system

Abstract

We provide the system. [Solution] A means for collecting a list of technical terms and related information, A means of cleansing the collected data and storing it in a database, A method for learning technical terms and their context using natural language processing models, A means of receiving and analyzing user inquiries, A means for searching a database based on the analysis results and generating an appropriate answer, A system that includes means for sending and displaying the generated response on the user's terminal.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 discloses a persona chatbot control method 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] In each department of an enterprise, different specialized terms and business processes are used, making it difficult for new employees, mid-career hires, and employees after transfer to quickly adapt to their work. Also, even existing employees may not be able to instantly obtain information they want to confirm or learn about their work. This problem may lead to a decrease in work efficiency and have an adverse impact on the productivity of the entire enterprise.

Means for Solving the Problems

[0005] The present invention solves the above-mentioned problems by providing a system comprising means for collecting a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their contexts using a natural language processing model, means for receiving and analyzing inquiries from users, means for searching the database based on the analysis results and generating appropriate answers, and means for transmitting and displaying the generated answers on a user terminal.

[0006] Furthermore, by including means for receiving feedback and using it to improve the model, the system's accuracy and usefulness will be continuously enhanced. Additionally, by incorporating a customized natural language processing model based on the specialized terminology and related information of each department, it will be possible to address the specific needs of each department and provide more accurate information.

[0007] A "glossary" is a list of specific terms and phrases used in each department of a company.

[0008] "Related information" refers to definitions of technical terms, examples of their usage, contextual information, and other related documents and data.

[0009] "Means of collection" refers to the procedures for gathering a list of technical terms and related information from each department.

[0010] "Cleaning" refers to the process of removing duplicates from collected data, correcting incomplete entries, and standardizing the format.

[0011] A "database" is a collection of structured data, a system for efficiently storing and managing a list of technical terms and related information.

[0012] A "natural language processing model" is a model that uses machine learning algorithms to understand and process human language.

[0013] "Learning method" refers to the process of analyzing and learning information about technical terms and their context using natural language processing models.

[0014] A "user" refers to a person who uses the system to ask questions or make inquiries related to their own work.

[0015] "Inquiry receiving mechanism" refers to the interface through which the system receives questions and inquiries from users.

[0016] "Analysis method" refers to the process of analyzing the information contained in a query using a natural language processing model to identify appropriate technical terms and related information.

[0017] "Database search method" refers to the process of searching for relevant information within a database based on the analysis results.

[0018] "Answer generation method" refers to the process of generating appropriate answers for users based on the results of database searches.

[0019] "Transmission method" refers to the procedure for sending the generated response to the user's device.

[0020] "Display means" refers to a function that allows the user terminal to visually display the received response to the user.

[0021] "Feedback receiving means" refers to an interface for receiving feedback from users and transmitting it to the system.

[0022] "Model improvement methods" refer to the process of retraining a natural language processing model based on received feedback to improve its accuracy and usefulness.

[0023] A "customized natural language processing model" refers to a natural language processing model that has been individually tailored and trained based on the specialized terminology and related information of a specific department.

Brief Description of the Drawings

[0024] [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] It is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] It 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] It shows an emotion map to which a plurality of emotions are mapped. [Figure 10] It shows an emotion map to which a plurality of emotions are mapped. [Figure 11] It is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12] It is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] It is a sequence diagram showing the processing flow of the data processing system in Example 2 when an emotion engine is combined. [Figure 14] It is a sequence diagram showing the processing flow of the data processing system in Application Example 2 when an emotion engine is combined.

Modes for Carrying Out the Invention

[0025] 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.

[0026] First, let's explain the terminology used in the following explanation.

[0027] 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).

[0028] 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.

[0029] 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.

[0030] 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).

[0031] 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."

[0032] [First Embodiment]

[0033] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0034] 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.

[0035] 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).

[0036] 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.

[0037] 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.

[0038] 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.

[0039] 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.

[0040] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0041] 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.

[0042] 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.

[0043] 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.

[0044] 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".

[0045] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. This system is implemented through a series of programmatic processes described below.

[0046] Data collection phase

[0047] First, the server collects a list of specialized terms and related information provided by each department of the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. The collected data is cleansed by the server to ensure consistency and accuracy. Specifically, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in the database.

[0048] Data processing and learning phases

[0049] Next, the server prepares a natural language processing (NLP) model and trains it based on the collected data. The server prepares the training data using specialized terminology and data used in its context within the database, and then trains the NLP model. For example, it teaches the model that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[0050] Inquiry processing phase

[0051] The user enters a question into the system using a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this entered question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[0052] Database search and response generation phase

[0053] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[0054] Result display phase

[0055] The generated response is sent from the server to the terminal. The terminal then displays the received response to the user.

[0056] Feedback processing phase

[0057] Users can provide feedback on the displayed answers. For example, they can input feedback such as "This answer was helpful" or "This answer was inappropriate." The device forwards this feedback to the server, which uses the received feedback to improve the NLP model. Inappropriate answers are used to retrain the model.

[0058] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, a feedback function continuously improves the accuracy and usefulness of the system. By using NLP models customized for each department, it achieves highly accurate information provision tailored to the specific needs of each department.

[0059] The following describes the processing flow.

[0060] Data collection phase

[0061] Step 1:

[0062] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[0063] Step 2:

[0064] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[0065] Step 3:

[0066] The server stores the cleansed data in the database.

[0067] Data processing and learning phases

[0068] Step 1:

[0069] The server prepares a natural language processing (NLP) model.

[0070] Step 2:

[0071] The server prepares the training data using specialized terminology and data used in the context of the database.

[0072] Step 3:

[0073] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[0074] Step 4:

[0075] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[0076] Inquiry processing phase

[0077] Step 1:

[0078] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[0079] Step 2:

[0080] The terminal forwards the entered question to the server.

[0081] Step 3:

[0082] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[0083] Database search and response generation phase

[0084] Step 1:

[0085] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[0086] Step 2:

[0087] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[0088] Step 3:

[0089] The server transfers the generated response to the terminal.

[0090] Result display phase

[0091] Step 1:

[0092] The terminal receives the response sent from the server.

[0093] Step 2:

[0094] The device displays the received response to the user.

[0095] Feedback processing phase

[0096] Step 1:

[0097] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[0098] Step 2:

[0099] The device forwards user feedback to the server.

[0100] Step 3:

[0101] The server analyzes the received feedback and uses it to improve the NLP model. For example, if there is feedback indicating an inappropriate response, the model is retrained based on that information.

[0102] Step 4:

[0103] The server adjusts the parameters of the NLP model based on the feedback, providing more accurate and useful answers to subsequent queries.

[0104] Through these steps, the system effectively manages the specialized terminology of each department and enables the rapid and accurate provision of information to users.

[0105] (Example 1)

[0106] 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."

[0107] Traditional enterprise systems faced the challenge of requiring significant time and effort to adapt to the different jargon and processes of each department. This hindered operational efficiency and could lead to decreased productivity. Furthermore, inaccurate responses to user inquiries prevented proper feedback from being obtained, hindering system improvement. In addition, insufficient performance of natural language processing models could lead to errors in analyzing complex jargon and context.

[0108] 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.

[0109] In this invention, the server includes means for collecting a list of technical terms and related information from each department, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for sending and displaying the generated answers on a user terminal, means for receiving user feedback and using it to improve the model, and means for providing a natural language processing model customized based on the technical terms and related information of each department. This enables the rapid and accurate provision of information that meets the specific needs of each department, resulting in increased operational efficiency and continuous system improvement.

[0110] "Each department of a company"

[0111] This refers to organizational units within a company that are responsible for different tasks or functions. Examples include the marketing department and the human resources department.

[0112] "List of Technical Terms"

[0113] This refers to a list of terms and phrases commonly used in a particular industry or field.

[0114] Related Information

[0115] This refers to supplementary data provided alongside technical terms, such as definitions of terms, usage examples, process documents, manuals, and project reports.

[0116] "cleansing"

[0117] This refers to operations performed on collected data, such as removing duplicates, correcting incomplete entries, and standardizing the format.

[0118] "Database"

[0119] This refers to a system for organizing and efficiently managing collected data.

[0120] "Natural language processing models"

[0121] This refers to artificial intelligence models designed to understand, analyze, and generate human language. Machine learning techniques are used in this process.

[0122] "Means for receiving and analyzing inquiries"

[0123] This refers to the process of receiving questions submitted by users, analyzing their content, and extracting relevant information.

[0124] "A means of searching a database and generating appropriate answers."

[0125] This refers to a function that searches a database based on the analyzed results and creates an answer that is appropriate for the user's question.

[0126] "User terminal"

[0127] This refers to a device that a user uses to access a system and input or retrieve information.

[0128] "feedback"

[0129] This refers to user-provided ratings and feedback, which are used to improve the system.

[0130] "Generative AI Model"

[0131] This refers to an artificial intelligence model trained for natural language processing.

[0132] "Prompt message"

[0133] This refers to the questions or instructions that users input into the system.

[0134] "Customized natural language processing models"

[0135] This refers to artificial intelligence models that have been specifically tailored to a particular department or task.

[0136] "Parameter tuning"

[0137] This refers to the process of adjusting a model to improve its performance. Examples include setting appropriate hyperparameters and adjusting the learning rate.

[0138] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. The system utilizes servers and terminals as hardware, and NLP (Natural Language Processing) technology and database management software as software. Typical software used includes Python and its libraries (e.g., spaCy, NLTK, TENSORFLOW®, PyTorch).

[0139] The server collects terminology lists and related information provided by various departments within the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. To cleanse this collected data, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in a database.

[0140] Next, the server prepares and trains an NLP model based on the collected data. The training data is prepared using specialized terminology and the context in which it is used within the database. For example, the model is trained to understand that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[0141] The user enters a question into the system using their own device. For example, they might enter, "What are the latest lead generation results?" This device forwards the user's question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[0142] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that you acquired 100 leads last week." The generated response is sent from the server to the terminal, which then displays it to the user.

[0143] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." This feedback is sent from the device to the server. The server uses the received feedback to improve the NLP model. Inappropriate answers are used as new training data for retraining.

[0144] The following are specific examples of prompts to enter into the system:

[0145] 1. "Please tell me your latest lead generation results."

[0146] 2. "What are the new trends in SEO?"

[0147] 3. "Please tell me about tools to increase conversion rates."

[0148] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, the system's accuracy and usefulness can be continuously improved through a feedback function. By using NLP models customized for each department, it achieves the provision of highly accurate information that meets the specific needs of each department.

[0149] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0150] Step 1: Data Collection

[0151] The server collects a list of specialized terms and related information from each department within the company. Specifically, it gathers data based on Excel files and documents entered or uploaded by departmental staff. The collected data includes definitions of terms, usage examples, process documents, manuals, and project reports.

[0152] Input: A list of specialized terms and related information provided by the department head.

[0153] Output: Collected raw data

[0154] Step 2: Data Cleansing

[0155] The server cleanses the collected data. This includes removing duplicate data, correcting incomplete entries (such as filling in blanks or correcting typos), and standardizing the format (for example, consistently converting date formats to YYYY-MM-DD).

[0156] Input: Collected raw data

[0157] Output: Cleansed data

[0158] Step 3: Database Storage

[0159] The server stores the cleansed data in the database. This ensures that data with guaranteed consistency and accuracy can be managed.

[0160] Input: Cleansed data

[0161] Output: Data stored in the database

[0162] Step 4: Training the NLP Model

[0163] The server trains an NLP model using the collected and cleansed data. It uses Python natural language processing libraries (e.g., spaCy, NLTK) or machine learning libraries (e.g., TensorFlow, PyTorch). The data is split into training and test data and fed to the NLP model. For example, the model is trained to understand that "lead" means "potential customer". After training, the model's performance is evaluated, and hyperparameters are tuned as needed.

[0164] Input: Cleansed data

[0165] Output: Trained NLP model

[0166] Step 5: Inquiry reception and analysis

[0167] The user enters a question into the system via a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this question to the server. The server analyzes the received question using an NLP model to identify keywords ("lead generation," "results") and context.

[0168] Input: User question (prompt)

[0169] Output: Results of question analysis (keywords and context)

[0170] Step 6: Database search and answer generation

[0171] The server searches the database based on the analysis results. For example, it searches for the latest project reports and statistics on lead generation and generates specific answers such as, "The latest lead generation results show that 100 leads were acquired last week." The answers are formatted using automated templates.

[0172] Input: Results of the analyzed question

[0173] Output: Generated answer

[0174] Step 7: Display Results

[0175] The generated response is sent from the server to the terminal. The terminal displays this response to the user through a GUI (Graphical User Interface). For example, it might be displayed on the dashboard of a web application.

[0176] Input: Response sent from the server

[0177] Output: Answer displayed on the user's terminal

[0178] Step 8: Feedback

[0179] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." The device sends this feedback to the server. The server uses the received feedback to improve the performance of the NLP model. Inappropriate answers are used as new training data for retraining.

[0180] Input: User feedback

[0181] Output: Improved NLP model

[0182] (Application Example 1)

[0183] 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."

[0184] In conventional systems, factory robots had difficulty quickly understanding specialized terminology and processes when performing new tasks. Furthermore, because different departments had different specialized terminology and processes, there was a lack of means to accurately acquire that information and efficiently perform tasks. In addition, there was a need for robots to acquire necessary information in real time while in operation and to improve their movements based on appropriate feedback.

[0185] 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.

[0186] In this invention, the server includes means for collecting a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their contexts using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for transmitting and displaying the generated answers on a user terminal, means for querying technical terms and processes while a factory robot is performing a new task and obtaining necessary information, and means for displaying the results via the robot's display or audio output. This enables the factory robot to quickly acquire information on specific technical terms and processes and perform tasks efficiently. Furthermore, the model can be continuously improved based on the robot's operation results, thereby improving the accuracy and efficiency of its operation.

[0187] A "glossary of technical terms" is a list that compiles terms used in a specific field or industry, their definitions, and related information.

[0188] "Related information" refers to materials such as usage examples related to technical terms, process documents, manuals, and project reports.

[0189] "Data cleansing" is an operation that involves removing duplicate data, correcting incomplete entries, and standardizing the format in order to ensure the consistency and accuracy of collected data.

[0190] A "database" is a collection of information that stores a list of specialized terms and related information, making it searchable and retrievable.

[0191] A "natural language processing model" is a machine learning model used to understand, analyze, and generate human language.

[0192] "Analysis" is the process of understanding user inquiries using natural language processing models and grasping their intent.

[0193] An "inquiry" refers to a question or request for information that a user makes to a system.

[0194] A "user terminal" is a device used by a user to access a system, input information, and receive results.

[0195] A "factory robot" is a mechanical device used in industrial settings to automatically perform specific tasks.

[0196] A "display" is a monitor or screen used to visually display information.

[0197] "Voice output" is a function that conveys information and instructions from the system to the user as voice.

[0198] "Feedback" refers to evaluations and comments provided by users regarding the system's responses and actions.

[0199] "Model improvement" is the process of improving the performance of a natural language processing model based on the feedback received.

[0200] This invention relates to a system that helps factory robots quickly understand and efficiently perform new tasks. The system is realized through the following processes. The main components of the system consist of a server, a robot as an edge device, and a user (human operator).

[0201] Data collection phase

[0202] The server collects a glossary of terms and related information from each department within the factory. The glossary includes related information such as process documents, manuals, and project reports. Apache® NiFi can be used for this collection. The collected data is cleansed by the server. The Pandas library is used to remove duplicate data, correct incomplete entries, and standardize the format. The cleansed data is then stored in a database (e.g., PostgreSQL).

[0203] Data processing and learning phases

[0204] The server prepares a natural language processing (NLP) model and trains it based on the collected data. It uses the Hugging Face's Transformers library to select an appropriate generative AI model (e.g., BERT) and then trains it. The following is an example of a prompt:

[0205] Example of a prompt:

[0206] Question: What are the latest lead generation results?

[0207] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[0208] Inquiry processing phase

[0209] When a factory robot performs a new task, it queries the server for technical terms and processes as needed. The robot receives instructions and questions from the user (operator) via its built-in display or voice output. These queries are transmitted to the server through a web framework such as Flask.

[0210] Database search and response generation phase

[0211] The server analyzes the received query using an NLP model, retrieves appropriate information from the database, and generates a response. The generated response is sent to the robot and communicated to the operator via display or voice output. gTTS (Google® Text-to-Speech) can be used for voice output.

[0212] Feedback processing phase

[0213] The robot sends feedback to the server based on its performance. This feedback includes user ratings and comments. The server uses the received feedback to continuously improve the performance of the NLP model. TensorFlow or PyTorch can be used to retrain the model.

[0214] Specific example

[0215] For example, consider a scenario where a factory robot asks for the "latest lead generation result." In this case, the server performs the analysis using the following prompt:

[0216] Example of a prompt:

[0217] Question: What are the latest lead generation results?

[0218] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[0219] In this way, it becomes possible to acquire information immediately and perform tasks efficiently, which is expected to improve the overall system performance.

[0220] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0221] Step 1: Data Collection Phase

[0222] The server collects terminology lists and related information from each department within the factory. Inputs include terminology lists, process documents, manuals, and project reports provided by each department. The server uses Apache NiFi to retrieve and relay this data. Outputs are stored on the server as raw data for cleansing.

[0223] Step 2: Data Cleansing

[0224] The server cleanses the collected data. Specifically, it uses the Pandas library to remove duplicate data, correct incomplete entries, and standardize the format. The input is raw data, and the output is cleaned data. This data is stored in a database.

[0225] Step 3: Training the natural language processing model

[0226] The server prepares a natural language processing model and trains it on cleansed data. It uses the Hugging Face's Transformers library to train a generative AI model (e.g., BERT). The input is a categorized list of terms and their usage contexts, and the trained output is an NLP model with high suitability for specific technical terms and processes.

[0227] Step 4: Receiving User Inquiries

[0228] When a factory robot performs a new task, it queries a server for technical terms and processes. The terminal (robot) receives user questions through its built-in display or voice input and transmits them to the server using the Flask framework. The input is the query content, and the output is the query request to the server.

[0229] Step 5: Analyze the inquiry

[0230] The server analyzes incoming queries using an NLP model. The input data is the user's question, and the output, as a result of the analysis, is information related to the appropriate technical terms and processes. The server then converts this into a query to be sent to the database.

[0231] Step 6: Database search and answer generation

[0232] The server uses the generated query to search the database and produce an appropriate response. The input is the parsed query, and the output is the generated response. For example, in the case of a query about the latest lead generation results, a response such as "According to the latest report from the lead generation department, 100 leads were acquired last week" would be generated.

[0233] Step 7: Submit and view results

[0234] The server sends the generated response to the terminal (robot). The terminal displays the result on its screen or communicates it to the user via voice output using gTTS. The input is the generated response text, and the output is the displayed information or voice.

[0235] Step 8: Receiving Feedback

[0236] The user provides feedback on the robot's performance. The terminal collects the user's feedback and transmits it to the server. The input is the user's evaluation and comments, and the output is the feedback data sent to the server.

[0237] Step 9: Model Improvement

[0238] The server uses the received feedback to improve the performance of the NLP model. The input is the feedback data, and the output is the updated and improved natural language processing model. TensorFlow or PyTorch is used for retraining.

[0239] This allows the system to continuously learn and improve, supporting the efficient task execution of factory robots.

[0240] 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.

[0241] This invention not only provides a system tailored to the specialized terminology and processes used in various departments of a company, but also aims to provide more appropriate information and improve the user experience by combining it with an emotion engine that recognizes user emotions. This system consists of a process of collecting, cleansing, and storing a list of specialized terms and related information in a database, and learning using a natural language processing (NLP) model. It also receives user inquiries, analyzes them, generates and displays answers, and collects and analyzes feedback to improve the model. Here, we will describe a specific embodiment that combines the system with an emotion engine that recognizes user emotions.

[0242] Data collection phase

[0243] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This information is then cleansed by the server to ensure consistency and accuracy before being stored in the database.

[0244] Data processing and learning phases

[0245] Next, the server prepares an NLP model and trains it using data from the database containing specialized terminology and the context in which it is used. As a result of the training, the model will be able to understand the context in which specific terms are used and process them appropriately. The performance of the trained model is also evaluated, and parameters are adjusted as needed.

[0246] Inquiry processing phase

[0247] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal transmits the entered question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[0248] Database search and response generation phase

[0249] The server searches the database based on the analysis results and collects the necessary information. From the collected information, it generates an appropriate response. For example, it might create a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[0250] Emotion recognition phase

[0251] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adapt the tone and level of detail of the response it provides.

[0252] Result display phase

[0253] The server sends the generated response to the terminal. The terminal displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[0254] Feedback processing phase

[0255] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[0256] Model and emotion engine improvement phase

[0257] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. Specifically, if there is inappropriate feedback, it analyzes its content and uses it to retrain the model. It is also used as data to improve the accuracy of emotion recognition.

[0258] In this embodiment, the system of the present invention not only provides information on technical terms and processes quickly and accurately, but also recognizes the user's emotions and provides appropriate information based on those emotions. This makes it possible to improve the user's work efficiency and enhance the quality of the user experience.

[0259] The following describes the processing flow.

[0260] Data collection phase

[0261] Step 1:

[0262] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[0263] Step 2:

[0264] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[0265] Step 3:

[0266] The server stores the cleansed data in the database.

[0267] Data processing and learning phases

[0268] Step 1:

[0269] The server prepares a natural language processing (NLP) model.

[0270] Step 2:

[0271] The server prepares the training data using specialized terminology and data used in the context of the database.

[0272] Step 3:

[0273] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[0274] Step 4:

[0275] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[0276] Inquiry processing phase

[0277] Step 1:

[0278] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[0279] Step 2:

[0280] The terminal forwards the entered question to the server.

[0281] Step 3:

[0282] The server analyzes the received question using the NLP model and identifies relevant technical terms and their context information.

[0283] Database Search and Answer Generation Phase

[0284] Step 1:

[0285] The server searches the database based on the analysis results. Identify and collect the necessary information.

[0286] Step 2:

[0287] Based on the search results, the server generates an appropriate answer in an easy-to-understand format for the user. For example, prepare a specific answer such as "The latest lead generation results obtained 100 leads last week."

[0288] Step 3:

[0289] The server transfers the generated answer to the terminal.

[0290] Sentiment Recognition Phase

[0291] Step 1:

[0292] The server passes the input content from the user to the sentiment engine to recognize the sentiment. For example, when the input question is "Please tell me about this setting," the sentiment engine analyzes whether the text contains doubt or anxiety.

[0293] Step 2:

[0294] The server adjusts the tone of the answer based on the results of the sentiment engine. For example, if the user is feeling anxious, the answer should be made more friendly and detailed.

[0295] Result Display Phase

[0296] Step 1:

[0297] The terminal receives the answer sent from the server.

[0298] Step 2:

[0299] The terminal displays the received answer to the user. The displayed answer is presented in an appropriate tone and level of detail based on the user's sentiment recognized by the sentiment engine.

[0300] Feedback Processing Phase

[0301] Step 1:

[0302] The user inputs feedback on the provided answer. For example, the user inputs feedback such as "This answer was helpful" or "The answer was inappropriate".

[0303] Step 2:

[0304] The terminal transfers the user's feedback to the server.

[0305] Model and Sentiment Engine Improvement Phase

[0306] Step 1:

[0307] The server analyzes the received feedback and improves both the NLP model and the sentiment engine. For example, if there is feedback on an inappropriate answer, the content is analyzed and used for retraining the model.

[0308] Step 2:

[0309] The server adjusts the parameters of the NLP model and the sentiment engine based on the feedback and provides more accurate and useful answers for future inquiries.

[0310] (Example 2)

[0311] 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".

[0312] The specialized terminology and processes used in various departments of a company are diverse, requiring advanced knowledge and rapid response to understand them and provide appropriate information. However, conventional systems have not adequately understood the context of specialized terminology or recognized user emotions, making it difficult to provide appropriate information and improve the user experience. Therefore, there is a need to develop systems that can provide information in a way that suits user needs.

[0313] The specific processing performed by the specific 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 a list of technical terms and related information from each department of a company, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing inquiries from users, means for searching the database based on the analysis results and generating an appropriate answer, means for sending and displaying the generated answer on a user terminal, and means for recognizing emotions from the content of the user's inquiry. This not only enables the provision of information specific to the technical terms and processes of each department, but also allows for adjustment of the tone and level of detail of the answer according to the user's emotions, which is expected to improve the user experience.

[0314] "Each department of a company" refers to multiple departments within a company, each with different tasks and roles.

[0315] A "list of technical terms" is a list that compiles specific terms and expressions used in a particular industry or field.

[0316] "Related information" refers to information such as definitions, usage examples, process documents, manuals, and project reports associated with technical terms.

[0317] A "server" is a computer system that provides information and services over a network.

[0318] "Means of collection" refers to the methods and techniques used to consolidate dispersed data into a single entity.

[0319] "Data cleansing methods" refer to techniques for removing duplicates, errors, and incomplete information from collected data, making it accurate and consistent.

[0320] "Storing in a database" means saving processed data in a structured format within a database.

[0321] A "natural language processing model" is a type of machine learning model used to understand and process human language.

[0322] "Methods for learning technical terms and their contexts" refers to methods of training models to understand the situations and contexts in which specific terms are used.

[0323] "Means for receiving and analyzing inquiries" refers to methods and technologies for receiving questions and requests from users and analyzing their content.

[0324] "Methods for searching a database based on analysis results" refers to methods for obtaining relevant data from a database based on the analyzed information.

[0325] "Means of generating appropriate answers" refers to methods of creating answers that are useful to the user based on search results.

[0326] A "user terminal" is a device (e.g., a personal computer, a smartphone) that a user uses to access information and services.

[0327] "Means of recognizing emotions" refers to technologies and methods for identifying emotions from user input.

[0328] "Methods for receiving feedback and using it to improve the model" refers to methods of receiving evaluations and opinions from users, updating the model based on them, and improving its performance.

[0329] A "customized natural language processing model" is a natural language processing model that has been tailored to specific needs or applications.

[0330] An "emotion recognition engine" is a specialized system or algorithm used to evaluate and analyze a user's emotions.

[0331] This invention is a system that specializes in the specialized terminology and processes used in each department of a company, and further combines this with an emotion engine that recognizes user emotions, thereby providing more appropriate information and improving the user experience. This system is implemented through the following series of processes.

[0332] Data collection phase

[0333] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This data collection is performed using API-based integration. The collected information is then cleansed by the server to ensure consistency and accuracy before being stored in the database. Data cleansing tools such as OpenRefine are used for this data cleansing process.

[0334] Data processing and learning phases

[0335] Next, the server prepares a natural language processing (NLP) model and trains it using data from the database containing specialized terms and the context in which they are used. NLP models used include natural language processing libraries such as Spacy and Hugging Face. This training enables the model to understand the context in which specific terms are used and to process them appropriately. After training, the model's performance is evaluated, and parameters are adjusted as needed.

[0336] Inquiry processing phase

[0337] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal forwards this question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[0338] Database search and response generation phase

[0339] The server searches the database based on the analysis results and collects the necessary information. It then generates an appropriate response. For example, it might provide a specific response such as, "The latest lead generation result shows that 100 leads were acquired last week."

[0340] Emotion recognition phase

[0341] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adjust the tone and level of detail of the response it provides.

[0342] Result display phase

[0343] The server sends the generated response to the device, and the device displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[0344] Feedback processing phase

[0345] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[0346] Model and emotion engine improvement phase

[0347] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. If there is feedback regarding inappropriate responses, the analysis results are used to retrain the model. The feedback is also utilized as data to improve the accuracy of emotion recognition.

[0348] Examples of specific cases and prompts for generative AI models.

[0349] As a concrete example, in response to a question about the latest lead generation results in the marketing department, the system provides the answer, "The latest lead generation results show that we acquired 100 leads last week." If the emotion engine detects that the user is in a hurry, it will quickly display a concise response such as, "Here are the latest lead generation results. We acquired 100 leads last week."

[0350] Examples of prompts for a generative AI model:

[0351] "Could you please tell me the latest lead generation results for your marketing department?"

[0352] "Could you please tell me the latest project progress in the XX department?"

[0353] "Please use the emotion engine to evaluate user satisfaction."

[0354] This system aims to improve operational efficiency and user experience by integrating the interpretation of technical terms with the recognition of user emotions.

[0355] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0356] Step 1:

[0357] The server collects a list of technical terms and related information from each department of the company. The input requires access information to each department's data source. Specifically, the server uses an API to send HTTP requests to each department's data source to retrieve the technical term list and related information. The output is the collected raw data.

[0358] Step 2:

[0359] The server cleanses the collected data. The raw data obtained in step 1 is used as input. Specifically, a data cleansing tool such as OpenRefine is used to remove duplicate data, fill in incomplete data, and ensure consistency. The output is the cleansed data.

[0360] Step 3:

[0361] The server stores the cleansed data in the database. The cleansed data generated in step 2 is required as input. Specifically, a database management system such as MySQL® is used, and the data is saved to the database using SQL queries. The output is structured data stored in the database.

[0362] Step 4:

[0363] The server prepares a natural language processing (NLP) model. The input requires an existing trained model and a list of specialized terminology for each field. Specifically, it imports and initializes natural language processing libraries such as Spacy or Hugging Face. The output is the NLP model prepared for training.

[0364] Step 5:

[0365] The server trains an NLP model using the specialized terminology and data used in its context within the database. The inputs required are the data stored in step 3 and the NLP model prepared in step 4. Specifically, the model is fed data, hyperparameters such as the number of epochs and learning rate are set, and training is performed. The output is the trained NLP model.

[0366] Step 6:

[0367] The server evaluates the trained model and adjusts the parameters as needed. The input requires a trained NLP model and an evaluation dataset. Specifically, it measures the model's accuracy and error and readjusts the hyperparameters. The output is an optimized NLP model.

[0368] Step 7:

[0369] The user enters a question into the system using a terminal. Input requires the user to enter a text box, such as "What are the latest lead generation results?". The output is the user's question stored in text format on the terminal.

[0370] Step 8:

[0371] The terminal forwards the user's question to the server. The input requires the user's question. Specifically, it sends an HTTP request to the server and transfers the data. The output is the server receiving the question.

[0372] Step 9:

[0373] The server uses an NLP model to analyze user questions. The input requires the user's question, transmitted from the terminal, and a trained NLP model. Specifically, the model analyzes the question and identifies relevant terminology and context. The output is the analysis result.

[0374] Step 10:

[0375] The server searches the database based on the analysis results. The analysis results and the database are required as input. Specifically, it generates an SQL query, sends it to the database, and retrieves the relevant data. The output will provide the necessary information.

[0376] Step 11:

[0377] The server generates appropriate answers from the collected information. It requires information retrieved from a database as input. Specifically, it integrates the information to create a user-friendly answer. The output is the generated answer.

[0378] Step 12:

[0379] The server uses an emotion engine to recognize emotions from user input. The input requires both the user's question and the emotion engine. Specifically, it performs text analysis to evaluate emotions based on tone and context. The output is the recognized emotion.

[0380] Step 13:

[0381] The server sends the generated response to the terminal, and the terminal displays the received response to the user. The required inputs are the generated response and the sentiment analysis results. Specifically, the response is sent to the terminal via an HTTP response, which the terminal then displays. The output is the response that the user can view.

[0382] Step 14:

[0383] The user provides feedback on the provided answer. The input requires the answer displayed by the system, and the user inputs feedback such as "This answer was helpful" or "This answer was inappropriate." The output is the feedback stored on the device.

[0384] Step 15:

[0385] The terminal forwards user feedback to the server. The input requires user feedback. Specifically, it sends the feedback to the server using an HTTP request. The output is the server receiving the feedback.

[0386] Step 16:

[0387] The server analyzes the received feedback and improves the NLP model and emotion engine. Input requires user feedback and training data. Specifically, it retrains the model based on the feedback to improve performance. The output is an improved NLP model and emotion engine.

[0388] (Application Example 2)

[0389] 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".

[0390] In traditional brick-and-mortar stores, communication between employees and customers often lacked the speed and accuracy of providing specialized information, and it was difficult to respond appropriately to customers' emotions. This resulted in decreased customer satisfaction and reduced operational efficiency. This invention aims to solve these problems and enable the provision of higher-quality service in customer interactions at physical stores.

[0391] The specific processing performed by the specific 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 a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating an appropriate response, means for transmitting and displaying the generated response on a user terminal, means for recognizing the user's emotions and adjusting the response tone and level of detail based on the recognition results, means for receiving feedback and using it to improve the model, and a natural language processing model customized based on the technical terms and related information of each department. This enables employees in physical stores to provide information quickly and accurately, as well as to respond appropriately to customer emotions.

[0392] A "glossary" is a list that compiles specific technical terms and expertise used within a company or department.

[0393] "Related information" refers to all information related to technical terms, including definitions of technical terms, examples of usage, process documents, manuals, and project reports.

[0394] "Means of collection" refers to the methods and devices for gathering a list of technical terms and related information.

[0395] "Cleansing" refers to the process of organizing and correcting collected data to ensure consistency and accuracy.

[0396] A "database" is a collection of data that is efficiently stored and managed in a way that makes it searchable and usable.

[0397] A "natural language processing model" is a model used to train algorithms and methods for analyzing and understanding human language using computers.

[0398] "Means of learning" refers to methods and devices for training natural language processing models to learn specialized terminology and its context.

[0399] "Means of analysis" refers to methods and devices for analyzing user inquiries and understanding their content.

[0400] "Searching means" refers to methods and devices for finding relevant information within a database based on analysis results.

[0401] "Means for generating answers" refers to methods and devices for creating appropriate answers for users based on retrieved information.

[0402] "User terminal" refers to devices or systems that a user directly operates, such as smartphones or computers.

[0403] "Means for recognizing emotions and adjusting response tone and level of detail based on the recognition results" refers to methods and apparatus for analyzing a user's emotions and adapting the tone and level of detail of the response accordingly.

[0404] "Means for receiving feedback and using it to improve the model" refers to methods and devices for receiving evaluations and opinions from users and incorporating them into improvements to natural language processing models and sentiment engines.

[0405] A "customized natural language processing model" refers to a natural language processing model that has been specifically tailored to a particular company or department.

[0406] This invention relates to an information provision system for improving the quality of customer service in physical stores. Specific embodiments thereof are described below.

[0407] 1. Data collection and cleansing

[0408] First, the server collects a list of technical terms and related information (e.g., product features, process descriptions, manuals, and usage examples) from various departments within the company. The collected data is then cleansed by the server and stored in a database, ensuring consistency and accuracy. This database is then used as a source of information for subsequent queries.

[0409] 2. Training a natural language processing model

[0410] The server prepares a natural language processing (NLP) model and trains it based on specialized terminology and its contextual information stored in a database. Through this training, the model learns to understand the context in which specific terms are used and generates appropriate responses accordingly.

[0411] 3. Inquiry analysis and response generation

[0412] The user (in this case, a store employee) enters an inquiry into the system using a device such as a smartphone. For example, the question might be, "Please tell me about the features of this new product." The device transmits the entered question to the server, which analyzes the question using an NLP model. Based on the analysis results, it searches the database and generates an appropriate answer. In the process, it also analyzes the customer's emotions and adjusts the response tone and level of detail accordingly.

[0413] 4. Emotion Recognition and Response Adjustment

[0414] The server uses an emotion engine to recognize emotions from user input. For example, it analyzes the tone and context of the entered question and adopts a comforting response tone if the user is dissatisfied. This allows users to receive more appropriate and satisfying service.

[0415] 5. Feedback and Model Improvement

[0416] The system also includes a feature that allows users to provide feedback on the answers they receive. For example, users can input feedback such as "This answer was helpful" or "This answer was inappropriate." The user's device forwards this feedback to the server, which then uses it to improve both the NLP model and the emotion engine. This enables continuous model improvement based on feedback.

[0417] 6. Usage example

[0418] For example, if a customer asks, "What are the features of the new product?", the server will provide information such as, "This is a new leather bag made from high-quality cowhide." Also, if a customer asks in a dissatisfied tone, "Is it possible to return this item?", the server will respond with something like, "Returns are accepted within 30 days of purchase. Please contact us if you have any problems; we will be happy to assist you."

[0419] Example of a prompt

[0420] "Could you please tell me the specifications of this product?"

[0421] "Is this item returnable?"

[0422] In this way, the present invention facilitates communication between employees and customers in physical stores, enabling the rapid and accurate provision of specialized information and appropriate responses that respond to customer emotions. This is expected to improve customer satisfaction and operational efficiency.

[0423] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0424] Step 1:

[0425] The server collects terminology lists and related information from each department. Specifically, the server uses APIs and file transfer protocols to retrieve data from each department's databases and document management systems. The input here is the terminology lists and related information provided by each department, and the output is the collected data.

[0426] Step 2:

[0427] The server cleanses the collected data, removing unnecessary and incorrect data to maintain consistency and accuracy. Specifically, it performs data cleaning operations using libraries such as Python's pandas library. The input here is the data collected in the previous step, and the output is the cleansed data.

[0428] Step 3:

[0429] The server stores the cleansed data in a database. Specifically, it inserts the data into an SQL database or a NoSQL database. Here, the input is the cleansed data, and the output is the data stored in the database.

[0430] Step 4:

[0431] The server trains a natural language processing (NLP) model. Specifically, it trains an NLP model (e.g., BERT or GPT) using technical terms and their contexts from a database. The input here is the technical terms and contextual information from the database, and the output is the trained NLP model.

[0432] Step 5:

[0433] The user (employee) enters a question using a terminal. For example, they might enter, "Please tell me the features of this new product." Here, the input is the user's question, and the output is the user's input data sent to the terminal.

[0434] Step 6:

[0435] The terminal forwards user input to the server. Specifically, it sends questions to the server using HTTP requests. Here, input refers to the user's input data, and output refers to the data forwarded to the server.

[0436] Step 7:

[0437] The server uses an NLP model to analyze the user's question. Specifically, it understands the context of the question and identifies relevant technical terms. The input here is the user's question, and the output is the analysis result.

[0438] Step 8:

[0439] The server searches the database based on the analysis results and collects the appropriate answers. The input here is the analysis results from the NLP model, and the output is the information collected from the database.

[0440] Step 9:

[0441] The server uses an emotion engine to recognize the user's emotions. Specifically, it analyzes the tone and context of the input question to identify the emotions the user is experiencing. Here, the input is the user's question, and the output is the emotion recognition result.

[0442] Step 10:

[0443] The server adjusts the tone and level of detail of its response based on the recognized emotion. For example, if the user is dissatisfied, it adopts a comforting tone. The input here is the emotion recognition result and collected information, and the output is the adjusted response.

[0444] Step 11:

[0445] The server sends the adjusted response to the terminal. Specifically, it sends the response to the terminal using an HTTP response. Here, the input is the adjusted response, and the output is the response sent to the terminal.

[0446] Step 12:

[0447] The terminal displays the received response to the user. The input here is the response sent to the terminal, and the output is the response displayed to the user.

[0448] Step 13:

[0449] Users provide feedback on the provided answers. For example, they might rate the answer as "This answer was helpful" or "This answer was inappropriate." The input here is the user's feedback, and the output is the feedback entered into the device.

[0450] Step 14:

[0451] The terminal forwards user feedback to the server. Specifically, it sends feedback to the server using an HTTP request. Here, the input is the feedback entered into the terminal, and the output is the feedback forwarded to the server.

[0452] Step 15:

[0453] The server improves the NLP model and emotion engine based on the feedback it receives. Specifically, it analyzes the content of the feedback and retrains the model as needed. The input here is the user's feedback, and the output is the improved NLP model and emotion engine.

[0454] These steps enable employees in physical stores to provide customers with quick and accurate information, as well as to respond appropriately to customers' emotions.

[0455] 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.

[0456] 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.

[0457] 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.

[0458] [Second Embodiment]

[0459] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0460] 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.

[0461] 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).

[0462] 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.

[0463] 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.

[0464] 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).

[0465] 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.

[0466] 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.

[0467] 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.

[0468] 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.

[0469] 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.

[0470] 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".

[0471] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. This system is implemented through a series of programmatic processes described below.

[0472] Data collection phase

[0473] First, the server collects a list of specialized terms and related information provided by each department of the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. The collected data is cleansed by the server to ensure consistency and accuracy. Specifically, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in the database.

[0474] Data processing and learning phases

[0475] Next, the server prepares a natural language processing (NLP) model and trains it based on the collected data. The server prepares the training data using specialized terminology and data used in its context within the database, and then trains the NLP model. For example, it teaches the model that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[0476] Inquiry processing phase

[0477] The user enters a question into the system using a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this entered question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[0478] Database search and response generation phase

[0479] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[0480] Result display phase

[0481] The generated response is sent from the server to the terminal. The terminal then displays the received response to the user.

[0482] Feedback processing phase

[0483] Users can provide feedback on the displayed answers. For example, they can input feedback such as "This answer was helpful" or "This answer was inappropriate." The device forwards this feedback to the server, which uses the received feedback to improve the NLP model. Inappropriate answers are used to retrain the model.

[0484] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, a feedback function continuously improves the accuracy and usefulness of the system. By using NLP models customized for each department, it achieves highly accurate information provision tailored to the specific needs of each department.

[0485] The following describes the processing flow.

[0486] Data collection phase

[0487] Step 1:

[0488] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[0489] Step 2:

[0490] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[0491] Step 3:

[0492] The server stores the cleansed data in the database.

[0493] Data processing and learning phases

[0494] Step 1:

[0495] The server prepares a natural language processing (NLP) model.

[0496] Step 2:

[0497] The server prepares the training data using specialized terminology and data used in the context of the database.

[0498] Step 3:

[0499] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[0500] Step 4:

[0501] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[0502] Inquiry processing phase

[0503] Step 1:

[0504] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[0505] Step 2:

[0506] The terminal forwards the entered question to the server.

[0507] Step 3:

[0508] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[0509] Database search and response generation phase

[0510] Step 1:

[0511] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[0512] Step 2:

[0513] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[0514] Step 3:

[0515] The server transfers the generated response to the terminal.

[0516] Result display phase

[0517] Step 1:

[0518] The terminal receives the response sent from the server.

[0519] Step 2:

[0520] The device displays the received response to the user.

[0521] Feedback processing phase

[0522] Step 1:

[0523] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[0524] Step 2:

[0525] The device forwards user feedback to the server.

[0526] Step 3:

[0527] The server analyzes the received feedback and uses it to improve the NLP model. For example, if there is feedback indicating an inappropriate response, the model is retrained based on that information.

[0528] Step 4:

[0529] The server adjusts the parameters of the NLP model based on the feedback, providing more accurate and useful answers to subsequent queries.

[0530] Through these steps, the system effectively manages the specialized terminology of each department and enables the rapid and accurate provision of information to users.

[0531] (Example 1)

[0532] 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."

[0533] Traditional enterprise systems faced the challenge of requiring significant time and effort to adapt to the different jargon and processes of each department. This hindered operational efficiency and could lead to decreased productivity. Furthermore, inaccurate responses to user inquiries prevented proper feedback from being obtained, hindering system improvement. In addition, insufficient performance of natural language processing models could lead to errors in analyzing complex jargon and context.

[0534] 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.

[0535] In this invention, the server includes means for collecting a list of technical terms and related information from each department, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for sending and displaying the generated answers on a user terminal, means for receiving user feedback and using it to improve the model, and means for providing a natural language processing model customized based on the technical terms and related information of each department. This enables the rapid and accurate provision of information that meets the specific needs of each department, resulting in increased operational efficiency and continuous system improvement.

[0536] "Each department of a company"

[0537] This refers to organizational units within a company that are responsible for different tasks or functions. Examples include the marketing department and the human resources department.

[0538] "List of Technical Terms"

[0539] This refers to a list of terms and phrases commonly used in a particular industry or field.

[0540] Related Information

[0541] This refers to supplementary data provided alongside technical terms, such as definitions of terms, usage examples, process documents, manuals, and project reports.

[0542] "cleansing"

[0543] This refers to operations performed on collected data, such as removing duplicates, correcting incomplete entries, and standardizing the format.

[0544] "Database"

[0545] This refers to a system for organizing and efficiently managing collected data.

[0546] "Natural language processing models"

[0547] This refers to artificial intelligence models designed to understand, analyze, and generate human language. Machine learning techniques are used in this process.

[0548] "Means for receiving and analyzing inquiries"

[0549] This refers to the process of receiving questions submitted by users, analyzing their content, and extracting relevant information.

[0550] "A means of searching a database and generating appropriate answers."

[0551] This refers to a function that searches a database based on the analyzed results and creates an answer that is appropriate for the user's question.

[0552] "User terminal"

[0553] This refers to a device that a user uses to access a system and input or retrieve information.

[0554] "feedback"

[0555] This refers to user-provided ratings and feedback, which are used to improve the system.

[0556] "Generative AI Model"

[0557] This refers to an artificial intelligence model trained for natural language processing.

[0558] "Prompt message"

[0559] This refers to the questions or instructions that users input into the system.

[0560] "Customized natural language processing models"

[0561] This refers to artificial intelligence models that have been specifically tailored to a particular department or task.

[0562] "Parameter tuning"

[0563] This refers to the process of adjusting a model to improve its performance. Examples include setting appropriate hyperparameters and adjusting the learning rate.

[0564] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. The system utilizes servers and terminals as hardware, and employs NLP (Natural Language Processing) technology and database management software as software. Typical software used includes Python and its libraries (e.g., spaCy, NLTK, TensorFlow, PyTorch).

[0565] The server collects terminology lists and related information provided by various departments within the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. To cleanse this collected data, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in a database.

[0566] Next, the server prepares and trains an NLP model based on the collected data. The training data is prepared using specialized terminology and the context in which it is used within the database. For example, the model is trained to understand that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[0567] The user enters a question into the system using their own device. For example, they might enter, "What are the latest lead generation results?" This device forwards the user's question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[0568] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that you acquired 100 leads last week." The generated response is sent from the server to the terminal, which then displays it to the user.

[0569] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." This feedback is sent from the device to the server. The server uses the received feedback to improve the NLP model. Inappropriate answers are used as new training data for retraining.

[0570] The following are specific examples of prompts to enter into the system:

[0571] 1. "Please tell me your latest lead generation results."

[0572] 2. "What are the new trends in SEO?"

[0573] 3. "Please tell me about tools to increase conversion rates."

[0574] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, the system's accuracy and usefulness can be continuously improved through a feedback function. By using NLP models customized for each department, it achieves the provision of highly accurate information that meets the specific needs of each department.

[0575] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0576] Step 1: Data Collection

[0577] The server collects a list of specialized terms and related information from each department within the company. Specifically, it gathers data based on Excel files and documents entered or uploaded by departmental staff. The collected data includes definitions of terms, usage examples, process documents, manuals, and project reports.

[0578] Input: A list of specialized terms and related information provided by the department head.

[0579] Output: Collected raw data

[0580] Step 2: Data Cleansing

[0581] The server cleanses the collected data. This includes removing duplicate data, correcting incomplete entries (such as filling in blanks or correcting typos), and standardizing the format (for example, consistently converting date formats to YYYY-MM-DD).

[0582] Input: Collected raw data

[0583] Output: Cleansed data

[0584] Step 3: Database Storage

[0585] The server stores the cleansed data in the database. This ensures that data with guaranteed consistency and accuracy can be managed.

[0586] Input: Cleansed data

[0587] Output: Data stored in the database

[0588] Step 4: Training the NLP Model

[0589] The server trains an NLP model using the collected and cleansed data. It uses Python natural language processing libraries (e.g., spaCy, NLTK) or machine learning libraries (e.g., TensorFlow, PyTorch). The data is split into training and test data and fed to the NLP model. For example, the model is trained to understand that "lead" means "potential customer". After training, the model's performance is evaluated, and hyperparameters are tuned as needed.

[0590] Input: Cleansed data

[0591] Output: Trained NLP model

[0592] Step 5: Inquiry reception and analysis

[0593] The user enters a question into the system via a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this question to the server. The server analyzes the received question using an NLP model to identify keywords ("lead generation," "results") and context.

[0594] Input: User question (prompt)

[0595] Output: Results of question analysis (keywords and context)

[0596] Step 6: Database search and answer generation

[0597] The server searches the database based on the analysis results. For example, it searches for the latest project reports and statistics on lead generation and generates specific answers such as, "The latest lead generation results show that 100 leads were acquired last week." The answers are formatted using automated templates.

[0598] Input: Results of the analyzed question

[0599] Output: Generated answer

[0600] Step 7: Display Results

[0601] The generated response is sent from the server to the terminal. The terminal displays this response to the user through a GUI (Graphical User Interface). For example, it might be displayed on the dashboard of a web application.

[0602] Input: Response sent from the server

[0603] Output: Answer displayed on the user's terminal

[0604] Step 8: Feedback

[0605] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." The device sends this feedback to the server. The server uses the received feedback to improve the performance of the NLP model. Inappropriate answers are used as new training data for retraining.

[0606] Input: User feedback

[0607] Output: Improved NLP model

[0608] (Application Example 1)

[0609] 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."

[0610] In conventional systems, factory robots had difficulty quickly understanding specialized terminology and processes when performing new tasks. Furthermore, because different departments had different specialized terminology and processes, there was a lack of means to accurately acquire that information and efficiently perform tasks. In addition, there was a need for robots to acquire necessary information in real time while in operation and to improve their movements based on appropriate feedback.

[0611] 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.

[0612] In this invention, the server includes means for collecting a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their contexts using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for transmitting and displaying the generated answers on a user terminal, means for querying technical terms and processes while a factory robot is performing a new task and obtaining necessary information, and means for displaying the results via the robot's display or audio output. This enables the factory robot to quickly acquire information on specific technical terms and processes and perform tasks efficiently. Furthermore, the model can be continuously improved based on the robot's operation results, thereby improving the accuracy and efficiency of its operation.

[0613] A "glossary of technical terms" is a list that compiles terms used in a specific field or industry, their definitions, and related information.

[0614] "Related information" refers to materials such as usage examples related to technical terms, process documents, manuals, and project reports.

[0615] "Data cleansing" is an operation that involves removing duplicate data, correcting incomplete entries, and standardizing the format in order to ensure the consistency and accuracy of collected data.

[0616] A "database" is a collection of information that stores a list of specialized terms and related information, making it searchable and retrievable.

[0617] A "natural language processing model" is a machine learning model used to understand, analyze, and generate human language.

[0618] "Analysis" is the process of understanding user inquiries using natural language processing models and grasping their intent.

[0619] An "inquiry" refers to a question or request for information that a user makes to a system.

[0620] A "user terminal" is a device used by a user to access a system, input information, and receive results.

[0621] A "factory robot" is a mechanical device used in industrial settings to automatically perform specific tasks.

[0622] A "display" is a monitor or screen used to visually display information.

[0623] "Voice output" is a function that conveys information and instructions from the system to the user as voice.

[0624] "Feedback" refers to evaluations and comments provided by users regarding the system's responses and actions.

[0625] "Model improvement" is the process of improving the performance of a natural language processing model based on the feedback received.

[0626] This invention relates to a system that helps factory robots quickly understand and efficiently perform new tasks. The system is realized through the following processes. The main components of the system consist of a server, a robot as an edge device, and a user (human operator).

[0627] Data collection phase

[0628] The server collects a glossary of terms and related information from each department within the factory. The glossary includes related information such as process documents, manuals, and project reports. Apache NiFi can be used for this collection. The collected data is cleansed by the server. The Pandas library is used to remove duplicate data, correct incomplete entries, and standardize the format. The cleansed data is then stored in a database (e.g., PostgreSQL).

[0629] Data processing and learning phases

[0630] The server prepares a natural language processing (NLP) model and trains it based on the collected data. It uses the Hugging Face's Transformers library to select an appropriate generative AI model (e.g., BERT) and then trains it. The following is an example of a prompt:

[0631] Example of a prompt:

[0632] Question: What are the latest lead generation results?

[0633] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[0634] Inquiry processing phase

[0635] When a factory robot performs a new task, it queries the server for technical terms and processes as needed. The robot receives instructions and questions from the user (operator) via its built-in display or voice output. These queries are transmitted to the server through a web framework such as Flask.

[0636] Database search and response generation phase

[0637] The server analyzes the received query using an NLP model, retrieves appropriate information from the database, and generates a response. The generated response is sent to the robot and communicated to the operator via display or voice output. gTTS (Google Text-to-Speech) can be used for voice output.

[0638] Feedback processing phase

[0639] The robot sends feedback to the server based on its performance. This feedback includes user ratings and comments. The server uses the received feedback to continuously improve the performance of the NLP model. TensorFlow or PyTorch can be used to retrain the model.

[0640] Specific example

[0641] For example, consider a scenario where a factory robot asks for the "latest lead generation result." In this case, the server performs the analysis using the following prompt:

[0642] Example of a prompt:

[0643] Question: What are the latest lead generation results?

[0644] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[0645] In this way, it becomes possible to acquire information immediately and perform tasks efficiently, which is expected to improve the overall system performance.

[0646] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0647] Step 1: Data Collection Phase

[0648] The server collects terminology lists and related information from each department within the factory. Inputs include terminology lists, process documents, manuals, and project reports provided by each department. The server uses Apache NiFi to retrieve and relay this data. Outputs are stored on the server as raw data for cleansing.

[0649] Step 2: Data Cleansing

[0650] The server cleanses the collected data. Specifically, it uses the Pandas library to remove duplicate data, correct incomplete entries, and standardize the format. The input is raw data, and the output is cleaned data. This data is stored in a database.

[0651] Step 3: Training the natural language processing model

[0652] The server prepares a natural language processing model and trains it on cleansed data. It uses the Hugging Face's Transformers library to train a generative AI model (e.g., BERT). The input is a categorized list of terms and their usage contexts, and the trained output is an NLP model with high suitability for specific technical terms and processes.

[0653] Step 4: Receiving User Inquiries

[0654] When a factory robot performs a new task, it queries a server for technical terms and processes. The terminal (robot) receives user questions through its built-in display or voice input and transmits them to the server using the Flask framework. The input is the query content, and the output is the query request to the server.

[0655] Step 5: Analyze the inquiry

[0656] The server analyzes incoming queries using an NLP model. The input data is the user's question, and the output, as a result of the analysis, is information related to the appropriate technical terms and processes. The server then converts this into a query to be sent to the database.

[0657] Step 6: Database search and answer generation

[0658] The server uses the generated query to search the database and produce an appropriate response. The input is the parsed query, and the output is the generated response. For example, in the case of a query about the latest lead generation results, a response such as "According to the latest report from the lead generation department, 100 leads were acquired last week" would be generated.

[0659] Step 7: Submit and view results

[0660] The server sends the generated response to the terminal (robot). The terminal displays the result on its screen or communicates it to the user via voice output using gTTS. The input is the generated response text, and the output is the displayed information or voice.

[0661] Step 8: Receiving Feedback

[0662] The user provides feedback on the robot's performance. The terminal collects the user's feedback and transmits it to the server. The input is the user's evaluation and comments, and the output is the feedback data sent to the server.

[0663] Step 9: Model Improvement

[0664] The server uses the received feedback to improve the performance of the NLP model. The input is the feedback data, and the output is the updated and improved natural language processing model. TensorFlow or PyTorch is used for retraining.

[0665] This allows the system to continuously learn and improve, supporting the efficient task execution of factory robots.

[0666] 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.

[0667] This invention not only provides a system tailored to the specialized terminology and processes used in various departments of a company, but also aims to provide more appropriate information and improve the user experience by combining it with an emotion engine that recognizes user emotions. This system consists of a process of collecting, cleansing, and storing a list of specialized terms and related information in a database, and learning using a natural language processing (NLP) model. It also receives user inquiries, analyzes them, generates and displays answers, and collects and analyzes feedback to improve the model. Here, we will describe a specific embodiment that combines the system with an emotion engine that recognizes user emotions.

[0668] Data collection phase

[0669] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This information is then cleansed by the server to ensure consistency and accuracy before being stored in the database.

[0670] Data processing and learning phases

[0671] Next, the server prepares an NLP model and trains it using data from the database containing specialized terminology and the context in which it is used. As a result of the training, the model will be able to understand the context in which specific terms are used and process them appropriately. The performance of the trained model is also evaluated, and parameters are adjusted as needed.

[0672] Inquiry processing phase

[0673] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal transmits the entered question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[0674] Database search and response generation phase

[0675] The server searches the database based on the analysis results and collects the necessary information. From the collected information, it generates an appropriate response. For example, it might create a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[0676] Emotion recognition phase

[0677] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adapt the tone and level of detail of the response it provides.

[0678] Result display phase

[0679] The server sends the generated response to the terminal. The terminal displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[0680] Feedback processing phase

[0681] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[0682] Model and emotion engine improvement phase

[0683] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. Specifically, if there is inappropriate feedback, it analyzes its content and uses it to retrain the model. It is also used as data to improve the accuracy of emotion recognition.

[0684] In this embodiment, the system of the present invention not only provides information on technical terms and processes quickly and accurately, but also recognizes the user's emotions and provides appropriate information based on those emotions. This makes it possible to improve the user's work efficiency and enhance the quality of the user experience.

[0685] The following describes the processing flow.

[0686] Data collection phase

[0687] Step 1:

[0688] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[0689] Step 2:

[0690] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[0691] Step 3:

[0692] The server stores the cleansed data in the database.

[0693] Data processing and learning phases

[0694] Step 1:

[0695] The server prepares a natural language processing (NLP) model.

[0696] Step 2:

[0697] The server prepares the training data using specialized terminology and data used in the context of the database.

[0698] Step 3:

[0699] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[0700] Step 4:

[0701] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[0702] Inquiry processing phase

[0703] Step 1:

[0704] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[0705] Step 2:

[0706] The terminal forwards the entered question to the server.

[0707] Step 3:

[0708] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[0709] Database search and response generation phase

[0710] Step 1:

[0711] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[0712] Step 2:

[0713] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[0714] Step 3:

[0715] The server transfers the generated response to the terminal.

[0716] Emotion recognition phase

[0717] Step 1:

[0718] The server passes the user's input to the sentiment engine, which then recognizes the emotions associated with it. For example, if the entered question is "Please tell me about this setting," the sentiment engine analyzes whether the sentence contains doubt or anxiety.

[0719] Step 2:

[0720] The server adjusts the tone of its responses based on the results of the emotion engine. For example, if the user is feeling anxious, the responses will become more helpful and detailed.

[0721] Result display phase

[0722] Step 1:

[0723] The terminal receives the response sent from the server.

[0724] Step 2:

[0725] The device displays the received responses to the user. The displayed responses are presented in an appropriate tone and level of detail based on the user's emotions as recognized by the emotion engine.

[0726] Feedback processing phase

[0727] Step 1:

[0728] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[0729] Step 2:

[0730] The device forwards user feedback to the server.

[0731] Model and emotion engine improvement phase

[0732] Step 1:

[0733] The server analyzes the received feedback and improves both the NLP model and the emotion engine. For example, if there is feedback indicating an inappropriate response, it analyzes the content and uses it to retrain the model.

[0734] Step 2:

[0735] The server adjusts the parameters of its NLP model and emotion engine based on the feedback, providing more accurate and useful answers to subsequent inquiries.

[0736] (Example 2)

[0737] 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".

[0738] The specialized terminology and processes used in various departments of a company are diverse, requiring advanced knowledge and rapid response to understand them and provide appropriate information. However, conventional systems have not adequately understood the context of specialized terminology or recognized user emotions, making it difficult to provide appropriate information and improve the user experience. Therefore, there is a need to develop systems that can provide information in a way that suits user needs.

[0739] The specific processing performed by the specific 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 a list of technical terms and related information from each department of a company, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing inquiries from users, means for searching the database based on the analysis results and generating an appropriate answer, means for sending and displaying the generated answer on a user terminal, and means for recognizing emotions from the content of the user's inquiry. This not only enables the provision of information specific to the technical terms and processes of each department, but also allows for adjustment of the tone and level of detail of the answer according to the user's emotions, which is expected to improve the user experience.

[0740] "Each department of a company" refers to multiple departments within a company, each with different tasks and roles.

[0741] A "list of technical terms" is a list that compiles specific terms and expressions used in a particular industry or field.

[0742] "Related information" refers to information such as definitions, usage examples, process documents, manuals, and project reports associated with technical terms.

[0743] A "server" is a computer system that provides information and services over a network.

[0744] "Means of collection" refers to the methods and techniques used to consolidate dispersed data into a single entity.

[0745] "Data cleansing methods" refer to techniques for removing duplicates, errors, and incomplete information from collected data, making it accurate and consistent.

[0746] "Storing in a database" means saving processed data in a structured format within a database.

[0747] A "natural language processing model" is a type of machine learning model used to understand and process human language.

[0748] "Methods for learning technical terms and their contexts" refers to methods of training models to understand the situations and contexts in which specific terms are used.

[0749] "Means for receiving and analyzing inquiries" refers to methods and technologies for receiving questions and requests from users and analyzing their content.

[0750] "Methods for searching a database based on analysis results" refers to methods for obtaining relevant data from a database based on the analyzed information.

[0751] "Means of generating appropriate answers" refers to methods of creating answers that are useful to the user based on search results.

[0752] A "user terminal" is a device (e.g., a personal computer, a smartphone) that a user uses to access information and services.

[0753] "Means of recognizing emotions" refers to technologies and methods for identifying emotions from user input.

[0754] "Methods for receiving feedback and using it to improve the model" refers to methods of receiving evaluations and opinions from users, updating the model based on them, and improving its performance.

[0755] A "customized natural language processing model" is a natural language processing model that has been tailored to specific needs or applications.

[0756] An "emotion recognition engine" is a specialized system or algorithm used to evaluate and analyze a user's emotions.

[0757] This invention is a system that specializes in the specialized terminology and processes used in each department of a company, and further combines this with an emotion engine that recognizes user emotions, thereby providing more appropriate information and improving the user experience. This system is implemented through the following series of processes.

[0758] Data collection phase

[0759] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This data collection is performed using API-based integration. The collected information is then cleansed by the server to ensure consistency and accuracy before being stored in the database. Data cleansing tools such as OpenRefine are used for this data cleansing process.

[0760] Data processing and learning phases

[0761] Next, the server prepares a natural language processing (NLP) model and trains it using data from the database containing specialized terms and the context in which they are used. NLP models used include natural language processing libraries such as Spacy and Hugging Face. This training enables the model to understand the context in which specific terms are used and to process them appropriately. After training, the model's performance is evaluated, and parameters are adjusted as needed.

[0762] Inquiry processing phase

[0763] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal forwards this question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[0764] Database search and response generation phase

[0765] The server searches the database based on the analysis results and collects the necessary information. It then generates an appropriate response. For example, it might provide a specific response such as, "The latest lead generation result shows that 100 leads were acquired last week."

[0766] Emotion recognition phase

[0767] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adjust the tone and level of detail of the response it provides.

[0768] Result display phase

[0769] The server sends the generated response to the device, and the device displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[0770] Feedback processing phase

[0771] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[0772] Model and emotion engine improvement phase

[0773] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. If there is feedback regarding inappropriate responses, the analysis results are used to retrain the model. The feedback is also utilized as data to improve the accuracy of emotion recognition.

[0774] Examples of specific cases and prompts for generative AI models.

[0775] As a concrete example, in response to a question about the latest lead generation results in the marketing department, the system provides the answer, "The latest lead generation results show that we acquired 100 leads last week." If the emotion engine detects that the user is in a hurry, it will quickly display a concise response such as, "Here are the latest lead generation results. We acquired 100 leads last week."

[0776] Examples of prompts for a generative AI model:

[0777] "Could you please tell me the latest lead generation results for your marketing department?"

[0778] "Could you please tell me the latest project progress in the XX department?"

[0779] "Please use the emotion engine to evaluate user satisfaction."

[0780] This system aims to improve operational efficiency and user experience by integrating the interpretation of technical terms with the recognition of user emotions.

[0781] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0782] Step 1:

[0783] The server collects a list of technical terms and related information from each department of the company. The input requires access information to each department's data source. Specifically, the server uses an API to send HTTP requests to each department's data source to retrieve the technical term list and related information. The output is the collected raw data.

[0784] Step 2:

[0785] The server cleanses the collected data. The raw data obtained in step 1 is used as input. Specifically, a data cleansing tool such as OpenRefine is used to remove duplicate data, fill in incomplete data, and ensure consistency. The output is the cleansed data.

[0786] Step 3:

[0787] The server stores the cleansed data in the database. The cleansed data generated in step 2 is required as input. Specifically, a database management system such as MySQL is used, and the data is saved to the database using SQL queries. The output is structured data stored in the database.

[0788] Step 4:

[0789] The server prepares a natural language processing (NLP) model. The input requires an existing trained model and a list of specialized terminology for each field. Specifically, it imports and initializes natural language processing libraries such as Spacy or Hugging Face. The output is the NLP model prepared for training.

[0790] Step 5:

[0791] The server trains an NLP model using the specialized terminology and data used in its context within the database. The inputs required are the data stored in step 3 and the NLP model prepared in step 4. Specifically, the model is fed data, hyperparameters such as the number of epochs and learning rate are set, and training is performed. The output is the trained NLP model.

[0792] Step 6:

[0793] The server evaluates the trained model and adjusts the parameters as needed. The input requires a trained NLP model and an evaluation dataset. Specifically, it measures the model's accuracy and error and readjusts the hyperparameters. The output is an optimized NLP model.

[0794] Step 7:

[0795] The user enters a question into the system using a terminal. Input requires the user to enter a text box, such as "What are the latest lead generation results?". The output is the user's question stored in text format on the terminal.

[0796] Step 8:

[0797] The terminal forwards the user's question to the server. The input requires the user's question. Specifically, it sends an HTTP request to the server and transfers the data. The output is the server receiving the question.

[0798] Step 9:

[0799] The server uses an NLP model to analyze user questions. The input requires the user's question, transmitted from the terminal, and a trained NLP model. Specifically, the model analyzes the question and identifies relevant terminology and context. The output is the analysis result.

[0800] Step 10:

[0801] The server searches the database based on the analysis results. The analysis results and the database are required as input. Specifically, it generates an SQL query, sends it to the database, and retrieves the relevant data. The output will provide the necessary information.

[0802] Step 11:

[0803] The server generates appropriate answers from the collected information. It requires information retrieved from a database as input. Specifically, it integrates the information to create a user-friendly answer. The output is the generated answer.

[0804] Step 12:

[0805] The server uses an emotion engine to recognize emotions from user input. The input requires both the user's question and the emotion engine. Specifically, it performs text analysis to evaluate emotions based on tone and context. The output is the recognized emotion.

[0806] Step 13:

[0807] The server sends the generated response to the terminal, and the terminal displays the received response to the user. The required inputs are the generated response and the sentiment analysis results. Specifically, the response is sent to the terminal via an HTTP response, which the terminal then displays. The output is the response that the user can view.

[0808] Step 14:

[0809] The user provides feedback on the provided answer. The input requires the answer displayed by the system, and the user inputs feedback such as "This answer was helpful" or "This answer was inappropriate." The output is the feedback stored on the device.

[0810] Step 15:

[0811] The terminal forwards user feedback to the server. The input requires user feedback. Specifically, it sends the feedback to the server using an HTTP request. The output is the server receiving the feedback.

[0812] Step 16:

[0813] The server analyzes the received feedback and improves the NLP model and emotion engine. Input requires user feedback and training data. Specifically, it retrains the model based on the feedback to improve performance. The output is an improved NLP model and emotion engine.

[0814] (Application Example 2)

[0815] 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."

[0816] In traditional brick-and-mortar stores, communication between employees and customers often lacked the speed and accuracy of providing specialized information, and it was difficult to respond appropriately to customers' emotions. This resulted in decreased customer satisfaction and reduced operational efficiency. This invention aims to solve these problems and enable the provision of higher-quality service in customer interactions at physical stores.

[0817] The specific processing performed by the specific 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 a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating an appropriate response, means for transmitting and displaying the generated response on a user terminal, means for recognizing the user's emotions and adjusting the response tone and level of detail based on the recognition results, means for receiving feedback and using it to improve the model, and a natural language processing model customized based on the technical terms and related information of each department. This enables employees in physical stores to provide information quickly and accurately, as well as to respond appropriately to customer emotions.

[0818] A "glossary" is a list that compiles specific technical terms and expertise used within a company or department.

[0819] "Related information" refers to all information related to technical terms, including definitions of technical terms, examples of usage, process documents, manuals, and project reports.

[0820] "Means of collection" refers to the methods and devices for gathering a list of technical terms and related information.

[0821] "Cleansing" refers to the process of organizing and correcting collected data to ensure consistency and accuracy.

[0822] A "database" is a collection of data that is efficiently stored and managed in a way that makes it searchable and usable.

[0823] A "natural language processing model" is a model used to train algorithms and methods for analyzing and understanding human language using computers.

[0824] "Means of learning" refers to methods and devices for training natural language processing models to learn specialized terminology and its context.

[0825] "Means of analysis" refers to methods and devices for analyzing user inquiries and understanding their content.

[0826] "Searching means" refers to methods and devices for finding relevant information within a database based on analysis results.

[0827] "Means for generating answers" refers to methods and devices for creating appropriate answers for users based on retrieved information.

[0828] "User terminal" refers to devices or systems that a user directly operates, such as smartphones or computers.

[0829] "Means for recognizing emotions and adjusting response tone and level of detail based on the recognition results" refers to methods and apparatus for analyzing a user's emotions and adapting the tone and level of detail of the response accordingly.

[0830] "Means for receiving feedback and using it to improve the model" refers to methods and devices for receiving evaluations and opinions from users and incorporating them into improvements to natural language processing models and sentiment engines.

[0831] A "customized natural language processing model" refers to a natural language processing model that has been specifically tailored to a particular company or department.

[0832] This invention relates to an information provision system for improving the quality of customer service in physical stores. Specific embodiments thereof are described below.

[0833] 1. Data collection and cleansing

[0834] First, the server collects a list of technical terms and related information (e.g., product features, process descriptions, manuals, and usage examples) from various departments within the company. The collected data is then cleansed by the server and stored in a database, ensuring consistency and accuracy. This database is then used as a source of information for subsequent queries.

[0835] 2. Training a natural language processing model

[0836] The server prepares a natural language processing (NLP) model and trains it based on specialized terminology and its contextual information stored in a database. Through this training, the model learns to understand the context in which specific terms are used and generates appropriate responses accordingly.

[0837] 3. Inquiry analysis and response generation

[0838] The user (in this case, a store employee) enters an inquiry into the system using a device such as a smartphone. For example, the question might be, "Please tell me about the features of this new product." The device transmits the entered question to the server, which analyzes the question using an NLP model. Based on the analysis results, it searches the database and generates an appropriate answer. In the process, it also analyzes the customer's emotions and adjusts the response tone and level of detail accordingly.

[0839] 4. Emotion Recognition and Response Adjustment

[0840] The server uses an emotion engine to recognize emotions from user input. For example, it analyzes the tone and context of the entered question and adopts a comforting response tone if the user is dissatisfied. This allows users to receive more appropriate and satisfying service.

[0841] 5. Feedback and Model Improvement

[0842] The system also includes a feature that allows users to provide feedback on the answers they receive. For example, users can input feedback such as "This answer was helpful" or "This answer was inappropriate." The user's device forwards this feedback to the server, which then uses it to improve both the NLP model and the emotion engine. This enables continuous model improvement based on feedback.

[0843] 6. Usage example

[0844] For example, if a customer asks, "What are the features of the new product?", the server will provide information such as, "This is a new leather bag made from high-quality cowhide." Also, if a customer asks in a dissatisfied tone, "Is it possible to return this item?", the server will respond with something like, "Returns are accepted within 30 days of purchase. Please contact us if you have any problems; we will be happy to assist you."

[0845] Example of a prompt

[0846] "Could you please tell me the specifications of this product?"

[0847] "Is this item returnable?"

[0848] In this way, the present invention facilitates communication between employees and customers in physical stores, enabling the rapid and accurate provision of specialized information and appropriate responses that respond to customer emotions. This is expected to improve customer satisfaction and operational efficiency.

[0849] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0850] Step 1:

[0851] The server collects terminology lists and related information from each department. Specifically, the server uses APIs and file transfer protocols to retrieve data from each department's databases and document management systems. The input here is the terminology lists and related information provided by each department, and the output is the collected data.

[0852] Step 2:

[0853] The server cleanses the collected data, removing unnecessary and incorrect data to maintain consistency and accuracy. Specifically, it performs data cleaning operations using libraries such as Python's pandas library. The input here is the data collected in the previous step, and the output is the cleansed data.

[0854] Step 3:

[0855] The server stores the cleansed data in a database. Specifically, it inserts the data into an SQL database or a NoSQL database. Here, the input is the cleansed data, and the output is the data stored in the database.

[0856] Step 4:

[0857] The server trains a natural language processing (NLP) model. Specifically, it trains an NLP model (e.g., BERT or GPT) using technical terms and their contexts from a database. The input here is the technical terms and contextual information from the database, and the output is the trained NLP model.

[0858] Step 5:

[0859] The user (employee) enters a question using a terminal. For example, they might enter, "Please tell me the features of this new product." Here, the input is the user's question, and the output is the user's input data sent to the terminal.

[0860] Step 6:

[0861] The terminal forwards user input to the server. Specifically, it sends questions to the server using HTTP requests. Here, input refers to the user's input data, and output refers to the data forwarded to the server.

[0862] Step 7:

[0863] The server uses an NLP model to analyze the user's question. Specifically, it understands the context of the question and identifies relevant technical terms. The input here is the user's question, and the output is the analysis result.

[0864] Step 8:

[0865] The server searches the database based on the analysis results and collects the appropriate answers. The input here is the analysis results from the NLP model, and the output is the information collected from the database.

[0866] Step 9:

[0867] The server uses an emotion engine to recognize the user's emotions. Specifically, it analyzes the tone and context of the input question to identify the emotions the user is experiencing. Here, the input is the user's question, and the output is the emotion recognition result.

[0868] Step 10:

[0869] The server adjusts the tone and level of detail of its response based on the recognized emotion. For example, if the user is dissatisfied, it adopts a comforting tone. The input here is the emotion recognition result and collected information, and the output is the adjusted response.

[0870] Step 11:

[0871] The server sends the adjusted response to the terminal. Specifically, it sends the response to the terminal using an HTTP response. Here, the input is the adjusted response, and the output is the response sent to the terminal.

[0872] Step 12:

[0873] The terminal displays the received response to the user. The input here is the response sent to the terminal, and the output is the response displayed to the user.

[0874] Step 13:

[0875] Users provide feedback on the provided answers. For example, they might rate the answer as "This answer was helpful" or "This answer was inappropriate." The input here is the user's feedback, and the output is the feedback entered into the device.

[0876] Step 14:

[0877] The terminal forwards user feedback to the server. Specifically, it sends feedback to the server using an HTTP request. Here, the input is the feedback entered into the terminal, and the output is the feedback forwarded to the server.

[0878] Step 15:

[0879] The server improves the NLP model and emotion engine based on the feedback it receives. Specifically, it analyzes the content of the feedback and retrains the model as needed. The input here is the user's feedback, and the output is the improved NLP model and emotion engine.

[0880] These steps enable employees in physical stores to provide customers with quick and accurate information, as well as to respond appropriately to customers' emotions.

[0881] 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.

[0882] The data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of the data generation model 58 is ChatGPT (Internet Search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">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.

[0883] 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.

[0884] [Third Embodiment]

[0885] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0886] 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.

[0887] 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).

[0888] 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.

[0889] 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.

[0890] 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).

[0891] 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.

[0892] 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.

[0893] 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.

[0894] 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.

[0895] 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.

[0896] 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".

[0897] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. This system is implemented through a series of programmatic processes described below.

[0898] Data collection phase

[0899] First, the server collects a list of specialized terms and related information provided by each department of the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. The collected data is cleansed by the server to ensure consistency and accuracy. Specifically, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in the database.

[0900] Data processing and learning phases

[0901] Next, the server prepares a natural language processing (NLP) model and trains it based on the collected data. The server prepares the training data using specialized terminology and data used in its context within the database, and then trains the NLP model. For example, it teaches the model that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[0902] Inquiry processing phase

[0903] The user enters a question into the system using a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this entered question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[0904] Database search and response generation phase

[0905] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[0906] Result display phase

[0907] The generated response is sent from the server to the terminal. The terminal then displays the received response to the user.

[0908] Feedback processing phase

[0909] Users can provide feedback on the displayed answers. For example, they can input feedback such as "This answer was helpful" or "This answer was inappropriate." The device forwards this feedback to the server, which uses the received feedback to improve the NLP model. Inappropriate answers are used to retrain the model.

[0910] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, a feedback function continuously improves the accuracy and usefulness of the system. By using NLP models customized for each department, it achieves highly accurate information provision tailored to the specific needs of each department.

[0911] The following describes the processing flow.

[0912] Data collection phase

[0913] Step 1:

[0914] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[0915] Step 2:

[0916] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[0917] Step 3:

[0918] The server stores the cleansed data in the database.

[0919] Data processing and learning phases

[0920] Step 1:

[0921] The server prepares a natural language processing (NLP) model.

[0922] Step 2:

[0923] The server prepares the training data using specialized terminology and data used in the context of the database.

[0924] Step 3:

[0925] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[0926] Step 4:

[0927] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[0928] Inquiry processing phase

[0929] Step 1:

[0930] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[0931] Step 2:

[0932] The terminal forwards the entered question to the server.

[0933] Step 3:

[0934] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[0935] Database search and response generation phase

[0936] Step 1:

[0937] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[0938] Step 2:

[0939] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[0940] Step 3:

[0941] The server transfers the generated response to the terminal.

[0942] Result display phase

[0943] Step 1:

[0944] The terminal receives the response sent from the server.

[0945] Step 2:

[0946] The device displays the received response to the user.

[0947] Feedback processing phase

[0948] Step 1:

[0949] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[0950] Step 2:

[0951] The device forwards user feedback to the server.

[0952] Step 3:

[0953] The server analyzes the received feedback and uses it to improve the NLP model. For example, if there is feedback indicating an inappropriate response, the model is retrained based on that information.

[0954] Step 4:

[0955] The server adjusts the parameters of the NLP model based on the feedback, providing more accurate and useful answers to subsequent queries.

[0956] Through these steps, the system effectively manages the specialized terminology of each department and enables the rapid and accurate provision of information to users.

[0957] (Example 1)

[0958] 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."

[0959] Traditional enterprise systems faced the challenge of requiring significant time and effort to adapt to the different jargon and processes of each department. This hindered operational efficiency and could lead to decreased productivity. Furthermore, inaccurate responses to user inquiries prevented proper feedback from being obtained, hindering system improvement. In addition, insufficient performance of natural language processing models could lead to errors in analyzing complex jargon and context.

[0960] 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.

[0961] In this invention, the server includes means for collecting a list of technical terms and related information from each department, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for sending and displaying the generated answers on a user terminal, means for receiving user feedback and using it to improve the model, and means for providing a natural language processing model customized based on the technical terms and related information of each department. This enables the rapid and accurate provision of information that meets the specific needs of each department, resulting in increased operational efficiency and continuous system improvement.

[0962] "Each department of a company"

[0963] This refers to organizational units within a company that are responsible for different tasks or functions. Examples include the marketing department and the human resources department.

[0964] "List of Technical Terms"

[0965] This refers to a list of terms and phrases commonly used in a particular industry or field.

[0966] Related Information

[0967] This refers to supplementary data provided alongside technical terms, such as definitions of terms, usage examples, process documents, manuals, and project reports.

[0968] "cleansing"

[0969] This refers to operations performed on collected data, such as removing duplicates, correcting incomplete entries, and standardizing the format.

[0970] "Database"

[0971] This refers to a system for organizing and efficiently managing collected data.

[0972] "Natural language processing models"

[0973] This refers to artificial intelligence models designed to understand, analyze, and generate human language. Machine learning techniques are used in this process.

[0974] "Means for receiving and analyzing inquiries"

[0975] This refers to the process of receiving questions submitted by users, analyzing their content, and extracting relevant information.

[0976] "A means of searching a database and generating appropriate answers."

[0977] This refers to a function that searches a database based on the analyzed results and creates an answer that is appropriate for the user's question.

[0978] "User terminal"

[0979] This refers to a device that a user uses to access a system and input or retrieve information.

[0980] "feedback"

[0981] This refers to user-provided ratings and feedback, which are used to improve the system.

[0982] "Generative AI Model"

[0983] This refers to an artificial intelligence model trained for natural language processing.

[0984] "Prompt message"

[0985] This refers to the questions or instructions that users input into the system.

[0986] "Customized natural language processing models"

[0987] This refers to artificial intelligence models that have been specifically tailored to a particular department or task.

[0988] "Parameter tuning"

[0989] This refers to the process of adjusting a model to improve its performance. Examples include setting appropriate hyperparameters and adjusting the learning rate.

[0990] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. The system utilizes servers and terminals as hardware, and employs NLP (Natural Language Processing) technology and database management software as software. Typical software used includes Python and its libraries (e.g., spaCy, NLTK, TensorFlow, PyTorch).

[0991] The server collects terminology lists and related information provided by various departments within the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. To cleanse this collected data, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in a database.

[0992] Next, the server prepares and trains an NLP model based on the collected data. The training data is prepared using specialized terminology and the context in which it is used within the database. For example, the model is trained to understand that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[0993] The user enters a question into the system using their own device. For example, they might enter, "What are the latest lead generation results?" This device forwards the user's question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[0994] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that you acquired 100 leads last week." The generated response is sent from the server to the terminal, which then displays it to the user.

[0995] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." This feedback is sent from the device to the server. The server uses the received feedback to improve the NLP model. Inappropriate answers are used as new training data for retraining.

[0996] The following are specific examples of prompts to enter into the system:

[0997] 1. "Please tell me your latest lead generation results."

[0998] 2. "What are the new trends in SEO?"

[0999] 3. "Please tell me about tools to increase conversion rates."

[1000] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, the system's accuracy and usefulness can be continuously improved through a feedback function. By using NLP models customized for each department, it achieves the provision of highly accurate information that meets the specific needs of each department.

[1001] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1002] Step 1: Data Collection

[1003] The server collects a list of specialized terms and related information from each department within the company. Specifically, it gathers data based on Excel files and documents entered or uploaded by departmental staff. The collected data includes definitions of terms, usage examples, process documents, manuals, and project reports.

[1004] Input: A list of specialized terms and related information provided by the department head.

[1005] Output: Collected raw data

[1006] Step 2: Data Cleansing

[1007] The server cleanses the collected data. This includes removing duplicate data, correcting incomplete entries (such as filling in blanks or correcting typos), and standardizing the format (for example, consistently converting date formats to YYYY-MM-DD).

[1008] Input: Collected raw data

[1009] Output: Cleansed data

[1010] Step 3: Database Storage

[1011] The server stores the cleansed data in the database. This ensures that data with guaranteed consistency and accuracy can be managed.

[1012] Input: Cleansed data

[1013] Output: Data stored in the database

[1014] Step 4: Training the NLP Model

[1015] The server trains an NLP model using the collected and cleansed data. It uses Python natural language processing libraries (e.g., spaCy, NLTK) or machine learning libraries (e.g., TensorFlow, PyTorch). The data is split into training and test data and fed to the NLP model. For example, the model is trained to understand that "lead" means "potential customer". After training, the model's performance is evaluated, and hyperparameters are tuned as needed.

[1016] Input: Cleansed data

[1017] Output: Trained NLP model

[1018] Step 5: Inquiry reception and analysis

[1019] The user enters a question into the system via a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this question to the server. The server analyzes the received question using an NLP model to identify keywords ("lead generation," "results") and context.

[1020] Input: User question (prompt)

[1021] Output: Results of question analysis (keywords and context)

[1022] Step 6: Database search and answer generation

[1023] The server searches the database based on the analysis results. For example, it searches for the latest project reports and statistics on lead generation and generates specific answers such as, "The latest lead generation results show that 100 leads were acquired last week." The answers are formatted using automated templates.

[1024] Input: Results of the analyzed question

[1025] Output: Generated answer

[1026] Step 7: Display Results

[1027] The generated response is sent from the server to the terminal. The terminal displays this response to the user through a GUI (Graphical User Interface). For example, it might be displayed on the dashboard of a web application.

[1028] Input: Response sent from the server

[1029] Output: Answer displayed on the user's terminal

[1030] Step 8: Feedback

[1031] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." The device sends this feedback to the server. The server uses the received feedback to improve the performance of the NLP model. Inappropriate answers are used as new training data for retraining.

[1032] Input: User feedback

[1033] Output: Improved NLP model

[1034] (Application Example 1)

[1035] 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."

[1036] In conventional systems, factory robots had difficulty quickly understanding specialized terminology and processes when performing new tasks. Furthermore, because different departments had different specialized terminology and processes, there was a lack of means to accurately acquire that information and efficiently perform tasks. In addition, there was a need for robots to acquire necessary information in real time while in operation and to improve their movements based on appropriate feedback.

[1037] 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.

[1038] In this invention, the server includes means for collecting a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their contexts using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for transmitting and displaying the generated answers on a user terminal, means for querying technical terms and processes while a factory robot is performing a new task and obtaining necessary information, and means for displaying the results via the robot's display or audio output. This enables the factory robot to quickly acquire information on specific technical terms and processes and perform tasks efficiently. Furthermore, the model can be continuously improved based on the robot's operation results, thereby improving the accuracy and efficiency of its operation.

[1039] A "glossary of technical terms" is a list that compiles terms used in a specific field or industry, their definitions, and related information.

[1040] "Related information" refers to materials such as usage examples related to technical terms, process documents, manuals, and project reports.

[1041] "Data cleansing" is an operation that involves removing duplicate data, correcting incomplete entries, and standardizing the format in order to ensure the consistency and accuracy of collected data.

[1042] A "database" is a collection of information that stores a list of specialized terms and related information, making it searchable and retrievable.

[1043] A "natural language processing model" is a machine learning model used to understand, analyze, and generate human language.

[1044] "Analysis" is the process of understanding user inquiries using natural language processing models and grasping their intent.

[1045] An "inquiry" refers to a question or request for information that a user makes to a system.

[1046] A "user terminal" is a device used by a user to access a system, input information, and receive results.

[1047] A "factory robot" is a mechanical device used in industrial settings to automatically perform specific tasks.

[1048] A "display" is a monitor or screen used to visually display information.

[1049] "Voice output" is a function that conveys information and instructions from the system to the user as voice.

[1050] "Feedback" refers to evaluations and comments provided by users regarding the system's responses and actions.

[1051] "Model improvement" is the process of improving the performance of a natural language processing model based on the feedback received.

[1052] This invention relates to a system that helps factory robots quickly understand and efficiently perform new tasks. The system is realized through the following processes. The main components of the system consist of a server, a robot as an edge device, and a user (human operator).

[1053] Data collection phase

[1054] The server collects a glossary of terms and related information from each department within the factory. The glossary includes related information such as process documents, manuals, and project reports. Apache NiFi can be used for this collection. The collected data is cleansed by the server. The Pandas library is used to remove duplicate data, correct incomplete entries, and standardize the format. The cleansed data is then stored in a database (e.g., PostgreSQL).

[1055] Data processing and learning phases

[1056] The server prepares a natural language processing (NLP) model and trains it based on the collected data. It uses the Hugging Face's Transformers library to select an appropriate generative AI model (e.g., BERT) and then trains it. The following is an example of a prompt:

[1057] Example of a prompt:

[1058] Question: What are the latest lead generation results?

[1059] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[1060] Inquiry processing phase

[1061] When a factory robot performs a new task, it queries the server for technical terms and processes as needed. The robot receives instructions and questions from the user (operator) via its built-in display or voice output. These queries are transmitted to the server through a web framework such as Flask.

[1062] Database search and response generation phase

[1063] The server analyzes the received query using an NLP model, retrieves appropriate information from the database, and generates a response. The generated response is sent to the robot and communicated to the operator via display or voice output. gTTS (Google Text-to-Speech) can be used for voice output.

[1064] Feedback processing phase

[1065] The robot sends feedback to the server based on its performance. This feedback includes user ratings and comments. The server uses the received feedback to continuously improve the performance of the NLP model. TensorFlow or PyTorch can be used to retrain the model.

[1066] Specific example

[1067] For example, consider a scenario where a factory robot asks for the "latest lead generation result." In this case, the server performs the analysis using the following prompt:

[1068] Example of a prompt:

[1069] Question: What are the latest lead generation results?

[1070] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[1071] In this way, it becomes possible to acquire information immediately and perform tasks efficiently, which is expected to improve the overall system performance.

[1072] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1073] Step 1: Data Collection Phase

[1074] The server collects terminology lists and related information from each department within the factory. Inputs include terminology lists, process documents, manuals, and project reports provided by each department. The server uses Apache NiFi to retrieve and relay this data. Outputs are stored on the server as raw data for cleansing.

[1075] Step 2: Data Cleansing

[1076] The server cleanses the collected data. Specifically, it uses the Pandas library to remove duplicate data, correct incomplete entries, and standardize the format. The input is raw data, and the output is cleaned data. This data is stored in a database.

[1077] Step 3: Training the natural language processing model

[1078] The server prepares a natural language processing model and trains it on cleansed data. It uses the Hugging Face's Transformers library to train a generative AI model (e.g., BERT). The input is a categorized list of terms and their usage contexts, and the trained output is an NLP model with high suitability for specific technical terms and processes.

[1079] Step 4: Receiving User Inquiries

[1080] When a factory robot performs a new task, it queries a server for technical terms and processes. The terminal (robot) receives user questions through its built-in display or voice input and transmits them to the server using the Flask framework. The input is the query content, and the output is the query request to the server.

[1081] Step 5: Analyze the inquiry

[1082] The server analyzes incoming queries using an NLP model. The input data is the user's question, and the output, as a result of the analysis, is information related to the appropriate technical terms and processes. The server then converts this into a query to be sent to the database.

[1083] Step 6: Database search and answer generation

[1084] The server uses the generated query to search the database and produce an appropriate response. The input is the parsed query, and the output is the generated response. For example, in the case of a query about the latest lead generation results, a response such as "According to the latest report from the lead generation department, 100 leads were acquired last week" would be generated.

[1085] Step 7: Submit and view results

[1086] The server sends the generated response to the terminal (robot). The terminal displays the result on its screen or communicates it to the user via voice output using gTTS. The input is the generated response text, and the output is the displayed information or voice.

[1087] Step 8: Receiving Feedback

[1088] The user provides feedback on the robot's performance. The terminal collects the user's feedback and transmits it to the server. The input is the user's evaluation and comments, and the output is the feedback data sent to the server.

[1089] Step 9: Model Improvement

[1090] The server uses the received feedback to improve the performance of the NLP model. The input is the feedback data, and the output is the updated and improved natural language processing model. TensorFlow or PyTorch is used for retraining.

[1091] This allows the system to continuously learn and improve, supporting the efficient task execution of factory robots.

[1092] 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.

[1093] This invention not only provides a system tailored to the specialized terminology and processes used in various departments of a company, but also aims to provide more appropriate information and improve the user experience by combining it with an emotion engine that recognizes user emotions. This system consists of a process of collecting, cleansing, and storing a list of specialized terms and related information in a database, and learning using a natural language processing (NLP) model. It also receives user inquiries, analyzes them, generates and displays answers, and collects and analyzes feedback to improve the model. Here, we will describe a specific embodiment that combines the system with an emotion engine that recognizes user emotions.

[1094] Data collection phase

[1095] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This information is then cleansed by the server to ensure consistency and accuracy before being stored in the database.

[1096] Data processing and learning phases

[1097] Next, the server prepares an NLP model and trains it using data from the database containing specialized terminology and the context in which it is used. As a result of the training, the model will be able to understand the context in which specific terms are used and process them appropriately. The performance of the trained model is also evaluated, and parameters are adjusted as needed.

[1098] Inquiry processing phase

[1099] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal transmits the entered question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[1100] Database search and response generation phase

[1101] The server searches the database based on the analysis results and collects the necessary information. From the collected information, it generates an appropriate response. For example, it might create a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[1102] Emotion recognition phase

[1103] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adapt the tone and level of detail of the response it provides.

[1104] Result display phase

[1105] The server sends the generated response to the terminal. The terminal displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[1106] Feedback processing phase

[1107] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[1108] Model and emotion engine improvement phase

[1109] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. Specifically, if there is inappropriate feedback, it analyzes its content and uses it to retrain the model. It is also used as data to improve the accuracy of emotion recognition.

[1110] In this embodiment, the system of the present invention not only provides information on technical terms and processes quickly and accurately, but also recognizes the user's emotions and provides appropriate information based on those emotions. This makes it possible to improve the user's work efficiency and enhance the quality of the user experience.

[1111] The following describes the processing flow.

[1112] Data collection phase

[1113] Step 1:

[1114] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[1115] Step 2:

[1116] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[1117] Step 3:

[1118] The server stores the cleansed data in the database.

[1119] Data processing and learning phases

[1120] Step 1:

[1121] The server prepares a natural language processing (NLP) model.

[1122] Step 2:

[1123] The server prepares the training data using specialized terminology and data used in the context of the database.

[1124] Step 3:

[1125] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[1126] Step 4:

[1127] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[1128] Inquiry processing phase

[1129] Step 1:

[1130] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[1131] Step 2:

[1132] The terminal forwards the entered question to the server.

[1133] Step 3:

[1134] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[1135] Database search and response generation phase

[1136] Step 1:

[1137] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[1138] Step 2:

[1139] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[1140] Step 3:

[1141] The server transfers the generated response to the terminal.

[1142] Emotion recognition phase

[1143] Step 1:

[1144] The server passes the user's input to the sentiment engine, which then recognizes the emotions associated with it. For example, if the entered question is "Please tell me about this setting," the sentiment engine analyzes whether the sentence contains doubt or anxiety.

[1145] Step 2:

[1146] The server adjusts the tone of its responses based on the results of the emotion engine. For example, if the user is feeling anxious, the responses will become more helpful and detailed.

[1147] Result display phase

[1148] Step 1:

[1149] The terminal receives the response sent from the server.

[1150] Step 2:

[1151] The device displays the received responses to the user. The displayed responses are presented in an appropriate tone and level of detail based on the user's emotions as recognized by the emotion engine.

[1152] Feedback processing phase

[1153] Step 1:

[1154] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[1155] Step 2:

[1156] The device forwards user feedback to the server.

[1157] Model and emotion engine improvement phase

[1158] Step 1:

[1159] The server analyzes the received feedback and improves both the NLP model and the emotion engine. For example, if there is feedback indicating an inappropriate response, it analyzes the content and uses it to retrain the model.

[1160] Step 2:

[1161] The server adjusts the parameters of its NLP model and emotion engine based on the feedback, providing more accurate and useful answers to subsequent inquiries.

[1162] (Example 2)

[1163] 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."

[1164] The specialized terminology and processes used in various departments of a company are diverse, requiring advanced knowledge and rapid response to understand them and provide appropriate information. However, conventional systems have not adequately understood the context of specialized terminology or recognized user emotions, making it difficult to provide appropriate information and improve the user experience. Therefore, there is a need to develop systems that can provide information in a way that suits user needs.

[1165] The specific processing performed by the specific 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 a list of technical terms and related information from each department of a company, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing inquiries from users, means for searching the database based on the analysis results and generating an appropriate answer, means for sending and displaying the generated answer on a user terminal, and means for recognizing emotions from the content of the user's inquiry. This not only enables the provision of information specific to the technical terms and processes of each department, but also allows for adjustment of the tone and level of detail of the answer according to the user's emotions, which is expected to improve the user experience.

[1166] "Each department of a company" refers to multiple departments within a company, each with different tasks and roles.

[1167] A "list of technical terms" is a list that compiles specific terms and expressions used in a particular industry or field.

[1168] "Related information" refers to information such as definitions, usage examples, process documents, manuals, and project reports associated with technical terms.

[1169] A "server" is a computer system that provides information and services over a network.

[1170] "Means of collection" refers to the methods and techniques used to consolidate dispersed data into a single entity.

[1171] "Data cleansing methods" refer to techniques for removing duplicates, errors, and incomplete information from collected data, making it accurate and consistent.

[1172] "Storing in a database" means saving processed data in a structured format within a database.

[1173] A "natural language processing model" is a type of machine learning model used to understand and process human language.

[1174] "Methods for learning technical terms and their contexts" refers to methods of training models to understand the situations and contexts in which specific terms are used.

[1175] "Means for receiving and analyzing inquiries" refers to methods and technologies for receiving questions and requests from users and analyzing their content.

[1176] "Methods for searching a database based on analysis results" refers to methods for obtaining relevant data from a database based on the analyzed information.

[1177] "Means of generating appropriate answers" refers to methods of creating answers that are useful to the user based on search results.

[1178] A "user terminal" is a device (e.g., a personal computer, a smartphone) that a user uses to access information and services.

[1179] "Means of recognizing emotions" refers to technologies and methods for identifying emotions from user input.

[1180] "Methods for receiving feedback and using it to improve the model" refers to methods of receiving evaluations and opinions from users, updating the model based on them, and improving its performance.

[1181] A "customized natural language processing model" is a natural language processing model that has been tailored to specific needs or applications.

[1182] An "emotion recognition engine" is a specialized system or algorithm used to evaluate and analyze a user's emotions.

[1183] This invention is a system that specializes in the specialized terminology and processes used in each department of a company, and further combines this with an emotion engine that recognizes user emotions, thereby providing more appropriate information and improving the user experience. This system is implemented through the following series of processes.

[1184] Data collection phase

[1185] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This data collection is performed using API-based integration. The collected information is then cleansed by the server to ensure consistency and accuracy before being stored in the database. Data cleansing tools such as OpenRefine are used for this data cleansing process.

[1186] Data processing and learning phases

[1187] Next, the server prepares a natural language processing (NLP) model and trains it using data from the database containing specialized terms and the context in which they are used. NLP models used include natural language processing libraries such as Spacy and Hugging Face. This training enables the model to understand the context in which specific terms are used and to process them appropriately. After training, the model's performance is evaluated, and parameters are adjusted as needed.

[1188] Inquiry processing phase

[1189] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal forwards this question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[1190] Database search and response generation phase

[1191] The server searches the database based on the analysis results and collects the necessary information. It then generates an appropriate response. For example, it might provide a specific response such as, "The latest lead generation result shows that 100 leads were acquired last week."

[1192] Emotion recognition phase

[1193] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adjust the tone and level of detail of the response it provides.

[1194] Result display phase

[1195] The server sends the generated response to the device, and the device displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[1196] Feedback processing phase

[1197] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[1198] Model and emotion engine improvement phase

[1199] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. If there is feedback regarding inappropriate responses, the analysis results are used to retrain the model. The feedback is also utilized as data to improve the accuracy of emotion recognition.

[1200] Examples of specific cases and prompts for generative AI models.

[1201] As a concrete example, in response to a question about the latest lead generation results in the marketing department, the system provides the answer, "The latest lead generation results show that we acquired 100 leads last week." If the emotion engine detects that the user is in a hurry, it will quickly display a concise response such as, "Here are the latest lead generation results. We acquired 100 leads last week."

[1202] Examples of prompts for a generative AI model:

[1203] "Could you please tell me the latest lead generation results for your marketing department?"

[1204] "Could you please tell me the latest project progress in the XX department?"

[1205] "Please use the emotion engine to evaluate user satisfaction."

[1206] This system aims to improve operational efficiency and user experience by integrating the interpretation of technical terms with the recognition of user emotions.

[1207] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1208] Step 1:

[1209] The server collects a list of technical terms and related information from each department of the company. The input requires access information to each department's data source. Specifically, the server uses an API to send HTTP requests to each department's data source to retrieve the technical term list and related information. The output is the collected raw data.

[1210] Step 2:

[1211] The server cleanses the collected data. The raw data obtained in step 1 is used as input. Specifically, a data cleansing tool such as OpenRefine is used to remove duplicate data, fill in incomplete data, and ensure consistency. The output is the cleansed data.

[1212] Step 3:

[1213] The server stores the cleansed data in the database. The cleansed data generated in step 2 is required as input. Specifically, a database management system such as MySQL is used, and the data is saved to the database using SQL queries. The output is structured data stored in the database.

[1214] Step 4:

[1215] The server prepares a natural language processing (NLP) model. The input requires an existing trained model and a list of specialized terminology for each field. Specifically, it imports and initializes natural language processing libraries such as Spacy or Hugging Face. The output is the NLP model prepared for training.

[1216] Step 5:

[1217] The server trains an NLP model using the specialized terminology and data used in its context within the database. The inputs required are the data stored in step 3 and the NLP model prepared in step 4. Specifically, the model is fed data, hyperparameters such as the number of epochs and learning rate are set, and training is performed. The output is the trained NLP model.

[1218] Step 6:

[1219] The server evaluates the trained model and adjusts the parameters as needed. The input requires a trained NLP model and an evaluation dataset. Specifically, it measures the model's accuracy and error and readjusts the hyperparameters. The output is an optimized NLP model.

[1220] Step 7:

[1221] The user enters a question into the system using a terminal. Input requires the user to enter a text box, such as "What are the latest lead generation results?". The output is the user's question stored in text format on the terminal.

[1222] Step 8:

[1223] The terminal forwards the user's question to the server. The input requires the user's question. Specifically, it sends an HTTP request to the server and transfers the data. The output is the server receiving the question.

[1224] Step 9:

[1225] The server uses an NLP model to analyze user questions. The input requires the user's question, transmitted from the terminal, and a trained NLP model. Specifically, the model analyzes the question and identifies relevant terminology and context. The output is the analysis result.

[1226] Step 10:

[1227] The server searches the database based on the analysis results. The analysis results and the database are required as input. Specifically, it generates an SQL query, sends it to the database, and retrieves the relevant data. The output will provide the necessary information.

[1228] Step 11:

[1229] The server generates appropriate answers from the collected information. It requires information retrieved from a database as input. Specifically, it integrates the information to create a user-friendly answer. The output is the generated answer.

[1230] Step 12:

[1231] The server uses an emotion engine to recognize emotions from user input. The input requires both the user's question and the emotion engine. Specifically, it performs text analysis to evaluate emotions based on tone and context. The output is the recognized emotion.

[1232] Step 13:

[1233] The server sends the generated response to the terminal, and the terminal displays the received response to the user. The required inputs are the generated response and the sentiment analysis results. Specifically, the response is sent to the terminal via an HTTP response, which the terminal then displays. The output is the response that the user can view.

[1234] Step 14:

[1235] The user provides feedback on the provided answer. The input requires the answer displayed by the system, and the user inputs feedback such as "This answer was helpful" or "This answer was inappropriate." The output is the feedback stored on the device.

[1236] Step 15:

[1237] The terminal forwards user feedback to the server. The input requires user feedback. Specifically, it sends the feedback to the server using an HTTP request. The output is the server receiving the feedback.

[1238] Step 16:

[1239] The server analyzes the received feedback and improves the NLP model and emotion engine. Input requires user feedback and training data. Specifically, it retrains the model based on the feedback to improve performance. The output is an improved NLP model and emotion engine.

[1240] (Application Example 2)

[1241] 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."

[1242] In traditional brick-and-mortar stores, communication between employees and customers often lacked the speed and accuracy of providing specialized information, and it was difficult to respond appropriately to customers' emotions. This resulted in decreased customer satisfaction and reduced operational efficiency. This invention aims to solve these problems and enable the provision of higher-quality service in customer interactions at physical stores.

[1243] The specific processing performed by the specific 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 a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating an appropriate response, means for transmitting and displaying the generated response on a user terminal, means for recognizing the user's emotions and adjusting the response tone and level of detail based on the recognition results, means for receiving feedback and using it to improve the model, and a natural language processing model customized based on the technical terms and related information of each department. This enables employees in physical stores to provide information quickly and accurately, as well as to respond appropriately to customer emotions.

[1244] A "glossary" is a list that compiles specific technical terms and expertise used within a company or department.

[1245] "Related information" refers to all information related to technical terms, including definitions of technical terms, examples of usage, process documents, manuals, and project reports.

[1246] "Means of collection" refers to the methods and devices for gathering a list of technical terms and related information.

[1247] "Cleansing" refers to the process of organizing and correcting collected data to ensure consistency and accuracy.

[1248] A "database" is a collection of data that is efficiently stored and managed in a way that makes it searchable and usable.

[1249] A "natural language processing model" is a model used to train algorithms and methods for analyzing and understanding human language using computers.

[1250] "Means of learning" refers to methods and devices for training natural language processing models to learn specialized terminology and its context.

[1251] "Means of analysis" refers to methods and devices for analyzing user inquiries and understanding their content.

[1252] "Searching means" refers to methods and devices for finding relevant information within a database based on analysis results.

[1253] "Means for generating answers" refers to methods and devices for creating appropriate answers for users based on retrieved information.

[1254] "User terminal" refers to devices or systems that a user directly operates, such as smartphones or computers.

[1255] "Means for recognizing emotions and adjusting response tone and level of detail based on the recognition results" refers to methods and apparatus for analyzing a user's emotions and adapting the tone and level of detail of the response accordingly.

[1256] "Means for receiving feedback and using it to improve the model" refers to methods and devices for receiving evaluations and opinions from users and incorporating them into improvements to natural language processing models and sentiment engines.

[1257] A "customized natural language processing model" refers to a natural language processing model that has been specifically tailored to a particular company or department.

[1258] This invention relates to an information provision system for improving the quality of customer service in physical stores. Specific embodiments thereof are described below.

[1259] 1. Data collection and cleansing

[1260] First, the server collects a list of technical terms and related information (e.g., product features, process descriptions, manuals, and usage examples) from various departments within the company. The collected data is then cleansed by the server and stored in a database, ensuring consistency and accuracy. This database is then used as a source of information for subsequent queries.

[1261] 2. Training a natural language processing model

[1262] The server prepares a natural language processing (NLP) model and trains it based on specialized terminology and its contextual information stored in a database. Through this training, the model learns to understand the context in which specific terms are used and generates appropriate responses accordingly.

[1263] 3. Inquiry analysis and response generation

[1264] The user (in this case, a store employee) enters an inquiry into the system using a device such as a smartphone. For example, the question might be, "Please tell me about the features of this new product." The device transmits the entered question to the server, which analyzes the question using an NLP model. Based on the analysis results, it searches the database and generates an appropriate answer. In the process, it also analyzes the customer's emotions and adjusts the response tone and level of detail accordingly.

[1265] 4. Emotion Recognition and Response Adjustment

[1266] The server uses an emotion engine to recognize emotions from user input. For example, it analyzes the tone and context of the entered question and adopts a comforting response tone if the user is dissatisfied. This allows users to receive more appropriate and satisfying service.

[1267] 5. Feedback and Model Improvement

[1268] The system also includes a feature that allows users to provide feedback on the answers they receive. For example, users can input feedback such as "This answer was helpful" or "This answer was inappropriate." The user's device forwards this feedback to the server, which then uses it to improve both the NLP model and the emotion engine. This enables continuous model improvement based on feedback.

[1269] 6. Usage example

[1270] For example, if a customer asks, "What are the features of the new product?", the server will provide information such as, "This is a new leather bag made from high-quality cowhide." Also, if a customer asks in a dissatisfied tone, "Is it possible to return this item?", the server will respond with something like, "Returns are accepted within 30 days of purchase. Please contact us if you have any problems; we will be happy to assist you."

[1271] Example of a prompt

[1272] "Could you please tell me the specifications of this product?"

[1273] "Is this item returnable?"

[1274] In this way, the present invention facilitates communication between employees and customers in physical stores, enabling the rapid and accurate provision of specialized information and appropriate responses that respond to customer emotions. This is expected to improve customer satisfaction and operational efficiency.

[1275] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1276] Step 1:

[1277] The server collects terminology lists and related information from each department. Specifically, the server uses APIs and file transfer protocols to retrieve data from each department's databases and document management systems. The input here is the terminology lists and related information provided by each department, and the output is the collected data.

[1278] Step 2:

[1279] The server cleanses the collected data, removing unnecessary and incorrect data to maintain consistency and accuracy. Specifically, it performs data cleaning operations using libraries such as Python's pandas library. The input here is the data collected in the previous step, and the output is the cleansed data.

[1280] Step 3:

[1281] The server stores the cleansed data in a database. Specifically, it inserts the data into an SQL database or a NoSQL database. Here, the input is the cleansed data, and the output is the data stored in the database.

[1282] Step 4:

[1283] The server trains a natural language processing (NLP) model. Specifically, it trains an NLP model (e.g., BERT or GPT) using technical terms and their contexts from a database. The input here is the technical terms and contextual information from the database, and the output is the trained NLP model.

[1284] Step 5:

[1285] The user (employee) enters a question using a terminal. For example, they might enter, "Please tell me the features of this new product." Here, the input is the user's question, and the output is the user's input data sent to the terminal.

[1286] Step 6:

[1287] The terminal forwards user input to the server. Specifically, it sends questions to the server using HTTP requests. Here, input refers to the user's input data, and output refers to the data forwarded to the server.

[1288] Step 7:

[1289] The server uses an NLP model to analyze the user's question. Specifically, it understands the context of the question and identifies relevant technical terms. The input here is the user's question, and the output is the analysis result.

[1290] Step 8:

[1291] The server searches the database based on the analysis results and collects the appropriate answers. The input here is the analysis results from the NLP model, and the output is the information collected from the database.

[1292] Step 9:

[1293] The server uses an emotion engine to recognize the user's emotions. Specifically, it analyzes the tone and context of the input question to identify the emotions the user is experiencing. Here, the input is the user's question, and the output is the emotion recognition result.

[1294] Step 10:

[1295] The server adjusts the tone and level of detail of its response based on the recognized emotion. For example, if the user is dissatisfied, it adopts a comforting tone. The input here is the emotion recognition result and collected information, and the output is the adjusted response.

[1296] Step 11:

[1297] The server sends the adjusted response to the terminal. Specifically, it sends the response to the terminal using an HTTP response. Here, the input is the adjusted response, and the output is the response sent to the terminal.

[1298] Step 12:

[1299] The terminal displays the received response to the user. The input here is the response sent to the terminal, and the output is the response displayed to the user.

[1300] Step 13:

[1301] Users provide feedback on the provided answers. For example, they might rate the answer as "This answer was helpful" or "This answer was inappropriate." The input here is the user's feedback, and the output is the feedback entered into the device.

[1302] Step 14:

[1303] The terminal forwards user feedback to the server. Specifically, it sends feedback to the server using an HTTP request. Here, the input is the feedback entered into the terminal, and the output is the feedback forwarded to the server.

[1304] Step 15:

[1305] The server improves the NLP model and emotion engine based on the feedback it receives. Specifically, it analyzes the content of the feedback and retrains the model as needed. The input here is the user's feedback, and the output is the improved NLP model and emotion engine.

[1306] These steps enable employees in physical stores to provide customers with quick and accurate information, as well as to respond appropriately to customers' emotions.

[1307] 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.

[1308] The data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of the data generation model 58 is ChatGPT (Internet Search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">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.

[1309] 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.

[1310] [Fourth Embodiment]

[1311] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[1312] 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.

[1313] 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).

[1314] 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.

[1315] 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.

[1316] 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).

[1317] 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.

[1318] 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.

[1319] 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.

[1320] 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.

[1321] 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.

[1322] 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.

[1323] 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".

[1324] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. This system is implemented through a series of programmatic processes described below.

[1325] Data collection phase

[1326] First, the server collects a list of specialized terms and related information provided by each department of the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. The collected data is cleansed by the server to ensure consistency and accuracy. Specifically, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in the database.

[1327] Data processing and learning phases

[1328] Next, the server prepares a natural language processing (NLP) model and trains it based on the collected data. The server prepares the training data using specialized terminology and data used in its context within the database, and then trains the NLP model. For example, it teaches the model that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[1329] Inquiry processing phase

[1330] The user enters a question into the system using a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this entered question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[1331] Database search and response generation phase

[1332] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[1333] Result display phase

[1334] The generated response is sent from the server to the terminal. The terminal then displays the received response to the user.

[1335] Feedback processing phase

[1336] Users can provide feedback on the displayed answers. For example, they can input feedback such as "This answer was helpful" or "This answer was inappropriate." The device forwards this feedback to the server, which uses the received feedback to improve the NLP model. Inappropriate answers are used to retrain the model.

[1337] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, a feedback function continuously improves the accuracy and usefulness of the system. By using NLP models customized for each department, it achieves highly accurate information provision tailored to the specific needs of each department.

[1338] The following describes the processing flow.

[1339] Data collection phase

[1340] Step 1:

[1341] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[1342] Step 2:

[1343] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[1344] Step 3:

[1345] The server stores the cleansed data in the database.

[1346] Data processing and learning phases

[1347] Step 1:

[1348] The server prepares a natural language processing (NLP) model.

[1349] Step 2:

[1350] The server prepares the training data using specialized terminology and data used in the context of the database.

[1351] Step 3:

[1352] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[1353] Step 4:

[1354] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[1355] Inquiry processing phase

[1356] Step 1:

[1357] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[1358] Step 2:

[1359] The terminal forwards the entered question to the server.

[1360] Step 3:

[1361] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[1362] Database search and response generation phase

[1363] Step 1:

[1364] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[1365] Step 2:

[1366] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[1367] Step 3:

[1368] The server transfers the generated response to the terminal.

[1369] Result display phase

[1370] Step 1:

[1371] The terminal receives the response sent from the server.

[1372] Step 2:

[1373] The device displays the received response to the user.

[1374] Feedback processing phase

[1375] Step 1:

[1376] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[1377] Step 2:

[1378] The device forwards user feedback to the server.

[1379] Step 3:

[1380] The server analyzes the received feedback and uses it to improve the NLP model. For example, if there is feedback indicating an inappropriate response, the model is retrained based on that information.

[1381] Step 4:

[1382] The server adjusts the parameters of the NLP model based on the feedback, providing more accurate and useful answers to subsequent queries.

[1383] Through these steps, the system effectively manages the specialized terminology of each department and enables the rapid and accurate provision of information to users.

[1384] (Example 1)

[1385] 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".

[1386] Traditional enterprise systems faced the challenge of requiring significant time and effort to adapt to the different jargon and processes of each department. This hindered operational efficiency and could lead to decreased productivity. Furthermore, inaccurate responses to user inquiries prevented proper feedback from being obtained, hindering system improvement. In addition, insufficient performance of natural language processing models could lead to errors in analyzing complex jargon and context.

[1387] 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.

[1388] In this invention, the server includes means for collecting a list of technical terms and related information from each department, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for sending and displaying the generated answers on a user terminal, means for receiving user feedback and using it to improve the model, and means for providing a natural language processing model customized based on the technical terms and related information of each department. This enables the rapid and accurate provision of information that meets the specific needs of each department, resulting in increased operational efficiency and continuous system improvement.

[1389] "Each department of a company"

[1390] This refers to organizational units within a company that are responsible for different tasks or functions. Examples include the marketing department and the human resources department.

[1391] "List of Technical Terms"

[1392] This refers to a list of terms and phrases commonly used in a particular industry or field.

[1393] Related Information

[1394] This refers to supplementary data provided alongside technical terms, such as definitions of terms, usage examples, process documents, manuals, and project reports.

[1395] "cleansing"

[1396] This refers to operations performed on collected data, such as removing duplicates, correcting incomplete entries, and standardizing the format.

[1397] "Database"

[1398] This refers to a system for organizing and efficiently managing collected data.

[1399] "Natural language processing models"

[1400] This refers to artificial intelligence models designed to understand, analyze, and generate human language. Machine learning techniques are used in this process.

[1401] "Means for receiving and analyzing inquiries"

[1402] This refers to the process of receiving questions submitted by users, analyzing their content, and extracting relevant information.

[1403] "A means of searching a database and generating appropriate answers."

[1404] This refers to a function that searches a database based on the analyzed results and creates an answer that is appropriate for the user's question.

[1405] "User terminal"

[1406] This refers to a device that a user uses to access a system and input or retrieve information.

[1407] "feedback"

[1408] This refers to user-provided ratings and feedback, which are used to improve the system.

[1409] "Generative AI Model"

[1410] This refers to an artificial intelligence model trained for natural language processing.

[1411] "Prompt message"

[1412] This refers to the questions or instructions that users input into the system.

[1413] "Customized natural language processing models"

[1414] This refers to artificial intelligence models that have been specifically tailored to a particular department or task.

[1415] "Parameter tuning"

[1416] This refers to the process of adjusting a model to improve its performance. Examples include setting appropriate hyperparameters and adjusting the learning rate.

[1417] This invention provides a system tailored to the specialized terminology and processes used in each department of a company, designed to enable personnel to quickly adapt to their work. The system utilizes servers and terminals as hardware, and employs NLP (Natural Language Processing) technology and database management software as software. Typical software used includes Python and its libraries (e.g., spaCy, NLTK, TensorFlow, PyTorch).

[1418] The server collects terminology lists and related information provided by various departments within the company. This information includes definitions of terms, usage examples, process documents, manuals, and project reports. To cleanse this collected data, operations such as removing duplicate data, correcting incomplete entries, and standardizing the format are performed. The cleansed data is then stored in a database.

[1419] Next, the server prepares and trains an NLP model based on the collected data. The training data is prepared using specialized terminology and the context in which it is used within the database. For example, the model is trained to understand that the term "lead" means "potential customer." During this process, the model's performance is evaluated, and parameters are tuned as needed.

[1420] The user enters a question into the system using their own device. For example, they might enter, "What are the latest lead generation results?" This device forwards the user's question to the server. The server analyzes the received question using an NLP model to identify relevant technical terms and contextual information.

[1421] The server searches the database based on the analysis results and generates an appropriate response. For example, it might generate a specific response such as, "The latest lead generation result is that you acquired 100 leads last week." The generated response is sent from the server to the terminal, which then displays it to the user.

[1422] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." This feedback is sent from the device to the server. The server uses the received feedback to improve the NLP model. Inappropriate answers are used as new training data for retraining.

[1423] The following are specific examples of prompts to enter into the system:

[1424] 1. "Please tell me your latest lead generation results."

[1425] 2. "What are the new trends in SEO?"

[1426] 3. "Please tell me about tools to increase conversion rates."

[1427] Thus, the system of the present invention collects, cleanses, and stores a list of technical terms and related information in a database, providing appropriate answers to user inquiries. Furthermore, the system's accuracy and usefulness can be continuously improved through a feedback function. By using NLP models customized for each department, it achieves the provision of highly accurate information that meets the specific needs of each department.

[1428] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1429] Step 1: Data Collection

[1430] The server collects a list of specialized terms and related information from each department within the company. Specifically, it gathers data based on Excel files and documents entered or uploaded by departmental staff. The collected data includes definitions of terms, usage examples, process documents, manuals, and project reports.

[1431] Input: A list of specialized terms and related information provided by the department head.

[1432] Output: Collected raw data

[1433] Step 2: Data Cleansing

[1434] The server cleanses the collected data. This includes removing duplicate data, correcting incomplete entries (such as filling in blanks or correcting typos), and standardizing the format (for example, consistently converting date formats to YYYY-MM-DD).

[1435] Input: Collected raw data

[1436] Output: Cleansed data

[1437] Step 3: Database Storage

[1438] The server stores the cleansed data in the database. This ensures that data with guaranteed consistency and accuracy can be managed.

[1439] Input: Cleansed data

[1440] Output: Data stored in the database

[1441] Step 4: Training the NLP Model

[1442] The server trains an NLP model using the collected and cleansed data. It uses Python natural language processing libraries (e.g., spaCy, NLTK) or machine learning libraries (e.g., TensorFlow, PyTorch). The data is split into training and test data and fed to the NLP model. For example, the model is trained to understand that "lead" means "potential customer". After training, the model's performance is evaluated, and hyperparameters are tuned as needed.

[1443] Input: Cleansed data

[1444] Output: Trained NLP model

[1445] Step 5: Inquiry reception and analysis

[1446] The user enters a question into the system via a terminal. For example, they might enter, "What are the latest lead generation results?" The terminal forwards this question to the server. The server analyzes the received question using an NLP model to identify keywords ("lead generation," "results") and context.

[1447] Input: User question (prompt)

[1448] Output: Results of question analysis (keywords and context)

[1449] Step 6: Database search and answer generation

[1450] The server searches the database based on the analysis results. For example, it searches for the latest project reports and statistics on lead generation and generates specific answers such as, "The latest lead generation results show that 100 leads were acquired last week." The answers are formatted using automated templates.

[1451] Input: Results of the analyzed question

[1452] Output: Generated answer

[1453] Step 7: Display Results

[1454] The generated response is sent from the server to the terminal. The terminal displays this response to the user through a GUI (Graphical User Interface). For example, it might be displayed on the dashboard of a web application.

[1455] Input: Response sent from the server

[1456] Output: Answer displayed on the user's terminal

[1457] Step 8: Feedback

[1458] Users can provide feedback on the displayed answers. For example, they can enter feedback such as "This answer was helpful" or "This answer was inaccurate." The device sends this feedback to the server. The server uses the received feedback to improve the performance of the NLP model. Inappropriate answers are used as new training data for retraining.

[1459] Input: User feedback

[1460] Output: Improved NLP model

[1461] (Application Example 1)

[1462] 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".

[1463] In conventional systems, factory robots had difficulty quickly understanding specialized terminology and processes when performing new tasks. Furthermore, because different departments had different specialized terminology and processes, there was a lack of means to accurately acquire that information and efficiently perform tasks. In addition, there was a need for robots to acquire necessary information in real time while in operation and to improve their movements based on appropriate feedback.

[1464] 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.

[1465] In this invention, the server includes means for collecting a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their contexts using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating appropriate answers, means for transmitting and displaying the generated answers on a user terminal, means for querying technical terms and processes while a factory robot is performing a new task and obtaining necessary information, and means for displaying the results via the robot's display or audio output. This enables the factory robot to quickly acquire information on specific technical terms and processes and perform tasks efficiently. Furthermore, the model can be continuously improved based on the robot's operation results, thereby improving the accuracy and efficiency of its operation.

[1466] A "glossary of technical terms" is a list that compiles terms used in a specific field or industry, their definitions, and related information.

[1467] "Related information" refers to materials such as usage examples related to technical terms, process documents, manuals, and project reports.

[1468] "Data cleansing" is an operation that involves removing duplicate data, correcting incomplete entries, and standardizing the format in order to ensure the consistency and accuracy of collected data.

[1469] A "database" is a collection of information that stores a list of specialized terms and related information, making it searchable and retrievable.

[1470] A "natural language processing model" is a machine learning model used to understand, analyze, and generate human language.

[1471] "Analysis" is the process of understanding user inquiries using natural language processing models and grasping their intent.

[1472] An "inquiry" refers to a question or request for information that a user makes to a system.

[1473] A "user terminal" is a device used by a user to access a system, input information, and receive results.

[1474] A "factory robot" is a mechanical device used in industrial settings to automatically perform specific tasks.

[1475] A "display" is a monitor or screen used to visually display information.

[1476] "Voice output" is a function that conveys information and instructions from the system to the user as voice.

[1477] "Feedback" refers to evaluations and comments provided by users regarding the system's responses and actions.

[1478] "Model improvement" is the process of improving the performance of a natural language processing model based on the feedback received.

[1479] This invention relates to a system that helps factory robots quickly understand and efficiently perform new tasks. The system is realized through the following processes. The main components of the system consist of a server, a robot as an edge device, and a user (human operator).

[1480] Data collection phase

[1481] The server collects a glossary of terms and related information from each department within the factory. The glossary includes related information such as process documents, manuals, and project reports. Apache NiFi can be used for this collection. The collected data is cleansed by the server. The Pandas library is used to remove duplicate data, correct incomplete entries, and standardize the format. The cleansed data is then stored in a database (e.g., PostgreSQL).

[1482] Data processing and learning phases

[1483] The server prepares a natural language processing (NLP) model and trains it based on the collected data. It uses the Hugging Face's Transformers library to select an appropriate generative AI model (e.g., BERT) and then trains it. The following is an example of a prompt:

[1484] Example of a prompt:

[1485] Question: What are the latest lead generation results?

[1486] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[1487] Inquiry processing phase

[1488] When a factory robot performs a new task, it queries the server for technical terms and processes as needed. The robot receives instructions and questions from the user (operator) via its built-in display or voice output. These queries are transmitted to the server through a web framework such as Flask.

[1489] Database search and response generation phase

[1490] The server analyzes the received query using an NLP model, retrieves appropriate information from the database, and generates a response. The generated response is sent to the robot and communicated to the operator via display or voice output. gTTS (Google Text-to-Speech) can be used for voice output.

[1491] Feedback processing phase

[1492] The robot sends feedback to the server based on its performance. This feedback includes user ratings and comments. The server uses the received feedback to continuously improve the performance of the NLP model. TensorFlow or PyTorch can be used to retrain the model.

[1493] Specific example

[1494] For example, consider a scenario where a factory robot asks for the "latest lead generation result." In this case, the server performs the analysis using the following prompt:

[1495] Example of a prompt:

[1496] Question: What are the latest lead generation results?

[1497] Text: According to the latest report from our lead generation department, we acquired 100 leads last week. The quality of the leads is very high, and our conversion rate is also improving.

[1498] In this way, it becomes possible to acquire information immediately and perform tasks efficiently, which is expected to improve the overall system performance.

[1499] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1500] Step 1: Data Collection Phase

[1501] The server collects terminology lists and related information from each department within the factory. Inputs include terminology lists, process documents, manuals, and project reports provided by each department. The server uses Apache NiFi to retrieve and relay this data. Outputs are stored on the server as raw data for cleansing.

[1502] Step 2: Data Cleansing

[1503] The server cleanses the collected data. Specifically, it uses the Pandas library to remove duplicate data, correct incomplete entries, and standardize the format. The input is raw data, and the output is cleaned data. This data is stored in a database.

[1504] Step 3: Training the natural language processing model

[1505] The server prepares a natural language processing model and trains it on cleansed data. It uses the Hugging Face's Transformers library to train a generative AI model (e.g., BERT). The input is a categorized list of terms and their usage contexts, and the trained output is an NLP model with high suitability for specific technical terms and processes.

[1506] Step 4: Receiving User Inquiries

[1507] When a factory robot performs a new task, it queries a server for technical terms and processes. The terminal (robot) receives user questions through its built-in display or voice input and transmits them to the server using the Flask framework. The input is the query content, and the output is the query request to the server.

[1508] Step 5: Analyze the inquiry

[1509] The server analyzes incoming queries using an NLP model. The input data is the user's question, and the output, as a result of the analysis, is information related to the appropriate technical terms and processes. The server then converts this into a query to be sent to the database.

[1510] Step 6: Database search and answer generation

[1511] The server uses the generated query to search the database and produce an appropriate response. The input is the parsed query, and the output is the generated response. For example, in the case of a query about the latest lead generation results, a response such as "According to the latest report from the lead generation department, 100 leads were acquired last week" would be generated.

[1512] Step 7: Submit and view results

[1513] The server sends the generated response to the terminal (robot). The terminal displays the result on its screen or communicates it to the user via voice output using gTTS. The input is the generated response text, and the output is the displayed information or voice.

[1514] Step 8: Receiving Feedback

[1515] The user provides feedback on the robot's performance. The terminal collects the user's feedback and transmits it to the server. The input is the user's evaluation and comments, and the output is the feedback data sent to the server.

[1516] Step 9: Model Improvement

[1517] The server uses the received feedback to improve the performance of the NLP model. The input is the feedback data, and the output is the updated and improved natural language processing model. TensorFlow or PyTorch is used for retraining.

[1518] This allows the system to continuously learn and improve, supporting the efficient task execution of factory robots.

[1519] 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.

[1520] This invention not only provides a system tailored to the specialized terminology and processes used in various departments of a company, but also aims to provide more appropriate information and improve the user experience by combining it with an emotion engine that recognizes user emotions. This system consists of a process of collecting, cleansing, and storing a list of specialized terms and related information in a database, and learning using a natural language processing (NLP) model. It also receives user inquiries, analyzes them, generates and displays answers, and collects and analyzes feedback to improve the model. Here, we will describe a specific embodiment that combines the system with an emotion engine that recognizes user emotions.

[1521] Data collection phase

[1522] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This information is then cleansed by the server to ensure consistency and accuracy before being stored in the database.

[1523] Data processing and learning phases

[1524] Next, the server prepares an NLP model and trains it using data from the database containing specialized terminology and the context in which it is used. As a result of the training, the model will be able to understand the context in which specific terms are used and process them appropriately. The performance of the trained model is also evaluated, and parameters are adjusted as needed.

[1525] Inquiry processing phase

[1526] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal transmits the entered question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[1527] Database search and response generation phase

[1528] The server searches the database based on the analysis results and collects the necessary information. From the collected information, it generates an appropriate response. For example, it might create a specific response such as, "The latest lead generation result is that 100 leads were acquired last week."

[1529] Emotion recognition phase

[1530] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adapt the tone and level of detail of the response it provides.

[1531] Result display phase

[1532] The server sends the generated response to the terminal. The terminal displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[1533] Feedback processing phase

[1534] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[1535] Model and emotion engine improvement phase

[1536] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. Specifically, if there is inappropriate feedback, it analyzes its content and uses it to retrain the model. It is also used as data to improve the accuracy of emotion recognition.

[1537] In this embodiment, the system of the present invention not only provides information on technical terms and processes quickly and accurately, but also recognizes the user's emotions and provides appropriate information based on those emotions. This makes it possible to improve the user's work efficiency and enhance the quality of the user experience.

[1538] The following describes the processing flow.

[1539] Data collection phase

[1540] Step 1:

[1541] The server collects lists of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company.

[1542] Step 2:

[1543] The server cleanses the collected data. Specifically, it removes duplicate data, corrects incomplete entries, and standardizes the format.

[1544] Step 3:

[1545] The server stores the cleansed data in the database.

[1546] Data processing and learning phases

[1547] Step 1:

[1548] The server prepares a natural language processing (NLP) model.

[1549] Step 2:

[1550] The server prepares the training data using specialized terminology and data used in the context of the database.

[1551] Step 3:

[1552] The server uses the prepared data to train an NLP model. For example, it learns that the term "lead" means "potential customer."

[1553] Step 4:

[1554] The server evaluates the performance of the trained model and tunes the parameters as needed based on metrics such as accuracy, recall, and F1 score.

[1555] Inquiry processing phase

[1556] Step 1:

[1557] The user uses a terminal to enter a question into the system. For example, they might enter, "What are the latest lead generation results?"

[1558] Step 2:

[1559] The terminal forwards the entered question to the server.

[1560] Step 3:

[1561] The server analyzes the received question using an NLP model to identify relevant technical terms and their contextual information.

[1562] Database search and response generation phase

[1563] Step 1:

[1564] The server searches the database based on the analysis results. It identifies and collects the necessary information.

[1565] Step 2:

[1566] The server generates appropriate responses for the user in an easy-to-understand format based on the search results. For example, it might prepare a specific response such as, "Your latest lead generation results show that you acquired 100 leads last week."

[1567] Step 3:

[1568] The server transfers the generated response to the terminal.

[1569] Emotion recognition phase

[1570] Step 1:

[1571] The server passes the user's input to the sentiment engine, which then recognizes the emotions associated with it. For example, if the entered question is "Please tell me about this setting," the sentiment engine analyzes whether the sentence contains doubt or anxiety.

[1572] Step 2:

[1573] The server adjusts the tone of its responses based on the results of the emotion engine. For example, if the user is feeling anxious, the responses will become more helpful and detailed.

[1574] Result display phase

[1575] Step 1:

[1576] The terminal receives the response sent from the server.

[1577] Step 2:

[1578] The device displays the received responses to the user. The displayed responses are presented in an appropriate tone and level of detail based on the user's emotions as recognized by the emotion engine.

[1579] Feedback processing phase

[1580] Step 1:

[1581] Users provide feedback on the answers they receive. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate."

[1582] Step 2:

[1583] The device forwards user feedback to the server.

[1584] Model and emotion engine improvement phase

[1585] Step 1:

[1586] The server analyzes the received feedback and improves both the NLP model and the emotion engine. For example, if there is feedback indicating an inappropriate response, it analyzes the content and uses it to retrain the model.

[1587] Step 2:

[1588] The server adjusts the parameters of its NLP model and emotion engine based on the feedback, providing more accurate and useful answers to subsequent inquiries.

[1589] (Example 2)

[1590] 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".

[1591] The specialized terminology and processes used in various departments of a company are diverse, requiring advanced knowledge and rapid response to understand them and provide appropriate information. However, conventional systems have not adequately understood the context of specialized terminology or recognized user emotions, making it difficult to provide appropriate information and improve the user experience. Therefore, there is a need to develop systems that can provide information in a way that suits user needs.

[1592] The specific processing performed by the specific 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 a list of technical terms and related information from each department of a company, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing inquiries from users, means for searching the database based on the analysis results and generating an appropriate answer, means for sending and displaying the generated answer on a user terminal, and means for recognizing emotions from the content of the user's inquiry. This not only enables the provision of information specific to the technical terms and processes of each department, but also allows for adjustment of the tone and level of detail of the answer according to the user's emotions, which is expected to improve the user experience.

[1593] "Each department of a company" refers to multiple departments within a company, each with different tasks and roles.

[1594] A "list of technical terms" is a list that compiles specific terms and expressions used in a particular industry or field.

[1595] "Related information" refers to information such as definitions, usage examples, process documents, manuals, and project reports associated with technical terms.

[1596] A "server" is a computer system that provides information and services over a network.

[1597] "Means of collection" refers to the methods and techniques used to consolidate dispersed data into a single entity.

[1598] "Data cleansing methods" refer to techniques for removing duplicates, errors, and incomplete information from collected data, making it accurate and consistent.

[1599] "Storing in a database" means saving processed data in a structured format within a database.

[1600] A "natural language processing model" is a type of machine learning model used to understand and process human language.

[1601] "Methods for learning technical terms and their contexts" refers to methods of training models to understand the situations and contexts in which specific terms are used.

[1602] "Means for receiving and analyzing inquiries" refers to methods and technologies for receiving questions and requests from users and analyzing their content.

[1603] "Methods for searching a database based on analysis results" refers to methods for obtaining relevant data from a database based on the analyzed information.

[1604] "Means of generating appropriate answers" refers to methods of creating answers that are useful to the user based on search results.

[1605] A "user terminal" is a device (e.g., a personal computer, a smartphone) that a user uses to access information and services.

[1606] "Means of recognizing emotions" refers to technologies and methods for identifying emotions from user input.

[1607] "Methods for receiving feedback and using it to improve the model" refers to methods of receiving evaluations and opinions from users, updating the model based on them, and improving its performance.

[1608] A "customized natural language processing model" is a natural language processing model that has been tailored to specific needs or applications.

[1609] An "emotion recognition engine" is a specialized system or algorithm used to evaluate and analyze a user's emotions.

[1610] This invention is a system that specializes in the specialized terminology and processes used in each department of a company, and further combines this with an emotion engine that recognizes user emotions, thereby providing more appropriate information and improving the user experience. This system is implemented through the following series of processes.

[1611] Data collection phase

[1612] First, the server collects a list of specialized terms and related information (definitions, usage examples, process documents, manuals, project reports, etc.) provided by various departments within the company. This data collection is performed using API-based integration. The collected information is then cleansed by the server to ensure consistency and accuracy before being stored in the database. Data cleansing tools such as OpenRefine are used for this data cleansing process.

[1613] Data processing and learning phases

[1614] Next, the server prepares a natural language processing (NLP) model and trains it using data from the database containing specialized terms and the context in which they are used. NLP models used include natural language processing libraries such as Spacy and Hugging Face. This training enables the model to understand the context in which specific terms are used and to process them appropriately. After training, the model's performance is evaluated, and parameters are adjusted as needed.

[1615] Inquiry processing phase

[1616] The user uses a terminal to input a question into the system. For example, they might input, "What are the latest lead generation results?" The terminal forwards this question to the server, which uses an NLP model to analyze the question and identify relevant technical terms and their context.

[1617] Database search and response generation phase

[1618] The server searches the database based on the analysis results and collects the necessary information. It then generates an appropriate response. For example, it might provide a specific response such as, "The latest lead generation result shows that 100 leads were acquired last week."

[1619] Emotion recognition phase

[1620] Here, the server uses an emotion engine to recognize emotions from the user's input. For example, it analyzes the tone and context of the entered text to determine what emotions the user is feeling. This allows the system to adjust the tone and level of detail of the response it provides.

[1621] Result display phase

[1622] The server sends the generated response to the device, and the device displays the received response to the user. The displayed response is shown in an appropriate tone and level of detail based on the user's emotions recognized by the emotion engine.

[1623] Feedback processing phase

[1624] Users provide feedback on the answers provided. For example, they might enter feedback such as "This answer was helpful" or "This answer was inappropriate." The device then forwards the user's feedback to the server.

[1625] Model and emotion engine improvement phase

[1626] The server analyzes the received feedback and the emotions associated with it, improving both the NLP model and the emotion engine. If there is feedback regarding inappropriate responses, the analysis results are used to retrain the model. The feedback is also utilized as data to improve the accuracy of emotion recognition.

[1627] Examples of specific cases and prompts for generative AI models.

[1628] As a concrete example, in response to a question about the latest lead generation results in the marketing department, the system provides the answer, "The latest lead generation results show that we acquired 100 leads last week." If the emotion engine detects that the user is in a hurry, it will quickly display a concise response such as, "Here are the latest lead generation results. We acquired 100 leads last week."

[1629] Examples of prompts for a generative AI model:

[1630] "Could you please tell me the latest lead generation results for your marketing department?"

[1631] "Could you please tell me the latest project progress in the XX department?"

[1632] "Please use the emotion engine to evaluate user satisfaction."

[1633] This system aims to improve operational efficiency and user experience by integrating the interpretation of technical terms with the recognition of user emotions.

[1634] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1635] Step 1:

[1636] The server collects a list of technical terms and related information from each department of the company. The input requires access information to each department's data source. Specifically, the server uses an API to send HTTP requests to each department's data source to retrieve the technical term list and related information. The output is the collected raw data.

[1637] Step 2:

[1638] The server cleanses the collected data. The raw data obtained in step 1 is used as input. Specifically, a data cleansing tool such as OpenRefine is used to remove duplicate data, fill in incomplete data, and ensure consistency. The output is the cleansed data.

[1639] Step 3:

[1640] The server stores the cleansed data in the database. The cleansed data generated in step 2 is required as input. Specifically, a database management system such as MySQL is used, and the data is saved to the database using SQL queries. The output is structured data stored in the database.

[1641] Step 4:

[1642] The server prepares a natural language processing (NLP) model. The input requires an existing trained model and a list of specialized terminology for each field. Specifically, it imports and initializes natural language processing libraries such as Spacy or Hugging Face. The output is the NLP model prepared for training.

[1643] Step 5:

[1644] The server trains an NLP model using the specialized terminology and data used in its context within the database. The inputs required are the data stored in step 3 and the NLP model prepared in step 4. Specifically, the model is fed data, hyperparameters such as the number of epochs and learning rate are set, and training is performed. The output is the trained NLP model.

[1645] Step 6:

[1646] The server evaluates the trained model and adjusts the parameters as needed. The input requires a trained NLP model and an evaluation dataset. Specifically, it measures the model's accuracy and error and readjusts the hyperparameters. The output is an optimized NLP model.

[1647] Step 7:

[1648] The user enters a question into the system using a terminal. Input requires the user to enter a text box, such as "What are the latest lead generation results?". The output is the user's question stored in text format on the terminal.

[1649] Step 8:

[1650] The terminal forwards the user's question to the server. The input requires the user's question. Specifically, it sends an HTTP request to the server and transfers the data. The output is the server receiving the question.

[1651] Step 9:

[1652] The server uses an NLP model to analyze user questions. The input requires the user's question, transmitted from the terminal, and a trained NLP model. Specifically, the model analyzes the question and identifies relevant terminology and context. The output is the analysis result.

[1653] Step 10:

[1654] The server searches the database based on the analysis results. The analysis results and the database are required as input. Specifically, it generates an SQL query, sends it to the database, and retrieves the relevant data. The output will provide the necessary information.

[1655] Step 11:

[1656] The server generates appropriate answers from the collected information. It requires information retrieved from a database as input. Specifically, it integrates the information to create a user-friendly answer. The output is the generated answer.

[1657] Step 12:

[1658] The server uses an emotion engine to recognize emotions from user input. The input requires both the user's question and the emotion engine. Specifically, it performs text analysis to evaluate emotions based on tone and context. The output is the recognized emotion.

[1659] Step 13:

[1660] The server sends the generated response to the terminal, and the terminal displays the received response to the user. The required inputs are the generated response and the sentiment analysis results. Specifically, the response is sent to the terminal via an HTTP response, which the terminal then displays. The output is the response that the user can view.

[1661] Step 14:

[1662] The user provides feedback on the provided answer. The input requires the answer displayed by the system, and the user inputs feedback such as "This answer was helpful" or "This answer was inappropriate." The output is the feedback stored on the device.

[1663] Step 15:

[1664] The terminal forwards user feedback to the server. The input requires user feedback. Specifically, it sends the feedback to the server using an HTTP request. The output is the server receiving the feedback.

[1665] Step 16:

[1666] The server analyzes the received feedback and improves the NLP model and emotion engine. Input requires user feedback and training data. Specifically, it retrains the model based on the feedback to improve performance. The output is an improved NLP model and emotion engine.

[1667] (Application Example 2)

[1668] 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".

[1669] In traditional brick-and-mortar stores, communication between employees and customers often lacked the speed and accuracy of providing specialized information, and it was difficult to respond appropriately to customers' emotions. This resulted in decreased customer satisfaction and reduced operational efficiency. This invention aims to solve these problems and enable the provision of higher-quality service in customer interactions at physical stores.

[1670] The specific processing performed by the specific 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 a list of technical terms and related information, means for cleansing the collected data and storing it in a database, means for learning technical terms and their context using a natural language processing model, means for receiving and analyzing user inquiries, means for searching the database based on the analysis results and generating an appropriate response, means for transmitting and displaying the generated response on a user terminal, means for recognizing the user's emotions and adjusting the response tone and level of detail based on the recognition results, means for receiving feedback and using it to improve the model, and a natural language processing model customized based on the technical terms and related information of each department. This enables employees in physical stores to provide information quickly and accurately, as well as to respond appropriately to customer emotions.

[1671] A "glossary" is a list that compiles specific technical terms and expertise used within a company or department.

[1672] "Related information" refers to all information related to technical terms, including definitions of technical terms, examples of usage, process documents, manuals, and project reports.

[1673] "Means of collection" refers to the methods and devices for gathering a list of technical terms and related information.

[1674] "Cleansing" refers to the process of organizing and correcting collected data to ensure consistency and accuracy.

[1675] A "database" is a collection of data that is efficiently stored and managed in a way that makes it searchable and usable.

[1676] A "natural language processing model" is a model used to train algorithms and methods for analyzing and understanding human language using computers.

[1677] "Means of learning" refers to methods and devices for training natural language processing models to learn specialized terminology and its context.

[1678] "Means of analysis" refers to methods and devices for analyzing user inquiries and understanding their content.

[1679] "Searching means" refers to methods and devices for finding relevant information within a database based on analysis results.

[1680] "Means for generating answers" refers to methods and devices for creating appropriate answers for users based on retrieved information.

[1681] "User terminal" refers to devices or systems that a user directly operates, such as smartphones or computers.

[1682] "Means for recognizing emotions and adjusting response tone and level of detail based on the recognition results" refers to methods and apparatus for analyzing a user's emotions and adapting the tone and level of detail of the response accordingly.

[1683] "Means for receiving feedback and using it to improve the model" refers to methods and devices for receiving evaluations and opinions from users and incorporating them into improvements to natural language processing models and sentiment engines.

[1684] A "customized natural language processing model" refers to a natural language processing model that has been specifically tailored to a particular company or department.

[1685] This invention relates to an information provision system for improving the quality of customer service in physical stores. Specific embodiments thereof are described below.

[1686] 1. Data collection and cleansing

[1687] First, the server collects a list of technical terms and related information (e.g., product features, process descriptions, manuals, and usage examples) from various departments within the company. The collected data is then cleansed by the server and stored in a database, ensuring consistency and accuracy. This database is then used as a source of information for subsequent queries.

[1688] 2. Training a natural language processing model

[1689] The server prepares a natural language processing (NLP) model and trains it based on specialized terminology and its contextual information stored in a database. Through this training, the model learns to understand the context in which specific terms are used and generates appropriate responses accordingly.

[1690] 3. Inquiry analysis and response generation

[1691] The user (in this case, a store employee) enters an inquiry into the system using a device such as a smartphone. For example, the question might be, "Please tell me about the features of this new product." The device transmits the entered question to the server, which analyzes the question using an NLP model. Based on the analysis results, it searches the database and generates an appropriate answer. In the process, it also analyzes the customer's emotions and adjusts the response tone and level of detail accordingly.

[1692] 4. Emotion Recognition and Response Adjustment

[1693] The server uses an emotion engine to recognize emotions from user input. For example, it analyzes the tone and context of the entered question and adopts a comforting response tone if the user is dissatisfied. This allows users to receive more appropriate and satisfying service.

[1694] 5. Feedback and Model Improvement

[1695] The system also includes a feature that allows users to provide feedback on the answers they receive. For example, users can input feedback such as "This answer was helpful" or "This answer was inappropriate." The user's device forwards this feedback to the server, which then uses it to improve both the NLP model and the emotion engine. This enables continuous model improvement based on feedback.

[1696] 6. Usage example

[1697] For example, if a customer asks, "What are the features of the new product?", the server will provide information such as, "This is a new leather bag made from high-quality cowhide." Also, if a customer asks in a dissatisfied tone, "Is it possible to return this item?", the server will respond with something like, "Returns are accepted within 30 days of purchase. Please contact us if you have any problems; we will be happy to assist you."

[1698] Example of a prompt

[1699] "Could you please tell me the specifications of this product?"

[1700] "Is this item returnable?"

[1701] In this way, the present invention facilitates communication between employees and customers in physical stores, enabling the rapid and accurate provision of specialized information and appropriate responses that respond to customer emotions. This is expected to improve customer satisfaction and operational efficiency.

[1702] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1703] Step 1:

[1704] The server collects terminology lists and related information from each department. Specifically, the server uses APIs and file transfer protocols to retrieve data from each department's databases and document management systems. The input here is the terminology lists and related information provided by each department, and the output is the collected data.

[1705] Step 2:

[1706] The server cleanses the collected data, removing unnecessary and incorrect data to maintain consistency and accuracy. Specifically, it performs data cleaning operations using libraries such as Python's pandas library. The input here is the data collected in the previous step, and the output is the cleansed data.

[1707] Step 3:

[1708] The server stores the cleansed data in a database. Specifically, it inserts the data into an SQL database or a NoSQL database. Here, the input is the cleansed data, and the output is the data stored in the database.

[1709] Step 4:

[1710] The server trains a natural language processing (NLP) model. Specifically, it trains an NLP model (e.g., BERT or GPT) using technical terms and their contexts from a database. The input here is the technical terms and contextual information from the database, and the output is the trained NLP model.

[1711] Step 5:

[1712] The user (employee) enters a question using a terminal. For example, they might enter, "Please tell me the features of this new product." Here, the input is the user's question, and the output is the user's input data sent to the terminal.

[1713] Step 6:

[1714] The terminal forwards user input to the server. Specifically, it sends questions to the server using HTTP requests. Here, input refers to the user's input data, and output refers to the data forwarded to the server.

[1715] Step 7:

[1716] The server uses an NLP model to analyze the user's question. Specifically, it understands the context of the question and identifies relevant technical terms. The input here is the user's question, and the output is the analysis result.

[1717] Step 8:

[1718] The server searches the database based on the analysis results and collects the appropriate answers. The input here is the analysis results from the NLP model, and the output is the information collected from the database.

[1719] Step 9:

[1720] The server uses an emotion engine to recognize the user's emotions. Specifically, it analyzes the tone and context of the input question to identify the emotions the user is experiencing. Here, the input is the user's question, and the output is the emotion recognition result.

[1721] Step 10:

[1722] The server adjusts the tone and level of detail of its response based on the recognized emotion. For example, if the user is dissatisfied, it adopts a comforting tone. The input here is the emotion recognition result and collected information, and the output is the adjusted response.

[1723] Step 11:

[1724] The server sends the adjusted response to the terminal. Specifically, it sends the response to the terminal using an HTTP response. Here, the input is the adjusted response, and the output is the response sent to the terminal.

[1725] Step 12:

[1726] The terminal displays the received response to the user. The input here is the response sent to the terminal, and the output is the response displayed to the user.

[1727] Step 13:

[1728] Users provide feedback on the provided answers. For example, they might rate the answer as "This answer was helpful" or "This answer was inappropriate." The input here is the user's feedback, and the output is the feedback entered into the device.

[1729] Step 14:

[1730] The terminal forwards user feedback to the server. Specifically, it sends feedback to the server using an HTTP request. Here, the input is the feedback entered into the terminal, and the output is the feedback forwarded to the server.

[1731] Step 15:

[1732] The server improves the NLP model and emotion engine based on the feedback it receives. Specifically, it analyzes the content of the feedback and retrains the model as needed. The input here is the user's feedback, and the output is the improved NLP model and emotion engine.

[1733] These steps enable employees in physical stores to provide customers with quick and accurate information, as well as to respond appropriately to customers' emotions.

[1734] 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.

[1735] The data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of the data generation model 58 is ChatGPT (Internet Search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">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.

[1736] 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 robot 414.

[1737] 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.

[1738] 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.

[1739] 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.

[1740] 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.

[1741] 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.

[1742] 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."

[1743] 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.

[1744] 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.

[1745] 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.

[1746] 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.

[1747] 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.

[1748] 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.

[1749] 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.

[1750] 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.

[1751] 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.

[1752] 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.

[1753] 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.

[1754] 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 as being incorporated by reference.

[1755] The following is further disclosed regarding the embodiments described above.

[1756] (Claim 1)

[1757] A list of technical terms and means of collecting related information,

[1758] A means of cleansing the collected data and storing it in a database,

[1759] A method for learning technical terms and their context using natural language processing models,

[1760] A means of receiving and analyzing user inquiries,

[1761] A means for searching a database based on the analysis results and generating an appropriate answer,

[1762] A system that includes means for sending and displaying the generated response on the user's terminal.

[1763] (Claim 2)

[1764] The system according to claim 1, comprising means for receiving feedback and using it to improve the model.

[1765] (Claim 3)

[1766] The system according to claim 1, comprising a natural language processing model customized based on the specialized terminology and related information of each department.

[1767] "Example 1"

[1768] (Claim 1)

[1769] A means of collecting a list of specialized terms and related information from each department of a company,

[1770] A means of cleansing the collected data and storing it in a database,

[1771] A method for learning technical terms and their context using natural language processing models,

[1772] A means of receiving and analyzing user inquiries,

[1773] A means for searching a database based on the analysis results and generating an appropriate answer,

[1774] A means of sending and displaying the generated response on the user's terminal,

[1775] A means of receiving user feedback and using it to improve the model,

[1776] A system that includes means for providing a natural language processing model customized based on the specialized terminology and related information of each department.

[1777] (Claim 2)

[1778] The system according to claim 1, which uses a generative AI model to analyze prompt sentences input by a user.

[1779] (Claim 3)

[1780] The system according to claim 1, comprising means for improving the performance of the model by tuning parameters.

[1781] "Application Example 1"

[1782] (Claim 1)

[1783] A list of technical terms and means of collecting related information,

[1784] A means of cleansing the collected data and storing it in a database,

[1785] A method for learning technical terms and their context using natural language processing models,

[1786] A means of receiving and analyzing user inquiries,

[1787] A means for searching a database based on the analysis results and generating an appropriate answer,

[1788] A means of sending and displaying the generated response on the user's terminal,

[1789] A means for factory robots to query technical terms and processes while performing new tasks and obtain necessary information,

[1790] A system including means for displaying results via a robot's display or audio output.

[1791] (Claim 2)

[1792] A means of receiving feedback and using it to improve the model,

[1793] The system according to claim 1, comprising means for enhancing a model based on the results of robot operation.

[1794] (Claim 3)

[1795] Equipped with a customized natural language processing model based on the specialized terminology and related information of each department,

[1796] The system according to claim 1, which enables a factory robot to quickly acquire information on specific technical terms and processes and to efficiently perform tasks.

[1797] "Example 2 of combining an emotion engine"

[1798] (Claim 1)

[1799] A means of collecting a list of specialized terms and related information from each department of a company,

[1800] A means of cleansing the collected data and storing it in a database,

[1801] A method for learning technical terms and their context using natural language processing models,

[1802] A means ...

Claims

1. A list of technical terms and means of collecting related information, A means of cleansing the collected data and storing it in a database, A method for learning technical terms and their context using natural language processing models, A means of receiving and analyzing user inquiries, A means for searching a database based on the analysis results and generating an appropriate answer, A system that includes means for sending and displaying the generated response on the user's terminal.

2. The system according to claim 1, comprising means for receiving feedback and using it to improve the model.

3. The system according to claim 1, comprising a natural language processing model customized based on the specialized terminology and related information of each department.

Images