., "DocFetch"). The collected documents are saved in various formats, including PDF, Word, and paper. Paper documents are scanned using a scanner (e.g., "ScanMaster") and converted into text data using OCR (Optical Character Recognition) technology (e.g., "TextExtractor"). PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[1011] 2. Generating vector data

[1012] The server converts digitized text data into vector data. Using natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec), it generates vector data that captures the semantic features of the text content. This vector data is then stored in a searchable database (e.g., "VectorDB").

[1013] 3. Loading and initializing the model

[1014] The server loads a large-scale language model (e.g., "GPT-3") and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[1015] 4. Introduction of an emotion recognition engine

[1016] The server also incorporates an emotion recognition engine (e.g., "EmotionDetect") that recognizes the user's emotions. This emotion recognition engine analyzes the word choices and sentence structure included in the user's questions to identify emotions. It also integrates voice input and facial recognition technology, enabling the analysis of voice and facial expression data.

[1017] 5. Question reception and sentiment analysis

[1018] The user enters a question using a terminal. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion recognition engine to identify the user's emotion (e.g., confusion, excitement, anger).

[1019] 6. Searching for related information

[1020] The server searches a searchable database (e.g., "VectorDB") to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[1021] 7. Emotion-based response generation

[1022] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the emotion recognition engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and helpful explanations to confused users.

[1023] 8. Providing responses and user feedback

[1024] The server provides the user with the generated response. The user reviews the response on their device and uses it to solve the problem. Furthermore, the emotion recognition engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[1025] Specific example

[1026] Example prompt:

[1027] User: "Please tell me how to create an invoice."

[1028] server:

[1029] Receive questions from users.

[1030] The emotion recognition engine analyzes the question and identifies that the user is confused.

[1031] Search for information related to invoice creation in the searchable database ("VectorDB").

[1032] Using a large-scale language model ("GPT-3"), detailed and thorough answers are generated.

[1033] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[1034] User: Check the answer on your device to help solve the problem.

[1035] In this way, the system of the present invention can achieve efficient management of internal company documents and prompt, emotionally responsive answers to user inquiries.

[1036] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1037] Step 1:

[1038] This system collects and digitizes internal company documents. The server uses manual or automated scripts (DocFetch) to collect documents such as work manuals, guidelines, and reports from various departments and systems. Paper documents are digitized using a scanner (ScanMaster) and converted into text data using OCR technology (TextExtractor). Input consists of various internal company documents, and output is in text format.

[1039] Specific operation: The server runs the "DocFetch" script to collect documents from shared folders on the network. It then uses the "ScanMaster" application to scan paper documents and converts them to text using the "TextExtractor" OCR system.

[1040] Step 2:

[1041] This process converts digitized text data into vector data. The server uses natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec) to generate vector data that captures the semantic features of the text content. The input is text data, and the output is vector data.

[1042] Specific operation: The server uses the "VectorTransform" module to convert text data into numerical vectors and insert them into "VectorDB".

[1043] Step 3:

[1044] This process loads and initializes a large-scale language model. The server loads and initializes a pre-trained large-scale language model (GPT-3), preparing it to generate answers to user questions. The input is the language model configuration information, and the output is the initialized model.

[1045] Specific operation: The server executes the "ModelLoader" script, loads the "GPT-3" model into memory, and performs initialization.

[1046] Step 4:

[1047] An emotion recognition engine will be implemented. The server will run an emotion recognition engine (EmotionDetect) that recognizes user emotions, adding a function to identify emotions by analyzing the word choices and sentence structure included in the user's questions. Voice input and facial recognition technology will also be integrated, enabling the analysis of voice and facial expression data. Input will consist of user questions and voice / facial expression data, and output will be data with emotion tags attached.

[1048] Specific operation: The server runs the "EmotionDetect" module and works in conjunction with the "FaceReader" and "VoiceAnalyzer" modules to analyze text, voice, and facial expression data in real time.

[1049] Step 5:

[1050] The system receives questions from users and performs sentiment analysis. Users input questions using a terminal, which the server receives and analyzes using a sentiment recognition engine. The input is the user's question, and the output is question data with analyzed sentiment tags.

[1051] Specific operation: The user enters a question using the "HelpDesk" application, and the server analyzes it using the "EmotionDetect" module and assigns an emotion tag.

[1052] Step 6:

[1053] This function searches for related information. The server searches "VectorDB" based on question data tagged with sentiment, and identifies information related to the question. The input is question data tagged with sentiment, and the output is the extracted related information.

[1054] Specific operation: The server uses the "SearchEngine" module to search "VectorDB" and retrieve relevant information.

[1055] Step 7:

[1056] It generates responses that respond to emotions. The server uses a large-scale language model based on relevant information to generate appropriate responses that reflect the user's emotions. The input is question data with relevant information and emotion tags, and the output is adjusted response data.

[1057] Specific operation: The server generates responses using the "GPT-3" model and adjusts the responses according to emotions using the "EmotionFilter" module.

[1058] Step 8:

[1059] The system provides answers and user feedback. The server delivers the generated answers to the user via the terminal. The user reviews the answers on the terminal and provides feedback if necessary. The emotion recognition engine monitors the user's emotional changes in real time and adjusts the answers as needed. The input is the generated answer data and user feedback, and the output is the answer, if possible, further adjusted.

[1060] Specific operation: The server uses the "ResponseSender" module to send the response to the terminal, and the user views the response in the "HelpDesk" application and enters feedback as needed. The "EmotionTracker" module analyzes the feedback in real time and readjusts the response as necessary.

[1061] (Application Example 2)

[1062] Next, we will explain application example 2. In the following explanation, 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."

[1063] Traditional information delivery systems lacked the ability to generate appropriate responses that reflected user emotions, and the technology to make the user experience personal and effective was absent. This resulted in users consistently receiving unsatisfactory responses. Furthermore, even content recommendation systems struggled to address individual user needs based on their emotions.

[1064] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users, analyzing the received questions with an emotion analysis engine to identify the user's emotions, means for generating answers to the questions using a large-scale language model based on relevant information, means for adjusting the generated answers to match the user's emotions and providing them, means for monitoring changes in the user's emotions in real time and readjusting the answers as necessary, and means for recommending content that matches the emotions. This makes it possible to provide appropriate answers and content recommendations that take emotions into consideration for each user.

[1065] "Internal documents" refer to documents such as reports, operational manuals, and guidelines that are created and managed within an organization.

[1066] "Digitization" refers to the technology of converting paper documents and other analog information into digital data.

[1067] "Vector data" is a data format that converts text data into a numerical representation, making it easier for computers to process and analyze.

[1068] A "database" is a system that structures and stores text data and other information so that it can be efficiently managed and searched.

[1069] "User questions" are input data that represent the information users want to know or the problems they want to solve when using the system.

[1070] "Receiving a question" means receiving a question submitted by a user and taking it in for processing.

[1071] An "emotion analysis engine" is an analytical tool that identifies emotions from user input data (text, voice, facial expressions).

[1072] A "large-scale language model" is a natural language processing model trained on vast amounts of text data, and is an algorithm capable of generating and understanding language.

[1073] "Answer generation" is the process by which a system creates an appropriate answer to a user's question based on relevant information.

[1074] "Adjusting to emotions" means changing the content and tone of the generated response according to the user's emotional state.

[1075] "Real-time monitoring" refers to constantly monitoring the user's emotional state and taking action as needed.

[1076] "Content recommendation" is a process that suggests entertainment and information, such as movies, music, and articles, based on the user's emotional state.

[1077] This invention is a system that provides appropriate responses and recommends content based on the user's emotions. Specific embodiments are shown below.

[1078] System program

[1079] 1. Collection and digitization of internal company documents

[1080] First, the server collects internal documents from each department and system. This is done manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[1081] 2. Generating vector data

[1082] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is transformed into numerical representations (vectors). This vector data captures the semantic features of the text content and is stored in a searchable database.

[1083] 3. Question reception and sentiment analysis

[1084] The user enters a question from their device. For example, they might send a question like, "I'm tired today, but I want to watch a relaxing movie." The server analyzes the received question using an emotion analysis engine to identify the user's emotions (e.g., fatigue, confusion, excitement, anger, etc.). The OpenAI API is used for this emotion analysis.

[1085] 4. Searching for related information and generating answers

[1086] The server searches a vector database to identify documents and information related to the question. Next, it uses a large-scale language model (GPT-3) based on this relevant information to generate an answer to the question. This large-scale language model may include OpenAI's GPT-3. The generated answer is then adjusted according to the user's sentiment.

[1087] 5. Facial Expression and Voice Analysis

[1088] Furthermore, the server utilizes cameras and microphones to analyze the user's facial expressions and voice. By using the OpenCV library to analyze facial expressions and the SpeechRecognition library to convert speech to text, the system captures the user's emotions from multiple perspectives.

[1089] 6. Content Recommendation and Delivery

[1090] Based on the results of sentiment analysis, the server recommends content. If the user is fatigued, relaxing movies or music articles will be recommended. For example, specific content might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[1091] Specific example

[1092] Prompt example:

[1093] 1. Sentiment Analysis: "Analyze the sentiment of the following sentence and label it: 'I'm tired today, but I want to watch a relaxing movie.'"

[1094] 2. Content Recommendation: "Please recommend three relaxing movies that you would suggest to a user who is feeling tired."

[1095] By following these specific steps, the server can provide appropriate responses and content recommendations that take each user's emotions into consideration.

[1096] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1097] Step 1:

[1098] The server collects internal company documents and stores them in electronic format. Specifically, documents are collected manually or using automated scripts from each department and system. If the collected documents are in PDF or Word format, they are directly converted into text data. Paper documents are scanned and converted into text data using OCR technology. As a result, all documents are temporarily stored in text format.

[1099] input:

[1100] PDF, Word, and paper-based internal documents

[1101] Data processing:

[1102] Digitization of paper documents using scanners

[1103] Text conversion using OCR technology

[1104] output:

[1105] Text-format internal documents

[1106] Step 2:

[1107] The server converts digitized text data into vector data. This uses natural language processing techniques and machine learning algorithms. Specifically, it uses document vectorizers such as the BERT model to convert text data into numerical representations. The converted vector data captures its semantic features and is stored in a searchable database.

[1108] input:

[1109] Text-format internal documents

[1110] Data calculation:

[1111] Vectorization using natural language processing techniques (such as the BERT model)

[1112] output:

[1113] Numerical representation data stored in a vector database

[1114] Step 3:

[1115] The user enters a question from their device. For example, they might enter a question like, "I'm tired today, so I'd like to watch a relaxing movie." The entered question is then sent to the server.

[1116] input:

[1117] User questions (text data)

[1118] Data calculation:

[1119] Receiving questions and preparing for analysis

[1120] output:

[1121] Text data on the server

[1122] Step 4:

[1123] The server analyzes the received question using an emotion analysis engine. It uses the OpenAI API and other tools to identify the user's emotions from the question. Specifically, it sends the question text to the API and retrieves an emotion label.

[1124] input:

[1125] User questions (text data)

[1126] Data calculation:

[1127] Sentiment analysis using OpenAI API, etc.

[1128] output:

[1129] Emotional labels (fatigue, confusion, excitement, anger, etc.)

[1130] Step 5:

[1131] The server searches a vector database to identify documents and information related to the question. Specifically, it uses vector data to effectively search for semantically relevant information. For example, it can find manuals and guidelines related to the user's question.

[1132] input:

[1133] Vector database

[1134] User Questions

[1135] Data calculation:

[1136] Search using vector data

[1137] output:

[1138] Information (documents) related to the question

[1139] Step 6:

[1140] The server generates answers to questions using large-scale language models based on relevant information. This may include OpenAI's GPT-3. Furthermore, it adjusts the answers according to the user's emotions based on sentiment analysis results.

[1141] input:

[1142] Related information (documents)

[1143] Emotional labels

[1144] Data calculation:

[1145] Response generation and adjustment using GPT-3, etc.

[1146] output:

[1147] Emotionally appropriate response

[1148] Step 7:

[1149] The server sends the generated response to the device and provides it to the user. Furthermore, the server monitors the user's emotional changes in real time and readjusts the response as needed. To do this, the device uses its camera and microphone to send the user's facial expressions and voice to the server.

[1150] input:

[1151] Adjusted answer

[1152] User facial expressions and voice data

[1153] Data calculation:

[1154] Real-time monitoring of emotions and response adjustment

[1155] output:

[1156] Answers provided to users

[1157] Step 8:

[1158] The server recommends content to the user based on the results of sentiment analysis. For example, if the user is tired, it will recommend relaxing movies or music. A specific example might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[1159] input:

[1160] Emotional labels

[1161] Data calculation:

[1162] Content recommendation using large-scale language models

[1163] output:

[1164] Recommended content provided to users

[1165] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1166] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1167] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[1168] [Fourth Embodiment]

[1169] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[1170] As shown in Figure 7, the 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.

[1171] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1172] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[1173] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[1174] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[1175] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[1176] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive 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 robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[1177] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[1178] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[1179] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1180] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[1181] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1182] This invention provides a system for collecting internal company documents, storing them in an electronic format, converting them into vector data, and saving them in a database. The aim is to generate rapid and accurate answers to user inquiries using a large-scale language model.

[1183] System program

[1184] 1. Data Collection

[1185] The server first collects internal company documents. These documents include various documents used within the company, such as operational manuals, reports, and guidelines. Because the collected documents are difficult to search in their original form, they are converted into a digital format. For example, paper manuals are scanned and converted into text data using OCR (Optical Character Recognition). The converted data is temporarily stored in memory or on the file system.

[1186] 2. Generating vector data

[1187] Next, the server converts the collected text data into vector data. This is the process of converting text data into a numerical representation, which is necessary for efficiently searching the content of the text. The vector data is generated and stored using natural language processing techniques. Specifically, each document is quantified as a feature vector and stored in a searchable database.

[1188] 3. Loading and initializing the model

[1189] The server loads a Large-Scale Language Model (LLM). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a large amount of text data. Once loaded, the model is ready to operate.

[1190] 4. Question reception and answer generation

[1191] The user sends a question from their terminal to the server. For example, a question like, "How do I create an invoice?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the manual on creating invoices.

[1192] The server, having retrieved relevant information, then uses a large-scale language model to generate an answer to the question. The model takes the question and related information as input and generates a natural language response based on it. After generating the answer, the server returns the result to the user. The user can view and refer to this answer through their terminal.

[1193] Specific example

[1194] User: "How do I create an invoice?"

[1195] server:

[1196] Receive questions from users.

[1197] Search for information on invoice creation on the Vector Store.

[1198] Use a large-scale language model to generate specific answers to questions.

[1199] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount," is returned to the user.

[1200] User: Check the answer on the device and understand the steps to create the invoice.

[1201] In this way, the system of the present invention enables efficient management of internal company documents and prompt responses to user inquiries.

[1202] The following describes the processing flow.

[1203] Step 1:

[1204] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper copies.

[1205] Step 2:

[1206] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR technology. PDF and Word documents are also converted into text data. As a result, all documents are temporarily stored in text format.

[1207] Step 3:

[1208] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert the text data into numerical representations (vectors). The vector data captures the semantic characteristics of the text content.

[1209] Step 4:

[1210] The server stores the vector data in a searchable database. The vector data is stored in a specific database (e.g., Elasticsearch) to allow for quick and efficient searching.

[1211] Step 5:

[1212] The server loads a Large-Scale Language Model (LLM). It loads a pre-trained model and prepares it for operation. This model is responsible for natural language processing in response to the question.

[1213] Step 6:

[1214] The user enters a question from their device. The question might be something like, "How do I create an invoice?"

[1215] Step 7:

[1216] The server receives a question from the user. The received question is then formatted into an appropriate format for analysis.

[1217] Step 8:

[1218] The server searches a vector database to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[1219] Step 9:

[1220] The server uses a large-scale language model to generate answers to questions based on relevant information. The model takes the question and search results as input and outputs answers in natural language.

[1221] Step 10:

[1222] The server provides the user with the generated response. The response is sent to the user via the terminal.

[1223] Step 11:

[1224] Users view the answers on their devices and obtain specific information related to their questions. For example, they can receive immediately useful information such as, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the products and services, 4. Calculate and enter the total amount."

[1225] (Example 1)

[1226] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1227] In modern businesses, a large volume of internal documents exists, and efficiently collecting, digitizing, and managing these documents is crucial. However, converting paper documents into digital data and saving that data in a searchable format is time-consuming and difficult to do efficiently. Furthermore, generating quick and accurate answers to user inquiries also requires a lot of manual work, making it inefficient.

[1228] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[1229] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting paper documents into text data using OCR technology; means for converting the stored documents into vector data using natural language processing technology and storing the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to questions using a generative AI model based on the related information; and means for providing the generated answers to users. This enables efficient management of internal company documents and the generation of quick and accurate answers to questions from users.

[1230] "Internal company documents" refer to various documents created and used within a company or organization, including operational manuals, reports, and guidelines.

[1231] "Means of collection" refers to the general term for methods and equipment used to collect and consolidate internal company documents in one place.

[1232] "Digitization methods" is a general term for technologies and devices used to convert paper documents into digital format, including scanners and OCR (optical character recognition) technology.

[1233] "OCR technology" is a technology that optically recognizes characters and converts paper documents into text data.

[1234] "Text data" refers to character information represented in digital format.

[1235] "Natural language processing technology" is a general term for technologies that process and understand human language using computers.

[1236] "Vector data" refers to text and other data that have been converted into numerical representations, making them easier to search and analyze.

[1237] A "searchable database" is a database system designed to allow for quick and efficient retrieval of stored data.

[1238] A "generative AI model" is a model that uses artificial intelligence (AI) to generate new information or answers based on input data.

[1239] "Question receiving means" refers to the general term for interfaces and functions used to receive questions submitted by users.

[1240] "Question analysis means" refers to technologies and methods for understanding the content of a received question and identifying the information necessary to generate an appropriate answer.

[1241] "Answer generation means" refers to the technologies and methods used to generate answers to questions, and often involves the use of AI models.

[1242] "User-provided means" refers to the collective technologies and functions used to provide generated answers to users.

[1243] This invention is a system that efficiently collects, digitizes, and manages a wide variety of internal company documents, and further generates quick and accurate answers to user inquiries. This system mainly consists of three elements: a server, terminals, and users.

[1244] The server first collects internal company documents. These documents include various types of documents used within the company, such as operational manuals, reports, and guidelines. The server uses a dedicated scanner to capture paper documents as digital images and converts these images into text data using OCR (Optical Character Recognition) software (e.g., Tesseract OCR). The converted text data is temporarily stored in memory or storage.

[1245] Next, the server converts the collected text data into vector data using natural language processing techniques (e.g., spaCy or the transformers library). This vector data is a numerical representation of the text data, enabling efficient searching. The converted vector data is stored in a searchable database (e.g., Elasticsearch).

[1246] The server also loads and initializes large-scale language models (e.g., the generative AI model OpenAI GPT-3), which are excellent models for natural language understanding and response generation. Once loaded, the models become operational on the server.

[1247] When a user sends a question from their device, the server receives and analyzes it. Specifically, suppose a user sends the question, "How do I create an invoice?" The server analyzes this question and searches for relevant information in its vector database. For example, it might use Elasticsearch to search for vector data containing information about creating invoices.

[1248] Based on the searched relevant information, the server uses a generative AI model to generate specific answers to the questions. The generated answers are returned to the user in the format of: "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount." The user can review and refer to these answers through their device.

[1249] In this way, the system of the present invention enables the efficient digitization and management of internal company documents, as well as the rapid and accurate generation of answers to user inquiries. Users can obtain detailed answers simply by asking questions, thereby improving the efficiency of their work.

[1250] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1251] Step 1:

[1252] The server collects internal company documents. It receives paper documents such as operational manuals, reports, and guidelines as input. The server uses a scanner to capture these paper documents as digital images and uses OCR (Optical Character Recognition) software to extract text data from the images. The extracted text data is temporarily stored in memory or storage as output. Specifically, the server operates a dedicated scanner to generate digital images. Then, it launches OCR software to analyze the character information within the images and convert it into text data.

[1253] Step 2:

[1254] The server converts collected text data into vector data using natural language processing techniques. It receives text data as input and uses natural language processing libraries (e.g., spaCy, transformers) to convert the text into numerical vectors. The generated vector data is stored in a searchable database (e.g., Elasticsearch). Specifically, the server loads text data, applies natural language processing algorithms to generate feature vectors, and then inserts the generated vectors into the database.

[1255] Step 3:

[1256] The server loads and initializes large-scale language models. It receives model files and configuration data as input. It imports the generative AI model library (e.g., OpenAI GPT-3), loads the model into memory, and initializes it. The output is a generative AI model ready to run. Specifically, the server loads the model files from the specified path, performs the initialization procedure, and makes the model usable.

[1257] Step 4:

[1258] The user enters a question from their terminal and sends it to the server. The server receives the user's question as input. For example, the user might send a question like, "How do I create an invoice?" The server analyzes the received question and queries a vector database using the question as input. The server then retrieves relevant information as output from this query. Specifically, the server analyzes the question content and extracts relevant keywords. It then sends queries to a search engine (e.g., Elasticsearch) to retrieve relevant information.

[1259] Step 5:

[1260] The server generates an answer using a generative AI model based on the relevant information it has acquired. The relevant information and the question are input to the generative AI model. The generative AI model processes the information and generates an appropriate answer. The output is the answer to the user's question. Specifically, the server provides the generative AI model with the question and relevant information as input and executes the process of generating an answer.

[1261] Step 6:

[1262] The server provides the user with the generated response. It receives the generated response as input and sends it to the terminal. The output is the response displayed on the user's terminal. Specifically, the server processes the generated response data and sends it to the user's terminal, which then displays this response on its screen.

[1263] The above outlines the specific processing flow of the system. This system is designed to achieve efficient document management and rapid response generation.

[1264] (Application Example 1)

[1265] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1266] In logistics centers, many workers perform their tasks based on various documents and manuals on a daily basis, but it is difficult to quickly search for this information and obtain answers. Furthermore, in order to proceed with work efficiently, there is a need for a means to immediately resolve questions and problems that arise during work. In this situation, a system is needed that efficiently manages internal documents and provides quick and accurate answers to questions from workers.

[1267] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[1268] In this invention, the server includes means for collecting internal company documents and storing them in an electronic format; means for converting the stored documents into vector data and saving the vector data in a searchable database; means for receiving questions from users and searching the database for information related to the received questions; means for generating answers to the questions using a large-scale language model based on the related information; means for providing the generated answers to users; and means for logistics center workers to input questions via smartphone and receive immediate answers to those questions. This enables workers to obtain quick and accurate answers to questions and problems that arise during their work.

[1269] "Internal company documents" refer to all types of documents used within a company, such as operational manuals, reports, and guidelines.

[1270] "Digitization" refers to the process of scanning paper documents and converting them into a digital format using optical character recognition (OCR) technology.

[1271] "Vector data" is a numerical representation of text data, and is used to efficiently search the contents of a document.

[1272] A "database" refers to an information system that stores and manages digital documents and vector data in a searchable format.

[1273] A "user" refers to a person or worker who uses the system to input questions.

[1274] A "question" refers to the information a user seeks from the system or the problem they want to solve.

[1275] A "large-scale language model" is a model that has been trained on a vast amount of text data based on natural language processing technology, and is used to generate answers to questions.

[1276] A "smartphone" is a type of portable mobile phone that allows for internet connectivity and advanced operations using applications.

[1277] A "logistics center" refers to a facility where logistics operations such as storing, organizing, and shipping goods are carried out.

[1278] This invention aims to improve operational efficiency in logistics centers, and to that end, provides a system that efficiently manages internal documents and responds quickly to questions from workers.

[1279] System program

[1280] 1. Data collection and digitization

[1281] The server first collects internal company documents within the logistics center. These include operational manuals, reports, and guidelines. If the collected documents are in paper format, they are converted to digital format by scanning and then converted into text data using optical character recognition (OCR) technology. The converted data is temporarily stored in memory or on the file system.

[1282] 2. Generating vector data

[1283] Next, the server converts the collected text data into vector data. This is a process that utilizes natural language processing techniques to convert text data into numerical representations. This vector data is then stored in a database in a searchable format. Specifically, each document is quantified as a feature vector and stored in the database.

[1284] 3. Loading and initializing the model

[1285] The server loads a large-scale language model (e.g., Hugging Face's Transformers library). This model excels at natural language understanding and generation and is used to generate answers to user questions. The model is pre-trained and has learned from a vast amount of text data. Once loaded, the model is ready to generate answers.

[1286] 4. Question reception and answer generation

[1287] An employee sends a question to the server via their smartphone. For example, they might send a question like, "How do I update the inventory list?" The server analyzes the received question and searches its vector database for information related to that question. This related information might include sections of the inventory management manual.

[1288] Based on the relevant information obtained, the server uses a large-scale language model (e.g., Hugging Face's Transformers) to generate an answer to the question. The generated answer might be something like, "The procedure for updating the inventory list is as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes." The server sends this answer back to the worker's smartphone, and the worker can use it as a reference to proceed with the task.

[1289] Hardware and software

[1290] Hardware: Primarily uses a standard smartphone and a cloud server (e.g., AWS EC2).

[1291] Software: We use Hugging Face Transformers for natural language processing, PostgreSQL as the database management system, and Django as the web framework.

[1292] Specific example

[1293] 1. The worker enters the question "How do I update the inventory list?" into the smartphone app.

[1294] 2. The app sends a question to the server.

[1295] 3. The server searches the database for relevant documents and generates answers using Hugging Face Transformers.

[1296] 4. The server sends the generated response back to the worker's smartphone.

[1297] Example of a prompt

[1298] "How do I update the inventory list?"

[1299] This allows logistics center workers to receive quick and accurate answers to questions that arise during their work, improving overall work efficiency.

[1300] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1301] Step 1:

[1302] The server collects and digitizes internal company documents within the logistics center. Specifically, it scans paper documents such as operational manuals, reports, and guidelines, and converts them into text data using optical character recognition (OCR) technology. The input is paper documents, and the output is digital text data. This data is temporarily stored in memory or the file system.

[1303] Step 2:

[1304] The server converts the collected text data into vector data. Specifically, it uses natural language processing techniques to convert the text data into numerical representations (feature vectors). The input is digital text data, and the output is vector data. This vector data is stored in a searchable database.

[1305] Step 3:

[1306] The server loads and initializes a large-scale language model. Here, the Hugging Face Transformers library is used. The input is pre-trained model data, and the output is the usable language model. This model is used for natural language understanding and generation.

[1307] Step 4:

[1308] The user (worker) inputs the question into the terminal via their smartphone and sends it to the server. The input is the text data of the question, and the output received by the server is the question content.

[1309] Step 5:

[1310] The server analyzes the received question and retrieves relevant information from the vector database based on that analysis. The input is the question's text data and the vector database, and the output is the relevant information (a specific section of the document).

[1311] Step 6:

[1312] The server generates answers to questions using a large-scale language model based on relevant information. The input consists of text data of the relevant information and the question, and the output generated by the model is a natural language answer.

[1313] Step 7:

[1314] The server sends the generated response back to the user's smartphone. The input is the text data of the generated response, and the output is the response displayed to the user.

[1315] Specific examples of operation:

[1316] 1. The worker types the question "How do I update the inventory list?" into their smartphone and sends it.

[1317] 2. The server receives the question and searches the database for relevant information on the corresponding internal document.

[1318] 3. The server uses a large-scale language model to generate an answer to the question, producing the following response: "The steps to update the inventory list are as follows: 1. Log in to the system. 2. Go to the inventory list page. 3. Edit the product information. 4. Save the changes."

[1319] 4. The server sends the generated response back to the worker's smartphone, allowing the worker to use it as a reference to proceed with their work.

[1320] Example prompt: "How do I update the inventory list?"

[1321] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1322] This invention provides a system that generates appropriate and emotion-responsive answers to user questions by combining a system that collects internal company documents, stores them in an electronic format, converts them into vector data, and stores them in a database with an emotion engine that recognizes user emotions. Specific embodiments are shown below.

[1323] System program

[1324] 1. Data collection (collection and digitization of internal company documents)

[1325] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, and other documents from various departments and systems using manual or automated scripts. The collected documents are stored in various formats, including PDF, Word, and paper copies. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[1326] 2. Generating vector data

[1327] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is converted into numerical representations (vectors). The vector data captures the semantic features of the text content and is stored in a searchable database.

[1328] 3. Loading and initializing the model

[1329] The server loads a Large-Scale Language Model (LLM) and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[1330] 4. Introduction of an emotional engine

[1331] The server will also incorporate an emotion engine that recognizes user emotions. This emotion engine analyzes the word choices and sentence structure in the user's questions to identify their emotions. In addition, by integrating voice input and facial recognition technology, it will also be able to analyze voice and facial expression data.

[1332] 5. Question reception and sentiment analysis

[1333] The user enters a question from their device. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion engine to identify the user's emotion (e.g., confusion, excitement, anger).

[1334] 6. Searching for related information

[1335] The server searches a vector database to identify documents and information related to the query. Using vector data allows for effective retrieval of semantically relevant information.

[1336] 7. Emotion-based response generation

[1337] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the sentiment engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and polite explanations to confused users.

[1338] 8. Providing responses and user feedback

[1339] The server provides the user with the generated response. The response is sent to the user via their device. The user views the response on their device and obtains specific information about the question. In addition, the sentiment engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[1340] Specific example

[1341] User: Typed "How do I create an invoice?"

[1342] server:

[1343] Receive questions from users.

[1344] The emotion engine analyzes the question content and identifies when the user is confused.

[1345] Search the vector database for information related to invoice creation.

[1346] We use large-scale language models to generate detailed and thorough answers.

[1347] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[1348] User: Check the answer on your device to help solve the problem.

[1349] In this way, the system of the present invention enables efficient management of internal company documents and prompt, emotionally responsive responses to user inquiries.

[1350] The following describes the processing flow.

[1351] Step 1:

[1352] The server collects internal company documents. Documents are collected from each department, either manually or using automated scripts. The collected documents are in various formats, such as PDF, Word, and paper copies.

[1353] Step 2:

[1354] The server digitizes collected internal company documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[1355] Step 3:

[1356] The server converts digitized text data into vector data. Natural language processing techniques and machine learning algorithms are used to convert text data into numerical representations (vectors). This allows for the capture of semantic features of the text content.

[1357] Step 4:

[1358] The server stores the vector data in a searchable database. The vector data is stored in a specific database so that it can be searched quickly and efficiently.

[1359] Step 5:

[1360] The server loads the Large-Scale Language Model (LLM) and prepares to operate. This is a crucial component for generating answers to user questions.

[1361] Step 6:

[1362] The server loads an emotion engine that recognizes the user's emotions. The emotion engine analyzes the words and sentence structure entered by the user to identify their emotions.

[1363] Step 7:

[1364] The user enters a question from their device. For example, they might send a question like "How do I create an invoice?" to the server.

[1365] Step 8:

[1366] The server receives a question from the user. The received question is analyzed by the sentiment engine, and the user's emotions are identified. For example, the wording and tone of the question can be used to determine that the user is confused.

[1367] Step 9:

[1368] The server searches the vector database to identify information related to the question. This extracts documents and information that are semantically relevant to the question.

[1369] Step 10:

[1370] The server generates answers to questions using a large-scale language model based on relevant information. The response is then adjusted to an appropriate tone and level of detail based on the user's emotions, as identified by the sentiment engine.

[1371] Step 11:

[1372] The server provides the user with a generated response. The response is sent to the user via their device. For example, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions."

[1373] Step 12:

[1374] The user views the answers on their device and obtains specific information regarding the question. Simultaneously, the sentiment engine monitors the user's emotional changes in real time and adjusts the answers as needed. For example, if the user remains confused, it provides further details.

[1375] Through the above series of steps, the system of the present invention enables efficient management of internal company documents and provides prompt and emotionally responsive answers to user inquiries.

[1376] (Example 2)

[1377] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1378] Traditional information management systems struggled to provide appropriate and timely answers to user inquiries. Furthermore, they lacked the ability to consider user emotions, resulting in a poor user experience. This often led to users being unable to quickly access the information they needed, ultimately reducing operational efficiency.

[1379] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users and searching the database for information related to the received questions, means for generating answers to questions using a large-scale language model based on the related information, means for providing the generated answers to users, means for analyzing the content of the questions and the emotional state of the users using an emotion recognition engine that recognizes the emotions of the users, and means for generating appropriate answers according to the emotions of the users based on the analysis results of the emotion recognition engine. This makes it possible to provide quick and appropriate answers according to the emotions of users to their questions.

[1380] "Internal documents" refer to documents created and used by a company or organization in connection with its operations, and include work manuals, guidelines, reports, etc.

[1381] "Digitization" refers to the process of converting physical paper documents into a digital format, which is achieved using scanners and OCR technology.

[1382] "Vector data" refers to data obtained by converting text data into numerical representations, and is generated using natural language processing techniques and machine learning algorithms.

[1383] A "searchable database" is a database system designed to store vector data and other types of data, and to enable efficient searching.

[1384] A "large-scale language model" is a machine learning model that excels at understanding and generating natural language, and is used to generate answers to user questions.

[1385] An "emotion recognition engine" is a software module that analyzes user input text, voice, facial expressions, etc., to identify their emotional state.

[1386] "Information related to the question" refers to the documents and data necessary to generate an answer to what the user asked.

[1387] An "appropriate answer" is one that provides accurate and useful information in response to a question, and is also tailored to facilitate communication that takes into account the user's emotional state.

[1388] This invention provides a system that digitizes internal company documents, converts them into vector data for storage, and generates answers to user questions using a large-scale language model. By further combining this with an emotion recognition engine, the system provides appropriate answers that correspond to the user's emotions.

[1389] 1. Data collection (collection and digitization of internal company documents)

[1390] The server first collects internal company documents. It gathers operational manuals, guidelines, reports, etc., from each department and system using manual or automated scripts (e.g., "DocFetch"). The collected documents are saved in various formats, including PDF, Word, and paper. Paper documents are scanned using a scanner (e.g., "ScanMaster") and converted into text data using OCR (Optical Character Recognition) technology (e.g., "TextExtractor"). PDF and Word documents are also converted to text data, and all documents are temporarily stored in text format.

[1391] 2. Generating vector data

[1392] The server converts digitized text data into vector data. Using natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec), it generates vector data that captures the semantic features of the text content. This vector data is then stored in a searchable database (e.g., "VectorDB").

[1393] 3. Loading and initializing the model

[1394] The server loads a large-scale language model (e.g., "GPT-3") and prepares it for operation. This model excels at natural language understanding and generation and is used to generate answers to user questions.

[1395] 4. Introduction of an emotion recognition engine

[1396] The server also incorporates an emotion recognition engine (e.g., "EmotionDetect") that recognizes the user's emotions. This emotion recognition engine analyzes the word choices and sentence structure included in the user's questions to identify emotions. It also integrates voice input and facial recognition technology, enabling the analysis of voice and facial expression data.

[1397] 5. Question reception and sentiment analysis

[1398] The user enters a question using a terminal. For example, they might send a question like, "How do I create an invoice?" The server analyzes the received question using an emotion recognition engine to identify the user's emotion (e.g., confusion, excitement, anger).

[1399] 6. Searching for related information

[1400] The server searches a searchable database (e.g., "VectorDB") to identify documents and information related to the question. Using vector data allows for effective retrieval of semantically relevant information.

[1401] 7. Emotion-based response generation

[1402] The server generates answers to questions using a large-scale language model based on relevant information. Based on the analysis results of the emotion recognition engine, it generates appropriate answers that correspond to the user's emotions. For example, it provides detailed and helpful explanations to confused users.

[1403] 8. Providing responses and user feedback

[1404] The server provides the user with the generated response. The user reviews the response on their device and uses it to solve the problem. Furthermore, the emotion recognition engine monitors the user's emotional changes in real time and adjusts the response if necessary.

[1405] Specific example

[1406] Example prompt:

[1407] User: "Please tell me how to create an invoice."

[1408] server:

[1409] Receive questions from users.

[1410] The emotion recognition engine analyzes the question and identifies that the user is confused.

[1411] Search for information related to invoice creation in the searchable database ("VectorDB").

[1412] Using a large-scale language model ("GPT-3"), detailed and thorough answers are generated.

[1413] The generated response, "Here's how to create an invoice: 1. Open the template, 2. Enter customer information, 3. Fill in the details of the goods and services, 4. Calculate and enter the total amount. If you have any questions, please let us know and we will provide you with more specific instructions," is provided to the user.

[1414] User: Check the answer on your device to help solve the problem.

[1415] In this way, the system of the present invention can achieve efficient management of internal company documents and prompt, emotionally responsive answers to user inquiries.

[1416] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1417] Step 1:

[1418] This system collects and digitizes internal company documents. The server uses manual or automated scripts (DocFetch) to collect documents such as work manuals, guidelines, and reports from various departments and systems. Paper documents are digitized using a scanner (ScanMaster) and converted into text data using OCR technology (TextExtractor). Input consists of various internal company documents, and output is in text format.

[1419] Specific operation: The server runs the "DocFetch" script to collect documents from shared folders on the network. It then uses the "ScanMaster" application to scan paper documents and converts them to text using the "TextExtractor" OCR system.

[1420] Step 2:

[1421] This process converts digitized text data into vector data. The server uses natural language processing techniques and machine learning algorithms (e.g., BERT and Word2Vec) to generate vector data that captures the semantic features of the text content. The input is text data, and the output is vector data.

[1422] Specific operation: The server uses the "VectorTransform" module to convert text data into numerical vectors and insert them into "VectorDB".

[1423] Step 3:

[1424] This process loads and initializes a large-scale language model. The server loads and initializes a pre-trained large-scale language model (GPT-3), preparing it to generate answers to user questions. The input is the language model configuration information, and the output is the initialized model.

[1425] Specific operation: The server executes the "ModelLoader" script, loads the "GPT-3" model into memory, and performs initialization.

[1426] Step 4:

[1427] An emotion recognition engine will be implemented. The server will run an emotion recognition engine (EmotionDetect) that recognizes user emotions, adding a function to identify emotions by analyzing the word choices and sentence structure included in the user's questions. Voice input and facial recognition technology will also be integrated, enabling the analysis of voice and facial expression data. Input will consist of user questions and voice / facial expression data, and output will be data with emotion tags attached.

[1428] Specific operation: The server runs the "EmotionDetect" module and works in conjunction with the "FaceReader" and "VoiceAnalyzer" modules to analyze text, voice, and facial expression data in real time.

[1429] Step 5:

[1430] The system receives questions from users and performs sentiment analysis. Users input questions using a terminal, which the server receives and analyzes using a sentiment recognition engine. The input is the user's question, and the output is question data with analyzed sentiment tags.

[1431] Specific operation: The user enters a question using the "HelpDesk" application, and the server analyzes it using the "EmotionDetect" module and assigns an emotion tag.

[1432] Step 6:

[1433] This function searches for related information. The server searches "VectorDB" based on question data tagged with sentiment, and identifies information related to the question. The input is question data tagged with sentiment, and the output is the extracted related information.

[1434] Specific operation: The server uses the "SearchEngine" module to search "VectorDB" and retrieve relevant information.

[1435] Step 7:

[1436] It generates responses that respond to emotions. The server uses a large-scale language model based on relevant information to generate appropriate responses that reflect the user's emotions. The input is question data with relevant information and emotion tags, and the output is adjusted response data.

[1437] Specific operation: The server generates responses using the "GPT-3" model and adjusts the responses according to emotions using the "EmotionFilter" module.

[1438] Step 8:

[1439] The system provides answers and user feedback. The server delivers the generated answers to the user via the terminal. The user reviews the answers on the terminal and provides feedback if necessary. The emotion recognition engine monitors the user's emotional changes in real time and adjusts the answers as needed. The input is the generated answer data and user feedback, and the output is the answer, if possible, further adjusted.

[1440] Specific operation: The server uses the "ResponseSender" module to send the response to the terminal, and the user views the response in the "HelpDesk" application and enters feedback as needed. The "EmotionTracker" module analyzes the feedback in real time and readjusts the response as necessary.

[1441] (Application Example 2)

[1442] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1443] Traditional information delivery systems lacked the ability to generate appropriate responses that reflected user emotions, and the technology to make the user experience personal and effective was absent. This resulted in users consistently receiving unsatisfactory responses. Furthermore, even content recommendation systems struggled to address individual user needs based on their emotions.

[1444] The identification processing performed by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting internal company documents and storing them in an electronic format, means for converting the stored documents into vector data and storing the vector data in a searchable database, means for receiving questions from users, analyzing the received questions with an emotion analysis engine to identify the user's emotions, means for generating answers to the questions using a large-scale language model based on relevant information, means for adjusting the generated answers to match the user's emotions and providing them, means for monitoring changes in the user's emotions in real time and readjusting the answers as necessary, and means for recommending content that matches the emotions. This makes it possible to provide appropriate answers and content recommendations that take emotions into consideration for each user.

[1445] "Internal documents" refer to documents such as reports, operational manuals, and guidelines that are created and managed within an organization.

[1446] "Digitization" refers to the technology of converting paper documents and other analog information into digital data.

[1447] "Vector data" is a data format that converts text data into a numerical representation, making it easier for computers to process and analyze.

[1448] A "database" is a system that structures and stores text data and other information so that it can be efficiently managed and searched.

[1449] "User questions" are input data that represent the information users want to know or the problems they want to solve when using the system.

[1450] "Receiving a question" means receiving a question submitted by a user and taking it in for processing.

[1451] An "emotion analysis engine" is an analytical tool that identifies emotions from user input data (text, voice, facial expressions).

[1452] A "large-scale language model" is a natural language processing model trained on vast amounts of text data, and is an algorithm capable of generating and understanding language.

[1453] "Answer generation" is the process by which a system creates an appropriate answer to a user's question based on relevant information.

[1454] "Adjusting to emotions" means changing the content and tone of the generated response according to the user's emotional state.

[1455] "Real-time monitoring" refers to constantly monitoring the user's emotional state and taking action as needed.

[1456] "Content recommendation" is a process that suggests entertainment and information, such as movies, music, and articles, based on the user's emotional state.

[1457] This invention is a system that provides appropriate responses and recommends content based on the user's emotions. Specific embodiments are shown below.

[1458] System program

[1459] 1. Collection and digitization of internal company documents

[1460] First, the server collects internal documents from each department and system. This is done manually or using automated scripts. The collected documents are saved in various formats, such as PDF, Word, and paper documents. Paper documents are scanned and converted into text data using OCR (Optical Character Recognition) technology. PDF and Word documents are also converted into text data, and all documents are temporarily stored in text format.

[1461] 2. Generating vector data

[1462] Next, the server converts the digitized text data into vector data. Using natural language processing techniques and machine learning algorithms, the text data is transformed into numerical representations (vectors). This vector data captures the semantic features of the text content and is stored in a searchable database.

[1463] 3. Question reception and sentiment analysis

[1464] The user enters a question from their device. For example, they might send a question like, "I'm tired today, but I want to watch a relaxing movie." The server analyzes the received question using an emotion analysis engine to identify the user's emotions (e.g., fatigue, confusion, excitement, anger, etc.). The OpenAI API is used for this emotion analysis.

[1465] 4. Searching for related information and generating answers

[1466] The server searches a vector database to identify documents and information related to the question. Next, it uses a large-scale language model (GPT-3) based on this relevant information to generate an answer to the question. This large-scale language model may include OpenAI's GPT-3. The generated answer is then adjusted according to the user's sentiment.

[1467] 5. Facial Expression and Voice Analysis

[1468] Furthermore, the server utilizes cameras and microphones to analyze the user's facial expressions and voice. By using the OpenCV library to analyze facial expressions and the SpeechRecognition library to convert speech to text, the system captures the user's emotions from multiple perspectives.

[1469] 6. Content Recommendation and Delivery

[1470] Based on the results of sentiment analysis, the server recommends content. If the user is fatigued, relaxing movies or music articles will be recommended. For example, specific content might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[1471] Specific example

[1472] Prompt example:

[1473] 1. Sentiment Analysis: "Analyze the sentiment of the following sentence and label it: 'I'm tired today, but I want to watch a relaxing movie.'"

[1474] 2. Content Recommendation: "Please recommend three relaxing movies that you would suggest to a user who is feeling tired."

[1475] By following these specific steps, the server can provide appropriate responses and content recommendations that take each user's emotions into consideration.

[1476] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1477] Step 1:

[1478] The server collects internal company documents and stores them in electronic format. Specifically, documents are collected manually or using automated scripts from each department and system. If the collected documents are in PDF or Word format, they are directly converted into text data. Paper documents are scanned and converted into text data using OCR technology. As a result, all documents are temporarily stored in text format.

[1479] input:

[1480] PDF, Word, and paper-based internal documents

[1481] Data processing:

[1482] Digitization of paper documents using scanners

[1483] Text conversion using OCR technology

[1484] output:

[1485] Text-format internal documents

[1486] Step 2:

[1487] The server converts digitized text data into vector data. This uses natural language processing techniques and machine learning algorithms. Specifically, it uses document vectorizers such as the BERT model to convert text data into numerical representations. The converted vector data captures its semantic features and is stored in a searchable database.

[1488] input:

[1489] Text-format internal documents

[1490] Data calculation:

[1491] Vectorization using natural language processing techniques (such as the BERT model)

[1492] output:

[1493] Numerical representation data stored in a vector database

[1494] Step 3:

[1495] The user enters a question from their device. For example, they might enter a question like, "I'm tired today, so I'd like to watch a relaxing movie." The entered question is then sent to the server.

[1496] input:

[1497] User questions (text data)

[1498] Data calculation:

[1499] Receiving questions and preparing for analysis

[1500] output:

[1501] Text data on the server

[1502] Step 4:

[1503] The server analyzes the received question using an emotion analysis engine. It uses the OpenAI API and other tools to identify the user's emotions from the question. Specifically, it sends the question text to the API and retrieves an emotion label.

[1504] input:

[1505] User questions (text data)

[1506] Data calculation:

[1507] Sentiment analysis using OpenAI API, etc.

[1508] output:

[1509] Emotional labels (fatigue, confusion, excitement, anger, etc.)

[1510] Step 5:

[1511] The server searches a vector database to identify documents and information related to the question. Specifically, it uses vector data to effectively search for semantically relevant information. For example, it can find manuals and guidelines related to the user's question.

[1512] input:

[1513] Vector database

[1514] User Questions

[1515] Data calculation:

[1516] Search using vector data

[1517] output:

[1518] Information (documents) related to the question

[1519] Step 6:

[1520] The server generates answers to questions using large-scale language models based on relevant information. This may include OpenAI's GPT-3. Furthermore, it adjusts the answers according to the user's emotions based on sentiment analysis results.

[1521] input:

[1522] Related information (documents)

[1523] Emotional labels

[1524] Data calculation:

[1525] Response generation and adjustment using GPT-3, etc.

[1526] output:

[1527] Emotionally appropriate response

[1528] Step 7:

[1529] The server sends the generated response to the device and provides it to the user. Furthermore, the server monitors the user's emotional changes in real time and readjusts the response as needed. To do this, the device uses its camera and microphone to send the user's facial expressions and voice to the server.

[1530] input:

[1531] Adjusted answer

[1532] User facial expressions and voice data

[1533] Data calculation:

[1534] Real-time monitoring of emotions and response adjustment

[1535] output:

[1536] Answers provided to users

[1537] Step 8:

[1538] The server recommends content to the user based on the results of sentiment analysis. For example, if the user is tired, it will recommend relaxing movies or music. A specific example might be presented such as, "Here are some recommended relaxing movies: 'Movie A', 'Movie B', 'Movie C'."

[1539] input:

[1540] Emotional labels

[1541] Data calculation:

[1542] Content recommendation using large-scale language models

[1543] output:

[1544] Recommended content provided to users

[1545] 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 controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1546] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1547] 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 this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[1548] Furthermore, the emotion identification model 59, acting 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 a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1549] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[1550] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[1551] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[1552] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[1553] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is 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 the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[1554] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[1555] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[1556] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[1557] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[1558] 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.

[1559] Furthermore, it is not necessary to store the entirety 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 the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[1560] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[1561] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of 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). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[1562] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[1563] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[1564] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[1565] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[1566] The following is further disclosed regarding the embodiments described above.

[1567] (Claim 1)

[1568] A means for collecting internal company documents and storing them in an electronic format,

[1569] A means for converting stored documents into vector data and saving the vector data in a searchable database,

[1570] A means for receiving questions from users and retrieving information related to the received questions from the database,

[1571] A means of generating answers to questions using a large-scale language model based on relevant information,

[1572] A system that includes means for providing the generated response to the user.

[1573] (Claim 2)

[1574] The system according to claim 1, wherein the means for generating an answer to a received question includes means for loading a large-scale language model and means for using the large-scale language model.

[1575] (Claim 3)

[1576] The system according to claim 1, wherein the means for generating vector data uses an algorithm that converts text data into a numerical representation.

[1577] "Example 1"

[1578] (Claim 1)

[1579] A means for collecting internal company documents and storing them in an electronic format,

[1580] A means of converting paper documents into text data using OCR technology,

[1581] A means for converting stored documents into vector data using natural language processing technology and storing the vector data in a searchable database,

[1582] A means for receiving questions from users and retrieving information related to the received questions from the database,

[1583] A means of generating an answer to a question using an AI model based on relevant information,

[1584] A system that includes means for providing the generated response to the user.

[1585] (Claim 2)

[1586] The system according to claim 1, comprising means for loading and initializing a generated AI model, and means for using the generated AI model.

[1587] (Claim 3)

[1588] The system according to claim 1, which uses an algorithm that converts text data into numerical representations using natural language processing technology.

[1589] "Application Example 1"

[1590] (Claim 1)

[1591] A means for collecting internal company documents and storing them in an electronic format,

[1592] A means for converting stored documents into vector data and saving the vector data in a searchable database,

[1593] A means for receiving questions from users and retrieving information related to the received questions from the database,

[1594] A means of generating answers to questions using a large-scale language model based on relevant information,

[1595] A means of providing the generated answer to the user,

[1596] A method by which logistics center workers can input questions via smartphone and receive immediate answers to those questions,

[1597] A system that includes this.

[1598] (Claim 2)

[1599] The system according to claim 1, wherein the means for generating an answer to a received question includes means for loading a large-scale language model and means for using the large-scale language model.

[1600] (Claim 3)

[1601] The system according to claim 1, wherein the means for generating vector data uses an algorithm that converts text data into a numerical representation.

[1602] "Example 2 of combining an emotion engine"

[1603] (Claim 1)

[1604] A means for collecting internal company documents and storing them in an electronic format,

[1605] A means for converting stored documents into vector data and saving the vector data in a searchable database,

[1606] A means for receiving questions from users and retrieving information related to the received questions from the database,

[1607] A means of generating answers to questions using a large-scale language model based on relevant information,

[1608] A means of providing the generated answer to the user,

[1609] A means for analyzing the question content and the user's emotional state using an emotion recognition engine that recognizes the user's emotions,

[1610] A system that includes means for generating appropriate responses that correspond to the user's emotions based on the analysis results of an emotion recognition engine.

[1611] (Claim 2)

[1612] The system according to claim 1, wherein the means for generating an answer to a received question includes means for loading a large-scale language model and means for using the large-scale language model.

[1613] (Claim 3)

[1614] The system according to claim 1, wherein the means for generating vector data uses an algorithm that converts text data into a numerical representation.

[1615] "Application example 2 when combining with an emotional engine"

[1616] (Claim 1)

[1617] A means for collecting internal company documents and storing them in an electronic format,

[1618] A means for converting stored documents into vector data and saving the vector data in a searchable database,

[1619] A means for receiving questions from users and retrieving information related to the received questions from the database,

[1620] A means of identifying the user's emotions by analyzing the received questions with an emotion analysis engine,

[1621] A means of generating answers to questions using a large-scale language model based on relevant information,

[1622] A means of adjusting and providing generated responses to match the user's emotions,

[1623] A means to monitor changes in user emotions in real time and readjust responses as needed,

[1624] A system that includes means for recommending content based on emotions.

[1625] (Claim 2)

[1626] The system according to claim 1, wherein the means for generating an answer to a received question includes means for loading a large-scale language model and means for using the large-scale language model.

[1627] (Claim 3)

[1628] The system according to claim 1, wherein the means for generating vector data uses an algorithm that converts text data into a numerical representation. [Explanation of Symbols]

[1629] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means for collecting internal company documents and storing them in an electronic format, A means for converting stored documents into vector data and saving the vector data in a searchable database, A means for receiving questions from users and retrieving information related to the received questions from the database, A means of generating answers to questions using a large-scale language model based on relevant information, A system that includes means for providing the generated response to the user.

2. The system according to claim 1, wherein the means for generating an answer to a received question includes means for loading a large-scale language model and means for using the large-scale language model.

3. The system according to claim 1, wherein the means for generating vector data uses an algorithm that converts text data into a numerical representation.

Images