system

Abstract

Provide a system. 【Solution means】 Means for acquiring data, Means for performing preprocessing, Means for training a machine learning model, Means for evaluating the accuracy of the model, Means for performing retraining, Means for applying the result based on the provided data, Means for recommending information based on the viewing history, A system including means for presenting information to the user.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 discloses a 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] With the increase and diversification of data volume, there is a demand for an AI system that can efficiently and accurately process different forms of information such as images and texts and enable real - time adaptation. However, conventional systems have problems in integrated data processing and improving the accuracy of learning models, and it has been particularly difficult to integrate and utilize different data forms. As a result, there is a problem that the speed of application and practical use in various fields is limited.

Means for Solving the Problems

[0005] This invention provides a means for acquiring data, preprocessing it, training a machine learning model, evaluating its accuracy, and retraining it as needed. Furthermore, it has a function to adjust hyperparameters based on the evaluation results and automates a series of processes, including noise reduction of image and text data, thereby enabling integrated processing of different data formats. As a result, it can be applied in various fields and the accuracy and efficiency of the system are improved.

[0006] "Data" refers to a collection of information, and specifically to the formats such as numbers, characters, and images used when computer systems process, store, and transmit information.

[0007] "Means of acquisition" refers to methods and devices for collecting data from external sources and making it accessible.

[0008] "Means for preprocessing" refers to methods or devices that perform initial processing to convert raw data into an analyzable format.

[0009] A "machine learning model" refers to a mathematical model that uses algorithms to learn patterns from data and is used to perform tasks such as prediction and classification.

[0010] "Means of training" refers to methods or devices that perform the process of optimizing the parameters of a machine learning model using data.

[0011] "Means for evaluating accuracy" refers to methods and devices for measuring and evaluating the performance and prediction accuracy of machine learning models.

[0012] "Means of retraining" refers to methods or devices for retraining existing machine learning models to improve their performance.

[0013] "Noise reduction" refers to the process of removing irrelevant parts or erroneous components from data.

[0014] The "hyperparameter" refers to a parameter that is set externally to adjust the structure of a machine learning model and its behavior during training.

[0015] The "means of application" refers to a method or device that uses the learned model and its output to solve real-world problems and provide services.

Brief Explanation of Drawings

[0016] [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 Embodiment 2 when the 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 the emotion engine is combined.

Mode for Carrying Out the Invention

[0017] Hereinafter, an example of an embodiment of the system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0018] First, the terms used in the following description will be explained.

[0019] In the following embodiments, the numbered processor (hereinafter simply referred to as "processor") may be one arithmetic unit or a combination of multiple arithmetic units. Also, the processor may be one type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), and the like.

[0020] In the following embodiments, the numbered RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a work memory by the processor.

[0021] In the following embodiments, the numbered storage is one or more non-volatile storage devices that store various programs and various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes, etc.

[0022] 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).

[0023] 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."

[0024] [First Embodiment]

[0025] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0026] 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.

[0027] 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).

[0028] 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.

[0029] 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.

[0030] 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.

[0031] 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.

[0032] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] 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.

[0034] 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.

[0035] 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.

[0036] 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".

[0037] The system of this invention is composed of multiple components that work together to automate everything from data acquisition and processing to model training and application. The operation of each component is described below.

[0038] First, the server is responsible for acquiring data from external data sources. This includes video and text data from surveillance cameras, sensors, and databases on the internet. The acquired data is temporarily stored and then pre-processed.

[0039] The server removes noise associated with the data during the preprocessing stage and standardizes the data format. For example, for video data, it standardizes the frame rate and resolution, and for text data, it performs tokenization and removes stop words.

[0040] Next, the server trains a machine learning model using the preprocessed data. The model learns using both video and text data to improve its ability to extract patterns and features. This model is designed to handle diverse data formats in an integrated manner.

[0041] After training is complete, the device evaluates the model's performance using new data. The evaluation uses accuracy metrics that indicate how well the model performs. Users can receive these evaluation results and check the system's adaptability and efficiency.

[0042] If the evaluation results do not meet expectations, the server initiates a retraining process. This adjusts the model's parameters and architecture based on the evaluation results, iteratively improving the model.

[0043] Ultimately, the trained AI models are implemented in various applications. For example, in security systems, they automatically detect suspicious behavior and issue alerts. In the medical field, they analyze image data to assist in diagnosis. In this way, users can utilize the developed systems for a wide range of purposes.

[0044] As a concrete example, in the entertainment industry, content recommendations can be made based on viewer preferences through video content analysis. Users can receive customized content suggestions based on their individual viewing history.

[0045] Thus, the system of the present invention enables efficient operation in a variety of application fields by automating the entire data processing process.

[0046] The following describes the processing flow.

[0047] Step 1:

[0048] The server retrieves video and text data from external data sources. This includes various data formats such as surveillance cameras, online databases, and APIs. The retrieved data is temporarily stored in storage for later processing.

[0049] Step 2:

[0050] The server performs preprocessing on the acquired data. Image data undergoes noise reduction and resolution adjustment, while text data is subjected to tokenization and stop word removal. This prepares the data for analysis.

[0051] Step 3:

[0052] The server starts training a machine learning model using the pre-processed data. It selects an appropriate algorithm and executes the training process so that the model can learn specific patterns and features using the data.

[0053] Step 4:

[0054] The device evaluates the accuracy of the trained model. This evaluation uses metrics such as accuracy, recall, and F1 score, which indicate how well the model can respond to new data.

[0055] Step 5:

[0056] The user reviews the evaluation results of the model sent from their device. This allows them to determine whether the model meets practical standards or requires further adjustments.

[0057] Step 6:

[0058] If the evaluation results show that the accuracy does not meet the target, the server will perform a retraining process. This involves adjusting the model's hyperparameters and retraining it on a different dataset to improve the model's accuracy.

[0059] Step 7:

[0060] Ultimately, if the training and evaluation yield satisfactory results, users apply the model to real-world applications. They analyze data in real time and utilize it in various application areas, such as security systems and medical diagnostic support systems.

[0061] (Example 1)

[0062] 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."

[0063] Currently, many fields require the efficient and automated execution of a series of processes, from information gathering and processing to model training, evaluation, and retraining. However, conventional systems suffer from insufficient coordination between individual processing steps, and processing performance is reduced due to differences in data formats and noise. Therefore, there is a need for a versatile system that can solve these problems all at once with high efficiency.

[0064] 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.

[0065] In this invention, the server includes means for collecting information, means for performing preliminary processing on the collected information, and means for training a machine learning algorithm. This makes it possible to efficiently collect and process various forms of data and build highly accurate machine learning models.

[0066] "Information" refers to all kinds of data obtained from a data source, including, for example, images, text, and numerical data.

[0067] "Means of collection" refers to methods and techniques for obtaining the desired information from external data sources.

[0068] "Pre-processing" refers to preparatory operations such as removing noise from acquired information and standardizing data formats.

[0069] A "machine learning algorithm" refers to a mathematical model or program that learns patterns based on given data and uses that learning to make decisions or predictions.

[0070] "Training" refers to the process of providing data to a machine learning algorithm to allow it to learn and improve its performance.

[0071] "Means of measuring performance" refers to methods for evaluating the predictive accuracy and reproducibility of a trained algorithm.

[0072] "Retraining" refers to the process of learning the algorithm again based on the feedback received after the initial training.

[0073] "Means of executing results" refers to methods for making decisions and taking actions based on collected information using pre-trained models.

[0074] The system in this invention is primarily composed of three entities working together: a server, a terminal, and a user. First, the server is responsible for collecting information from external data sources. Specifically, it acquires image and text data from monitoring devices, sensors, and network databases. This allows for real-time information collection and support for diverse data formats.

[0075] Subsequently, the server performs preliminary processing on the acquired data. The software used here includes natural language processing tools such as NLTK and spaCy, as well as image processing libraries such as OpenCV. Specifically, noise reduction and format standardization are performed, and the data is handed over to the next stage in a high-quality state.

[0076] Next, the server is trained using machine learning algorithms. Frameworks such as TENSORFLOW® and PyTorch are used to construct a neural network, which is then trained on pre-processed data. This model is designed to identify diverse patterns and features of information, enabling highly accurate predictions.

[0077] After training, the device measures the model's performance using new data. In this step, the model's prediction accuracy and recall are evaluated and the results are presented in an easily understandable format for the user by quantifying them.

[0078] The user reviews these measurement results, and the server retrains the model if necessary. Retraining involves adjusting parameters and utilizing additional data to improve the model's performance. This process allows the model to gain the ability to make more accurate and reliable predictions.

[0079] Finally, the trained AI model can be used in a variety of application fields. Specific examples include detecting suspicious behavior in security systems and assisting in diagnosis in the medical field. In the entertainment industry, it can be used to recommend content based on viewing data. An example of a prompt would be, "Create a list of movies tailored to the user's preferences." In this way, the present invention utilizes a generative AI model to process and apply data quickly and automatically, thereby achieving efficient system operation.

[0080] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0081] Step 1:

[0082] The server collects information from external data sources. Inputs include image data from monitoring devices and sensors, and text data obtained from network APIs. In terms of operation, the server initiates the data collection process periodically according to a set schedule. The output is the raw, collected data.

[0083] Step 2:

[0084] The server performs preliminary processing on the collected data. The input is the raw data obtained in step 1, and the output is the processed, clean data. Specifically, for image data, OpenCV is used to adjust the resolution and remove noise, and for text data, NLTK and spaCy are used to tokenize and remove stop words.

[0085] Step 3:

[0086] The server trains a machine learning algorithm using preprocessed data. The input is the clean data obtained in step 2, and the output is the trained model. Specifically, it builds a neural network using TensorFlow or PyTorch and learns features from the data.

[0087] Step 4:

[0088] The device measures the performance of the trained model using newly collected data. The input is the model obtained in step 3 and the new evaluation data, and the output is the evaluation metric (e.g., accuracy and recall). Specifically, it runs an evaluation script to check whether the model's predictions are accurate.

[0089] Step 5:

[0090] The user reviews the evaluation results, and the server retrains the model if necessary. The input is the evaluation results from step 4, and the output is the improved model. In practice, the model's hyperparameters are adjusted, and new data is added for further training.

[0091] Step 6:

[0092] Ultimately, users utilize the trained AI model in real-world applications. Inputs are user prompts or new use cases, while outputs are results and recommendations tailored to the user's needs. For example, based on the prompt "Create a list of movies tailored to the user's preferences," the system generates a customized list of movie recommendations.

[0093] (Application Example 1)

[0094] 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."

[0095] In today's information-saturated world, efficiently providing information tailored to individual user preferences is challenging. In particular, content distribution services need to accurately analyze users' viewing history and preferences and automate personalized recommendations. Traditional methods have struggled to provide optimal information to individual users quickly and effectively.

[0096] 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.

[0097] In this invention, the server includes means for acquiring data, means for preprocessing data, means for training a machine learning model, means for evaluating the accuracy of the model, means for retraining the model, means for applying the results based on the provided data, means for recommending information based on viewing history, and means for presenting information to the user. This enables automatic and effective recommendation of information based on each user's preferences.

[0098] "Means for acquiring data" refers to devices or methods for collecting information from external data sources.

[0099] "Means for preprocessing" refers to a processing device or method for removing noise or standardizing the format of acquired data.

[0100] "Means for training a machine learning model" refers to an apparatus or method for training a machine learning algorithm using preprocessed data.

[0101] "Means for evaluating the accuracy of a model" refers to a device or method for verifying the performance of a trained machine learning model with new data.

[0102] "Means for retraining" refers to a device or method for adjusting parameters based on the accuracy evaluation of the model and repeating the training process.

[0103] "Means for applying results based on provided data" refers to devices or methods for applying learned knowledge to real-world applications.

[0104] "Means for recommending information based on viewing history" refers to a device or method for analyzing a user's past viewing data and presenting relevant information.

[0105] "Means of presenting information to users" refers to a device or method for displaying or notifying users of recommended information.

[0106] The embodiments for carrying out the present invention will now be described. This system is a comprehensive data processing system equipped with the functions of data acquisition, preprocessing, model training, accuracy evaluation, retraining, information recommendation, and result presentation. The specific configuration and operation of the system are described below.

[0107] The server first collects information from external data sources. Specifically, it retrieves video and text data from the internet and local databases. Programming languages ​​such as Python can be used for this process, and scripts are applied to automate data collection.

[0108] Next, the server performs preprocessing on the acquired data, such as noise reduction and formatting standardization. For example, in the case of image data, it standardizes the resolution, and in the case of text data, it uses Natural Language Processing (NLP) techniques to remove unnecessary words.

[0109] Next, the server trains a machine learning model based on the pre-processed data. During this process, it uses machine learning frameworks such as TensorFlow and PyTorch to execute algorithms like collaborative filtering and deep learning, analyzing the user's viewing history and interests.

[0110] The model's accuracy is evaluated on the device using new data, and retraining is performed as needed. This process improves the model's accuracy by adjusting hyperparameters.

[0111] Based on learned knowledge, the system generates recommendations for users based on their viewing history and presents them via smartphones or smart glasses. For example, a user who has watched a lot of cooking-related content in the past will be recommended newly released cooking shows.

[0112] As a concrete example, by entering a prompt such as, "Please suggest cooking videos I want to watch next based on my viewing history," the generating AI model selects videos that match the user's preferences and provides the results. This allows users to quickly obtain information based on their interests.

[0113] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0114] Step 1:

[0115] The server acquires video and text data from external data sources. The input for this data acquisition process is the internet or a local database. The server uses Python scripts to automate data collection and temporarily store the acquired data. The output is the stored raw data.

[0116] Step 2:

[0117] The server performs preprocessing on the acquired data, including noise reduction and formatting standardization. The input for this step is raw data. Data processing includes resolution standardization for image data and removal of unnecessary words for text data. Text is cleaned up using NLP techniques and standardized as a dataset. The output is preprocessed data.

[0118] Step 3:

[0119] The server trains a machine learning model using preprocessed data. The input for this step is a cleaned-up dataset. The server uses TensorFlow or PyTorch to apply collaborative filtering algorithms to train the model. The resulting output is a trained machine learning model.

[0120] Step 4:

[0121] The device evaluates the model's accuracy using newly acquired data. The input is the new test data. The device measures the model's prediction accuracy and evaluates the result as an accuracy metric. The output is the evaluation result.

[0122] Step 5:

[0123] The server initiates the retraining process as needed. The input for this step is the evaluation results. Based on the evaluation, the server adjusts the hyperparameters and retrains the model with necessary data processing. The output is a more accurate, retrained model.

[0124] Step 6:

[0125] The server generates recommendation information based on the user's viewing history. The inputs are the model and the viewing history. Using the generative AI model, it selects the most suitable information for the user and generates prompt messages. The output is the recommendation information.

[0126] Step 7:

[0127] Users receive recommendations through their smartphones or smart glasses. The input is the generated recommendations. The user's device displays the information, presenting content tailored to their personal preferences. The output is the presented information.

[0128] 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.

[0129] This invention combines a system that handles everything from data acquisition and preprocessing to training, evaluation, and retraining of machine learning models, and the application of the results, with an emotion engine that recognizes user emotions. Specific embodiments of this invention are described below.

[0130] First, the server acquires video, text, and audio data from external sources. The types of data acquired vary depending on the application, but are primarily data related to user interaction. The acquired data is temporarily stored in storage.

[0131] Next, the server performs preprocessing on this data. For image data, it performs resolution standardization and noise reduction; for text data, it performs tokenization and removal of irrelevant words; and for audio data, it performs noise filtering and feature extraction for sentiment recognition.

[0132] Furthermore, the server utilizes the pre-processed data to train machine learning models. These models are designed to learn not only patterns in the data but also emotional information obtained from the emotion engine.

[0133] The trained model is evaluated using new data via the device. This evaluation process verifies not only the model's predictive accuracy but also its emotion recognition accuracy. The user reviews the evaluation results from the device and provides instructions for adjustments to improve the system as needed.

[0134] If necessary, the server will retrain and update the system to improve weaknesses identified in the previous evaluation. The retraining process will focus particularly on optimizing hyperparameters in sentiment analysis.

[0135] Ultimately, if the model's performance meets the criteria, users will put the system into practical use. For example, in online meeting systems, providing real-time sentiment analysis results during meetings allows for facilitation tailored to the emotional state of participants. In the field of education, teachers can efficiently understand students' emotional states, supporting improvements in the quality of individualized instruction.

[0136] Thus, this invention enables comprehensive analysis and application of diverse user data, including emotions, and is expected to be used in a wide range of fields, from business settings to daily life.

[0137] The following describes the processing flow.

[0138] Step 1:

[0139] The server acquires data from surveillance cameras, microphones, and text input devices. This includes video, audio, and text data. The acquired data is stored in temporary storage before processing.

[0140] Step 2:

[0141] The server receives video data and performs preprocessing such as adjusting the frame rate and removing noise. For audio data, noise filtering and speech feature extraction for sentiment analysis are performed. Text data is cleaned through tokenization and removal of unnecessary words.

[0142] Step 3:

[0143] The server trains a machine learning model using pre-processed data. Here, it learns multidimensional patterns, including user emotions, based on integrated information from video, audio, and text. During the training process, an emotion engine is incorporated so that the model can recognize various emotional states.

[0144] Step 4:

[0145] The device feeds new data to the trained model and evaluates the model's accuracy and performance. This evaluation process uses metrics such as confusion matrix, accuracy, recall, and F1 score to check the accuracy of sentiment recognition.

[0146] Step 5:

[0147] The user reviews the evaluation results received from the terminal and determines whether the system's performance has reached a practical level. Based on the evaluation, they provide feedback to the server for further functional improvements if necessary.

[0148] Step 6:

[0149] Based on the evaluation results, the server retrains the model as needed. This retraining process includes adjusting hyperparameters and selecting new datasets to improve the accuracy of emotion recognition.

[0150] Step 7:

[0151] After the model's performance meets the required standards through retraining, users can utilize the model in their daily work. For example, in customer service, the model can analyze the user's emotional state in real time and optimize service responses. This enables efficient and responsive customer support.

[0152] (Example 2)

[0153] 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".

[0154] Conventional information analysis systems lacked the means to perform consistent analysis in environments where multiple information formats coexisted. Furthermore, the effective utilization of acquired information and the retraining processes for improving machine learning accuracy were not adequately developed. As a result, efficiency and accuracy were compromised throughout the process from information acquisition to application.

[0155] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0156] In this invention, the server includes means for acquiring information from an external source, means for performing initial processing based on the acquired information, and means for training a learning algorithm using the initially processed information. This enables the effective integration of multiple forms of information, improving the accuracy of machine learning and expanding its range of applications.

[0157] "Means of acquiring information from external sources" refers to methods comprised of processes and technologies that a system uses to collect data from external information sources.

[0158] "Means of performing initial processing based on acquired information" refers to processes such as noise reduction, standardization, and feature extraction in order to convert the collected data into a format that can be analyzed and learned from.

[0159] "Means of training a learning algorithm using pre-processed information" refers to the process of building or improving a machine learning model using processed data to enhance its pattern recognition and predictive capabilities.

[0160] "Means of evaluating performance" are methods or metrics used to measure how accurately a trained machine learning model operates based on input data.

[0161] A "means of retraining" refers to a mechanism that, based on evaluation results, reviews the training dataset and parameter settings to improve the model's performance, and then runs the training process again.

[0162] "Methods for utilizing conclusions" refer to the process of using analytical results and predictive information to take concrete actions and make decisions in real-world application situations.

[0163] This invention is a system that consistently handles everything from data acquisition and initial processing to training, evaluation, retraining, and application of machine learning models. In addition, it is equipped with an emotion engine for comprehensive analysis, including sentiment analysis. This system is realized through the cooperation of the server, terminal, and user elements.

[0164] First, the server obtains necessary information from external sources. This includes using APIs via the internet and accessing various databases. The data primarily consists of video, text, and audio, and this information is temporarily stored in the server's internal storage.

[0165] The acquired data is then initially processed on the server. Specifically, image data is subjected to resolution standardization and noise reduction using an image processing library. Text data is processed using a natural language processing library, such as NLTK, for tokenization and stop word removal. Audio data is subjected to noise filtering and feature extraction, such as MFCC, using a speech analysis library.

[0166] The server then uses this pre-processed data to train a learning algorithm. This process uses a machine learning framework, such as TensorFlow, to train a model for pattern recognition. A sentiment engine is also built in, and sentiment patterns are learned simultaneously at this stage.

[0167] Once the model is trained, the device evaluates its performance with new data. A machine learning evaluation library is used to measure how accurately the model makes predictions and recognizes sentiments. The results are visualized and reported to the user.

[0168] Users can review the evaluation results reported from their terminals and, if they determine that system improvements are needed, they can instruct the server to retrain the system. Retraining is performed to address weaknesses highlighted in the evaluation and to adjust hyperparameters.

[0169] Ultimately, models that meet the criteria will be made available for application by users. A concrete example is real-time sentiment analysis in online meetings, enabling facilitation that takes into account the emotional state of meeting participants. In the field of education, it is expected that understanding students' emotional states in real time will improve the quality of individualized instruction.

[0170] An example of a prompt message would be, "Detect emotions from the user's voice data and suggest appropriate facilitation based on the emotional state of the meeting participants." This allows for the integrated analysis of diverse data, enabling its use in a variety of application scenarios.

[0171] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0172] Step 1:

[0173] The server acquires information from external sources. The input consists of three types: audio data, video data, and text data, collected using APIs and database access. The output is raw, unprocessed data temporarily stored on the server's internal storage. In this procedure, the data collection module retrieves data via API endpoints and streaming protocols.

[0174] Step 2:

[0175] The server performs initial processing on the acquired information. The input is the raw data obtained in step 1. For image data, an image processing library is used to perform resolution standardization and noise reduction. For text data, a natural language processing library is used for tokenization and stop word removal. For speech data, a speech analysis library is used for noise filtering and feature extraction, and processed data in a unified format is generated as output.

[0176] Step 3:

[0177] The server uses the pre-processed data to train the learning algorithm. The input is the unified format data obtained in step 2. A machine learning framework is used to train the pattern recognition model and the sentiment engine. In this process, the model is trained over multiple epochs using the training dataset, and the trained model is obtained as output.

[0178] Step 4:

[0179] The device evaluates the trained model with new data. The inputs are the test dataset and the trained model obtained in step 3. Evaluation uses metrics to check the model's predictive accuracy and sentiment recognition accuracy. The evaluation output is a model performance report, which is visualized and reported to the user.

[0180] Step 5:

[0181] The user reviews the evaluation results via their terminal and determines the need for retraining. The input is the performance report presented in step 4. If necessary, the user instructs the server to retrain, particularly by adjusting the hyperparameters. The output is an improved model.

[0182] Step 6:

[0183] The user ultimately applies the applicable model to real-world situations. The input consists of a performance-approved, trained model and real-time application data. For example, the system analyzes participants' emotions during an online meeting and adjusts the meeting's facilitation based on the results. The output of this step is improved meeting efficiency and enhanced instruction quality in educational settings.

[0184] (Application Example 2)

[0185] 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".

[0186] In smart cities, improving the quality of citizen services requires understanding residents' emotions and stress levels in real time and optimizing public services based on that information. However, conventional systems have the challenge of not being able to efficiently collect and analyze individual emotional data and immediately utilize it for service improvement.

[0187] 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.

[0188] In this invention, the server includes means for acquiring data, means for pre-processing data, and means for analyzing user emotions. This enables real-time analysis of residents' emotional states and the optimization of public services by utilizing the results.

[0189] "Means of acquiring data" refers to the function of carrying out the process of gathering necessary information from external sources.

[0190] "Means of preprocessing" refers to functions that remove unnecessary data and eliminate noise in order to improve the accuracy of acquired data.

[0191] "Methods for training machine learning models" refer to functions that train models to recognize certain patterns or features based on pre-processed data.

[0192] "Means for evaluating model accuracy" refers to functions that measure the predictive accuracy and generalization ability of machine learning models and determine their superiority or inferiority.

[0193] A "means of retraining" refers to a function that retrains a model based on evaluations to improve its performance.

[0194] "Means of applying results based on provided data" refers to a function that applies analysis results to realistic uses and situations to achieve optimized results.

[0195] "Means of analyzing user emotions" refers to a function that identifies and analyzes individual emotional states from data such as audio, video, and text.

[0196] "Means for visualizing the results of sentiment analysis" refers to a function that displays the analyzed sentiment information in an easy-to-understand manner using graphs, charts, etc.

[0197] "Means for optimizing public services based on analysis results" refers to a function that improves the content and methods of service provision to citizens based on information obtained from sentiment analysis.

[0198] This invention is a system for optimizing public services in smart cities using citizen sentiment data. The system first uses a server to acquire video, audio, and text data from smartphones and other devices. Streaming technologies such as Kafka are used for data acquisition to enable real-time processing.

[0199] The acquired data is cleaned up in the preprocessing step. Video data is denoised using OpenCV and adjusted to a standard resolution. Tokenization and removal of irrelevant words from text data are performed using the Python NLTK library. Furthermore, features are extracted from audio data using librosa in preparation for sentiment analysis.

[0200] The server uses this preprocessed data to train a machine learning model built with TensorFlow. This model is designed to comprehensively recognize not only data patterns but also the user's emotional state. Each time new data is provided, the terminal evaluates the model's accuracy and the accuracy of emotion recognition, and instructs the model to retrain if performance improvement is needed. The retraining process uses H2O.ai to tune the model's hyperparameters.

[0201] As a result, users can view a dashboard on their smartphones that visualizes the analyzed sentiment data. This information is also provided to smart city administrators, making it possible, for example, to prioritize the provision of relevant services or events in areas with a high concentration of stressed residents.

[0202] For example, if data is analyzed showing that a large number of residents are experiencing stress during a specific time period, the government can then hold relaxation-related events during that time. Another example of a prompt for using such a generative AI model is, "Citizens in a certain area are experiencing stress; please suggest solutions."

[0203] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0204] Step 1:

[0205] The server acquires video, audio, and text data in real time from the user's smartphone or digital device. Input is raw data sent from the device, and output is raw data temporarily stored on the server's storage. This data is streamed using Kafka.

[0206] Step 2:

[0207] The server preprocesses the acquired raw data. Specifically, video data is denoised and adjusted to standard resolution using OpenCV. Text data is tokenized and irrelevant words are removed using the Python NLTK library. For audio data, librosa is used to extract features necessary for sentiment recognition. The input is raw data, and the output is cleaned-up data ready for analysis.

[0208] Step 3:

[0209] The server trains a machine learning model built with TensorFlow based on preprocessed data. The input is the preprocessed dataset, and the output is the trained model with updated sentiment recognition. This gives the server the ability to analyze data patterns and sentiment states in an integrated manner.

[0210] Step 4:

[0211] The device evaluates the model using new data. Here, it checks the model's prediction accuracy and sentiment recognition accuracy and sends feedback to the server based on these results. The input is new data for evaluation, and the output is the model's evaluation result. The device collects evaluation feedback and instructs retraining if necessary.

[0212] Step 5:

[0213] Based on the evaluation results, the server retrains the model using H2O.ai if necessary, particularly optimizing the hyperparameters. The input consists of evaluation feedback and existing model data, and the output is the retrained model with improved accuracy.

[0214] Step 6:

[0215] Users view the analyzed sentiment data as a visualized dashboard on their devices. Specifically, they can use tools like Power BI to view their emotional state in real time using graphs and charts. The input is the analysis results from an improved model, and the output is visualized sentiment data. This information can be supplied to citizens and smart city administrators and used to generate prompts for AI models, such as "Citizens in a certain area are experiencing stress; please suggest solutions."

[0216] 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.

[0217] 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.

[0218] 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.

[0219] [Second Embodiment]

[0220] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0221] 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.

[0222] 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).

[0223] 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.

[0224] 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.

[0225] 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).

[0226] 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.

[0227] 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.

[0228] 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.

[0229] 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.

[0230] 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.

[0231] 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".

[0232] The system of this invention is composed of multiple components that work together to automate everything from data acquisition and processing to model training and application. The operation of each component is described below.

[0233] First, the server is responsible for acquiring data from external data sources. This includes video and text data from surveillance cameras, sensors, and databases on the internet. The acquired data is temporarily stored and then pre-processed.

[0234] The server removes noise associated with the data during the preprocessing stage and standardizes the data format. For example, for video data, it standardizes the frame rate and resolution, and for text data, it performs tokenization and removes stop words.

[0235] Next, the server trains a machine learning model using the preprocessed data. The model learns using both video and text data to improve its ability to extract patterns and features. This model is designed to handle diverse data formats in an integrated manner.

[0236] After training is complete, the device evaluates the model's performance using new data. The evaluation uses accuracy metrics that indicate how well the model performs. Users can receive these evaluation results and check the system's adaptability and efficiency.

[0237] If the evaluation results do not meet expectations, the server initiates a retraining process. This adjusts the model's parameters and architecture based on the evaluation results, iteratively improving the model.

[0238] Ultimately, the trained AI models are implemented in various applications. For example, in security systems, they automatically detect suspicious behavior and issue alerts. In the medical field, they analyze image data to assist in diagnosis. In this way, users can utilize the developed systems for a wide range of purposes.

[0239] As a concrete example, in the entertainment industry, content recommendations can be made based on viewer preferences through video content analysis. Users can receive customized content suggestions based on their individual viewing history.

[0240] Thus, the system of the present invention enables efficient operation in a variety of application fields by automating the entire data processing process.

[0241] The following describes the processing flow.

[0242] Step 1:

[0243] The server retrieves video and text data from external data sources. This includes various data formats such as surveillance cameras, online databases, and APIs. The retrieved data is temporarily stored in storage for later processing.

[0244] Step 2:

[0245] The server performs preprocessing on the acquired data. Image data undergoes noise reduction and resolution adjustment, while text data is subjected to tokenization and stop word removal. This prepares the data for analysis.

[0246] Step 3:

[0247] The server starts training a machine learning model using the pre-processed data. It selects an appropriate algorithm and executes the training process so that the model can learn specific patterns and features using the data.

[0248] Step 4:

[0249] The device evaluates the accuracy of the trained model. This evaluation uses metrics such as accuracy, recall, and F1 score, which indicate how well the model can respond to new data.

[0250] Step 5:

[0251] The user reviews the evaluation results of the model sent from their device. This allows them to determine whether the model meets practical standards or requires further adjustments.

[0252] Step 6:

[0253] If the evaluation results show that the accuracy does not meet the target, the server will perform a retraining process. This involves adjusting the model's hyperparameters and retraining it on a different dataset to improve the model's accuracy.

[0254] Step 7:

[0255] Ultimately, if the training and evaluation yield satisfactory results, users apply the model to real-world applications. They analyze data in real time and utilize it in various application areas, such as security systems and medical diagnostic support systems.

[0256] (Example 1)

[0257] 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."

[0258] Currently, many fields require the efficient and automated execution of a series of processes, from information gathering and processing to model training, evaluation, and retraining. However, conventional systems suffer from insufficient coordination between individual processing steps, and processing performance is reduced due to differences in data formats and noise. Therefore, there is a need for a versatile system that can solve these problems all at once with high efficiency.

[0259] 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.

[0260] In this invention, the server includes means for collecting information, means for performing preliminary processing on the collected information, and means for training a machine learning algorithm. This makes it possible to efficiently collect and process various forms of data and build highly accurate machine learning models.

[0261] "Information" refers to all kinds of data obtained from a data source, including, for example, images, text, and numerical data.

[0262] "Means of collection" refers to methods and techniques for obtaining the desired information from external data sources.

[0263] "Pre-processing" refers to preparatory operations such as removing noise from acquired information and standardizing data formats.

[0264] A "machine learning algorithm" refers to a mathematical model or program that learns patterns based on given data and uses that learning to make decisions or predictions.

[0265] "Training" refers to the process of providing data to a machine learning algorithm to allow it to learn and improve its performance.

[0266] "Means of measuring performance" refers to methods for evaluating the predictive accuracy and reproducibility of a trained algorithm.

[0267] "Retraining" refers to the process of learning the algorithm again based on the feedback received after the initial training.

[0268] "Means of executing results" refers to methods for making decisions and taking actions based on collected information using pre-trained models.

[0269] The system in this invention is primarily composed of three entities working together: a server, a terminal, and a user. First, the server is responsible for collecting information from external data sources. Specifically, it acquires image and text data from monitoring devices, sensors, and network databases. This allows for real-time information collection and support for diverse data formats.

[0270] Subsequently, the server performs preliminary processing on the acquired data. The software used here includes natural language processing tools such as NLTK and spaCy, as well as image processing libraries such as OpenCV. Specifically, noise reduction and format standardization are performed, and the data is handed over to the next stage in a high-quality state.

[0271] Next, the server is trained using machine learning algorithms. Frameworks such as TensorFlow and PyTorch are used to build a neural network, which is then trained on pre-processed data. This model is designed to identify diverse patterns and features of information, enabling highly accurate predictions.

[0272] After training, the device measures the model's performance using new data. In this step, the model's prediction accuracy and recall are evaluated and the results are presented in an easily understandable format for the user by quantifying them.

[0273] The user reviews these measurement results, and the server retrains the model if necessary. Retraining involves adjusting parameters and utilizing additional data to improve the model's performance. This process allows the model to gain the ability to make more accurate and reliable predictions.

[0274] Finally, the trained AI model can be used in a variety of application fields. Specific examples include detecting suspicious behavior in security systems and assisting in diagnosis in the medical field. In the entertainment industry, it can be used to recommend content based on viewing data. An example of a prompt would be, "Create a list of movies tailored to the user's preferences." In this way, the present invention utilizes a generative AI model to process and apply data quickly and automatically, thereby achieving efficient system operation.

[0275] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0276] Step 1:

[0277] The server collects information from external data sources. The inputs include image data from monitoring devices and sensors, and text data obtained from APIs on the network. In specific operations, the data collection process is started according to a periodically set schedule. The output is the raw data obtained.

[0278] Step 2:

[0279] The server performs pre - processing on the collected data. The input is the raw data obtained in Step 1, and the output is the processed clean data. As specific processing, for image data, resolution adjustment and noise removal are performed using OpenCV, and for text data, tokenization and stop - word removal are performed using NLTK or spaCy.

[0280] Step 3:

[0281] The server trains a machine - learning algorithm using the pre - processed data. The input is the clean data obtained in Step 2, and the output is the trained model. Specifically, a neural network is constructed using TensorFlow or PyTorch to learn features from the data.

[0282] Step 4:

[0283] The terminal measures the performance of the trained model using newly collected data. The input is the model obtained in Step 3 and new evaluation data, and the output is evaluation metrics (e.g., accuracy and recall). Specifically, an evaluation script is executed to check whether the prediction results of the model are accurate.

[0284] Step 5:

[0285] The user checks the evaluation results, and if necessary, the server performs retraining. The input is the evaluation result of step 4, and the output is an improved model. In a specific operation, the hyperparameters of the model are adjusted, new data is added, and training is performed again.

[0286] Step 6:

[0287] Finally, the user applies the trained AI model to actual applications. The input is the prompt text or new use cases from the user, and the output is the results or recommendations suitable for the user's needs. As a specific example, based on the prompt text "Create a list of movies that match the user's preferences", the system generates a customized movie recommendation list.

[0288] (Application Example 1)

[0289] Next, Application Example 1 will be described. In the following description, the data processing device 12 is referred to as the "server", and the smart glasses 214 are referred to as the "terminal".

[0290] In the current era of information overload, it is difficult to efficiently provide information that suits individual users' preferences. In particular, in content delivery services, it is required to accurately analyze users' viewing histories and preferences and automate personalized recommendations. There is an issue that it was difficult to quickly and effectively provide the most suitable information for individual users with conventional methods.

[0291] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0292] In this invention, the server includes means for acquiring data, means for preprocessing data, means for training a machine learning model, means for evaluating the accuracy of the model, means for retraining the model, means for applying the results based on the provided data, means for recommending information based on viewing history, and means for presenting information to the user. This enables automatic and effective recommendation of information based on each user's preferences.

[0293] "Means for acquiring data" refers to devices or methods for collecting information from external data sources.

[0294] "Means for preprocessing" refers to a processing device or method for removing noise or standardizing the format of acquired data.

[0295] "Means for training a machine learning model" refers to an apparatus or method for training a machine learning algorithm using preprocessed data.

[0296] "Means for evaluating the accuracy of a model" refers to a device or method for verifying the performance of a trained machine learning model with new data.

[0297] "Means for retraining" refers to a device or method for adjusting parameters based on the accuracy evaluation of the model and repeating the training process.

[0298] "Means for applying results based on provided data" refers to devices or methods for applying learned knowledge to real-world applications.

[0299] "Means for recommending information based on viewing history" refers to a device or method for analyzing a user's past viewing data and presenting relevant information.

[0300] "Means of presenting information to users" refers to a device or method for displaying or notifying users of recommended information.

[0301] The embodiments for carrying out the present invention will now be described. This system is a comprehensive data processing system equipped with the functions of data acquisition, preprocessing, model training, accuracy evaluation, retraining, information recommendation, and result presentation. The specific configuration and operation of the system are described below.

[0302] The server first collects information from external data sources. Specifically, it retrieves video and text data from the internet and local databases. Programming languages ​​such as Python can be used for this process, and scripts are applied to automate data collection.

[0303] Next, the server performs preprocessing on the acquired data, such as noise reduction and formatting standardization. For example, in the case of image data, it standardizes the resolution, and in the case of text data, it uses Natural Language Processing (NLP) techniques to remove unnecessary words.

[0304] Next, the server trains a machine learning model based on the pre-processed data. During this process, it uses machine learning frameworks such as TensorFlow and PyTorch to execute algorithms like collaborative filtering and deep learning, analyzing the user's viewing history and interests.

[0305] The model's accuracy is evaluated on the device using new data, and retraining is performed as needed. This process improves the model's accuracy by adjusting hyperparameters.

[0306] Based on learned knowledge, the system generates recommendations for users based on their viewing history and presents them via smartphones or smart glasses. For example, a user who has watched a lot of cooking-related content in the past will be recommended newly released cooking shows.

[0307] As a specific example, by inputting a prompt sentence such as "Please propose a cooking video that I would like to watch next based on my viewing history", the generative AI model selects a video that matches the user's preferences and provides the result. This enables the user to quickly obtain information based on their interests.

[0308] The flow of the specific process in Application Example 1 will be described using FIG. 12.

[0309] Step 1:

[0310] The server acquires video data and text data from an external data source. The input for this data acquisition process is the Internet or a local database. The server automates data collection using a Python script and operates to temporarily store the acquired data. The output is the stored raw data.

[0311] Step 2:

[0312] The server performs preprocessing on the acquired data, including noise removal and format unification. The input for this step is the raw data. Data processing includes unifying the resolution for image data and removing unnecessary words for text data. The text is cleaned up using NLP technology and unified into a dataset. The output is the preprocessed data.

[0313] Step 3:

[0314] The server trains a machine learning model using the preprocessed data. The input for this step is the cleaned-up dataset. The server uses TensorFlow or PyTorch and applies a collaborative filtering algorithm to train the model. The output as a result is the trained machine learning model.

[0315] Step 4:

[0316] The device evaluates the model's accuracy using newly acquired data. The input is the new test data. The device measures the model's prediction accuracy and evaluates the result as an accuracy metric. The output is the evaluation result.

[0317] Step 5:

[0318] The server initiates the retraining process as needed. The input for this step is the evaluation results. Based on the evaluation, the server adjusts the hyperparameters and retrains the model with necessary data processing. The output is a more accurate, retrained model.

[0319] Step 6:

[0320] The server generates recommendation information based on the user's viewing history. The inputs are the model and the viewing history. Using the generative AI model, it selects the most suitable information for the user and generates prompt messages. The output is the recommendation information.

[0321] Step 7:

[0322] Users receive recommendations through their smartphones or smart glasses. The input is the generated recommendations. The user's device displays the information, presenting content tailored to their personal preferences. The output is the presented information.

[0323] 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.

[0324] This invention combines a system that handles everything from data acquisition and preprocessing to training, evaluation, and retraining of machine learning models, and the application of the results, with an emotion engine that recognizes user emotions. Specific embodiments of this invention are described below.

[0325] First, the server acquires video, text, and audio data from external sources. The types of data acquired vary depending on the application, but are primarily data related to user interaction. The acquired data is temporarily stored in storage.

[0326] Next, the server performs preprocessing on this data. For image data, it performs resolution standardization and noise reduction; for text data, it performs tokenization and removal of irrelevant words; and for audio data, it performs noise filtering and feature extraction for sentiment recognition.

[0327] Furthermore, the server utilizes the pre-processed data to train machine learning models. These models are designed to learn not only patterns in the data but also emotional information obtained from the emotion engine.

[0328] The trained model is evaluated using new data via the device. This evaluation process verifies not only the model's predictive accuracy but also its emotion recognition accuracy. The user reviews the evaluation results from the device and provides instructions for adjustments to improve the system as needed.

[0329] If necessary, the server will retrain and update the system to improve weaknesses identified in the previous evaluation. The retraining process will focus particularly on optimizing hyperparameters in sentiment analysis.

[0330] Ultimately, if the model's performance meets the criteria, users will put the system into practical use. For example, in online meeting systems, providing real-time sentiment analysis results during meetings allows for facilitation tailored to the emotional state of participants. In the field of education, teachers can efficiently understand students' emotional states, supporting improvements in the quality of individualized instruction.

[0331] Thus, this invention enables comprehensive analysis and application of diverse user data, including emotions, and is expected to be used in a wide range of fields, from business settings to daily life.

[0332] The following describes the processing flow.

[0333] Step 1:

[0334] The server acquires data from surveillance cameras, microphones, and text input devices. This includes video, audio, and text data. The acquired data is stored in temporary storage before processing.

[0335] Step 2:

[0336] The server receives video data and performs preprocessing such as adjusting the frame rate and removing noise. For audio data, noise filtering and speech feature extraction for sentiment analysis are performed. Text data is cleaned through tokenization and removal of unnecessary words.

[0337] Step 3:

[0338] The server trains a machine learning model using pre-processed data. Here, it learns multidimensional patterns, including user emotions, based on integrated information from video, audio, and text. During the training process, an emotion engine is incorporated so that the model can recognize various emotional states.

[0339] Step 4:

[0340] The device feeds new data to the trained model and evaluates the model's accuracy and performance. This evaluation process uses metrics such as confusion matrix, accuracy, recall, and F1 score to check the accuracy of sentiment recognition.

[0341] Step 5:

[0342] The user reviews the evaluation results received from the terminal and determines whether the system's performance has reached a practical level. Based on the evaluation, they provide feedback to the server for further functional improvements if necessary.

[0343] Step 6:

[0344] Based on the evaluation results, the server retrains the model as needed. This retraining process includes adjusting hyperparameters and selecting new datasets to improve the accuracy of emotion recognition.

[0345] Step 7:

[0346] After the model's performance meets the required standards through retraining, users can utilize the model in their daily work. For example, in customer service, the model can analyze the user's emotional state in real time and optimize service responses. This enables efficient and responsive customer support.

[0347] (Example 2)

[0348] 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".

[0349] Conventional information analysis systems lacked the means to perform consistent analysis in environments where multiple information formats coexisted. Furthermore, the effective utilization of acquired information and the retraining processes for improving machine learning accuracy were not adequately developed. As a result, efficiency and accuracy were compromised throughout the process from information acquisition to application.

[0350] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0351] In this invention, the server includes means for acquiring information from an external source, means for performing initial processing based on the acquired information, and means for training a learning algorithm using the initially processed information. This enables the effective integration of multiple forms of information, improving the accuracy of machine learning and expanding its range of applications.

[0352] "Means of acquiring information from external sources" refers to methods comprised of processes and technologies that a system uses to collect data from external information sources.

[0353] "Means of performing initial processing based on acquired information" refers to processes such as noise reduction, standardization, and feature extraction in order to convert the collected data into a format that can be analyzed and learned from.

[0354] "Means of training a learning algorithm using pre-processed information" refers to the process of building or improving a machine learning model using processed data to enhance its pattern recognition and predictive capabilities.

[0355] "Means of evaluating performance" are methods or metrics used to measure how accurately a trained machine learning model operates based on input data.

[0356] A "means of retraining" refers to a mechanism that, based on evaluation results, reviews the training dataset and parameter settings to improve the model's performance, and then runs the training process again.

[0357] "Methods for utilizing conclusions" refer to the process of using analytical results and predictive information to take concrete actions and make decisions in real-world application situations.

[0358] This invention is a system that consistently handles everything from data acquisition and initial processing to training, evaluation, retraining, and application of machine learning models. In addition, it is equipped with an emotion engine for comprehensive analysis, including sentiment analysis. This system is realized through the cooperation of the server, terminal, and user elements.

[0359] First, the server obtains necessary information from external sources. This includes using APIs via the internet and accessing various databases. The data primarily consists of video, text, and audio, and this information is temporarily stored in the server's internal storage.

[0360] The acquired data is then initially processed on the server. Specifically, image data is subjected to resolution standardization and noise reduction using an image processing library. Text data is processed using a natural language processing library, such as NLTK, for tokenization and stop word removal. Audio data is subjected to noise filtering and feature extraction, such as MFCC, using a speech analysis library.

[0361] The server then uses this pre-processed data to train a learning algorithm. This process uses a machine learning framework, such as TensorFlow, to train a model for pattern recognition. A sentiment engine is also built in, and sentiment patterns are learned simultaneously at this stage.

[0362] Once the model is trained, the device evaluates its performance with new data. A machine learning evaluation library is used to measure how accurately the model makes predictions and recognizes sentiments. The results are visualized and reported to the user.

[0363] Users can review the evaluation results reported from their terminals and, if they determine that system improvements are needed, they can instruct the server to retrain the system. Retraining is performed to address weaknesses highlighted in the evaluation and to adjust hyperparameters.

[0364] Ultimately, models that meet the criteria will be made available for application by users. A concrete example is real-time sentiment analysis in online meetings, enabling facilitation that takes into account the emotional state of meeting participants. In the field of education, it is expected that understanding students' emotional states in real time will improve the quality of individualized instruction.

[0365] An example of a prompt message would be, "Detect emotions from the user's voice data and suggest appropriate facilitation based on the emotional state of the meeting participants." This allows for the integrated analysis of diverse data, enabling its use in a variety of application scenarios.

[0366] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0367] Step 1:

[0368] The server acquires information from external sources. The input consists of three types: audio data, video data, and text data, collected using APIs and database access. The output is raw, unprocessed data temporarily stored on the server's internal storage. In this procedure, the data collection module retrieves data via API endpoints and streaming protocols.

[0369] Step 2:

[0370] The server performs initial processing on the acquired information. The input is the raw data obtained in step 1. For image data, an image processing library is used to perform resolution standardization and noise reduction. For text data, a natural language processing library is used for tokenization and stop word removal. For speech data, a speech analysis library is used for noise filtering and feature extraction, and processed data in a unified format is generated as output.

[0371] Step 3:

[0372] The server uses the pre-processed data to train the learning algorithm. The input is the unified format data obtained in step 2. A machine learning framework is used to train the pattern recognition model and the sentiment engine. In this process, the model is trained over multiple epochs using the training dataset, and the trained model is obtained as output.

[0373] Step 4:

[0374] The device evaluates the trained model with new data. The inputs are the test dataset and the trained model obtained in step 3. Evaluation uses metrics to check the model's predictive accuracy and sentiment recognition accuracy. The evaluation output is a model performance report, which is visualized and reported to the user.

[0375] Step 5:

[0376] The user reviews the evaluation results via their terminal and determines the need for retraining. The input is the performance report presented in step 4. If necessary, the user instructs the server to retrain, particularly by adjusting the hyperparameters. The output is an improved model.

[0377] Step 6:

[0378] The user ultimately applies the applicable model to real-world situations. The input consists of a performance-approved, trained model and real-time application data. For example, the system analyzes participants' emotions during an online meeting and adjusts the meeting's facilitation based on the results. The output of this step is improved meeting efficiency and enhanced instruction quality in educational settings.

[0379] (Application Example 2)

[0380] 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."

[0381] In smart cities, improving the quality of citizen services requires understanding residents' emotions and stress levels in real time and optimizing public services based on that information. However, conventional systems have the challenge of not being able to efficiently collect and analyze individual emotional data and immediately utilize it for service improvement.

[0382] 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.

[0383] In this invention, the server includes means for acquiring data, means for pre-processing data, and means for analyzing user emotions. This enables real-time analysis of residents' emotional states and the optimization of public services by utilizing the results.

[0384] "Means of acquiring data" refers to the function of carrying out the process of gathering necessary information from external sources.

[0385] "Means of preprocessing" refers to functions that remove unnecessary data and eliminate noise in order to improve the accuracy of acquired data.

[0386] "Methods for training machine learning models" refer to functions that train models to recognize certain patterns or features based on pre-processed data.

[0387] "Means for evaluating model accuracy" refers to functions that measure the predictive accuracy and generalization ability of machine learning models and determine their superiority or inferiority.

[0388] A "means of retraining" refers to a function that retrains a model based on evaluations to improve its performance.

[0389] "Means of applying results based on provided data" refers to a function that applies analysis results to realistic uses and situations to achieve optimized results.

[0390] "Means of analyzing user emotions" refers to a function that identifies and analyzes individual emotional states from data such as audio, video, and text.

[0391] "Means for visualizing the results of sentiment analysis" refers to a function that displays the analyzed sentiment information in an easy-to-understand manner using graphs, charts, etc.

[0392] "Means for optimizing public services based on analysis results" refers to a function that improves the content and methods of service provision to citizens based on information obtained from sentiment analysis.

[0393] This invention is a system for optimizing public services in smart cities using citizen sentiment data. The system first uses a server to acquire video, audio, and text data from smartphones and other devices. Streaming technologies such as Kafka are used for data acquisition to enable real-time processing.

[0394] The acquired data is cleaned up in the preprocessing step. Video data is denoised using OpenCV and adjusted to a standard resolution. Tokenization and removal of irrelevant words from text data are performed using the Python NLTK library. Furthermore, features are extracted from audio data using librosa in preparation for sentiment analysis.

[0395] The server uses this preprocessed data to train a machine learning model built with TensorFlow. This model is designed to comprehensively recognize not only data patterns but also the user's emotional state. Each time new data is provided, the terminal evaluates the model's accuracy and the accuracy of emotion recognition, and instructs the model to retrain if performance improvement is needed. The retraining process uses H2O.ai to tune the model's hyperparameters.

[0396] As a result, users can view a dashboard on their smartphones that visualizes the analyzed sentiment data. This information is also provided to smart city administrators, making it possible, for example, to prioritize the provision of relevant services or events in areas with a high concentration of stressed residents.

[0397] For example, if data is analyzed showing that a large number of residents are experiencing stress during a specific time period, the government can then hold relaxation-related events during that time. Another example of a prompt for using such a generative AI model is, "Citizens in a certain area are experiencing stress; please suggest solutions."

[0398] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0399] Step 1:

[0400] The server acquires video, audio, and text data in real time from the user's smartphone or digital device. Input is raw data sent from the device, and output is raw data temporarily stored on the server's storage. This data is streamed using Kafka.

[0401] Step 2:

[0402] The server preprocesses the acquired raw data. Specifically, video data is denoised and adjusted to standard resolution using OpenCV. Text data is tokenized and irrelevant words are removed using the Python NLTK library. For audio data, librosa is used to extract features necessary for sentiment recognition. The input is raw data, and the output is cleaned-up data ready for analysis.

[0403] Step 3:

[0404] The server trains a machine learning model built with TensorFlow based on preprocessed data. The input is the preprocessed dataset, and the output is the trained model with updated sentiment recognition. This gives the server the ability to analyze data patterns and sentiment states in an integrated manner.

[0405] Step 4:

[0406] The device evaluates the model using new data. Here, it checks the model's prediction accuracy and sentiment recognition accuracy and sends feedback to the server based on these results. The input is new data for evaluation, and the output is the model's evaluation result. The device collects evaluation feedback and instructs retraining if necessary.

[0407] Step 5:

[0408] Based on the evaluation results, the server retrains the model using H2O.ai if necessary, particularly optimizing the hyperparameters. The input consists of evaluation feedback and existing model data, and the output is the retrained model with improved accuracy.

[0409] Step 6:

[0410] Users view the analyzed sentiment data as a visualized dashboard on their devices. Specifically, they can use tools like Power BI to view their emotional state in real time using graphs and charts. The input is the analysis results from an improved model, and the output is visualized sentiment data. This information can be supplied to citizens and smart city administrators and used to generate prompts for AI models, such as "Citizens in a certain area are experiencing stress; please suggest solutions."

[0411] 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.

[0412] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0413] 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.

[0414] [Third Embodiment]

[0415] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0416] 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.

[0417] 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).

[0418] 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.

[0419] 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.

[0420] 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).

[0421] 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.

[0422] 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.

[0423] 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.

[0424] 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.

[0425] 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.

[0426] 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".

[0427] The system of this invention is composed of multiple components that work together to automate everything from data acquisition and processing to model training and application. The operation of each component is described below.

[0428] First, the server is responsible for acquiring data from external data sources. This includes video and text data from surveillance cameras, sensors, and databases on the internet. The acquired data is temporarily stored and then pre-processed.

[0429] The server removes noise associated with the data during the preprocessing stage and standardizes the data format. For example, for video data, it standardizes the frame rate and resolution, and for text data, it performs tokenization and removes stop words.

[0430] Next, the server trains a machine learning model using the preprocessed data. The model learns using both video and text data to improve its ability to extract patterns and features. This model is designed to handle diverse data formats in an integrated manner.

[0431] After training is complete, the device evaluates the model's performance using new data. The evaluation uses accuracy metrics that indicate how well the model performs. Users can receive these evaluation results and check the system's adaptability and efficiency.

[0432] If the evaluation results do not meet expectations, the server initiates a retraining process. This adjusts the model's parameters and architecture based on the evaluation results, iteratively improving the model.

[0433] Ultimately, the trained AI models are implemented in various applications. For example, in security systems, they automatically detect suspicious behavior and issue alerts. In the medical field, they analyze image data to assist in diagnosis. In this way, users can utilize the developed systems for a wide range of purposes.

[0434] As a concrete example, in the entertainment industry, content recommendations can be made based on viewer preferences through video content analysis. Users can receive customized content suggestions based on their individual viewing history.

[0435] Thus, the system of the present invention enables efficient operation in a variety of application fields by automating the entire data processing process.

[0436] The following describes the processing flow.

[0437] Step 1:

[0438] The server retrieves video and text data from external data sources. This includes various data formats such as surveillance cameras, online databases, and APIs. The retrieved data is temporarily stored in storage for later processing.

[0439] Step 2:

[0440] The server performs preprocessing on the acquired data. Image data undergoes noise reduction and resolution adjustment, while text data is subjected to tokenization and stop word removal. This prepares the data for analysis.

[0441] Step 3:

[0442] The server starts training a machine learning model using the pre-processed data. It selects an appropriate algorithm and executes the training process so that the model can learn specific patterns and features using the data.

[0443] Step 4:

[0444] The device evaluates the accuracy of the trained model. This evaluation uses metrics such as accuracy, recall, and F1 score, which indicate how well the model can respond to new data.

[0445] Step 5:

[0446] The user reviews the evaluation results of the model sent from their device. This allows them to determine whether the model meets practical standards or requires further adjustments.

[0447] Step 6:

[0448] If the evaluation results show that the accuracy does not meet the target, the server will perform a retraining process. This involves adjusting the model's hyperparameters and retraining it on a different dataset to improve the model's accuracy.

[0449] Step 7:

[0450] Ultimately, if the training and evaluation yield satisfactory results, users apply the model to real-world applications. They analyze data in real time and utilize it in various application areas, such as security systems and medical diagnostic support systems.

[0451] (Example 1)

[0452] 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."

[0453] Currently, many fields require the efficient and automated execution of a series of processes, from information gathering and processing to model training, evaluation, and retraining. However, conventional systems suffer from insufficient coordination between individual processing steps, and processing performance is reduced due to differences in data formats and noise. Therefore, there is a need for a versatile system that can solve these problems all at once with high efficiency.

[0454] 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.

[0455] In this invention, the server includes means for collecting information, means for performing preliminary processing on the collected information, and means for training a machine learning algorithm. This makes it possible to efficiently collect and process various forms of data and build highly accurate machine learning models.

[0456] "Information" refers to all kinds of data obtained from a data source, including, for example, images, text, and numerical data.

[0457] "Means of collection" refers to methods and techniques for obtaining the desired information from external data sources.

[0458] "Pre-processing" refers to preparatory operations such as removing noise from acquired information and standardizing data formats.

[0459] A "machine learning algorithm" refers to a mathematical model or program that learns patterns based on given data and uses that learning to make decisions or predictions.

[0460] "Training" refers to the process of providing data to a machine learning algorithm to allow it to learn and improve its performance.

[0461] "Means of measuring performance" refers to methods for evaluating the predictive accuracy and reproducibility of a trained algorithm.

[0462] "Retraining" refers to the process of learning the algorithm again based on the feedback received after the initial training.

[0463] "Means of executing results" refers to methods for making decisions and taking actions based on collected information using pre-trained models.

[0464] The system in this invention is primarily composed of three entities working together: a server, a terminal, and a user. First, the server is responsible for collecting information from external data sources. Specifically, it acquires image and text data from monitoring devices, sensors, and network databases. This allows for real-time information collection and support for diverse data formats.

[0465] Subsequently, the server performs preliminary processing on the acquired data. The software used here includes natural language processing tools such as NLTK and spaCy, as well as image processing libraries such as OpenCV. Specifically, noise reduction and format standardization are performed, and the data is handed over to the next stage in a high-quality state.

[0466] Next, the server is trained using machine learning algorithms. Frameworks such as TensorFlow and PyTorch are used to build a neural network, which is then trained on pre-processed data. This model is designed to identify diverse patterns and features of information, enabling highly accurate predictions.

[0467] After training, the device measures the model's performance using new data. In this step, the model's prediction accuracy and recall are evaluated and the results are presented in an easily understandable format for the user by quantifying them.

[0468] The user reviews these measurement results, and the server retrains the model if necessary. Retraining involves adjusting parameters and utilizing additional data to improve the model's performance. This process allows the model to gain the ability to make more accurate and reliable predictions.

[0469] Finally, the trained AI model can be used in a variety of application fields. Specific examples include detecting suspicious behavior in security systems and assisting in diagnosis in the medical field. In the entertainment industry, it can be used to recommend content based on viewing data. An example of a prompt would be, "Create a list of movies tailored to the user's preferences." In this way, the present invention utilizes a generative AI model to process and apply data quickly and automatically, thereby achieving efficient system operation.

[0470] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0471] Step 1:

[0472] The server collects information from external data sources. Inputs include image data from monitoring devices and sensors, and text data obtained from network APIs. In terms of operation, the server initiates the data collection process periodically according to a set schedule. The output is the raw, collected data.

[0473] Step 2:

[0474] The server performs preliminary processing on the collected data. The input is the raw data obtained in step 1, and the output is the processed, clean data. Specifically, for image data, OpenCV is used to adjust the resolution and remove noise, and for text data, NLTK and spaCy are used to tokenize and remove stop words.

[0475] Step 3:

[0476] The server trains a machine learning algorithm using preprocessed data. The input is the clean data obtained in step 2, and the output is the trained model. Specifically, it builds a neural network using TensorFlow or PyTorch and learns features from the data.

[0477] Step 4:

[0478] The device measures the performance of the trained model using newly collected data. The input is the model obtained in step 3 and the new evaluation data, and the output is the evaluation metric (e.g., accuracy and recall). Specifically, it runs an evaluation script to check whether the model's predictions are accurate.

[0479] Step 5:

[0480] The user reviews the evaluation results, and the server retrains the model if necessary. The input is the evaluation results from step 4, and the output is the improved model. In practice, the model's hyperparameters are adjusted, and new data is added for further training.

[0481] Step 6:

[0482] Ultimately, users utilize the trained AI model in real-world applications. Inputs are user prompts or new use cases, while outputs are results and recommendations tailored to the user's needs. For example, based on the prompt "Create a list of movies tailored to the user's preferences," the system generates a customized list of movie recommendations.

[0483] (Application Example 1)

[0484] 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."

[0485] In today's information-saturated world, efficiently providing information tailored to individual user preferences is challenging. In particular, content distribution services need to accurately analyze users' viewing history and preferences and automate personalized recommendations. Traditional methods have struggled to provide optimal information to individual users quickly and effectively.

[0486] 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.

[0487] In this invention, the server includes means for acquiring data, means for preprocessing data, means for training a machine learning model, means for evaluating the accuracy of the model, means for retraining the model, means for applying the results based on the provided data, means for recommending information based on viewing history, and means for presenting information to the user. This enables automatic and effective recommendation of information based on each user's preferences.

[0488] "Means for acquiring data" refers to devices or methods for collecting information from external data sources.

[0489] "Means for preprocessing" refers to a processing device or method for removing noise or standardizing the format of acquired data.

[0490] "Means for training a machine learning model" refers to an apparatus or method for training a machine learning algorithm using preprocessed data.

[0491] "Means for evaluating the accuracy of a model" refers to a device or method for verifying the performance of a trained machine learning model with new data.

[0492] "Means for retraining" refers to a device or method for adjusting parameters based on the accuracy evaluation of the model and repeating the training process.

[0493] "Means for applying results based on provided data" refers to devices or methods for applying learned knowledge to real-world applications.

[0494] "Means for recommending information based on viewing history" refers to a device or method for analyzing a user's past viewing data and presenting relevant information.

[0495] "Means of presenting information to users" refers to a device or method for displaying or notifying users of recommended information.

[0496] The embodiments for carrying out the present invention will now be described. This system is a comprehensive data processing system equipped with the functions of data acquisition, preprocessing, model training, accuracy evaluation, retraining, information recommendation, and result presentation. The specific configuration and operation of the system are described below.

[0497] The server first collects information from external data sources. Specifically, it retrieves video and text data from the internet and local databases. Programming languages ​​such as Python can be used for this process, and scripts are applied to automate data collection.

[0498] Next, the server performs preprocessing on the acquired data, such as noise reduction and formatting standardization. For example, in the case of image data, it standardizes the resolution, and in the case of text data, it uses Natural Language Processing (NLP) techniques to remove unnecessary words.

[0499] Next, the server trains a machine learning model based on the pre-processed data. During this process, it uses machine learning frameworks such as TensorFlow and PyTorch to execute algorithms like collaborative filtering and deep learning, analyzing the user's viewing history and interests.

[0500] The model's accuracy is evaluated on the device using new data, and retraining is performed as needed. This process improves the model's accuracy by adjusting hyperparameters.

[0501] Based on learned knowledge, the system generates recommendations for users based on their viewing history and presents them via smartphones or smart glasses. For example, a user who has watched a lot of cooking-related content in the past will be recommended newly released cooking shows.

[0502] As a concrete example, by entering a prompt such as, "Please suggest cooking videos I want to watch next based on my viewing history," the generating AI model selects videos that match the user's preferences and provides the results. This allows users to quickly obtain information based on their interests.

[0503] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0504] Step 1:

[0505] The server acquires video and text data from external data sources. The input for this data acquisition process is the internet or a local database. The server uses Python scripts to automate data collection and temporarily store the acquired data. The output is the stored raw data.

[0506] Step 2:

[0507] The server performs preprocessing on the acquired data, including noise reduction and formatting standardization. The input for this step is raw data. Data processing includes resolution standardization for image data and removal of unnecessary words for text data. Text is cleaned up using NLP techniques and standardized as a dataset. The output is preprocessed data.

[0508] Step 3:

[0509] The server trains a machine learning model using preprocessed data. The input for this step is a cleaned-up dataset. The server uses TensorFlow or PyTorch to apply collaborative filtering algorithms to train the model. The resulting output is a trained machine learning model.

[0510] Step 4:

[0511] The device evaluates the model's accuracy using newly acquired data. The input is the new test data. The device measures the model's prediction accuracy and evaluates the result as an accuracy metric. The output is the evaluation result.

[0512] Step 5:

[0513] The server initiates the retraining process as needed. The input for this step is the evaluation results. Based on the evaluation, the server adjusts the hyperparameters and retrains the model with necessary data processing. The output is a more accurate, retrained model.

[0514] Step 6:

[0515] The server generates recommendation information based on the user's viewing history. The inputs are the model and the viewing history. Using the generative AI model, it selects the most suitable information for the user and generates prompt messages. The output is the recommendation information.

[0516] Step 7:

[0517] Users receive recommendations through their smartphones or smart glasses. The input is the generated recommendations. The user's device displays the information, presenting content tailored to their personal preferences. The output is the presented information.

[0518] 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.

[0519] This invention combines a system that handles everything from data acquisition and preprocessing to training, evaluation, and retraining of machine learning models, and the application of the results, with an emotion engine that recognizes user emotions. Specific embodiments of this invention are described below.

[0520] First, the server acquires video, text, and audio data from external sources. The types of data acquired vary depending on the application, but are primarily data related to user interaction. The acquired data is temporarily stored in storage.

[0521] Next, the server performs preprocessing on this data. For image data, it performs resolution standardization and noise reduction; for text data, it performs tokenization and removal of irrelevant words; and for audio data, it performs noise filtering and feature extraction for sentiment recognition.

[0522] Furthermore, the server utilizes the pre-processed data to train machine learning models. These models are designed to learn not only patterns in the data but also emotional information obtained from the emotion engine.

[0523] The trained model is evaluated using new data via the device. This evaluation process verifies not only the model's predictive accuracy but also its emotion recognition accuracy. The user reviews the evaluation results from the device and provides instructions for adjustments to improve the system as needed.

[0524] If necessary, the server will retrain and update the system to improve weaknesses identified in the previous evaluation. The retraining process will focus particularly on optimizing hyperparameters in sentiment analysis.

[0525] Ultimately, if the model's performance meets the criteria, users will put the system into practical use. For example, in online meeting systems, providing real-time sentiment analysis results during meetings allows for facilitation tailored to the emotional state of participants. In the field of education, teachers can efficiently understand students' emotional states, supporting improvements in the quality of individualized instruction.

[0526] Thus, this invention enables comprehensive analysis and application of diverse user data, including emotions, and is expected to be used in a wide range of fields, from business settings to daily life.

[0527] The following describes the processing flow.

[0528] Step 1:

[0529] The server acquires data from surveillance cameras, microphones, and text input devices. This includes video, audio, and text data. The acquired data is stored in temporary storage before processing.

[0530] Step 2:

[0531] The server receives video data and performs preprocessing such as adjusting the frame rate and removing noise. For audio data, noise filtering and speech feature extraction for sentiment analysis are performed. Text data is cleaned through tokenization and removal of unnecessary words.

[0532] Step 3:

[0533] The server trains a machine learning model using pre-processed data. Here, it learns multidimensional patterns, including user emotions, based on integrated information from video, audio, and text. During the training process, an emotion engine is incorporated so that the model can recognize various emotional states.

[0534] Step 4:

[0535] The device feeds new data to the trained model and evaluates the model's accuracy and performance. This evaluation process uses metrics such as confusion matrix, accuracy, recall, and F1 score to check the accuracy of sentiment recognition.

[0536] Step 5:

[0537] The user reviews the evaluation results received from the terminal and determines whether the system's performance has reached a practical level. Based on the evaluation, they provide feedback to the server for further functional improvements if necessary.

[0538] Step 6:

[0539] Based on the evaluation results, the server retrains the model as needed. This retraining process includes adjusting hyperparameters and selecting new datasets to improve the accuracy of emotion recognition.

[0540] Step 7:

[0541] After the model's performance meets the required standards through retraining, users can utilize the model in their daily work. For example, in customer service, the model can analyze the user's emotional state in real time and optimize service responses. This enables efficient and responsive customer support.

[0542] (Example 2)

[0543] 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."

[0544] Conventional information analysis systems lacked the means to perform consistent analysis in environments where multiple information formats coexisted. Furthermore, the effective utilization of acquired information and the retraining processes for improving machine learning accuracy were not adequately developed. As a result, efficiency and accuracy were compromised throughout the process from information acquisition to application.

[0545] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0546] In this invention, the server includes means for acquiring information from an external source, means for performing initial processing based on the acquired information, and means for training a learning algorithm using the initially processed information. This enables the effective integration of multiple forms of information, improving the accuracy of machine learning and expanding its range of applications.

[0547] "Means of acquiring information from external sources" refers to methods comprised of processes and technologies that a system uses to collect data from external information sources.

[0548] "Means of performing initial processing based on acquired information" refers to processes such as noise reduction, standardization, and feature extraction in order to convert the collected data into a format that can be analyzed and learned from.

[0549] "Means of training a learning algorithm using pre-processed information" refers to the process of building or improving a machine learning model using processed data to enhance its pattern recognition and predictive capabilities.

[0550] "Means of evaluating performance" are methods or metrics used to measure how accurately a trained machine learning model operates based on input data.

[0551] A "means of retraining" refers to a mechanism that, based on evaluation results, reviews the training dataset and parameter settings to improve the model's performance, and then runs the training process again.

[0552] "Methods for utilizing conclusions" refer to the process of using analytical results and predictive information to take concrete actions and make decisions in real-world application situations.

[0553] This invention is a system that consistently handles everything from data acquisition and initial processing to training, evaluation, retraining, and application of machine learning models. In addition, it is equipped with an emotion engine for comprehensive analysis, including sentiment analysis. This system is realized through the cooperation of the server, terminal, and user elements.

[0554] First, the server obtains necessary information from external sources. This includes using APIs via the internet and accessing various databases. The data primarily consists of video, text, and audio, and this information is temporarily stored in the server's internal storage.

[0555] The acquired data is then initially processed on the server. Specifically, image data is subjected to resolution standardization and noise reduction using an image processing library. Text data is processed using a natural language processing library, such as NLTK, for tokenization and stop word removal. Audio data is subjected to noise filtering and feature extraction, such as MFCC, using a speech analysis library.

[0556] The server then uses this pre-processed data to train a learning algorithm. This process uses a machine learning framework, such as TensorFlow, to train a model for pattern recognition. A sentiment engine is also built in, and sentiment patterns are learned simultaneously at this stage.

[0557] Once the model is trained, the device evaluates its performance with new data. A machine learning evaluation library is used to measure how accurately the model makes predictions and recognizes sentiments. The results are visualized and reported to the user.

[0558] Users can review the evaluation results reported from their terminals and, if they determine that system improvements are needed, they can instruct the server to retrain the system. Retraining is performed to address weaknesses highlighted in the evaluation and to adjust hyperparameters.

[0559] Ultimately, models that meet the criteria will be made available for application by users. A concrete example is real-time sentiment analysis in online meetings, enabling facilitation that takes into account the emotional state of meeting participants. In the field of education, it is expected that understanding students' emotional states in real time will improve the quality of individualized instruction.

[0560] An example of a prompt message would be, "Detect emotions from the user's voice data and suggest appropriate facilitation based on the emotional state of the meeting participants." This allows for the integrated analysis of diverse data, enabling its use in a variety of application scenarios.

[0561] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0562] Step 1:

[0563] The server acquires information from external sources. The input consists of three types: audio data, video data, and text data, collected using APIs and database access. The output is raw, unprocessed data temporarily stored on the server's internal storage. In this procedure, the data collection module retrieves data via API endpoints and streaming protocols.

[0564] Step 2:

[0565] The server performs initial processing on the acquired information. The input is the raw data obtained in step 1. For image data, an image processing library is used to perform resolution standardization and noise reduction. For text data, a natural language processing library is used for tokenization and stop word removal. For speech data, a speech analysis library is used for noise filtering and feature extraction, and processed data in a unified format is generated as output.

[0566] Step 3:

[0567] The server uses the pre-processed data to train the learning algorithm. The input is the unified format data obtained in step 2. A machine learning framework is used to train the pattern recognition model and the sentiment engine. In this process, the model is trained over multiple epochs using the training dataset, and the trained model is obtained as output.

[0568] Step 4:

[0569] The device evaluates the trained model with new data. The inputs are the test dataset and the trained model obtained in step 3. Evaluation uses metrics to check the model's predictive accuracy and sentiment recognition accuracy. The evaluation output is a model performance report, which is visualized and reported to the user.

[0570] Step 5:

[0571] The user reviews the evaluation results via their terminal and determines the need for retraining. The input is the performance report presented in step 4. If necessary, the user instructs the server to retrain, particularly by adjusting the hyperparameters. The output is an improved model.

[0572] Step 6:

[0573] The user ultimately applies the applicable model to real-world situations. The input consists of a performance-approved, trained model and real-time application data. For example, the system analyzes participants' emotions during an online meeting and adjusts the meeting's facilitation based on the results. The output of this step is improved meeting efficiency and enhanced instruction quality in educational settings.

[0574] (Application Example 2)

[0575] 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."

[0576] In smart cities, improving the quality of citizen services requires understanding residents' emotions and stress levels in real time and optimizing public services based on that information. However, conventional systems have the challenge of not being able to efficiently collect and analyze individual emotional data and immediately utilize it for service improvement.

[0577] 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.

[0578] In this invention, the server includes means for acquiring data, means for pre-processing data, and means for analyzing user emotions. This enables real-time analysis of residents' emotional states and the optimization of public services by utilizing the results.

[0579] "Means of acquiring data" refers to the function of carrying out the process of gathering necessary information from external sources.

[0580] "Means of preprocessing" refers to functions that remove unnecessary data and eliminate noise in order to improve the accuracy of acquired data.

[0581] "Methods for training machine learning models" refer to functions that train models to recognize certain patterns or features based on pre-processed data.

[0582] "Means for evaluating model accuracy" refers to functions that measure the predictive accuracy and generalization ability of machine learning models and determine their superiority or inferiority.

[0583] A "means of retraining" refers to a function that retrains a model based on evaluations to improve its performance.

[0584] "Means of applying results based on provided data" refers to a function that applies analysis results to realistic uses and situations to achieve optimized results.

[0585] "Means of analyzing user emotions" refers to a function that identifies and analyzes individual emotional states from data such as audio, video, and text.

[0586] "Means for visualizing the results of sentiment analysis" refers to a function that displays the analyzed sentiment information in an easy-to-understand manner using graphs, charts, etc.

[0587] "Means for optimizing public services based on analysis results" refers to a function that improves the content and methods of service provision to citizens based on information obtained from sentiment analysis.

[0588] This invention is a system for optimizing public services in smart cities using citizen sentiment data. The system first uses a server to acquire video, audio, and text data from smartphones and other devices. Streaming technologies such as Kafka are used for data acquisition to enable real-time processing.

[0589] The acquired data is cleaned up in the preprocessing step. Video data is denoised using OpenCV and adjusted to a standard resolution. Tokenization and removal of irrelevant words from text data are performed using the Python NLTK library. Furthermore, features are extracted from audio data using librosa in preparation for sentiment analysis.

[0590] The server uses this preprocessed data to train a machine learning model built with TensorFlow. This model is designed to comprehensively recognize not only data patterns but also the user's emotional state. Each time new data is provided, the terminal evaluates the model's accuracy and the accuracy of emotion recognition, and instructs the model to retrain if performance improvement is needed. The retraining process uses H2O.ai to tune the model's hyperparameters.

[0591] As a result, users can view a dashboard on their smartphones that visualizes the analyzed sentiment data. This information is also provided to smart city administrators, making it possible, for example, to prioritize the provision of relevant services or events in areas with a high concentration of stressed residents.

[0592] For example, if data is analyzed showing that a large number of residents are experiencing stress during a specific time period, the government can then hold relaxation-related events during that time. Another example of a prompt for using such a generative AI model is, "Citizens in a certain area are experiencing stress; please suggest solutions."

[0593] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0594] Step 1:

[0595] The server acquires video, audio, and text data in real time from the user's smartphone or digital device. Input is raw data sent from the device, and output is raw data temporarily stored on the server's storage. This data is streamed using Kafka.

[0596] Step 2:

[0597] The server preprocesses the acquired raw data. Specifically, video data is denoised and adjusted to standard resolution using OpenCV. Text data is tokenized and irrelevant words are removed using the Python NLTK library. For audio data, librosa is used to extract features necessary for sentiment recognition. The input is raw data, and the output is cleaned-up data ready for analysis.

[0598] Step 3:

[0599] The server trains a machine learning model built with TensorFlow based on preprocessed data. The input is the preprocessed dataset, and the output is the trained model with updated sentiment recognition. This gives the server the ability to analyze data patterns and sentiment states in an integrated manner.

[0600] Step 4:

[0601] The device evaluates the model using new data. Here, it checks the model's prediction accuracy and sentiment recognition accuracy and sends feedback to the server based on these results. The input is new data for evaluation, and the output is the model's evaluation result. The device collects evaluation feedback and instructs retraining if necessary.

[0602] Step 5:

[0603] Based on the evaluation results, the server retrains the model using H2O.ai if necessary, particularly optimizing the hyperparameters. The input consists of evaluation feedback and existing model data, and the output is the retrained model with improved accuracy.

[0604] Step 6:

[0605] Users view the analyzed sentiment data as a visualized dashboard on their devices. Specifically, they can use tools like Power BI to view their emotional state in real time using graphs and charts. The input is the analysis results from an improved model, and the output is visualized sentiment data. This information can be supplied to citizens and smart city administrators and used to generate prompts for AI models, such as "Citizens in a certain area are experiencing stress; please suggest solutions."

[0606] 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.

[0607] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0608] 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.

[0609] [Fourth Embodiment]

[0610] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0611] 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.

[0612] 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).

[0613] 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.

[0614] 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.

[0615] 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).

[0616] 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.

[0617] 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.

[0618] 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.

[0619] 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.

[0620] 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.

[0621] 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.

[0622] 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".

[0623] The system of this invention is composed of multiple components that work together to automate everything from data acquisition and processing to model training and application. The operation of each component is described below.

[0624] First, the server is responsible for acquiring data from external data sources. This includes video and text data from surveillance cameras, sensors, and databases on the internet. The acquired data is temporarily stored and then pre-processed.

[0625] The server removes noise associated with the data during the preprocessing stage and standardizes the data format. For example, for video data, it standardizes the frame rate and resolution, and for text data, it performs tokenization and removes stop words.

[0626] Next, the server trains a machine learning model using the preprocessed data. The model learns using both video and text data to improve its ability to extract patterns and features. This model is designed to handle diverse data formats in an integrated manner.

[0627] After training is complete, the device evaluates the model's performance using new data. The evaluation uses accuracy metrics that indicate how well the model performs. Users can receive these evaluation results and check the system's adaptability and efficiency.

[0628] If the evaluation results do not meet expectations, the server initiates a retraining process. This adjusts the model's parameters and architecture based on the evaluation results, iteratively improving the model.

[0629] Ultimately, the trained AI models are implemented in various applications. For example, in security systems, they automatically detect suspicious behavior and issue alerts. In the medical field, they analyze image data to assist in diagnosis. In this way, users can utilize the developed systems for a wide range of purposes.

[0630] As a concrete example, in the entertainment industry, content recommendations can be made based on viewer preferences through video content analysis. Users can receive customized content suggestions based on their individual viewing history.

[0631] Thus, the system of the present invention enables efficient operation in a variety of application fields by automating the entire data processing process.

[0632] The following describes the processing flow.

[0633] Step 1:

[0634] The server retrieves video and text data from external data sources. This includes various data formats such as surveillance cameras, online databases, and APIs. The retrieved data is temporarily stored in storage for later processing.

[0635] Step 2:

[0636] The server performs preprocessing on the acquired data. Image data undergoes noise reduction and resolution adjustment, while text data is subjected to tokenization and stop word removal. This prepares the data for analysis.

[0637] Step 3:

[0638] The server starts training a machine learning model using the pre-processed data. It selects an appropriate algorithm and executes the training process so that the model can learn specific patterns and features using the data.

[0639] Step 4:

[0640] The device evaluates the accuracy of the trained model. This evaluation uses metrics such as accuracy, recall, and F1 score, which indicate how well the model can respond to new data.

[0641] Step 5:

[0642] The user reviews the evaluation results of the model sent from their device. This allows them to determine whether the model meets practical standards or requires further adjustments.

[0643] Step 6:

[0644] If the evaluation results show that the accuracy does not meet the target, the server will perform a retraining process. This involves adjusting the model's hyperparameters and retraining it on a different dataset to improve the model's accuracy.

[0645] Step 7:

[0646] Ultimately, if the training and evaluation yield satisfactory results, users apply the model to real-world applications. They analyze data in real time and utilize it in various application areas, such as security systems and medical diagnostic support systems.

[0647] (Example 1)

[0648] 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".

[0649] Currently, many fields require the efficient and automated execution of a series of processes, from information gathering and processing to model training, evaluation, and retraining. However, conventional systems suffer from insufficient coordination between individual processing steps, and processing performance is reduced due to differences in data formats and noise. Therefore, there is a need for a versatile system that can solve these problems all at once with high efficiency.

[0650] 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.

[0651] In this invention, the server includes means for collecting information, means for performing preliminary processing on the collected information, and means for training a machine learning algorithm. This makes it possible to efficiently collect and process various forms of data and build highly accurate machine learning models.

[0652] "Information" refers to all kinds of data obtained from a data source, including, for example, images, text, and numerical data.

[0653] "Means of collection" refers to methods and techniques for obtaining the desired information from external data sources.

[0654] "Pre-processing" refers to preparatory operations such as removing noise from acquired information and standardizing data formats.

[0655] A "machine learning algorithm" refers to a mathematical model or program that learns patterns based on given data and uses that learning to make decisions or predictions.

[0656] "Training" refers to the process of providing data to a machine learning algorithm to allow it to learn and improve its performance.

[0657] "Means of measuring performance" refers to methods for evaluating the predictive accuracy and reproducibility of a trained algorithm.

[0658] "Retraining" refers to the process of learning the algorithm again based on the feedback received after the initial training.

[0659] "Means of executing results" refers to methods for making decisions and taking actions based on collected information using pre-trained models.

[0660] The system in this invention is primarily composed of three entities working together: a server, a terminal, and a user. First, the server is responsible for collecting information from external data sources. Specifically, it acquires image and text data from monitoring devices, sensors, and network databases. This allows for real-time information collection and support for diverse data formats.

[0661] Subsequently, the server performs preliminary processing on the acquired data. The software used here includes natural language processing tools such as NLTK and spaCy, as well as image processing libraries such as OpenCV. Specifically, noise reduction and format standardization are performed, and the data is handed over to the next stage in a high-quality state.

[0662] Next, the server is trained using machine learning algorithms. Frameworks such as TensorFlow and PyTorch are used to build a neural network, which is then trained on pre-processed data. This model is designed to identify diverse patterns and features of information, enabling highly accurate predictions.

[0663] After training, the device measures the model's performance using new data. In this step, the model's prediction accuracy and recall are evaluated and the results are presented in an easily understandable format for the user by quantifying them.

[0664] The user reviews these measurement results, and the server retrains the model if necessary. Retraining involves adjusting parameters and utilizing additional data to improve the model's performance. This process allows the model to gain the ability to make more accurate and reliable predictions.

[0665] Finally, the trained AI model can be used in a variety of application fields. Specific examples include detecting suspicious behavior in security systems and assisting in diagnosis in the medical field. In the entertainment industry, it can be used to recommend content based on viewing data. An example of a prompt would be, "Create a list of movies tailored to the user's preferences." In this way, the present invention utilizes a generative AI model to process and apply data quickly and automatically, thereby achieving efficient system operation.

[0666] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0667] Step 1:

[0668] The server collects information from external data sources. Inputs include image data from monitoring devices and sensors, and text data obtained from network APIs. In terms of operation, the server initiates the data collection process periodically according to a set schedule. The output is the raw, collected data.

[0669] Step 2:

[0670] The server performs preliminary processing on the collected data. The input is the raw data obtained in step 1, and the output is the processed, clean data. Specifically, for image data, OpenCV is used to adjust the resolution and remove noise, and for text data, NLTK and spaCy are used to tokenize and remove stop words.

[0671] Step 3:

[0672] The server trains a machine learning algorithm using preprocessed data. The input is the clean data obtained in step 2, and the output is the trained model. Specifically, it builds a neural network using TensorFlow or PyTorch and learns features from the data.

[0673] Step 4:

[0674] The device measures the performance of the trained model using newly collected data. The input is the model obtained in step 3 and the new evaluation data, and the output is the evaluation metric (e.g., accuracy and recall). Specifically, it runs an evaluation script to check whether the model's predictions are accurate.

[0675] Step 5:

[0676] The user reviews the evaluation results, and the server retrains the model if necessary. The input is the evaluation results from step 4, and the output is the improved model. In practice, the model's hyperparameters are adjusted, and new data is added for further training.

[0677] Step 6:

[0678] Ultimately, users utilize the trained AI model in real-world applications. Inputs are user prompts or new use cases, while outputs are results and recommendations tailored to the user's needs. For example, based on the prompt "Create a list of movies tailored to the user's preferences," the system generates a customized list of movie recommendations.

[0679] (Application Example 1)

[0680] 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".

[0681] In today's information-saturated world, efficiently providing information tailored to individual user preferences is challenging. In particular, content distribution services need to accurately analyze users' viewing history and preferences and automate personalized recommendations. Traditional methods have struggled to provide optimal information to individual users quickly and effectively.

[0682] 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.

[0683] In this invention, the server includes means for acquiring data, means for preprocessing data, means for training a machine learning model, means for evaluating the accuracy of the model, means for retraining the model, means for applying the results based on the provided data, means for recommending information based on viewing history, and means for presenting information to the user. This enables automatic and effective recommendation of information based on each user's preferences.

[0684] "Means for acquiring data" refers to devices or methods for collecting information from external data sources.

[0685] "Means for preprocessing" refers to a processing device or method for removing noise or standardizing the format of acquired data.

[0686] "Means for training a machine learning model" refers to an apparatus or method for training a machine learning algorithm using preprocessed data.

[0687] "Means for evaluating the accuracy of a model" refers to a device or method for verifying the performance of a trained machine learning model with new data.

[0688] "Means for retraining" refers to a device or method for adjusting parameters based on the accuracy evaluation of the model and repeating the training process.

[0689] "Means for applying results based on provided data" refers to devices or methods for applying learned knowledge to real-world applications.

[0690] "Means for recommending information based on viewing history" refers to a device or method for analyzing a user's past viewing data and presenting relevant information.

[0691] "Means of presenting information to users" refers to a device or method for displaying or notifying users of recommended information.

[0692] The embodiments for carrying out the present invention will now be described. This system is a comprehensive data processing system equipped with the functions of data acquisition, preprocessing, model training, accuracy evaluation, retraining, information recommendation, and result presentation. The specific configuration and operation of the system are described below.

[0693] The server first collects information from external data sources. Specifically, it retrieves video and text data from the internet and local databases. Programming languages ​​such as Python can be used for this process, and scripts are applied to automate data collection.

[0694] Next, the server performs preprocessing on the acquired data, such as noise reduction and formatting standardization. For example, in the case of image data, it standardizes the resolution, and in the case of text data, it uses Natural Language Processing (NLP) techniques to remove unnecessary words.

[0695] Next, the server trains a machine learning model based on the pre-processed data. During this process, it uses machine learning frameworks such as TensorFlow and PyTorch to execute algorithms like collaborative filtering and deep learning, analyzing the user's viewing history and interests.

[0696] The model's accuracy is evaluated on the device using new data, and retraining is performed as needed. This process improves the model's accuracy by adjusting hyperparameters.

[0697] Based on learned knowledge, the system generates recommendations for users based on their viewing history and presents them via smartphones or smart glasses. For example, a user who has watched a lot of cooking-related content in the past will be recommended newly released cooking shows.

[0698] As a concrete example, by entering a prompt such as, "Please suggest cooking videos I want to watch next based on my viewing history," the generating AI model selects videos that match the user's preferences and provides the results. This allows users to quickly obtain information based on their interests.

[0699] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0700] Step 1:

[0701] The server acquires video and text data from external data sources. The input for this data acquisition process is the internet or a local database. The server uses Python scripts to automate data collection and temporarily store the acquired data. The output is the stored raw data.

[0702] Step 2:

[0703] The server performs preprocessing on the acquired data, including noise reduction and formatting standardization. The input for this step is raw data. Data processing includes resolution standardization for image data and removal of unnecessary words for text data. Text is cleaned up using NLP techniques and standardized as a dataset. The output is preprocessed data.

[0704] Step 3:

[0705] The server trains a machine learning model using preprocessed data. The input for this step is a cleaned-up dataset. The server uses TensorFlow or PyTorch to apply collaborative filtering algorithms to train the model. The resulting output is a trained machine learning model.

[0706] Step 4:

[0707] The device evaluates the model's accuracy using newly acquired data. The input is the new test data. The device measures the model's prediction accuracy and evaluates the result as an accuracy metric. The output is the evaluation result.

[0708] Step 5:

[0709] The server initiates the retraining process as needed. The input for this step is the evaluation results. Based on the evaluation, the server adjusts the hyperparameters and retrains the model with necessary data processing. The output is a more accurate, retrained model.

[0710] Step 6:

[0711] The server generates recommendation information based on the user's viewing history. The inputs are the model and the viewing history. Using the generative AI model, it selects the most suitable information for the user and generates prompt messages. The output is the recommendation information.

[0712] Step 7:

[0713] Users receive recommendations through their smartphones or smart glasses. The input is the generated recommendations. The user's device displays the information, presenting content tailored to their personal preferences. The output is the presented information.

[0714] 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.

[0715] This invention combines a system that handles everything from data acquisition and preprocessing to training, evaluation, and retraining of machine learning models, and the application of the results, with an emotion engine that recognizes user emotions. Specific embodiments of this invention are described below.

[0716] First, the server acquires video, text, and audio data from external sources. The types of data acquired vary depending on the application, but are primarily data related to user interaction. The acquired data is temporarily stored in storage.

[0717] Next, the server performs preprocessing on this data. For image data, it performs resolution standardization and noise reduction; for text data, it performs tokenization and removal of irrelevant words; and for audio data, it performs noise filtering and feature extraction for sentiment recognition.

[0718] Furthermore, the server utilizes the pre-processed data to train machine learning models. These models are designed to learn not only patterns in the data but also emotional information obtained from the emotion engine.

[0719] The trained model is evaluated using new data via the device. This evaluation process verifies not only the model's predictive accuracy but also its emotion recognition accuracy. The user reviews the evaluation results from the device and provides instructions for adjustments to improve the system as needed.

[0720] If necessary, the server will retrain and update the system to improve weaknesses identified in the previous evaluation. The retraining process will focus particularly on optimizing hyperparameters in sentiment analysis.

[0721] Ultimately, if the model's performance meets the criteria, users will put the system into practical use. For example, in online meeting systems, providing real-time sentiment analysis results during meetings allows for facilitation tailored to the emotional state of participants. In the field of education, teachers can efficiently understand students' emotional states, supporting improvements in the quality of individualized instruction.

[0722] Thus, this invention enables comprehensive analysis and application of diverse user data, including emotions, and is expected to be used in a wide range of fields, from business settings to daily life.

[0723] The following describes the processing flow.

[0724] Step 1:

[0725] The server acquires data from surveillance cameras, microphones, and text input devices. This includes video, audio, and text data. The acquired data is stored in temporary storage before processing.

[0726] Step 2:

[0727] The server receives video data and performs preprocessing such as adjusting the frame rate and removing noise. For audio data, noise filtering and speech feature extraction for sentiment analysis are performed. Text data is cleaned through tokenization and removal of unnecessary words.

[0728] Step 3:

[0729] The server trains a machine learning model using pre-processed data. Here, it learns multidimensional patterns, including user emotions, based on integrated information from video, audio, and text. During the training process, an emotion engine is incorporated so that the model can recognize various emotional states.

[0730] Step 4:

[0731] The device feeds new data to the trained model and evaluates the model's accuracy and performance. This evaluation process uses metrics such as confusion matrix, accuracy, recall, and F1 score to check the accuracy of sentiment recognition.

[0732] Step 5:

[0733] The user reviews the evaluation results received from the terminal and determines whether the system's performance has reached a practical level. Based on the evaluation, they provide feedback to the server for further functional improvements if necessary.

[0734] Step 6:

[0735] Based on the evaluation results, the server retrains the model as needed. This retraining process includes adjusting hyperparameters and selecting new datasets to improve the accuracy of emotion recognition.

[0736] Step 7:

[0737] After the model's performance meets the required standards through retraining, users can utilize the model in their daily work. For example, in customer service, the model can analyze the user's emotional state in real time and optimize service responses. This enables efficient and responsive customer support.

[0738] (Example 2)

[0739] 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".

[0740] Conventional information analysis systems lacked the means to perform consistent analysis in environments where multiple information formats coexisted. Furthermore, the effective utilization of acquired information and the retraining processes for improving machine learning accuracy were not adequately developed. As a result, efficiency and accuracy were compromised throughout the process from information acquisition to application.

[0741] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0742] In this invention, the server includes means for acquiring information from an external source, means for performing initial processing based on the acquired information, and means for training a learning algorithm using the initially processed information. This enables the effective integration of multiple forms of information, improving the accuracy of machine learning and expanding its range of applications.

[0743] "Means of acquiring information from external sources" refers to methods comprised of processes and technologies that a system uses to collect data from external information sources.

[0744] "Means of performing initial processing based on acquired information" refers to processes such as noise reduction, standardization, and feature extraction in order to convert the collected data into a format that can be analyzed and learned from.

[0745] "Means of training a learning algorithm using pre-processed information" refers to the process of building or improving a machine learning model using processed data to enhance its pattern recognition and predictive capabilities.

[0746] "Means of evaluating performance" are methods or metrics used to measure how accurately a trained machine learning model operates based on input data.

[0747] A "means of retraining" refers to a mechanism that, based on evaluation results, reviews the training dataset and parameter settings to improve the model's performance, and then runs the training process again.

[0748] "Methods for utilizing conclusions" refer to the process of using analytical results and predictive information to take concrete actions and make decisions in real-world application situations.

[0749] This invention is a system that consistently handles everything from data acquisition and initial processing to training, evaluation, retraining, and application of machine learning models. In addition, it is equipped with an emotion engine for comprehensive analysis, including sentiment analysis. This system is realized through the cooperation of the server, terminal, and user elements.

[0750] First, the server obtains necessary information from external sources. This includes using APIs via the internet and accessing various databases. The data primarily consists of video, text, and audio, and this information is temporarily stored in the server's internal storage.

[0751] The acquired data is then initially processed on the server. Specifically, image data is subjected to resolution standardization and noise reduction using an image processing library. Text data is processed using a natural language processing library, such as NLTK, for tokenization and stop word removal. Audio data is subjected to noise filtering and feature extraction, such as MFCC, using a speech analysis library.

[0752] The server then uses this pre-processed data to train a learning algorithm. This process uses a machine learning framework, such as TensorFlow, to train a model for pattern recognition. A sentiment engine is also built in, and sentiment patterns are learned simultaneously at this stage.

[0753] Once the model is trained, the device evaluates its performance with new data. A machine learning evaluation library is used to measure how accurately the model makes predictions and recognizes sentiments. The results are visualized and reported to the user.

[0754] Users can review the evaluation results reported from their terminals and, if they determine that system improvements are needed, they can instruct the server to retrain the system. Retraining is performed to address weaknesses highlighted in the evaluation and to adjust hyperparameters.

[0755] Ultimately, models that meet the criteria will be made available for application by users. A concrete example is real-time sentiment analysis in online meetings, enabling facilitation that takes into account the emotional state of meeting participants. In the field of education, it is expected that understanding students' emotional states in real time will improve the quality of individualized instruction.

[0756] An example of a prompt message would be, "Detect emotions from the user's voice data and suggest appropriate facilitation based on the emotional state of the meeting participants." This allows for the integrated analysis of diverse data, enabling its use in a variety of application scenarios.

[0757] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0758] Step 1:

[0759] The server acquires information from external sources. The input consists of three types: audio data, video data, and text data, collected using APIs and database access. The output is raw, unprocessed data temporarily stored on the server's internal storage. In this procedure, the data collection module retrieves data via API endpoints and streaming protocols.

[0760] Step 2:

[0761] The server performs initial processing on the acquired information. The input is the raw data obtained in step 1. For image data, an image processing library is used to perform resolution standardization and noise reduction. For text data, a natural language processing library is used for tokenization and stop word removal. For speech data, a speech analysis library is used for noise filtering and feature extraction, and processed data in a unified format is generated as output.

[0762] Step 3:

[0763] The server uses the pre-processed data to train the learning algorithm. The input is the unified format data obtained in step 2. A machine learning framework is used to train the pattern recognition model and the sentiment engine. In this process, the model is trained over multiple epochs using the training dataset, and the trained model is obtained as output.

[0764] Step 4:

[0765] The device evaluates the trained model with new data. The inputs are the test dataset and the trained model obtained in step 3. Evaluation uses metrics to check the model's predictive accuracy and sentiment recognition accuracy. The evaluation output is a model performance report, which is visualized and reported to the user.

[0766] Step 5:

[0767] The user reviews the evaluation results via their terminal and determines the need for retraining. The input is the performance report presented in step 4. If necessary, the user instructs the server to retrain, particularly by adjusting the hyperparameters. The output is an improved model.

[0768] Step 6:

[0769] The user ultimately applies the applicable model to real-world situations. The input consists of a performance-approved, trained model and real-time application data. For example, the system analyzes participants' emotions during an online meeting and adjusts the meeting's facilitation based on the results. The output of this step is improved meeting efficiency and enhanced instruction quality in educational settings.

[0770] (Application Example 2)

[0771] 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".

[0772] In smart cities, improving the quality of citizen services requires understanding residents' emotions and stress levels in real time and optimizing public services based on that information. However, conventional systems have the challenge of not being able to efficiently collect and analyze individual emotional data and immediately utilize it for service improvement.

[0773] 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.

[0774] In this invention, the server includes means for acquiring data, means for pre-processing data, and means for analyzing user emotions. This enables real-time analysis of residents' emotional states and the optimization of public services by utilizing the results.

[0775] "Means of acquiring data" refers to the function of carrying out the process of gathering necessary information from external sources.

[0776] "Means of preprocessing" refers to functions that remove unnecessary data and eliminate noise in order to improve the accuracy of acquired data.

[0777] "Methods for training machine learning models" refer to functions that train models to recognize certain patterns or features based on pre-processed data.

[0778] "Means for evaluating model accuracy" refers to functions that measure the predictive accuracy and generalization ability of machine learning models and determine their superiority or inferiority.

[0779] A "means of retraining" refers to a function that retrains a model based on evaluations to improve its performance.

[0780] "Means of applying results based on provided data" refers to a function that applies analysis results to realistic uses and situations to achieve optimized results.

[0781] "Means of analyzing user emotions" refers to a function that identifies and analyzes individual emotional states from data such as audio, video, and text.

[0782] "Means for visualizing the results of sentiment analysis" refers to a function that displays the analyzed sentiment information in an easy-to-understand manner using graphs, charts, etc.

[0783] "Means for optimizing public services based on analysis results" refers to a function that improves the content and methods of service provision to citizens based on information obtained from sentiment analysis.

[0784] This invention is a system for optimizing public services in smart cities using citizen sentiment data. The system first uses a server to acquire video, audio, and text data from smartphones and other devices. Streaming technologies such as Kafka are used for data acquisition to enable real-time processing.

[0785] The acquired data is cleaned up in the preprocessing step. Video data is denoised using OpenCV and adjusted to a standard resolution. Tokenization and removal of irrelevant words from text data are performed using the Python NLTK library. Furthermore, features are extracted from audio data using librosa in preparation for sentiment analysis.

[0786] The server uses this preprocessed data to train a machine learning model built with TensorFlow. This model is designed to comprehensively recognize not only data patterns but also the user's emotional state. Each time new data is provided, the terminal evaluates the model's accuracy and the accuracy of emotion recognition, and instructs the model to retrain if performance improvement is needed. The retraining process uses H2O.ai to tune the model's hyperparameters.

[0787] As a result, users can view a dashboard on their smartphones that visualizes the analyzed sentiment data. This information is also provided to smart city administrators, making it possible, for example, to prioritize the provision of relevant services or events in areas with a high concentration of stressed residents.

[0788] For example, if data is analyzed showing that a large number of residents are experiencing stress during a specific time period, the government can then hold relaxation-related events during that time. Another example of a prompt for using such a generative AI model is, "Citizens in a certain area are experiencing stress; please suggest solutions."

[0789] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0790] Step 1:

[0791] The server acquires video, audio, and text data in real time from the user's smartphone or digital device. Input is raw data sent from the device, and output is raw data temporarily stored on the server's storage. This data is streamed using Kafka.

[0792] Step 2:

[0793] The server preprocesses the acquired raw data. Specifically, video data is denoised and adjusted to standard resolution using OpenCV. Text data is tokenized and irrelevant words are removed using the Python NLTK library. For audio data, librosa is used to extract features necessary for sentiment recognition. The input is raw data, and the output is cleaned-up data ready for analysis.

[0794] Step 3:

[0795] The server trains a machine learning model built with TensorFlow based on preprocessed data. The input is the preprocessed dataset, and the output is the trained model with updated sentiment recognition. This gives the server the ability to analyze data patterns and sentiment states in an integrated manner.

[0796] Step 4:

[0797] The device evaluates the model using new data. Here, it checks the model's prediction accuracy and sentiment recognition accuracy and sends feedback to the server based on these results. The input is new data for evaluation, and the output is the model's evaluation result. The device collects evaluation feedback and instructs retraining if necessary.

[0798] Step 5:

[0799] Based on the evaluation results, the server retrains the model using H2O.ai if necessary, particularly optimizing the hyperparameters. The input consists of evaluation feedback and existing model data, and the output is the retrained model with improved accuracy.

[0800] Step 6:

[0801] Users view the analyzed sentiment data as a visualized dashboard on their devices. Specifically, they can use tools like Power BI to view their emotional state in real time using graphs and charts. The input is the analysis results from an improved model, and the output is visualized sentiment data. This information can be supplied to citizens and smart city administrators and used to generate prompts for AI models, such as "Citizens in a certain area are experiencing stress; please suggest solutions."

[0802] 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.

[0803] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0804] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[0805] 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.

[0806] 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.

[0807] 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.

[0808] The inside of the Emotion Map 400 represents what's in your mind, while the outside represents what you're doing. Therefore, the further you go out the 400-coordinate scale, the more visible your emotions become (the more they manifest in your actions).

[0809] 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.

[0810] 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."

[0811] 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.

[0812] 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.

[0813] 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.

[0814] 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.

[0815] 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.

[0816] 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.

[0817] 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.

[0818] 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.

[0819] 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.

[0820] 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.

[0821] 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.

[0822] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[0823] The following is further disclosed regarding the embodiments described above.

[0824] (Claim 1)

[0825] Means of acquiring data,

[0826] Means for performing pre-processing,

[0827] Methods for training machine learning models,

[0828] A means of evaluating the accuracy of the model,

[0829] Methods for relearning,

[0830] A system that includes means for applying results based on the provided data.

[0831] (Claim 2)

[0832] The system according to claim 1, wherein the preprocessing means performs noise reduction on image data and text data.

[0833] (Claim 3)

[0834] The system according to claim 1, wherein the retraining means adjusts hyperparameters based on the evaluation results.

[0835] "Example 1"

[0836] (Claim 1)

[0837] Means of collecting information,

[0838] A means of performing preliminary processing on the collected information,

[0839] Methods for training machine learning algorithms,

[0840] A means of measuring the performance of an algorithm,

[0841] Methods for retraining,

[0842] A system that includes means for performing actions based on the information provided.

[0843] (Claim 2)

[0844] The system according to claim 1, wherein the pre-processing means removes noise from image information and text information.

[0845] (Claim 3)

[0846] The system according to claim 1, wherein the retraining means optimizes the setpoints based on the performance measurement results.

[0847] "Application Example 1"

[0848] (Claim 1)

[0849] Means of acquiring data,

[0850] Means for performing pre-processing,

[0851] Methods for training machine learning models,

[0852] A means of evaluating the accuracy of the model,

[0853] Methods for relearning,

[0854] Means for applying the results based on the provided data,

[0855] A means of recommending information based on viewing history,

[0856] A system that includes means of presenting information to users.

[0857] (Claim 2)

[0858] The system according to claim 1, wherein the preprocessing means performs noise reduction on image data and text data.

[0859] (Claim 3)

[0860] The system according to claim 1, wherein the retraining means adjusts hyperparameters based on evaluation results and utilizes viewing history.

[0861] "Example 2 of combining an emotion engine"

[0862] (Claim 1)

[0863] Means of obtaining information from external sources,

[0864] A means for performing initial processing based on acquired information,

[0865] A means of training a learning algorithm using initially processed information,

[0866] A means of evaluating the performance of a trained learning algorithm,

[0867] A means of repeating the learning process based on the evaluation results,

[0868] A system that includes means for utilizing conclusions based on the information provided.

[0869] (Claim 2)

[0870] The system according to claim 1, wherein the initial processing means performs noise reduction on visual and textual information.

[0871] (Claim 3)

[0872] The system according to claim 1, wherein the retraining means optimizes adjustable values ​​based on the evaluation results.

[0873] "Application example 2 when combining with an emotional engine"

[0874] (Claim 1)

[0875] Means of acquiring data,

[0876] Means for performing pre-processing,

[0877] Methods for training machine learning models,

[0878] A means of evaluating the accuracy of the model,

[0879] Methods for relearning,

[0880] Means for applying the results based on the provided data,

[0881] A means of analyzing user emotions,

[0882] A means of visualizing the results of emotion analysis,

[0883] A system that includes means for optimizing public services based on analysis results.

[0884] (Claim 2)

[0885] The system according to claim 1, wherein the preprocessing means performs noise reduction on image data and text data, and extracts feature quantities from audio data.

[0886] (Claim 3)

[0887] The system according to claim 1, wherein the retraining means adjusts hyperparameters and optimizes the sentiment analysis model based on the evaluation results. [Explanation of symbols]

[0888] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. Means of acquiring data, Means for performing pre-processing, Methods for training machine learning models, A means of evaluating the accuracy of the model, Methods for relearning, Means for applying the results based on the provided data, A means of recommending information based on viewing history, A system that includes means of presenting information to users.

2. The system according to claim 1, wherein the preprocessing means performs noise reduction on image data and text data.

3. The system according to claim 1, wherein the retraining means adjusts hyperparameters based on evaluation results and utilizes viewing history.

Images