system

Abstract

To provide a system.SOLUTION: A system includes: means for preparing a natural language processing (NLP) model on which basis learning has been performed in advance; means for collecting interaction on a portable terminal of a customer in a place where complete security is ensured; means for causing an NLP model only for the customer to relearn using the collected data; and means for recognizing a feeling of the user and adjusting the learning of the NLP model to correspond to a feeling analyzed by the feeling engine for analyzing.SELECTED DRAWING: Figure 1

Description

[Technical field]

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

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

[0003] [Patent Document 1] JP 2022-180282 A Summary of the Invention [Problem to be solved by the invention]

[0004] Currently, there is a demand to predict the thoughts and actions of individual customers and provide services based on those predictions. However, in order to accurately predict customer thoughts and actions, it is necessary to collect and analyze a large amount of behavioral data of individual customers, such as their interactions on their mobile phones. However, collecting such data may violate the privacy of customers, and security measures are also required to safely manage the data. [Means for solving the problem]

[0005] This invention prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on their mobile phone in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality. The NLP is then able to predict the customer's thoughts, and the customer is then able to "allow / reject" API connections to their own NLP from various apps around the world. This makes it possible to predict the customer's thoughts and actions, and provide services based on those predictions. It also ensures the protection of customer privacy and data security. [Brief description of the drawings]

[0006] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Diagram 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. FIG. [Diagram 3] FIG. 11 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Diagram 5] FIG. 13 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 13 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 13 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 2 is a sequence diagram showing a process flow of the data processing system in the first embodiment of the first form example. [Figure 12] FIG. 11 is a sequence diagram showing a process flow of the data processing system in the application example 1 of the first embodiment. [Figure 13] FIG. 11 is a sequence diagram showing a process flow of the data processing system in the second embodiment of the second embodiment. [Figure 14] FIG. 13 is a sequence diagram showing a process flow of the data processing system in application example 2 of embodiment 2. [Figure 15] FIG. 13 is a sequence diagram showing a process flow of a data processing system in a third embodiment of the present invention. [Figure 16] FIG. 13 is a sequence diagram showing a process flow of the data processing system in the third application example of the third embodiment. [Figure 17] FIG. 13 is a sequence diagram showing the flow of processing in the data processing system in the first embodiment of the first form example when an emotion engine is combined. [Figure 18] FIG. 13 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Form Example 1 when combined with an emotion engine. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] Hereinafter, an example of an embodiment of a system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0008] First, the terms used in the following description will be explained.

[0009] In the following embodiments, a signed processor (hereinafter simply referred to as a "processor") may be one arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be one type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), or a TPU (TENSOR PROCESSING UNIT (registered trademark)).

[0010] In the following embodiments, a signed RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by the processor.

[0011] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0012] In the following embodiments, a communication I / F (Interface) with a code is an interface including a communication processor and an antenna. The communication I / F controls communication between multiple computers. An example of a communication standard applied to the communication I / F is a wireless communication standard including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0013] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. In addition, in this specification, the same idea as "A and / or B" is also applied when three or more things are expressed by connecting them with "and / or."

[0014] [First embodiment]

[0015] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0016] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0017] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[0018] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0019] The reception device 38 includes a touch panel 38A and a microphone 38B, and receives user input. The touch panel 38A detects contact with a pointer (e.g., a pen or a finger) to receive user input by the touch of the pointer. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0020] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (e.g., voice 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 voice according to instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a Complementary Metal-Oxide-Semiconductor (CMOS) image sensor or a Charge Coupled Device (CCD) image sensor.

[0021] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54.

[0022] FIG. 2 shows an example of main functions of the data processing device 12 and the smart device 14.

[0023] As shown in Fig. 2, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32. The specific process program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific process program 56 from the storage 32, and executes the read specific process program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific process program 56 executed on the RAM 30.

[0024] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0025] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores a reception output program 60. The reception output program 60 is used together with the specific processing program 56 by the data processing system 10. The processor 46 reads out 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 the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0026] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0027] "Example 1"

[0028] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[0029] "Example 2"

[0030] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[0031] The process flow of each embodiment will be described below.

[0032] "Example 1"

[0033] Step 1: Prepare a pre-trained natural language processing (NLP) model. This NLP model is trained, for example, from a large amount of text data.

[0034] Step 2: In a fully secured location, enter all of the customer's mobile phone interactions, including, for example, the customer's emails, social media posts, search history, etc.

[0035] Step 3: Using this data, we build a personalized NLP that is trained on that person's personality and has the ability to predict what the customer will think and do.

[0036] Step 4: Allow customers to "allow / deny" API connections to their NLP from various apps around the world, giving customers the freedom to control which apps can use their data.

[0037] "Example 2"

[0038] Step 1: Prepare a pre-trained natural language processing (NLP) model. This NLP model is trained, for example, from a large amount of text data.

[0039] Step 2: In a fully secured location, enter all of the customer's mobile phone interactions, including, for example, the customer's emails, social media posts, search history, etc.

[0040] Step 3: Using this data, we build a personalized NLP that is trained on that person's personality and has the ability to predict what the customer will think and do.

[0041] Step 4: Allow customers to "allow / deny" API connections to their NLP from various apps around the world, giving customers the freedom to control which apps can use their data.

[0042] Example 1

[0043] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal".

[0044] Conventional natural language processing (NLP) systems have difficulty building personalized models that reflect the characteristics and behavior of individual customers. In addition, there was no established method for safely collecting customer data and using it while protecting privacy. Furthermore, there was a lack of means for customers to freely control which applications could use their data.

[0045] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0046] In this invention, the server includes a means for preparing a natural language processing (NLP) model that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure location and constructing an NLP model for each customer that has been trained with that person's personality, a means for enabling the NLP model to predict the customer's thoughts and actions, a means for collecting customer data, encrypting it and transmitting it to the server, a means for the server to fine-tune the NLP model using the customer data, and a means for the terminal to provide an interface for managing the customer's API connection settings. This makes it possible to construct a personalized NLP model that reflects the characteristics of the customer, safely collect and use customer data, and enable the customer to freely control the use of their own data.

[0047] A "natural language processing (NLP) model" is a machine learning model that is trained based on large amounts of text data and has the ability to understand and generate natural language.

[0048] "Complete security" means that advanced security measures are in place to prevent unauthorized access and information leaks during data collection, transmission, and storage.

[0049] "Communication equipment" refers to devices used by customers, such as mobile phones and tablets, through which data is entered and transmitted.

[0050] "Interactions" refers to all communication activities conducted by customers through communication devices, such as emails, social media posts, search history, etc.

[0051] A "personality-trained NLP model for each customer" is an NLP model trained to reflect the characteristics and behavioral patterns of each individual customer based on their communication activity data.

[0052] An "API connection" is a method of connecting different software to exchange data and functions through an application program interface.

[0053] "Allow / Deny" refers to the act of a customer choosing whether to allow or deny a particular application access to their NLP models.

[0054] "Data collection" is the process of obtaining interaction data from a customer's communications device.

[0055] "Encryption" is a security technology that converts data into a form that cannot be deciphered by third parties.

[0056] "Fine-tuning" is the process of fine-tuning an existing base model to a specific dataset to improve its performance.

[0057] An "interface" refers to the screen or operating means that allows the user to interact with the system, and in this case refers to the means used to set whether to allow or deny API connections.

[0058] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs all of the customer's interactions on communication devices, and builds an NLP model for each customer that is trained based on that person's personality. Furthermore, it enables this NLP model to predict the customer's thoughts and actions, and enables the customer to "allow / reject" API connections to their NLP model from various applications in the world.

[0059] Hardware and Software Used

[0060] server

[0061] The server uses machine learning frameworks such as TENSORFLOW® and PyTorch to train NLP models based on large amounts of text data. The models use advanced natural language processing techniques such as BERT and GPT-3®. The server saves the trained models and prepares them for later use.

[0062] Terminal

[0063] The device installs a dedicated app that allows the user to input interactions on their communication device. With the user's permission, the app collects data such as emails, social media posts, and search history. The collected data is encrypted and sent to a server. The device provides an interface for the user to manage API connections to their NLP models.

[0064] User

[0065] The user uses a dedicated app to input interactions on their communication device. The user's communication device transmits this data to the server. Through the device interface, the user can allow or deny the API connection to their NLP model.

[0066] Data processing and calculation

[0067] The server uses the received customer data to build an NLP model specific to that customer. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data preprocessing, model training, and evaluation. For example, the server analyzes the content of a customer's emails to learn what topics the customer is interested in.

[0068] Examples

[0069] For example, if a user allows their NLP model to be connected when using a shopping app, the app can recommend products that the user might be interested in based on the user's past purchase history and search history. On the other hand, they can protect their privacy by refusing to connect to SNS apps.

[0070] Examples of prompt statements

[0071] Examples of prompt sentences to be input to a generative AI model include the following:

[0072] "Predict the next topic that a user will be interested in based on their past emails and social media posts."

[0073] Using this prompt, the generative AI model can predict the user's interests and provide appropriate information.

[0074] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0075] Step 1:

[0076] The server prepares a natural language processing (NLP) model that has been pre-trained. Specifically, the server uses machine learning frameworks such as TensorFlow and PyTorch to train an NLP model based on a large amount of text data. This model utilizes advanced natural language processing techniques such as BERT and GPT-3. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[0077] Step 2:

[0078] A user installs a dedicated app to input communications on their communication device. With the user's permission, this app collects data such as emails, SNS posts, and search history. In concrete terms, the app obtains this data from the user's communication device, encrypts it, and sends it to a server. The input required is the user's communication data, and the output is the encrypted data sent to the server.

[0079] Step 3:

[0080] The server uses the received customer data to build a customer-specific NLP model. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data pre-processing, model training, and evaluation. As input, encrypted customer data is required, and as output, a customer-specific NLP model is obtained.

[0081] Step 4:

[0082] The terminal provides an interface for the user to manage API connections to their NLP models. In concrete terms, the terminal allows the user to select which applications are allowed or denied to use the NLP model. The terminal requires the user's selection information as input, and obtains the permission / denial setting for API connections as output.

[0083] Step 5:

[0084] The server controls the access of each application based on the user's API connection settings. Specifically, the server receives the user's setting information and allows or denies API access based on that information. The input is the user's API connection settings, and the output is access control for each application.

[0085] Step 6:

[0086] A user uses his / her own NLP model through an authorized application. Specifically, the user receives personalized services using his / her own NLP model in a shopping application, information providing application, etc. The input is a request from the application, and the output is the prediction result and recommendation information by the NLP model.

[0087] (Application example 1)

[0088] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0089] Conventional online shopping sites do not adequately recommend products based on customers' interests and purchase history, making it difficult to timely suggest products that meet customers' needs. In addition, there is a lack of systems that can respond to individual needs while protecting customers' privacy.

[0090] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means. In this invention, the server includes a means for preparing natural language processing (NLP) that has been previously trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that is trained with the personality of that person, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to his / her NLP from various applications in the world, a means for using the customer's NLP to recommend products based on the customer's interests and purchase history, and a means for predicting the customer's behavioral patterns and making timely product suggestions. This makes it possible to timely suggest products that meet the customer's needs.

[0091] "Natural Language Processing (NLP)" is a technology that uses algorithms trained from large amounts of text data to understand and generate human language.

[0092] "Full security" means that advanced protective measures are in place to prevent data leakage and unauthorized access.

[0093] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[0094] "Interaction" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile devices.

[0095] "Personality-trained NLP for each customer" is an NLP model that learns each individual customer's thinking and behavior patterns based on their communication activities.

[0096] An "API connection" is an interface for exchanging data and functions between different software programs.

[0097] "Allow / Deny" is the act of a Customer choosing to allow or deny an API connection to their NLP.

[0098] "Means of product recommendation" is a function that uses customer NLP to suggest products based on the customer's interests and purchasing history.

[0099] A "means for predicting behavioral patterns" is a function that predicts future behavior based on a customer's past behavioral data.

[0100] "Timely product proposals" refers to proposing products that meet customer needs at the right time.

[0101] The system for implementing this invention has the following configuration. First, the server prepares natural language processing (NLP) that has undergone base learning in advance. Next, in a completely secure location, all interactions on the customer's mobile device (emails, SNS posts, search history, etc.) are entered, and an NLP for each customer that is trained on that person's personality is constructed. This NLP is designed to be able to predict the customer's thoughts.

[0102] Customers can "allow / deny" API connections to their NLP to various applications in the world, giving them the freedom to control which applications can use their data.

[0103] In addition, the server uses NLP to recommend products based on the customer's interests and purchase history, making it possible to predict customer behavior patterns and provide timely product suggestions.

[0104] Hardware and Software Used

[0105] Hardware: Servers (for data processing), smartphones (for collecting customer data)

[0106] Software: Python, scikit-learn (machine learning library), TfidfVectorizer (vectorization of text data)

[0107] Data processing and calculation

[0108] 1. Data collection:

[0109] The server collects customer emails, social media posts, and search history, and stores this data in JSON format.

[0110] 2. Building the NLP model:

[0111] The server vectorizes the text data using TfidfVectorizer, and uses the vectorized data to build the client's NLP model.

[0112] 3. Product recommendation:

[0113] The server compares the customer's NLP model with the product list, calculates the cosine similarity, and recommends products with high similarity.

[0114] Examples

[0115] For example, if customer A mentions in an email that he / she "want a new smartphone," posts on social media that he / she is "interested in the latest gadgets," and his / her search history includes "smartphone reviews," the server will use this data to build an NLP model for customer A. The server will then use this NLP model to recommend the latest smartphones, smartwatches, gadget accessories, etc.

[0116] Examples of prompt statements

[0117] Collect customer emails, social media posts, and search history, and vectorize them using TfidfVectorizer. Then build an NLP model of the customer and compare it with the product list to calculate the cosine similarity. Recommend products with high similarity.

[0118] In this way, a personalized shopping assistant can be realized that utilizes individual NLP for each customer.

[0119] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0120] Step 1:

[0121] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and has basic language understanding capabilities. The input is the pre-trained NLP model, and the output is a usable NLP model.

[0122] Step 2:

[0123] The server, located in a completely secure location, collects all of the customer's mobile device interactions (emails, social media posts, search history, etc.). This data is encrypted and stored to protect the customer's privacy. The input is the customer's communication data, and the output is encrypted customer data.

[0124] Step 3:

[0125] The server uses the collected customer data to build an NLP model for each customer that is trained on that person's personality. Specifically, it uses TfidfVectorizer to vectorize the text data and generate an NLP model for the customer. The input is the encrypted customer data, and the output is an NLP model for each customer.

[0126] Step 4:

[0127] The server uses the customer's NLP model to predict the customer's thoughts and actions. To do this, it applies an algorithm that predicts future behavior based on past data. The input is the customer's NLP model, and the output is the predicted customer behavior pattern.

[0128] Step 5:

[0129] Users can "allow / deny" API connections to their NLP models from various applications in the world. The server controls API connections based on user permissions. The input is the user's permission information, and the output is a list of allowed applications.

[0130] Step 6:

[0131] The server uses the customer's NLP model to recommend products based on the customer's interests and purchase history. Specifically, it compares the customer's NLP model with the product list and calculates the cosine similarity. The inputs are the customer's NLP model and the product list, and the output is the recommended product list.

[0132] Step 7:

[0133] The server predicts customer behavior patterns and makes timely product suggestions. This makes it possible to suggest products that meet customer needs at the right time. The inputs are the predicted customer behavior patterns and a product list, and the output is a timely suggested product list.

[0134] Example 2

[0135] Next, a second embodiment of the second embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0136] Conventional natural language processing (NLP) systems have difficulty accurately predicting the thoughts and behavior of individual users, and lack the means to freely control which applications can use user data. As a result, there is a high risk of violating user privacy and a poor user experience.

[0137] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0138] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for collecting customer data and constructing an individual NLP model, and a means for the customer to manage API connections using a dedicated management application. This makes it possible to accurately predict the user's thoughts and actions and freely control which applications can use the user's data.

[0139] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0140] "Base learning" is the process of training a basic model using a large amount of data in advance.

[0141] "Complete Security" refers to the highest level of protection to prevent unauthorized access or leakage of data.

[0142] A "communication device" is a device used by a user, such as a mobile phone or tablet.

[0143] "Interaction" refers to activities such as emails, social media posts, and search history that users conduct through communication devices.

[0144] "Personality-trained NLP for each customer" is an NLP model that learns the characteristics and behavioral patterns of each individual customer based on their data.

[0145] An "API connection" is an interface that allows an application to communicate with other software or services.

[0146] "Allow / deny" is the act of a user choosing whether to allow or deny a particular application access to their data.

[0147] "Management Application" means dedicated software that allows a User to manage API connections to their NLP models.

[0148] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on communication devices in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality. This system makes it possible to predict the customer's thoughts, and enables the customer to "allow / reject" API connections to their own NLP from various applications in the world.

[0149] Hardware and Software Used

[0150] server

[0151] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. This NLP model uses a generative AI model such as GPT-4 (registered trademark). The server is responsible for collecting customer data and building individual NLP models.

[0152] Terminal

[0153] The terminal is a communication device (such as a mobile phone or tablet) used by the user. The terminal collects data such as the user's emails, SNS posts, search history, etc., and transmits it to the server.

[0154] User

[0155] Users enter their interactions on their communication devices and manage API connections to their NLP models using a dedicated management application, and users choose to allow or deny specific applications access to their data.

[0156] Data processing and calculation

[0157] The server analyzes the collected customer data and builds an individual NLP model. Specifically, the customer data is tokenized and input into the base NLP model. The model learns the customer's language patterns and behavioral patterns and generates an individual NLP model. This model has the ability to predict the customer's thoughts and behavior.

[0158] Examples

[0159] For example, if a user is using a social networking app, the app needs the user's permission to access the user's NLP model. If the user gives permission, the social networking app can provide content tailored to the user's interests based on the user's past posts and search history.

[0160] Examples of prompt statements

[0161] "Recommend new movies that you think might interest me based on my past social media posts and search history."

[0162] When this prompt is input into a generative AI model, the model analyzes the user's past data and recommends movies that might interest them.

[0163] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0164] Step 1:

[0165] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, it uses a generative AI model such as GPT-4 to analyze and tokenize the text data and train the model. The input of this process is a large amount of text data, and the output is a base NLP model.

[0166] Step 2:

[0167] A user inputs interactions on their communication device. Specifically, they open an email app, select past emails, and send them to the server. They also open a social networking app, select past posts and comments, and send them to the server. They also export their browser search history and send it to the server. The input to this process is the data on the user's communication device, and the output is customer data sent to the server.

[0168] Step 3:

[0169] The server analyzes the collected customer data and builds an individual NLP model. Specifically, it tokenizes the customer data and inputs it into the base NLP model. The model learns the customer's language and behavioral patterns and generates an individual NLP model. The input of this process is the customer data and the output is an individual NLP model.

[0170] Step 4:

[0171] Users manage API connections to their NLP models using a dedicated management application: they open the management application and choose whether to allow or deny a particular application access to their data. The input to this process is the user's choice, and the output is the setting to allow or deny the API connection.

[0172] Step 5:

[0173] The server controls API connections based on user settings, granting access to NLP models to authorized applications and denying access to denied applications. The input to this process is the user settings, and the output is the API connection control result.

[0174] Step 6:

[0175] A user inputs a prompt sentence into the generative AI model using a specific application, for example, "Please recommend new movies that might interest me based on my past SNS posts and search history." The input of this process is the prompt sentence, and the output is the recommendation result from the generative AI model.

[0176] Step 7:

[0177] The server provides the user with the recommendation results from the generative AI model. Specifically, it analyzes the user's past data and recommends movies that the user may be interested in. The input of this process is the recommendation results from the generative AI model, and the output is the recommendation information provided to the user.

[0178] (Application example 2)

[0179] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0180] Conventional online shopping sites did not adequately recommend products based on individual customer needs and preferences, making it difficult to improve customer satisfaction. It was also difficult to provide a personalized shopping experience while protecting customer privacy.

[0181] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0182] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their NLP from various apps in the world, a means for using the NLP for each customer to recommend products that are optimal for each customer, and a means for displaying the recommended products to the customer. This makes it possible to recommend products personalized based on individual needs and preferences while protecting the customer's privacy.

[0183] "Natural Language Processing (NLP)" is a technology that learns from large amounts of text data to understand and generate human language.

[0184] "Full Security" refers to advanced protection measures to prevent data leakage and unauthorized access.

[0185] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[0186] "Customer-specific NLP" is a natural language processing model that is trained based on individual customer data and has the ability to predict what those customers will think and do.

[0187] An "API connection" is an interface for exchanging data and functions between different software systems.

[0188] "Allow / deny" is the act of a customer choosing which apps can use their data.

[0189] "Means of product recommendation" refers to a method that uses NLP for each customer to suggest the most suitable products to each individual customer.

[0190] A "means for displaying products" is a method for visually presenting the recommended products to a customer.

[0191] As an embodiment of the present invention, a smartphone application for a mail-order site will be described as an example. This application has a function of recommending the best products to each individual customer by using natural language processing (NLP) for each customer.

[0192] System Configuration

[0193] 1. Pre-trained Natural Language Processing (NLP):

[0194] The server prepares a base NLP model trained from a large amount of text data, which serves as the basis for building individual NLP models based on the customer's interactions on their mobile phone.

[0195] 2. Fully secure data entry:

[0196] Customers' mobile phone interactions (emails, SNS posts, search history, etc.) are entered in a completely secure environment, allowing necessary data to be collected while protecting customer privacy.

[0197] 3. Build customer-specific NLP:

[0198] The server uses the collected data to build an NLP model that reflects each customer's personality, and this NLP model has the ability to predict the customer's thoughts and behavior.

[0199] 4. Allow / Deny API Connection:

[0200] Customers can "allow / deny" API connections to their NLP to various apps around the world, giving them the freedom to control which apps can use their data.

[0201] 5. Generate product recommendations:

[0202] The server uses per-customer NLP to recommend the best products for each individual customer, based on the customer's past behavior and preferences.

[0203] 6. Display of recommended products:

[0204] The recommended products are displayed to customers through a smartphone application, making it easy for them to find the products that best suit them.

[0205] Program processing explanation

[0206] The server uses Python and the requests library to collect customer data and send it to the NLP API, which generates an NLP model for each customer and makes product recommendations based on that model. The recommended products are displayed to the customer through a smartphone application.

[0207] As a concrete example, if a customer has recently searched for "new smartphone," an NLP model can use that information to recommend "latest smartphones" and "smartphone accessories."

[0208] Example of a prompt to be input to a generative AI model:

[0209] Your customer's recent search history includes "new smartphone." Use this information to recommend the best products to them.

[0210] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[0211] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0212] Step 1:

[0213] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and serves as the basis for building a customer-specific NLP model. The input is the pre-trained text data, and the output is the base NLP model.

[0214] Step 2:

[0215] The terminal collects the customer's mobile phone interactions (emails, SNS posts, search history, etc.) in a completely secure environment. The input is the customer's mobile phone interaction data, and the output is the collected customer data.

[0216] Step 3:

[0217] The server builds an NLP model that reflects the personality of each individual customer based on the collected customer data. The input is the collected customer data and the base NLP model, and the output is an NLP model for each customer. In concrete terms, the server sends customer data to the NLP API and generates an individual NLP model.

[0218] Step 4:

[0219] Users can "allow / deny" API connections to their NLP from various apps out there. The input is the user's choice of allow / deny, and the output is the allow / deny setting for API connections. In concrete terms, the user selects allow / deny on the application's settings screen.

[0220] Step 5:

[0221] The server uses an NLP model for each customer to recommend the best products to each individual customer. The input is the NLP model for each customer and the customer's past behavioral data, and the output is a list of recommended products. In concrete terms, the server uses the NLP model to analyze the customer's preferences and select the best products.

[0222] Step 6:

[0223] The terminal displays the recommended products to the customer. The input is a list of recommended products, and the output is product information that is displayed to the customer. In concrete terms, the terminal visually presents the product information through a smartphone application.

[0224] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[0225] Furthermore, an emotion engine that estimates the emotion of the user may be combined. That is, the identification processing unit 290 may estimate the emotion of the user using the emotion identification model 59, and perform identification processing using the emotion of the user.

[0226] "Example 1"

[0227] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[0228] "Example 2"

[0229] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[0230] The process flow of each embodiment will be described below.

[0231] "Example 1"

[0232] Step 1: The system prepares a pre-trained natural language processing (NLP) model.

[0233] Step 2: The system inputs all of the customer's mobile phone interactions in a fully secure location and builds an NLP model for each customer that is trained on that person's personality.

[0234] Step 3: The system uses the NLP model it built to predict what the customer will think.

[0235] Step 4: The system provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[0236] Step 5: The system combines an emotion engine to recognize the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text and voice inputs.

[0237] Step 6: The emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad".

[0238] "Example 2"

[0239] Step 1: The system prepares a pre-trained natural language processing (NLP) model.

[0240] Step 2: The system inputs all of the customer's mobile phone interactions in a fully secure location and builds an NLP model for each customer that is trained on that person's personality.

[0241] Step 3: The system uses the NLP model it built to predict what the customer will think.

[0242] Step 4: The system provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[0243] Step 5: The system combines an emotion engine to recognize the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text and voice inputs.

[0244] Step 6: The emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad".

[0245] Example 1

[0246] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal".

[0247] Conventional natural language processing (NLP) systems have difficulty in fully reflecting the individual personality and emotions of users, making it difficult to provide personalized experiences. They also lack the ability to safely manage user data and allow or deny the use of data for specific applications. Furthermore, they lack the ability to recognize user emotions and adjust NLP learning based on them.

[0248] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0249] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that has been trained with the personality of that person, a means for enabling the NLP to predict the thoughts of the customer, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This provides a personalized experience that reflects the individual personality and emotions of the user, and enables safe management and use control of data.

[0250] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0251] "Base training" is the process of training a model using a large amount of data in advance.

[0252] "Full Security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[0253] A "mobile terminal" is a portable electronic device such as a smartphone or tablet.

[0254] "Interactions" refers to user activity data such as emails, social media posts, and search history.

[0255] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[0256] An "API connection" is a way of exchanging data between systems through an application programming interface.

[0257] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input.

[0258] "Learning and tuning" is the process of changing the parameters of a model based on analyzed data to improve performance.

[0259] A "personalized experience" refers to providing services and features that are customized based on a user's individual characteristics and preferences.

[0260] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0261] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on a mobile device in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality.Furthermore, by combining an emotion engine that recognizes the user's emotions and adjusting the learning of the NLP based on the analyzed emotions, a personalized experience that reflects the user's individual personality and emotions is provided.

[0262] Hardware and Software Used

[0263] The server prepares natural language processing (NLP) models that have been trained in advance from large amounts of text data. For example, advanced language models such as GPT-3 and BERT are used for these NLP models. The server deploys these models on the cloud and makes them accessible via API.

[0264] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and transmits it to the server using a secure communication protocol (e.g., HTTPS).

[0265] The server analyzes the received data and builds an NLP model for each customer. Specifically, the server preprocesses the data, tokenizing the text, removing stop words, stemming, etc. Then, the server uses the data to retrain the NLP model.

[0266] Users can select which applications they want to allow or deny API connections to their NLP models from a dedicated management screen. The device sends this setting to the server, which then stores it.

[0267] The server receives the user's text or voice input and analyzes the emotion using the emotion engine. For example, if the user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" from this text.

[0268] The server adjusts the learning of the NLP model using the analyzed emotion data. For example, if the user frequently expresses the emotion "sad," the server adjusts the learning of the NLP model to correspond to the user's emotion "sad." This creates a personalized NLP that adapts to the user's emotions.

[0269] Examples

[0270] Example 1: Parsing an email

[0271] If a user sends an email saying "I'm very tired today," the device sends this text to the server, which uses an emotion engine to analyze the emotion of "tiredness" and incorporates it into the NLP model.

[0272] Example 2: Social media posts

[0273] If a user posts on social media, "I'm happy that a new cafe has opened," the device sends this text to the server. The server uses an emotion engine to analyze the emotion of "happiness" and reflects it in the NLP model.

[0274] Examples of prompt statements

[0275] "How can we analyze the sentiment of text posted by users on social media and adjust NLP learning based on that sentiment?"

[0276] The system allows users to safely manage their data and enjoy personalized experiences that respond to their emotions.

[0277] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0278] Step 1:

[0279] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, the server uses advanced language models such as GPT-3 and BERT, and deploys these models on the cloud. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[0280] Step 2:

[0281] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and sends it to the server using a secure communication protocol (e.g., HTTPS). As input, the user's activity data is required, and as output, the encrypted data is sent to the server.

[0282] Step 3:

[0283] The server analyzes the received data and builds a per-customer NLP model. Specifically, the server pre-processes the data, tokenizing the text, removing stop words, stemming, etc. Then it retrains the NLP model with these data. As input, it requires encrypted user data, and as output, it gets a per-customer NLP model.

[0284] Step 4:

[0285] From a dedicated management screen, the user selects the applications for which they want to allow or deny API connections to their NLP models. The device sends this setting to the server, which then stores the setting. As input, the user's permission setting is required, and as output, the setting for allowing or denying API connections is stored on the server.

[0286] Step 5:

[0287] The server receives the user's text or voice input and analyzes the emotion using an emotion engine. For example, if the user inputs "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. The input requires the user's text or voice data, and the output is the analyzed emotion data.

[0288] Step 6:

[0289] The server uses the analyzed emotion data to adjust the learning of the NLP model. For example, if the user frequently expresses the emotion "sad", the server adjusts the learning of the NLP model to correspond to the user's emotion "sad". As input, the analyzed emotion data is required, and as output, an emotion-adapted NLP model is obtained.

[0290] (Application example 1)

[0291] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0292] Conventional content delivery services have had difficulty providing personalized content based on users' emotions and behavior. They also lacked technology to accurately analyze users' thoughts and emotions while protecting their privacy. Furthermore, there were limited ways for users to freely control which applications could use their data.

[0293] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0294] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to their own NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, analyzing emotions from the user's text input or voice input, and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the user's emotions and behavior. This makes it possible to provide personalized content based on the user's emotions and behavior.

[0295] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[0296] "Full security" refers to a state in which advanced security measures are in place to prevent unauthorized access or leakage of data and protect customer privacy.

[0297] "Customer-specific NLP" is an individual natural language processing model that learns about a customer's personality and behavioral patterns based on their interactions on their mobile phone.

[0298] An "API connection" is a connection that exchanges data and functions between different software through an application program interface.

[0299] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input and adjusts the system's operation based on the results of that analysis.

[0300] "Content recommendation" is a function that selects and provides the most suitable content, such as videos, articles, and music, based on the user's emotions and behavior.

[0301] A system for implementing the present invention is configured as follows: First, a server prepares a natural language processing (NLP) model that has undergone base training in advance. This NLP model has been trained from a large amount of text data and has the ability to understand and generate the language of a user.

[0302] Next, the device (e.g., a smartphone or head-mounted display) inputs all of the user's mobile phone interactions in a fully secure environment. This input data includes the user's emails, social media posts, search history, etc. Based on this data, the server builds an NLP model for each user. This model learns the user's personality and behavioral patterns and has the ability to predict the user's thoughts.

[0303] Furthermore, the server combines an emotion engine. This emotion engine analyzes emotions from the user's text input or voice input, and adjusts the learning of the NLP model based on the analysis results. For example, if the user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. Also, if the user frequently shows the emotion of "sad," the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion of "sad."

[0304] The system also has the ability to recommend optimal content based on a user's emotions and behavior. For example, if a user posts on social media that they are "tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos.

[0305] Users can also "allow / deny" API connections to their NLP models to various applications out there, giving users the freedom to control which applications can use their data.

[0306] For example, if a user posts on social media that "I'm tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos. In this way, it is possible to provide personalized content based on the user's emotions and behavior.

[0307] An example of a prompt sentence might be:

[0308] "If a user posts on social media, 'I'm tired today,' what emotion will the emotion engine analyze, and what content will the NLP model recommend?"

[0309] This system will make it possible to provide personalized content based on a user's emotions and behavior, which is expected to improve the user experience.

[0310] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0311] Step 1:

[0312] The device inputs all of the user's mobile phone interactions. The input data includes the user's emails, social media posts, search history, etc. This data is collected in a completely secure environment. The input data becomes the basis for analyzing the user's behavior and thoughts.

[0313] Step 2:

[0314] The server builds an NLP model for each user based on the collected data. Specifically, it uses an NLP model that has undergone base training in advance to analyze the user's text data and learn the user's personality and behavioral patterns. In this process, the input data is analyzed and processed to extract the user's characteristics. The output is an NLP model for each user.

[0315] Step 3:

[0316] The server uses an emotion engine to analyze emotions from the user's text or voice input. For example, if the user inputs the text "Today was fun," the emotion engine analyzes the emotion "joy" from this text. The input is the user's text or voice data, and the output is the analyzed emotion data.

[0317] Step 4:

[0318] The server adjusts the learning of the NLP model based on the analysis result of the emotion engine. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion "sad". The input is the analyzed emotion data, and the output is the adjusted NLP model.

[0319] Step 5:

[0320] The server recommends optimal content based on the user's emotions and behavior. For example, if a user posts on social media that "I'm tired today," the emotion engine analyzes the emotion of "tiredness," and based on this information, the NLP model recommends relaxing music and refreshing videos. The input is the tuned NLP model and the analyzed emotion data, and the output is the recommended content.

[0321] Step 6:

[0322] Users can "allow / deny" API connections to their NLP models to various applications in the world. Users can control data usage for each application. The input is the user's permission settings, and the output is API connection control information.

[0323] In this way, a system is realized that provides personalized content based on the user's emotions and behavior.

[0324] Example 2

[0325] Next, a second embodiment of the second embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0326] Conventional natural language processing (NLP) systems have the ability to understand general language patterns, but they have difficulty generating personalized responses based on the characteristics and emotions of individual users. It is also a challenge to build individual NLP models while protecting the privacy of users. Furthermore, users lack the means to freely control which applications can use their data.

[0327] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0328] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile terminal in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes and analyzes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This enables highly personalized NLP based on the user's characteristics and emotions, and allows the user to freely control which applications can use their data by building individual NLP models while protecting the user's privacy.

[0329] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0330] "Base training" is an initial learning process that is carried out using a large amount of data in advance for a specific task.

[0331] "Full security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[0332] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[0333] "Interaction" refers to communication activities such as emails, social media posts, and search history that users conduct via their mobile devices.

[0334] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[0335] "API connection" refers to a means of connection that allows different software to share data and functions through an application program interface.

[0336] "Application" means a software program that provides specific functions or services.

[0337] An "emotion engine" refers to technology that analyzes emotions from a user's text or voice input.

[0338] "Learning tuning" refers to the process of changing the parameters of a model based on analyzed data to improve performance.

[0339] "Privacy protection" refers to protective measures to prevent personal information from being used improperly.

[0340] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs the customer's interactions on a mobile device in a completely secure environment, and builds an individual NLP model for each customer based on that data.Furthermore, by combining an emotion engine that recognizes and analyzes the user's emotions and adjusting the learning of the NLP based on the analysis results, personalized NLP that adapts to the user's characteristics and emotions is realized.

[0341] Hardware and Software Used

[0342] server

[0343] The server prepares an NLP model that has been trained in advance. Specifically, it uses existing NLP models such as GPT-3 and BERT and deploys them on the cloud. The server analyzes the collected data and executes machine learning algorithms to build individual NLP models for each customer.

[0344] Terminal

[0345] The terminal collects the customer's mobile device interactions and transmits them to the server in a fully secure environment. The terminal collects data through a dedicated application, encrypts the data using SSL / TLS protocol, and transmits it securely to the server.

[0346] Emotion Engine

[0347] The emotion engine analyzes emotions from the user's text and voice input. The emotion engine sends the analysis results to the server and uses them to adjust the learning of the NLP model. For example, if a user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" and reflects it in the NLP model.

[0348] Examples

[0349] For example, if a user posts on social media, "I'm so tired today," the emotion engine will analyze the emotion of "fatigue." The server will then feed this emotion data into the NLP model and adjust the learning to generate appropriate responses when the user is tired.

[0350] Examples of prompt statements

[0351] "If a user posts on social media, 'I'm so tired today,' please explain what emotion the emotion engine analyzes and how the NLP model adjusts its learning."

[0352] The system enables highly personalized NLP based on user characteristics and emotions, building individual NLP models while protecting user privacy and giving users full control over which applications can use their data.

[0353] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0354] Step 1:

[0355] The server prepares a natural language processing (NLP) model that has been pre-trained as a base.

[0356] Input: A large amount of text data

[0357] Data processing: Training NLP models using text data

[0358] Output: An NLP model with basic language understanding capabilities.

[0359] How it works: The server uses existing NLP models such as GPT-3 and BERT and deploys them on the cloud.

[0360] Step 2:

[0361] The terminal inputs all of the customer's mobile interactions in a completely secure environment.

[0362] Input: Customer emails, social media posts, search history, etc.

[0363] Data processing: Collect and encrypt data

[0364] Output: Encrypted customer data

[0365] How it works: The device collects data through a dedicated application, encrypts the data using the SSL / TLS protocol, and securely transmits it to the server.

[0366] Step 3:

[0367] The server uses the collected data to build an individual NLP model for each customer.

[0368] Input: Encrypted customer data

[0369] Data processing: Analyze data and learn customer language and behavior patterns

[0370] Output: A separate NLP model for each customer

[0371] What happens: The server uses machine learning algorithms to train the model on specific phrases and expressions used by the customer.

[0372] Step 4:

[0373] The server combines the NLP system with an emotion engine that recognizes and analyzes the user's emotions.

[0374] Input: User text and voice input

[0375] Data processing: Emotion engine analyzes emotions

[0376] Output: Parsed emotion data

[0377] How it works: The server integrates an emotion engine into the NLP model to analyze emotions from the user's text and voice input.

[0378] Step 5:

[0379] The server adjusts the NLP learning based on the emotions analyzed by the emotion engine.

[0380] Input: Parsed emotion data

[0381] Data processing: Adjusting parameters of NLP models

[0382] Output: Emotion-adapted NLP model

[0383] Specific operation: The server uses the emotion data obtained from the emotion engine to adjust the parameters of the NLP model.

[0384] Step 6:

[0385] Users can "allow / deny" API connections to their NLP from various applications out there.

[0386] Input: User permission settings

[0387] Data processing: Control API connections based on permission settings

[0388] Output: API connections to authorized applications

[0389] How it works: Through a dedicated admin screen, users configure which applications can access their NLP models.

[0390] (Application example 2)

[0391] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0392] Conventional content delivery services have the problem that they do not adequately recommend personalized content based on users' interests and emotions. In addition, there is no established method for effectively utilizing user data while protecting user privacy. In addition, there is a problem that it is difficult to gain users' trust because users lack the means to control the use of their data.

[0393] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0394] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their own NLP from various apps in the world, a means for analyzing the customer's emotions and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the customer's emotions and interests. This enables personalized content recommendations based on the user's emotions and interests, and allows the user's data to be effectively utilized while protecting the user's privacy. In addition, by providing a means for the user to control the use of their own data, the user's trust can be gained.

[0395] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[0396] "Full security" means that advanced protective measures are in place to prevent unauthorized access or leakage of data.

[0397] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[0398] "Personality-trained NLP for each customer" is a natural language processing model that learns the behavior and thought patterns of each individual customer and is specialized for that customer.

[0399] An "API connection" is an interface for exchanging data and functions between different software programs.

[0400] "Emotion analysis" is the process of identifying a user's current emotional state from their text or voice input.

[0401] "Tuning NLP learning" means changing the learning parameters of a natural language processing model based on analyzed sentiment data.

[0402] "Recommending content" means presenting information such as the most suitable movies, music, articles, etc. based on the user's interests and emotions.

[0403] The following system configuration will be described as an embodiment of the present invention.

[0404] System Programming

[0405] The system consists of the following main components:

[0406] 1. Natural Language Processing (NLP) Model: Prepare a pre-trained NLP model, which is trained from a large amount of text data.

[0407] 2. Data entry method: In a fully secure location, all of the customer's mobile phone interactions (emails, social media posts, search history, etc.) are entered.

[0408] 3. Individual NLP model construction method: Based on the input data, an NLP model trained on the personality of each customer is constructed.

[0409] 4. Mind-reading: The NLP model built can predict what customers will think.

[0410] 5. API connection control: Allows customers to "allow / deny" API connections to their NLP from various apps around the world.

[0411] 6. Sentiment analysis tools: Analyze emotions from customer text and voice input.

[0412] 7. Learning adjustment measures: Adjust the learning of the NLP model based on the analyzed sentiment data.

[0413] 8. Content Recommendation: Recommend the most relevant content based on customer sentiment and interests.

[0414] Processing Description

[0415] The server first prepares an NLP model that has already undergone base training. This NLP model is trained from a large amount of text data. Next, in a completely secure location, it inputs all of the customer's mobile phone interactions (emails, SNS posts, search history, etc.). Based on this data, it builds an NLP model trained on each customer's personality.

[0416] The NLP model that is built is tuned to be able to predict what the customer will think, and it also provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[0417] The emotion analysis means analyzes emotions from the customer's text input or voice input. For example, if a customer inputs the text "I had fun today," the emotion analysis means analyzes the emotion of "joy" from this text. Based on the analyzed emotion data, the learning of the NLP model is adjusted.

[0418] Finally, the system recommends the most suitable content based on the customer's emotions and interests. For example, if a customer sends an email saying "I had fun today" and searches for "latest movies," the system will analyze that the customer has the emotion of "joy" and recommend the latest comedy movies.

[0419] Examples of concrete examples and prompts

[0420] As a specific example, if a user sends an email saying "I had fun today" and searches for "latest movies," the system will analyze that the user has the emotion of "joy" and recommend the latest comedy movies.

[0421] An example of a prompt is as follows:

[0422] Develop an application that analyzes user sentiment based on user emails, SNS posts, and search history, and recommends the most suitable content. If a user sends an email saying "I had fun today" and searches for "latest movies," recommend the latest comedy movies.

[0423] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0424] Step 1:

[0425] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data. The existing text data is used as input, and the pre-trained NLP model is obtained as output.

[0426] Step 2:

[0427] The server is a completely secure location and inputs all of the customer's mobile phone interactions (emails, SNS posts, search history, etc.) The input is the customer's communication data, and the output is that this data is safely stored on the server.

[0428] Step 3:

[0429] The server builds an NLP model trained on the personality of each customer based on the input data. The customer's communication data is used as input, and the NLP model for each customer is obtained as output.

[0430] Step 4:

[0431] The server adjusts the constructed NLP model so that it can predict the customer's thoughts. The NLP model for each customer is used as input, and the output is an NLP model with a thought-predicting function.

[0432] Step 5:

[0433] The server provides a way for customers to "allow / deny" API connections to their NLP from various apps out there. The input is the customer's permission information, and the output is the allow / deny setting for the API connection.

[0434] Step 6:

[0435] The server analyzes emotions from the customer's text and voice input, using the customer's text and voice data as input and obtaining analyzed emotion data as output.

[0436] Step 7:

[0437] The server adjusts the learning of the NLP model based on the analyzed emotion data: the emotion data and the NLP model are used as inputs, and the emotion-adapted NLP model is obtained as output.

[0438] Step 8:

[0439] The server recommends optimal content based on the customer's emotions and interests. As input, emotion data and customer interest data are used, and as output, the recommended content is obtained.

[0440] 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 a voice indicating a user input for the result of the specific processing. The control unit 46A transmits the voice data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0441] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by making a neural network perform deep learning. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating a voice, text data indicating a text, and image data indicating an image is input. The data generation model 58 performs inference on the input inference data according to the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0442] Another example of generative AI is Gemini (registered trademark) (Internet search<url: https: gemini.google.com ?hl="ja">) are mentioned.

[0443] In the above embodiment, an example was given in which the specific process was performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0444] [Second embodiment]

[0445] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0446] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0447] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[0448] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0449] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs the voice according to instructions from the processor 46.

[0450] Camera 42 is a small digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (e.g., an imaging range defined by an angle of view equivalent to the width of the field of vision of an average healthy person).

[0451] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54. The transmission and reception of various types of information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is performed in a secure state.

[0452] Fig. 4 shows an example of main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0453] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32, and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0454] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0455] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0456] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0457] "Example 1"

[0458] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[0459] "Example 2"

[0460] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[0461] The process flow of each embodiment will be described below.

[0462] "Example 1"

[0463] Step 1: Prepare a pre-trained natural language processing (NLP) model. This NLP model is trained, for example, from a large amount of text data.

[0464] Step 2: In a fully secured location, enter all of the customer's mobile phone interactions, including, for example, the customer's emails, social media posts, search history, etc.

[0465] Step 3: Using this data, we build a personalized NLP that is trained on that person's personality and has the ability to predict what the customer will think and do.

[0466] Step 4: Allow customers to "allow / deny" API connections to their NLP from various apps around the world, giving customers the freedom to control which apps can use their data.

[0467] Example 1

[0468] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0469] Conventional natural language processing (NLP) systems have difficulty building personalized models that reflect the characteristics and behavior of individual customers. In addition, there was no established method for safely collecting customer data and using it while protecting privacy. Furthermore, there was a lack of means for customers to freely control which applications could use their data.

[0470] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0471] In this invention, the server includes a means for preparing a natural language processing (NLP) model that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure location and constructing an NLP model for each customer that has been trained with that person's personality, a means for enabling the NLP model to predict the customer's thoughts and actions, a means for collecting customer data, encrypting it and transmitting it to the server, a means for the server to fine-tune the NLP model using the customer data, and a means for the terminal to provide an interface for managing the customer's API connection settings. This makes it possible to construct a personalized NLP model that reflects the characteristics of the customer, safely collect and use customer data, and enable the customer to freely control the use of their own data.

[0472] A "natural language processing (NLP) model" is a machine learning model that is trained based on large amounts of text data and has the ability to understand and generate natural language.

[0473] "Complete security" means that advanced security measures are in place to prevent unauthorized access and information leaks during data collection, transmission, and storage.

[0474] "Communication equipment" refers to devices used by customers, such as mobile phones and tablets, through which data is entered and transmitted.

[0475] "Interactions" refers to all communication activities conducted by customers through communication devices, such as emails, social media posts, search history, etc.

[0476] A "personality-trained NLP model for each customer" is an NLP model trained to reflect the characteristics and behavioral patterns of each individual customer based on their communication activity data.

[0477] An "API connection" is a method of connecting different software to exchange data and functions through an application program interface.

[0478] "Allow / Deny" refers to the act of a customer choosing whether to allow or deny a particular application access to their NLP models.

[0479] "Data collection" is the process of obtaining interaction data from a customer's communications device.

[0480] "Encryption" is a security technology that converts data into a form that cannot be deciphered by third parties.

[0481] "Fine-tuning" is the process of fine-tuning an existing base model to a specific dataset to improve its performance.

[0482] An "interface" refers to the screen or operating means that allows the user to interact with the system, and in this case refers to the means used to set whether to allow or deny API connections.

[0483] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs all of the customer's interactions on communication devices, and builds an NLP model for each customer that is trained based on that person's personality. Furthermore, it enables this NLP model to predict the customer's thoughts and actions, and enables the customer to "allow / reject" API connections to their NLP model from various applications around the world.

[0484] Hardware and Software Used

[0485] server

[0486] The server uses machine learning frameworks such as TensorFlow or PyTorch to train NLP models on large amounts of text data. The models use advanced natural language processing techniques such as BERT and GPT-3. The server saves the trained models and prepares them for later processing.

[0487] Terminal

[0488] The device installs a dedicated app that allows the user to input interactions on their communication device. With the user's permission, the app collects data such as emails, social media posts, and search history. The collected data is encrypted and sent to a server. The device provides an interface for the user to manage API connections to their NLP models.

[0489] User

[0490] The user uses a dedicated app to input interactions on their communication device. The user's communication device transmits this data to the server. Through the device interface, the user can allow or deny the API connection to their NLP model.

[0491] Data processing and calculation

[0492] The server uses the received customer data to build an NLP model specific to that customer. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data preprocessing, model training, and evaluation. For example, the server analyzes the content of a customer's emails to learn what topics the customer is interested in.

[0493] Examples

[0494] For example, if a user allows their NLP model to be connected when using a shopping app, the app can recommend products that the user might be interested in based on the user's past purchase history and search history. On the other hand, they can protect their privacy by refusing to connect to SNS apps.

[0495] Examples of prompt statements

[0496] Examples of prompt sentences to be input to a generative AI model include the following:

[0497] "Predict the next topic that a user will be interested in based on their past emails and social media posts."

[0498] Using this prompt, the generative AI model can predict the user's interests and provide appropriate information.

[0499] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0500] Step 1:

[0501] The server prepares a natural language processing (NLP) model that has been pre-trained. Specifically, the server uses machine learning frameworks such as TensorFlow and PyTorch to train an NLP model based on a large amount of text data. This model utilizes advanced natural language processing techniques such as BERT and GPT-3. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[0502] Step 2:

[0503] A user installs a dedicated app to input communications on their communication device. With the user's permission, this app collects data such as emails, SNS posts, and search history. In concrete terms, the app obtains this data from the user's communication device, encrypts it, and sends it to a server. The input required is the user's communication data, and the output is the encrypted data sent to the server.

[0504] Step 3:

[0505] The server uses the received customer data to build a customer-specific NLP model. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data pre-processing, model training, and evaluation. As input, encrypted customer data is required, and as output, a customer-specific NLP model is obtained.

[0506] Step 4:

[0507] The terminal provides an interface for the user to manage API connections to their NLP models. In concrete terms, the terminal allows the user to select which applications are allowed or denied to use the NLP model. The terminal requires the user's selection information as input, and obtains the permission / denial setting for API connections as output.

[0508] Step 5:

[0509] The server controls the access of each application based on the user's API connection settings. Specifically, the server receives the user's setting information and allows or denies API access based on that information. The input is the user's API connection settings, and the output is access control for each application.

[0510] Step 6:

[0511] A user uses his / her own NLP model through an authorized application. Specifically, the user receives personalized services using his / her own NLP model in a shopping application, information providing application, etc. The input is a request from the application, and the output is the prediction result and recommendation information by the NLP model.

[0512] (Application example 1)

[0513] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0514] Conventional online shopping sites did not adequately recommend products based on customers' interests and purchase history, making it difficult to suggest products that matched customers' needs in a timely manner. In addition, there was a lack of a system that could respond to individual needs while protecting customers' privacy.

[0515] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means. In this invention, the server includes a means for preparing natural language processing (NLP) that has been previously trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that is trained with the personality of that person, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to his / her NLP from various applications in the world, a means for using the customer's NLP to recommend products based on the customer's interests and purchase history, and a means for predicting the customer's behavioral patterns and making timely product suggestions. This makes it possible to timely suggest products that meet the customer's needs.

[0516] "Natural Language Processing (NLP)" is a technology that uses algorithms trained from large amounts of text data to understand and generate human language.

[0517] "Full security" means that advanced protective measures are in place to prevent data leakage and unauthorized access.

[0518] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[0519] "Interaction" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile devices.

[0520] "Personality-trained NLP for each customer" is an NLP model that learns each individual customer's thinking and behavior patterns based on their communication activities.

[0521] An "API connection" is an interface for exchanging data and functions between different software programs.

[0522] "Allow / Deny" is the act of a Customer choosing to allow or deny an API connection to their NLP.

[0523] "Means of product recommendation" is a function that uses customer NLP to suggest products based on the customer's interests and purchasing history.

[0524] A "means for predicting behavioral patterns" is a function that predicts future behavior based on a customer's past behavioral data.

[0525] "Timely product proposals" refers to proposing products that meet customer needs at the right time.

[0526] The system for implementing this invention has the following configuration. First, the server prepares natural language processing (NLP) that has undergone base learning in advance. Next, in a completely secure location, all interactions on the customer's mobile device (emails, SNS posts, search history, etc.) are entered, and an NLP for each customer that is trained on that person's personality is constructed. This NLP is designed to be able to predict the customer's thoughts.

[0527] Customers can "allow / deny" API connections to their NLP to various applications in the world, giving them the freedom to control which applications can use their data.

[0528] In addition, the server uses NLP to recommend products based on the customer's interests and purchase history, making it possible to predict customer behavior patterns and provide timely product suggestions.

[0529] Hardware and Software Used

[0530] Hardware: Servers (for data processing), smartphones (for collecting customer data)

[0531] Software: Python, scikit-learn (machine learning library), TfidfVectorizer (vectorization of text data)

[0532] Data processing and calculation

[0533] 1. Data collection:

[0534] The server collects customer emails, social media posts, and search history, and stores this data in JSON format.

[0535] 2. Building the NLP model:

[0536] The server vectorizes the text data using TfidfVectorizer, and uses the vectorized data to build the client's NLP model.

[0537] 3. Product recommendation:

[0538] The server compares the customer's NLP model with the product list, calculates the cosine similarity, and recommends products with high similarity.

[0539] Examples

[0540] For example, if customer A mentions in an email that he / she "want a new smartphone," posts on social media that he / she is "interested in the latest gadgets," and his / her search history includes "smartphone reviews," the server will use this data to build an NLP model for customer A. The server will then use this NLP model to recommend the latest smartphones, smartwatches, gadget accessories, etc.

[0541] Examples of prompt statements

[0542] Collect customer emails, social media posts, and search history, and vectorize them using TfidfVectorizer. Then build an NLP model of the customer and compare it with the product list to calculate the cosine similarity. Recommend products with high similarity.

[0543] In this way, a personalized shopping assistant can be realized that utilizes individual NLP for each customer.

[0544] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0545] Step 1:

[0546] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and has basic language understanding capabilities. The input is the pre-trained NLP model, and the output is a usable NLP model.

[0547] Step 2:

[0548] The server, located in a completely secure location, collects all of the customer's mobile device interactions (emails, social media posts, search history, etc.). This data is encrypted and stored to protect the customer's privacy. The input is the customer's communication data, and the output is encrypted customer data.

[0549] Step 3:

[0550] The server uses the collected customer data to build an NLP model for each customer that is trained on that person's personality. Specifically, it uses TfidfVectorizer to vectorize the text data and generate an NLP model for the customer. The input is the encrypted customer data, and the output is an NLP model for each customer.

[0551] Step 4:

[0552] The server uses the customer's NLP model to predict the customer's thoughts and actions. To do this, it applies an algorithm that predicts future behavior based on past data. The input is the customer's NLP model, and the output is the predicted customer behavior pattern.

[0553] Step 5:

[0554] Users can "allow / deny" API connections to their NLP models from various applications in the world. The server controls API connections based on user permissions. The input is the user's permission information, and the output is a list of allowed applications.

[0555] Step 6:

[0556] The server uses the customer's NLP model to recommend products based on the customer's interests and purchase history. Specifically, it compares the customer's NLP model with the product list and calculates the cosine similarity. The inputs are the customer's NLP model and the product list, and the output is the recommended product list.

[0557] Step 7:

[0558] The server predicts customer behavior patterns and makes timely product suggestions. This makes it possible to suggest products that meet customer needs at the right time. The inputs are the predicted customer behavior patterns and a product list, and the output is a timely suggested product list.

[0559] Example 2

[0560] Next, a description will be given of Example 2 of the embodiment 2. In the following description, the data processing device 12 is referred to as a "server" and the smart glasses 214 are referred to as a "terminal".

[0561] Conventional natural language processing (NLP) systems have difficulty accurately predicting the thoughts and behavior of individual users, and lack the means to freely control which applications can use user data. As a result, there is a high risk of violating user privacy and a poor user experience.

[0562] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0563] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for collecting customer data and constructing an individual NLP model, and a means for the customer to manage API connections using a dedicated management application. This makes it possible to accurately predict the user's thoughts and actions and freely control which applications can use the user's data.

[0564] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0565] "Base learning" is the process of training a basic model using a large amount of data in advance.

[0566] "Complete Security" refers to the highest level of protection to prevent unauthorized access or leakage of data.

[0567] A "communication device" is a device used by a user, such as a mobile phone or tablet.

[0568] "Interaction" refers to activities such as emails, social media posts, and search history that users conduct through communication devices.

[0569] "Personality-trained NLP for each customer" is an NLP model that learns the characteristics and behavioral patterns of each individual customer based on their data.

[0570] An "API connection" is an interface that allows an application to communicate with other software or services.

[0571] "Allow / deny" is the act of a user choosing whether to allow or deny a particular application access to their data.

[0572] "Management Application" means dedicated software that allows a User to manage API connections to their NLP models.

[0573] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on communication devices in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality. This system makes it possible to predict the customer's thoughts, and enables the customer to "allow / reject" API connections to their own NLP from various applications in the world.

[0574] Hardware and Software Used

[0575] server

[0576] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. This NLP model uses a generative AI model such as GPT-4. The server is responsible for collecting customer data and building individual NLP models.

[0577] Terminal

[0578] The terminal is a communication device (such as a mobile phone or tablet) used by the user. The terminal collects data such as the user's emails, SNS posts, search history, etc., and transmits it to the server.

[0579] User

[0580] Users enter their interactions on their communication devices and manage API connections to their NLP models using a dedicated management application, and users choose to allow or deny specific applications access to their data.

[0581] Data processing and calculation

[0582] The server analyzes the collected customer data and builds an individual NLP model. Specifically, the customer data is tokenized and input into the base NLP model. The model learns the customer's language patterns and behavioral patterns and generates an individual NLP model. This model has the ability to predict the customer's thoughts and behavior.

[0583] Examples

[0584] For example, if a user is using a social networking app, the app needs the user's permission to access the user's NLP model. If the user gives permission, the social networking app can provide content tailored to the user's interests based on the user's past posts and search history.

[0585] Examples of prompt statements

[0586] "Recommend new movies that you think might interest me based on my past social media posts and search history."

[0587] When this prompt is input into a generative AI model, the model analyzes the user's past data and recommends movies that might interest them.

[0588] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0589] Step 1:

[0590] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, it uses a generative AI model such as GPT-4 to analyze and tokenize the text data and train the model. The input of this process is a large amount of text data, and the output is a base NLP model.

[0591] Step 2:

[0592] A user inputs interactions on their communication device. Specifically, they open an email app, select past emails, and send them to the server. They also open a social networking app, select past posts and comments, and send them to the server. They also export their browser search history and send it to the server. The input to this process is the data on the user's communication device, and the output is customer data sent to the server.

[0593] Step 3:

[0594] The server analyzes the collected customer data and builds an individual NLP model. Specifically, it tokenizes the customer data and inputs it into the base NLP model. The model learns the customer's language and behavioral patterns and generates an individual NLP model. The input of this process is the customer data and the output is an individual NLP model.

[0595] Step 4:

[0596] Users manage API connections to their NLP models using a dedicated management application: they open the management application and choose whether to allow or deny a particular application access to their data. The input to this process is the user's choice, and the output is the setting to allow or deny the API connection.

[0597] Step 5:

[0598] The server controls API connections based on user settings, granting access to NLP models to authorized applications and denying access to denied applications. The input to this process is the user settings, and the output is the API connection control result.

[0599] Step 6:

[0600] A user inputs a prompt sentence into the generative AI model using a specific application, for example, "Please recommend new movies that might interest me based on my past SNS posts and search history." The input of this process is the prompt sentence, and the output is the recommendation result from the generative AI model.

[0601] Step 7:

[0602] The server provides the user with the recommendation results from the generative AI model. Specifically, it analyzes the user's past data and recommends movies that the user may be interested in. The input of this process is the recommendation results from the generative AI model, and the output is the recommendation information provided to the user.

[0603] (Application example 2)

[0604] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the smart glasses 214 will be referred to as a "terminal".

[0605] Conventional online shopping sites did not adequately recommend products based on individual customer needs and preferences, making it difficult to improve customer satisfaction. It was also difficult to provide a personalized shopping experience while protecting customer privacy.

[0606] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0607] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their NLP from various apps in the world, a means for using the NLP for each customer to recommend products that are optimal for each customer, and a means for displaying the recommended products to the customer. This makes it possible to recommend products personalized based on individual needs and preferences while protecting the customer's privacy.

[0608] "Natural Language Processing (NLP)" is a technology that learns from large amounts of text data to understand and generate human language.

[0609] "Full Security" refers to advanced protection measures to prevent data leakage and unauthorized access.

[0610] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[0611] "Customer-specific NLP" is a natural language processing model that is trained based on individual customer data and has the ability to predict what those customers will think and do.

[0612] An "API connection" is an interface for exchanging data and functions between different software systems.

[0613] "Allow / deny" is the act of a customer choosing which apps can use their data.

[0614] "Means of product recommendation" refers to a method that uses NLP for each customer to suggest the most suitable products to each individual customer.

[0615] A "means for displaying products" is a method for visually presenting the recommended products to a customer.

[0616] As an embodiment of the present invention, a smartphone application for a mail-order site will be described as an example. This application has a function of recommending the best products to each individual customer by using natural language processing (NLP) for each customer.

[0617] System Configuration

[0618] 1. Pre-trained Natural Language Processing (NLP):

[0619] The server prepares a base NLP model trained from a large amount of text data, which serves as the basis for building individual NLP models based on the customer's interactions on their mobile phone.

[0620] 2. Fully secure data entry:

[0621] Customers' mobile phone interactions (emails, SNS posts, search history, etc.) are entered in a completely secure environment, allowing necessary data to be collected while protecting customer privacy.

[0622] 3. Build customer-specific NLP:

[0623] The server uses the collected data to build an NLP model that reflects each customer's personality, and this NLP model has the ability to predict the customer's thoughts and behavior.

[0624] 4. Allow / Deny API Connection:

[0625] Customers can "allow / deny" API connections to their NLP to various apps around the world, giving them the freedom to control which apps can use their data.

[0626] 5. Generate product recommendations:

[0627] The server uses per-customer NLP to recommend the best products for each individual customer, based on the customer's past behavior and preferences.

[0628] 6. Display of recommended products:

[0629] The recommended products are displayed to customers through a smartphone application, making it easy for them to find the products that best suit them.

[0630] Program processing explanation

[0631] The server uses Python and the requests library to collect customer data and send it to the NLP API, which generates an NLP model for each customer and makes product recommendations based on that model. The recommended products are displayed to the customer through a smartphone application.

[0632] As a concrete example, if a customer has recently searched for "new smartphone," an NLP model can use that information to recommend "latest smartphones" and "smartphone accessories."

[0633] Example of a prompt to be input to a generative AI model:

[0634] Your customer's recent search history includes "new smartphone." Use this information to recommend the best products to them.

[0635] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[0636] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0637] Step 1:

[0638] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and serves as the basis for building a customer-specific NLP model. The input is the pre-trained text data, and the output is the base NLP model.

[0639] Step 2:

[0640] The terminal collects the customer's mobile phone interactions (emails, SNS posts, search history, etc.) in a completely secure environment. The input is the customer's mobile phone interaction data, and the output is the collected customer data.

[0641] Step 3:

[0642] The server builds an NLP model that reflects the personality of each individual customer based on the collected customer data. The input is the collected customer data and the base NLP model, and the output is an NLP model for each customer. In concrete terms, the server sends customer data to the NLP API and generates an individual NLP model.

[0643] Step 4:

[0644] Users can "allow / deny" API connections to their NLP from various apps out there. The input is the user's choice of allow / deny, and the output is the allow / deny setting for API connections. In concrete terms, the user selects allow / deny on the application's settings screen.

[0645] Step 5:

[0646] The server uses an NLP model for each customer to recommend the best products to each individual customer. The input is the NLP model for each customer and the customer's past behavioral data, and the output is a list of recommended products. In concrete terms, the server uses the NLP model to analyze the customer's preferences and select the best products.

[0647] Step 6:

[0648] The terminal displays the recommended products to the customer. The input is a list of recommended products, and the output is product information that is displayed to the customer. In concrete terms, the terminal visually presents the product information through a smartphone application.

[0649] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[0650] In addition, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0651] "Example 1"

[0652] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[0653] "Example 2"

[0654] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[0655] The process flow of each embodiment will be described below.

[0656] "Example 1"

[0657] Step 1: The system prepares a pre-trained natural language processing (NLP) model.

[0658] Step 2: The system inputs all of the customer's mobile phone interactions in a fully secure location and builds an NLP model for each customer that is trained on that person's personality.

[0659] Step 3: The system uses the NLP model it built to predict what the customer will think.

[0660] Step 4: The system provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[0661] Step 5: The system combines an emotion engine to recognize the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text and voice inputs.

[0662] Step 6: The emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad".

[0663] Example 1

[0664] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0665] Conventional natural language processing (NLP) systems have difficulty in fully reflecting the individual personality and emotions of users, making it difficult to provide personalized experiences. They also lack the ability to safely manage user data and allow or deny the use of data for specific applications. Furthermore, they lack the ability to recognize user emotions and adjust NLP learning based on them.

[0666] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0667] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that has been trained with the personality of that person, a means for enabling the NLP to predict the thoughts of the customer, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This provides a personalized experience that reflects the individual personality and emotions of the user, and enables safe management and use control of data.

[0668] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0669] "Base training" is the process of training a model using a large amount of data in advance.

[0670] "Full Security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[0671] A "mobile terminal" is a portable electronic device such as a smartphone or tablet.

[0672] "Interactions" refers to user activity data such as emails, social media posts, and search history.

[0673] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[0674] An "API connection" is a way of exchanging data between systems through an application programming interface.

[0675] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input.

[0676] "Learning and tuning" is the process of changing the parameters of a model based on analyzed data to improve performance.

[0677] A "personalized experience" refers to providing services and features that are customized based on a user's individual characteristics and preferences.

[0678] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0679] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on a mobile device in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality.Furthermore, by combining an emotion engine that recognizes the user's emotions and adjusting the learning of the NLP based on the analyzed emotions, a personalized experience that reflects the user's individual personality and emotions is provided.

[0680] Hardware and Software Used

[0681] The server prepares natural language processing (NLP) models that have been trained in advance from large amounts of text data. For example, advanced language models such as GPT-3 and BERT are used for these NLP models. The server deploys these models on the cloud and makes them accessible via API.

[0682] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and transmits it to the server using a secure communication protocol (e.g., HTTPS).

[0683] The server analyzes the received data and builds an NLP model for each customer. Specifically, the server preprocesses the data, tokenizing the text, removing stop words, stemming, etc. Then, the server uses the data to retrain the NLP model.

[0684] Users can select which applications they want to allow or deny API connections to their NLP models from a dedicated management screen. The device sends this setting to the server, which then stores it.

[0685] The server receives the user's text or voice input and analyzes the emotion using the emotion engine. For example, if the user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" from this text.

[0686] The server adjusts the learning of the NLP model using the analyzed emotion data. For example, if the user frequently expresses the emotion "sad," the server adjusts the learning of the NLP model to correspond to the user's emotion "sad." This creates a personalized NLP that adapts to the user's emotions.

[0687] Examples

[0688] Example 1: Parsing an email

[0689] If a user sends an email saying "I'm very tired today," the device sends this text to the server, which uses an emotion engine to analyze the emotion of "tiredness" and incorporates it into the NLP model.

[0690] Example 2: Social media posts

[0691] If a user posts on social media, "I'm happy that a new cafe has opened," the device sends this text to the server. The server uses an emotion engine to analyze the emotion of "happiness" and reflects it in the NLP model.

[0692] Examples of prompt statements

[0693] "How can we analyze the sentiment of text posted by users on social media and adjust NLP learning based on that sentiment?"

[0694] The system allows users to safely manage their data and enjoy personalized experiences that respond to their emotions.

[0695] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0696] Step 1:

[0697] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, the server uses advanced language models such as GPT-3 and BERT, and deploys these models on the cloud. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[0698] Step 2:

[0699] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and sends it to the server using a secure communication protocol (e.g., HTTPS). As input, the user's activity data is required, and as output, the encrypted data is sent to the server.

[0700] Step 3:

[0701] The server analyzes the received data and builds a per-customer NLP model. Specifically, the server pre-processes the data, tokenizing the text, removing stop words, stemming, etc. Then it retrains the NLP model with these data. As input, it requires encrypted user data, and as output, it gets a per-customer NLP model.

[0702] Step 4:

[0703] From a dedicated management screen, the user selects the applications for which they want to allow or deny API connections to their NLP models. The device sends this setting to the server, which then stores the setting. As input, the user's permission setting is required, and as output, the setting for allowing or denying API connections is stored on the server.

[0704] Step 5:

[0705] The server receives the user's text or voice input and analyzes the emotion using an emotion engine. For example, if the user inputs "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. The input requires the user's text or voice data, and the output is the analyzed emotion data.

[0706] Step 6:

[0707] The server uses the analyzed emotion data to adjust the learning of the NLP model. For example, if the user frequently expresses the emotion "sad", the server adjusts the learning of the NLP model to correspond to the user's emotion "sad". As input, the analyzed emotion data is required, and as output, an emotion-adapted NLP model is obtained.

[0708] (Application example 1)

[0709] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal".

[0710] Conventional content delivery services have had difficulty providing personalized content based on users' emotions and behavior. They also lacked technology to accurately analyze users' thoughts and emotions while protecting their privacy. Furthermore, there were limited ways for users to freely control which applications could use their data.

[0711] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0712] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to their own NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, analyzing emotions from the user's text input or voice input, and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the user's emotions and behavior. This makes it possible to provide personalized content based on the user's emotions and behavior.

[0713] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[0714] "Full security" refers to a state in which advanced security measures are in place to prevent unauthorized access or leakage of data and protect customer privacy.

[0715] "Customer-specific NLP" is an individual natural language processing model that learns about a customer's personality and behavioral patterns based on their interactions on their mobile phone.

[0716] An "API connection" is a connection that exchanges data and functions between different software through an application program interface.

[0717] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input and adjusts the system's operation based on the results of that analysis.

[0718] "Content recommendation" is a function that selects and provides the most suitable content, such as videos, articles, and music, based on the user's emotions and behavior.

[0719] A system for implementing the present invention is configured as follows: First, a server prepares a natural language processing (NLP) model that has undergone base training in advance. This NLP model has been trained from a large amount of text data and has the ability to understand and generate the language of a user.

[0720] Next, the device (e.g., a smartphone or head-mounted display) inputs all of the user's mobile phone interactions in a fully secure environment. This input data includes the user's emails, social media posts, search history, etc. Based on this data, the server builds an NLP model for each user. This model learns the user's personality and behavioral patterns and has the ability to predict the user's thoughts.

[0721] Furthermore, the server combines an emotion engine. This emotion engine analyzes emotions from the user's text input or voice input, and adjusts the learning of the NLP model based on the analysis results. For example, if the user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. Also, if the user frequently shows the emotion of "sad," the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion of "sad."

[0722] The system also has the ability to recommend optimal content based on a user's emotions and behavior. For example, if a user posts on social media that they are "tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos.

[0723] Users can also "allow / deny" API connections to their NLP models to various applications out there, giving users the freedom to control which applications can use their data.

[0724] For example, if a user posts on social media that "I'm tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos. In this way, it is possible to provide personalized content based on the user's emotions and behavior.

[0725] An example of a prompt sentence might be:

[0726] "If a user posts on social media, 'I'm tired today,' what emotion will the emotion engine analyze, and what content will the NLP model recommend?"

[0727] This system will make it possible to provide personalized content based on a user's emotions and behavior, which is expected to improve the user experience.

[0728] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0729] Step 1:

[0730] The device inputs all of the user's mobile phone interactions. The input data includes the user's emails, social media posts, search history, etc. This data is collected in a completely secure environment. The input data becomes the basis for analyzing the user's behavior and thoughts.

[0731] Step 2:

[0732] The server builds an NLP model for each user based on the collected data. Specifically, it uses an NLP model that has undergone base training in advance to analyze the user's text data and learn the user's personality and behavioral patterns. In this process, the input data is analyzed and processed to extract the user's characteristics. The output is an NLP model for each user.

[0733] Step 3:

[0734] The server uses an emotion engine to analyze emotions from the user's text or voice input. For example, if the user inputs the text "Today was fun," the emotion engine analyzes the emotion "joy" from this text. The input is the user's text or voice data, and the output is the analyzed emotion data.

[0735] Step 4:

[0736] The server adjusts the learning of the NLP model based on the analysis result of the emotion engine. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion "sad". The input is the analyzed emotion data, and the output is the adjusted NLP model.

[0737] Step 5:

[0738] The server recommends optimal content based on the user's emotions and behavior. For example, if a user posts on social media that "I'm tired today," the emotion engine analyzes the emotion of "tiredness," and based on this information, the NLP model recommends relaxing music and refreshing videos. The input is the tuned NLP model and the analyzed emotion data, and the output is the recommended content.

[0739] Step 6:

[0740] Users can "allow / deny" API connections to their NLP models to various applications in the world. Users can control data usage for each application. The input is the user's permission settings, and the output is API connection control information.

[0741] In this way, a system is realized that provides personalized content based on the user's emotions and behavior.

[0742] Example 2

[0743] Next, a description will be given of Example 2 of the embodiment 2. In the following description, the data processing device 12 is referred to as a "server" and the smart glasses 214 are referred to as a "terminal".

[0744] Conventional natural language processing (NLP) systems have the ability to understand general language patterns, but they have difficulty generating personalized responses based on the characteristics and emotions of individual users. It is also a challenge to build individual NLP models while protecting the privacy of users. Furthermore, users lack the means to freely control which applications can use their data.

[0745] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0746] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile terminal in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes and analyzes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This enables highly personalized NLP based on the user's characteristics and emotions, and allows the user to freely control which applications can use their data by building individual NLP models while protecting the user's privacy.

[0747] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0748] "Base training" is an initial learning process that is carried out using a large amount of data in advance for a specific task.

[0749] "Full security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[0750] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[0751] "Interaction" refers to communication activities such as emails, social media posts, and search history that users conduct via their mobile devices.

[0752] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[0753] "API connection" refers to a means of connection that allows different software to share data and functions through an application program interface.

[0754] "Application" means a software program that provides specific functions or services.

[0755] An "emotion engine" refers to technology that analyzes emotions from a user's text or voice input.

[0756] "Learning tuning" refers to the process of changing the parameters of a model based on analyzed data to improve performance.

[0757] "Privacy protection" refers to protective measures to prevent personal information from being used improperly.

[0758] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs the customer's interactions on a mobile device in a completely secure environment, and builds an individual NLP model for each customer based on that data.Furthermore, by combining an emotion engine that recognizes and analyzes the user's emotions and adjusting the learning of the NLP based on the analysis results, personalized NLP that adapts to the user's characteristics and emotions is realized.

[0759] Hardware and Software Used

[0760] server

[0761] The server prepares an NLP model that has been trained in advance. Specifically, it uses existing NLP models such as GPT-3 and BERT and deploys them on the cloud. The server analyzes the collected data and executes machine learning algorithms to build individual NLP models for each customer.

[0762] Terminal

[0763] The terminal collects the customer's mobile device interactions and transmits them to the server in a fully secure environment. The terminal collects data through a dedicated application, encrypts the data using SSL / TLS protocol, and transmits it securely to the server.

[0764] Emotion Engine

[0765] The emotion engine analyzes emotions from the user's text and voice input. The emotion engine sends the analysis results to the server and uses them to adjust the learning of the NLP model. For example, if a user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" and reflects it in the NLP model.

[0766] Examples

[0767] For example, if a user posts on social media, "I'm so tired today," the emotion engine will analyze the emotion of "fatigue." The server will then feed this emotion data into the NLP model and adjust the learning to generate appropriate responses when the user is tired.

[0768] Examples of prompt statements

[0769] "If a user posts on social media, 'I'm so tired today,' please explain what emotion the emotion engine analyzes and how the NLP model adjusts its learning."

[0770] The system enables highly personalized NLP based on user characteristics and emotions, building individual NLP models while protecting user privacy and giving users full control over which applications can use their data.

[0771] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0772] Step 1:

[0773] The server prepares a natural language processing (NLP) model that has been pre-trained as a base.

[0774] Input: A large amount of text data

[0775] Data processing: Training NLP models using text data

[0776] Output: An NLP model with basic language understanding capabilities.

[0777] How it works: The server uses existing NLP models such as GPT-3 and BERT and deploys them on the cloud.

[0778] Step 2:

[0779] The terminal inputs all of the customer's mobile interactions in a completely secure environment.

[0780] Input: Customer emails, social media posts, search history, etc.

[0781] Data processing: Collect and encrypt data

[0782] Output: Encrypted customer data

[0783] How it works: The device collects data through a dedicated application, encrypts the data using the SSL / TLS protocol, and securely transmits it to the server.

[0784] Step 3:

[0785] The server uses the collected data to build an individual NLP model for each customer.

[0786] Input: Encrypted customer data

[0787] Data processing: Analyze data and learn customer language and behavior patterns

[0788] Output: A separate NLP model for each customer

[0789] What happens: The server uses machine learning algorithms to train the model on specific phrases and expressions used by the customer.

[0790] Step 4:

[0791] The server combines the NLP system with an emotion engine that recognizes and analyzes the user's emotions.

[0792] Input: User text and voice input

[0793] Data processing: Emotion engine analyzes emotions

[0794] Output: Parsed emotion data

[0795] How it works: The server integrates an emotion engine into the NLP model to analyze emotions from the user's text and voice input.

[0796] Step 5:

[0797] The server adjusts the NLP learning based on the emotions analyzed by the emotion engine.

[0798] Input: Parsed emotion data

[0799] Data processing: Adjusting parameters of NLP models

[0800] Output: Emotion-adapted NLP model

[0801] Specific operation: The server uses the emotion data obtained from the emotion engine to adjust the parameters of the NLP model.

[0802] Step 6:

[0803] Users can "allow / deny" API connections to their NLP from various applications out there.

[0804] Input: User permission settings

[0805] Data processing: Control API connections based on permission settings

[0806] Output: API connections to authorized applications

[0807] How it works: Through a dedicated admin screen, users configure which applications can access their NLP models.

[0808] (Application example 2)

[0809] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the smart glasses 214 will be referred to as a "terminal".

[0810] Conventional content delivery services have the problem that they do not adequately recommend personalized content based on users' interests and emotions. In addition, there is no established method for effectively utilizing user data while protecting user privacy. Furthermore, there is a problem that it is difficult to gain users' trust because users lack the means to control the use of their data.

[0811] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0812] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their own NLP from various apps in the world, a means for analyzing the customer's emotions and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the customer's emotions and interests. This enables personalized content recommendations based on the user's emotions and interests, and allows the user's data to be effectively utilized while protecting the user's privacy. In addition, by providing a means for the user to control the use of their own data, the user's trust can be gained.

[0813] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[0814] "Full security" means that advanced protective measures are in place to prevent unauthorized access or leakage of data.

[0815] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[0816] "Personality-trained NLP for each customer" is a natural language processing model that learns the behavior and thought patterns of each individual customer and is specialized for that customer.

[0817] An "API connection" is an interface for exchanging data and functions between different software programs.

[0818] "Emotion analysis" is the process of identifying a user's current emotional state from their text or voice input.

[0819] "Tuning NLP learning" means changing the learning parameters of a natural language processing model based on analyzed sentiment data.

[0820] "Recommending content" means presenting information such as the most suitable movies, music, articles, etc. based on the user's interests and emotions.

[0821] The following system configuration will be described as an embodiment of the present invention.

[0822] System Programming

[0823] The system consists of the following main components:

[0824] 1. Natural Language Processing (NLP) Model: Prepare a pre-trained NLP model, which is trained from a large amount of text data.

[0825] 2. Data entry method: In a fully secure location, all of the customer's mobile phone interactions (emails, social media posts, search history, etc.) are entered.

[0826] 3. Individual NLP model construction method: Based on the input data, an NLP model trained on the personality of each customer is constructed.

[0827] 4. Mind-reading: The NLP model built can predict what customers will think.

[0828] 5. API connection control: Allows customers to "allow / deny" API connections to their NLP from various apps around the world.

[0829] 6. Sentiment analysis tools: Analyze emotions from customer text and voice input.

[0830] 7. Learning adjustment measures: Adjust the learning of the NLP model based on the analyzed sentiment data.

[0831] 8. Content Recommendation: Recommend the most relevant content based on customer sentiment and interests.

[0832] Processing Description

[0833] The server first prepares an NLP model that has already undergone base training. This NLP model is trained from a large amount of text data. Next, in a completely secure location, it inputs all of the customer's mobile phone interactions (emails, SNS posts, search history, etc.). Based on this data, it builds an NLP model trained on each customer's personality.

[0834] The NLP model that is built is tuned to be able to predict what the customer will think, and it also provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[0835] The emotion analysis means analyzes emotions from the customer's text input or voice input. For example, if a customer inputs the text "I had fun today," the emotion analysis means analyzes the emotion of "joy" from this text. Based on the analyzed emotion data, the learning of the NLP model is adjusted.

[0836] Finally, the system recommends the most suitable content based on the customer's emotions and interests. For example, if a customer sends an email saying "I had fun today" and searches for "latest movies," the system will analyze that the customer has the emotion of "joy" and recommend the latest comedy movies.

[0837] Examples of concrete examples and prompts

[0838] As a specific example, if a user sends an email saying "I had fun today" and searches for "latest movies," the system will analyze that the user has the emotion of "joy" and recommend the latest comedy movies.

[0839] An example of a prompt is as follows:

[0840] Develop an application that analyzes user sentiment based on user emails, SNS posts, and search history, and recommends the most suitable content. If a user sends an email saying "I had fun today" and searches for "latest movies," recommend the latest comedy movies.

[0841] The flow of the specific process in the application example 2 will be described with reference to FIG.

[0842] Step 1:

[0843] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data. The existing text data is used as input, and the pre-trained NLP model is obtained as output.

[0844] Step 2:

[0845] The server is a completely secure location and inputs all of the customer's mobile phone interactions (emails, SNS posts, search history, etc.) The input is the customer's communication data, and the output is that this data is safely stored on the server.

[0846] Step 3:

[0847] The server builds an NLP model trained on the personality of each customer based on the input data. The customer's communication data is used as input, and the NLP model for each customer is obtained as output.

[0848] Step 4:

[0849] The server adjusts the constructed NLP model so that it can predict the customer's thoughts. The NLP model for each customer is used as input, and the output is an NLP model with a thought-predicting function.

[0850] Step 5:

[0851] The server provides a way for customers to "allow / deny" API connections to their NLP from various apps out there. The input is the customer's permission information, and the output is the allow / deny setting for the API connection.

[0852] Step 6:

[0853] The server analyzes emotions from the customer's text and voice input, using the customer's text and voice data as input and obtaining analyzed emotion data as output.

[0854] Step 7:

[0855] The server adjusts the learning of the NLP model based on the analyzed emotion data: the emotion data and the NLP model are used as inputs, and the emotion-adapted NLP model is obtained as output.

[0856] Step 8:

[0857] The server recommends optimal content based on the customer's emotions and interests. As input, emotion data and customer interest data are used, and as output, the recommended content is obtained.

[0858] 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 a voice indicating a user input for the result of the specific processing. The control unit 46A transmits the voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0859] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by making a neural network perform deep learning. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating a voice, text data indicating a text, and image data indicating an image is input. The data generation model 58 performs inference on the input inference data according to the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0860] Another example of generative AI is Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) are mentioned.

[0861] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0862] [Third embodiment]

[0863] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0864] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0865] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[0866] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0867] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs the voice according to instructions from the processor 46.

[0868] Camera 42 is a small digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (e.g., an imaging range defined by an angle of view equivalent to the width of the field of vision of an average healthy person).

[0869] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54. The transmission and reception of various types of information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is performed in a secure state.

[0870] Fig. 6 shows an example of main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0871] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32, and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0872] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0873] In the headset type terminal 314, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0874] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0875] "Example 1"

[0876] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[0877] "Example 2"

[0878] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[0879] The process flow of each embodiment will be described below.

[0880] "Example 1"

[0881] Step 1: Prepare a pre-trained natural language processing (NLP) model. This NLP model is trained, for example, from a large amount of text data.

[0882] Step 2: In a fully secured location, enter all of the customer's mobile phone interactions, including, for example, the customer's emails, social media posts, search history, etc.

[0883] Step 3: Using this data, we build a personalized NLP that is trained on that person's personality and has the ability to predict what the customer will think and do.

[0884] Step 4: Allow customers to "allow / deny" API connections to their NLP from various apps around the world, giving customers the freedom to control which apps can use their data.

[0885] Example 1

[0886] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[0887] Conventional natural language processing (NLP) systems have difficulty building personalized models that reflect the characteristics and behavior of individual customers. In addition, there was no established method for safely collecting customer data and using it while protecting privacy. Furthermore, there was a lack of means for customers to freely control which applications could use their data.

[0888] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0889] In this invention, the server includes a means for preparing a natural language processing (NLP) model that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure location and constructing an NLP model for each customer that has been trained with that person's personality, a means for enabling the NLP model to predict the customer's thoughts and actions, a means for collecting customer data, encrypting it and transmitting it to the server, a means for the server to fine-tune the NLP model using the customer data, and a means for the terminal to provide an interface for managing the customer's API connection settings. This makes it possible to construct a personalized NLP model that reflects the characteristics of the customer, safely collect and use customer data, and enable the customer to freely control the use of their own data.

[0890] A "natural language processing (NLP) model" is a machine learning model that is trained based on large amounts of text data and has the ability to understand and generate natural language.

[0891] "Complete security" means that advanced security measures are in place to prevent unauthorized access and information leaks during data collection, transmission, and storage.

[0892] "Communication equipment" refers to devices used by customers, such as mobile phones and tablets, through which data is entered and transmitted.

[0893] "Interactions" refers to all communication activities conducted by customers through communication devices, such as emails, social media posts, search history, etc.

[0894] A "personality-trained NLP model for each customer" is an NLP model trained to reflect the characteristics and behavioral patterns of each individual customer based on their communication activity data.

[0895] An "API connection" is a method of connecting different software to exchange data and functions through an application program interface.

[0896] "Allow / Deny" refers to the act of a customer choosing whether to allow or deny a particular application access to their NLP models.

[0897] "Data collection" is the process of obtaining interaction data from a customer's communications device.

[0898] "Encryption" is a security technology that converts data into a form that cannot be deciphered by third parties.

[0899] "Fine-tuning" is the process of fine-tuning an existing base model to a specific dataset to improve its performance.

[0900] An "interface" refers to the screen or operating means that allows the user to interact with the system, and in this case refers to the means used to set whether to allow or deny API connections.

[0901] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs all of the customer's interactions on communication devices, and builds an NLP model for each customer that is trained based on that person's personality. Furthermore, it enables this NLP model to predict the customer's thoughts and actions, and enables the customer to "allow / reject" API connections to their NLP model from various applications in the world.

[0902] Hardware and Software Used

[0903] server

[0904] The server uses machine learning frameworks such as TensorFlow or PyTorch to train NLP models on large amounts of text data. The models use advanced natural language processing techniques such as BERT and GPT-3. The server saves the trained models and prepares them for later processing.

[0905] Terminal

[0906] The device installs a dedicated app that allows the user to input interactions on their communication device. With the user's permission, the app collects data such as emails, social media posts, and search history. The collected data is encrypted and sent to a server. The device provides an interface for the user to manage API connections to their NLP models.

[0907] User

[0908] The user uses a dedicated app to input interactions on their communication device. The user's communication device transmits this data to the server. Through the device interface, the user can allow or deny the API connection to their NLP model.

[0909] Data processing and calculation

[0910] The server uses the received customer data to build an NLP model specific to that customer. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data preprocessing, model training, and evaluation. For example, the server analyzes the content of a customer's emails to learn what topics the customer is interested in.

[0911] Examples

[0912] For example, if a user allows their NLP model to be connected when using a shopping app, the app can recommend products that the user might be interested in based on the user's past purchase history and search history. On the other hand, they can protect their privacy by refusing to connect to SNS apps.

[0913] Examples of prompt statements

[0914] Examples of prompt sentences to be input to a generative AI model include the following:

[0915] "Predict the next topic that a user will be interested in based on their past emails and social media posts."

[0916] Using this prompt, the generative AI model can predict the user's interests and provide appropriate information.

[0917] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0918] Step 1:

[0919] The server prepares a natural language processing (NLP) model that has been pre-trained. Specifically, the server uses machine learning frameworks such as TensorFlow and PyTorch to train an NLP model based on a large amount of text data. This model utilizes advanced natural language processing techniques such as BERT and GPT-3. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[0920] Step 2:

[0921] A user installs a dedicated app to input communications on their communication device. With the user's permission, this app collects data such as emails, SNS posts, and search history. In concrete terms, the app obtains this data from the user's communication device, encrypts it, and sends it to a server. The input required is the user's communication data, and the output is the encrypted data sent to the server.

[0922] Step 3:

[0923] The server uses the received customer data to build a customer-specific NLP model. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data pre-processing, model training, and evaluation. As input, encrypted customer data is required, and as output, a customer-specific NLP model is obtained.

[0924] Step 4:

[0925] The terminal provides an interface for the user to manage API connections to their NLP models. In concrete terms, the terminal allows the user to select which applications are allowed or denied to use the NLP model. The terminal requires the user's selection information as input, and obtains the permission / denial setting for API connections as output.

[0926] Step 5:

[0927] The server controls the access of each application based on the user's API connection settings. Specifically, the server receives the user's setting information and allows or denies API access based on that information. The input is the user's API connection settings, and the output is access control for each application.

[0928] Step 6:

[0929] A user uses his / her own NLP model through an authorized application. Specifically, the user receives personalized services using his / her own NLP model in a shopping application, information providing application, etc. The input is a request from the application, and the output is the prediction result and recommendation information by the NLP model.

[0930] (Application example 1)

[0931] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[0932] Conventional online shopping sites did not adequately recommend products based on customers' interests and purchase history, making it difficult to suggest products that matched customers' needs in a timely manner. In addition, there was a lack of a system that could respond to individual needs while protecting customers' privacy.

[0933] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means. In this invention, the server includes a means for preparing natural language processing (NLP) that has been previously trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that is trained with the personality of that person, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to his / her NLP from various applications in the world, a means for using the customer's NLP to recommend products based on the customer's interests and purchase history, and a means for predicting the customer's behavioral patterns and making timely product suggestions. This makes it possible to timely suggest products that meet the customer's needs.

[0934] "Natural Language Processing (NLP)" is a technology that uses algorithms trained from large amounts of text data to understand and generate human language.

[0935] "Full security" means that advanced protective measures are in place to prevent data leakage and unauthorized access.

[0936] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[0937] "Interaction" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile devices.

[0938] "Personality-trained NLP for each customer" is an NLP model that learns each individual customer's thinking and behavior patterns based on their communication activities.

[0939] An "API connection" is an interface for exchanging data and functions between different software programs.

[0940] "Allow / Deny" is the act of a Customer choosing to allow or deny an API connection to their NLP.

[0941] "Means of product recommendation" is a function that uses customer NLP to suggest products based on the customer's interests and purchasing history.

[0942] A "means for predicting behavioral patterns" is a function that predicts future behavior based on a customer's past behavioral data.

[0943] "Timely product proposals" refers to proposing products that meet customer needs at the right time.

[0944] The system for implementing this invention has the following configuration. First, the server prepares natural language processing (NLP) that has undergone base learning in advance. Next, in a completely secure location, all interactions on the customer's mobile device (emails, SNS posts, search history, etc.) are entered, and an NLP for each customer that is trained on that person's personality is constructed. This NLP is designed to be able to predict the customer's thoughts.

[0945] Customers can "allow / deny" API connections to their NLP to various applications in the world, giving them the freedom to control which applications can use their data.

[0946] In addition, the server uses NLP to recommend products based on the customer's interests and purchase history, making it possible to predict customer behavior patterns and provide timely product suggestions.

[0947] Hardware and Software Used

[0948] Hardware: Servers (for data processing), smartphones (for collecting customer data)

[0949] Software: Python, scikit-learn (machine learning library), TfidfVectorizer (vectorization of text data)

[0950] Data processing and calculation

[0951] 1. Data collection:

[0952] The server collects customer emails, social media posts, and search history, and stores this data in JSON format.

[0953] 2. Building the NLP model:

[0954] The server vectorizes the text data using TfidfVectorizer, and uses the vectorized data to build the client's NLP model.

[0955] 3. Product recommendation:

[0956] The server compares the customer's NLP model with the product list, calculates the cosine similarity, and recommends products with high similarity.

[0957] Examples

[0958] For example, if customer A mentions in an email that he / she "want a new smartphone," posts on social media that he / she is "interested in the latest gadgets," and his / her search history includes "smartphone reviews," the server will use this data to build an NLP model for customer A. The server will then use this NLP model to recommend the latest smartphones, smartwatches, gadget accessories, etc.

[0959] Examples of prompt statements

[0960] Collect customer emails, social media posts, and search history, and vectorize them using TfidfVectorizer. Then build an NLP model of the customer and compare it with the product list to calculate the cosine similarity. Recommend products with high similarity.

[0961] In this way, a personalized shopping assistant can be realized that utilizes individual NLP for each customer.

[0962] The flow of the specific process in the application example 1 will be described with reference to FIG.

[0963] Step 1:

[0964] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and has basic language understanding capabilities. The input is the pre-trained NLP model, and the output is a usable NLP model.

[0965] Step 2:

[0966] The server, located in a completely secure location, collects all of the customer's mobile device interactions (emails, social media posts, search history, etc.). This data is encrypted and stored to protect the customer's privacy. The input is the customer's communication data, and the output is encrypted customer data.

[0967] Step 3:

[0968] The server uses the collected customer data to build an NLP model for each customer that is trained on that person's personality. Specifically, it uses TfidfVectorizer to vectorize the text data and generate an NLP model for the customer. The input is the encrypted customer data, and the output is an NLP model for each customer.

[0969] Step 4:

[0970] The server uses the customer's NLP model to predict the customer's thoughts and actions. To do this, it applies an algorithm that predicts future behavior based on past data. The input is the customer's NLP model, and the output is the predicted customer behavior pattern.

[0971] Step 5:

[0972] Users can "allow / deny" API connections to their NLP models from various applications in the world. The server controls API connections based on user permissions. The input is the user's permission information, and the output is a list of allowed applications.

[0973] Step 6:

[0974] The server uses the customer's NLP model to recommend products based on the customer's interests and purchase history. Specifically, it compares the customer's NLP model with the product list and calculates the cosine similarity. The inputs are the customer's NLP model and the product list, and the output is the recommended product list.

[0975] Step 7:

[0976] The server predicts customer behavior patterns and makes timely product suggestions. This makes it possible to suggest products that meet customer needs at the right time. The inputs are the predicted customer behavior patterns and a product list, and the output is a timely suggested product list.

[0977] Example 2

[0978] Next, a second embodiment of the second form will be described. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[0979] Conventional natural language processing (NLP) systems have difficulty accurately predicting the thoughts and behavior of individual users, and lack the means to freely control which applications can use user data. As a result, there is a high risk of violating user privacy and a poor user experience.

[0980] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0981] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for collecting customer data and constructing an individual NLP model, and a means for the customer to manage API connections using a dedicated management application. This makes it possible to accurately predict the user's thoughts and actions and freely control which applications can use the user's data.

[0982] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[0983] "Base learning" is the process of training a basic model using a large amount of data in advance.

[0984] "Complete Security" refers to the highest level of protection to prevent unauthorized access or leakage of data.

[0985] A "communication device" is a device used by a user, such as a mobile phone or tablet.

[0986] "Interaction" refers to activities such as emails, social media posts, and search history that users conduct through communication devices.

[0987] "Personality-trained NLP for each customer" is an NLP model that learns the characteristics and behavioral patterns of each individual customer based on their data.

[0988] An "API connection" is an interface that allows an application to communicate with other software or services.

[0989] "Allow / deny" is the act of a user choosing whether to allow or deny a particular application access to their data.

[0990] "Management Application" means dedicated software that allows a User to manage API connections to their NLP models.

[0991] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on communication devices in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality. This system makes it possible to predict the customer's thoughts, and enables the customer to "allow / reject" API connections to their own NLP from various applications in the world.

[0992] Hardware and Software Used

[0993] server

[0994] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. This NLP model uses a generative AI model such as GPT-4. The server is responsible for collecting customer data and building individual NLP models.

[0995] Terminal

[0996] The terminal is a communication device (such as a mobile phone or tablet) used by the user. The terminal collects data such as the user's emails, SNS posts, search history, etc., and transmits it to the server.

[0997] User

[0998] Users enter their interactions on their communication devices and manage API connections to their NLP models using a dedicated management application, and users choose to allow or deny specific applications access to their data.

[0999] Data processing and calculation

[1000] The server analyzes the collected customer data and builds an individual NLP model. Specifically, the customer data is tokenized and input into the base NLP model. The model learns the customer's language patterns and behavioral patterns and generates an individual NLP model. This model has the ability to predict the customer's thoughts and behavior.

[1001] Examples

[1002] For example, if a user is using a social networking app, the app needs the user's permission to access the user's NLP model. If the user gives permission, the social networking app can provide content tailored to the user's interests based on the user's past posts and search history.

[1003] Examples of prompt statements

[1004] "Recommend new movies that you think might interest me based on my past social media posts and search history."

[1005] When this prompt is input into a generative AI model, the model analyzes the user's past data and recommends movies that might interest them.

[1006] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1007] Step 1:

[1008] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, it uses a generative AI model such as GPT-4 to analyze and tokenize the text data and train the model. The input of this process is a large amount of text data, and the output is a base NLP model.

[1009] Step 2:

[1010] A user inputs interactions on their communication device. Specifically, they open an email app, select past emails, and send them to the server. They also open a social networking app, select past posts and comments, and send them to the server. They also export their browser search history and send it to the server. The input to this process is the data on the user's communication device, and the output is customer data sent to the server.

[1011] Step 3:

[1012] The server analyzes the collected customer data and builds an individual NLP model. Specifically, it tokenizes the customer data and inputs it into the base NLP model. The model learns the customer's language and behavioral patterns and generates an individual NLP model. The input of this process is the customer data and the output is an individual NLP model.

[1013] Step 4:

[1014] Users manage API connections to their NLP models using a dedicated management application: they open the management application and choose whether to allow or deny a particular application access to their data. The input to this process is the user's choice, and the output is the setting to allow or deny the API connection.

[1015] Step 5:

[1016] The server controls API connections based on user settings, granting access to NLP models to authorized applications and denying access to denied applications. The input to this process is the user settings, and the output is the API connection control result.

[1017] Step 6:

[1018] A user inputs a prompt sentence into the generative AI model using a specific application, for example, "Please recommend new movies that might interest me based on my past SNS posts and search history." The input of this process is the prompt sentence, and the output is the recommendation result from the generative AI model.

[1019] Step 7:

[1020] The server provides the user with the recommendation results from the generative AI model. Specifically, it analyzes the user's past data and recommends movies that the user may be interested in. The input of this process is the recommendation results from the generative AI model, and the output is the recommendation information provided to the user.

[1021] (Application example 2)

[1022] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1023] Conventional online shopping sites did not adequately recommend products based on individual customer needs and preferences, making it difficult to improve customer satisfaction. It was also difficult to provide a personalized shopping experience while protecting customer privacy.

[1024] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1025] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their NLP from various apps in the world, a means for using the NLP for each customer to recommend products that are optimal for each customer, and a means for displaying the recommended products to the customer. This makes it possible to recommend products personalized based on individual needs and preferences while protecting the customer's privacy.

[1026] "Natural Language Processing (NLP)" is a technology that learns from large amounts of text data to understand and generate human language.

[1027] "Full Security" refers to advanced protection measures to prevent data leakage and unauthorized access.

[1028] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[1029] "Customer-specific NLP" is a natural language processing model that is trained based on individual customer data and has the ability to predict what those customers will think and do.

[1030] An "API connection" is an interface for exchanging data and functions between different software systems.

[1031] "Allow / deny" is the act of a customer choosing which apps can use their data.

[1032] "Means of product recommendation" refers to a method that uses NLP for each customer to suggest the most suitable products to each individual customer.

[1033] A "means for displaying products" is a method for visually presenting the recommended products to a customer.

[1034] As an embodiment of the present invention, a smartphone application for a mail-order site will be described as an example. This application has a function of recommending the best products to each individual customer by using natural language processing (NLP) for each customer.

[1035] System Configuration

[1036] 1. Pre-trained Natural Language Processing (NLP):

[1037] The server prepares a base NLP model trained from a large amount of text data, which serves as the basis for building individual NLP models based on the customer's interactions on their mobile phone.

[1038] 2. Fully secure data entry:

[1039] Customers' mobile phone interactions (emails, SNS posts, search history, etc.) are entered in a completely secure environment, allowing necessary data to be collected while protecting customer privacy.

[1040] 3. Build customer-specific NLP:

[1041] The server uses the collected data to build an NLP model that reflects each customer's personality, and this NLP model has the ability to predict the customer's thoughts and behavior.

[1042] 4. Allow / Deny API Connection:

[1043] Customers can "allow / deny" API connections to their NLP to various apps around the world, giving them the freedom to control which apps can use their data.

[1044] 5. Generate product recommendations:

[1045] The server uses per-customer NLP to recommend the best products for each individual customer, based on the customer's past behavior and preferences.

[1046] 6. Display of recommended products:

[1047] The recommended products are displayed to customers through a smartphone application, making it easy for them to find the products that best suit them.

[1048] Program processing explanation

[1049] The server uses Python and the requests library to collect customer data and send it to the NLP API, which generates an NLP model for each customer and makes product recommendations based on that model. The recommended products are displayed to the customer through a smartphone application.

[1050] As a concrete example, if a customer has recently searched for "new smartphone," an NLP model can use that information to recommend "latest smartphones" and "smartphone accessories."

[1051] Example of a prompt to be input to a generative AI model:

[1052] Your customer's recent search history includes "new smartphone." Use this information to recommend the best products to them.

[1053] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[1054] The flow of the specific process in the application example 2 will be described with reference to FIG.

[1055] Step 1:

[1056] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and serves as the basis for building a customer-specific NLP model. The input is the pre-trained text data, and the output is the base NLP model.

[1057] Step 2:

[1058] The terminal collects the customer's mobile phone interactions (emails, SNS posts, search history, etc.) in a completely secure environment. The input is the customer's mobile phone interaction data, and the output is the collected customer data.

[1059] Step 3:

[1060] The server builds an NLP model that reflects the personality of each individual customer based on the collected customer data. The input is the collected customer data and the base NLP model, and the output is an NLP model for each customer. In concrete terms, the server sends customer data to the NLP API and generates an individual NLP model.

[1061] Step 4:

[1062] Users can "allow / deny" API connections to their NLP from various apps out there. The input is the user's choice of allow / deny, and the output is the allow / deny setting for API connections. In concrete terms, the user selects allow / deny on the application's settings screen.

[1063] Step 5:

[1064] The server uses an NLP model for each customer to recommend the best products to each individual customer. The input is the NLP model for each customer and the customer's past behavioral data, and the output is a list of recommended products. In concrete terms, the server uses the NLP model to analyze the customer's preferences and select the best products.

[1065] Step 6:

[1066] The terminal displays the recommended products to the customer. The input is a list of recommended products, and the output is product information that is displayed to the customer. In concrete terms, the terminal visually presents the product information through a smartphone application.

[1067] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[1068] In addition, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1069] "Example 1"

[1070] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[1071] "Example 2"

[1072] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[1073] The process flow of each embodiment will be described below.

[1074] "Example 1"

[1075] Step 1: The system prepares a pre-trained natural language processing (NLP) model.

[1076] Step 2: The system inputs all of the customer's mobile phone interactions in a fully secure location and builds an NLP model for each customer that is trained on that person's personality.

[1077] Step 3: The system uses the NLP model it built to predict what the customer will think.

[1078] Step 4: The system provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[1079] Step 5: The system combines an emotion engine to recognize the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text and voice inputs.

[1080] Step 6: The emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad".

[1081] Example 1

[1082] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[1083] Conventional natural language processing (NLP) systems have difficulty in fully reflecting the individual personality and emotions of users, making it difficult to provide personalized experiences. They also lack the ability to safely manage user data and allow or deny the use of data for specific applications. Furthermore, they lack the ability to recognize user emotions and adjust NLP learning based on them.

[1084] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1085] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that has been trained with the personality of that person, a means for enabling the NLP to predict the thoughts of the customer, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This provides a personalized experience that reflects the individual personality and emotions of the user, and enables safe management and use control of data.

[1086] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[1087] "Base training" is the process of training a model using a large amount of data in advance.

[1088] "Full Security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[1089] A "mobile terminal" is a portable electronic device such as a smartphone or tablet.

[1090] "Interactions" refers to user activity data such as emails, social media posts, and search history.

[1091] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[1092] An "API connection" is a way of exchanging data between systems through an application programming interface.

[1093] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input.

[1094] "Learning and tuning" is the process of changing the parameters of a model based on analyzed data to improve performance.

[1095] A "personalized experience" refers to providing services and features that are customized based on a user's individual characteristics and preferences.

[1096] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[1097] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on a mobile device in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality.Furthermore, by combining an emotion engine that recognizes the user's emotions and adjusting the learning of the NLP based on the analyzed emotions, a personalized experience that reflects the user's individual personality and emotions is provided.

[1098] Hardware and Software Used

[1099] The server prepares natural language processing (NLP) models that have been trained in advance from large amounts of text data. For example, advanced language models such as GPT-3 and BERT are used for these NLP models. The server deploys these models on the cloud and makes them accessible via API.

[1100] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and transmits it to the server using a secure communication protocol (e.g., HTTPS).

[1101] The server analyzes the received data and builds an NLP model for each customer. Specifically, the server preprocesses the data, tokenizing the text, removing stop words, stemming, etc. Then, the server uses the data to retrain the NLP model.

[1102] Users can select which applications they want to allow or deny API connections to their NLP models from a dedicated management screen. The device sends this setting to the server, which then stores it.

[1103] The server receives the user's text or voice input and analyzes the emotion using the emotion engine. For example, if the user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" from this text.

[1104] The server adjusts the learning of the NLP model using the analyzed emotion data. For example, if the user frequently expresses the emotion "sad," the server adjusts the learning of the NLP model to correspond to the user's emotion "sad." This creates a personalized NLP that adapts to the user's emotions.

[1105] Examples

[1106] Example 1: Parsing an email

[1107] If a user sends an email saying "I'm very tired today," the device sends this text to the server, which uses an emotion engine to analyze the emotion of "tiredness" and incorporates it into the NLP model.

[1108] Example 2: Social media posts

[1109] If a user posts on social media, "I'm happy that a new cafe has opened," the device sends this text to the server. The server uses an emotion engine to analyze the emotion of "happiness" and reflects it in the NLP model.

[1110] Examples of prompt statements

[1111] "How can we analyze the sentiment of text posted by users on social media and adjust NLP learning based on that sentiment?"

[1112] The system allows users to safely manage their data and enjoy personalized experiences that respond to their emotions.

[1113] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1114] Step 1:

[1115] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, the server uses advanced language models such as GPT-3 and BERT, and deploys these models on the cloud. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[1116] Step 2:

[1117] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and sends it to the server using a secure communication protocol (e.g., HTTPS). As input, the user's activity data is required, and as output, the encrypted data is sent to the server.

[1118] Step 3:

[1119] The server analyzes the received data and builds a per-customer NLP model. Specifically, the server pre-processes the data, tokenizing the text, removing stop words, stemming, etc. Then it retrains the NLP model with these data. As input, it requires encrypted user data, and as output, it gets a per-customer NLP model.

[1120] Step 4:

[1121] From a dedicated management screen, the user selects the applications for which they want to allow or deny API connections to their NLP models. The device sends this setting to the server, which then stores the setting. As input, the user's permission setting is required, and as output, the setting for allowing or denying API connections is stored on the server.

[1122] Step 5:

[1123] The server receives the user's text or voice input and analyzes the emotion using an emotion engine. For example, if the user inputs "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. The input requires the user's text or voice data, and the output is the analyzed emotion data.

[1124] Step 6:

[1125] The server uses the analyzed emotion data to adjust the learning of the NLP model. For example, if the user frequently expresses the emotion "sad", the server adjusts the learning of the NLP model to correspond to the user's emotion "sad". As input, the analyzed emotion data is required, and as output, an emotion-adapted NLP model is obtained.

[1126] (Application example 1)

[1127] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1128] Conventional content delivery services have had difficulty providing personalized content based on users' emotions and behavior. They also lacked technology to accurately analyze users' thoughts and emotions while protecting their privacy. Furthermore, there were limited ways for users to freely control which applications could use their data.

[1129] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1130] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to their own NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, analyzing emotions from the user's text input or voice input, and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the user's emotions and behavior. This makes it possible to provide personalized content based on the user's emotions and behavior.

[1131] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[1132] "Full security" refers to a state in which advanced security measures are in place to prevent unauthorized access or leakage of data and protect customer privacy.

[1133] "Customer-specific NLP" is an individual natural language processing model that learns about a customer's personality and behavioral patterns based on their interactions on their mobile phone.

[1134] An "API connection" is a connection that exchanges data and functions between different software through an application program interface.

[1135] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input and adjusts the system's operation based on the results of that analysis.

[1136] "Content recommendation" is a function that selects and provides the most suitable content, such as videos, articles, and music, based on the user's emotions and behavior.

[1137] A system for implementing the present invention is configured as follows: First, a server prepares a natural language processing (NLP) model that has undergone base training in advance. This NLP model has been trained from a large amount of text data and has the ability to understand and generate the language of a user.

[1138] Next, the device (e.g., a smartphone or head-mounted display) inputs all of the user's mobile phone interactions in a fully secure environment. This input data includes the user's emails, social media posts, search history, etc. Based on this data, the server builds an NLP model for each user. This model learns the user's personality and behavioral patterns and has the ability to predict the user's thoughts.

[1139] Furthermore, the server combines an emotion engine. This emotion engine analyzes emotions from the user's text input or voice input, and adjusts the learning of the NLP model based on the analysis results. For example, if the user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. Also, if the user frequently shows the emotion of "sad," the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion of "sad."

[1140] The system also has the ability to recommend optimal content based on a user's emotions and behavior. For example, if a user posts on social media that they are "tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos.

[1141] Users can also "allow / deny" API connections to their NLP models to various applications out there, giving users the freedom to control which applications can use their data.

[1142] For example, if a user posts on social media that "I'm tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos. In this way, it is possible to provide personalized content based on the user's emotions and behavior.

[1143] An example of a prompt sentence might be:

[1144] "If a user posts on social media, 'I'm tired today,' what emotion will the emotion engine analyze, and what content will the NLP model recommend?"

[1145] This system will make it possible to provide personalized content based on a user's emotions and behavior, which is expected to improve the user experience.

[1146] The flow of the specific process in the application example 1 will be described with reference to FIG.

[1147] Step 1:

[1148] The device inputs all of the user's mobile phone interactions. The input data includes the user's emails, social media posts, search history, etc. This data is collected in a completely secure environment. The input data becomes the basis for analyzing the user's behavior and thoughts.

[1149] Step 2:

[1150] The server builds an NLP model for each user based on the collected data. Specifically, it uses an NLP model that has undergone base training in advance to analyze the user's text data and learn the user's personality and behavioral patterns. In this process, the input data is analyzed and processed to extract the user's characteristics. The output is an NLP model for each user.

[1151] Step 3:

[1152] The server uses an emotion engine to analyze emotions from the user's text or voice input. For example, if the user inputs the text "Today was fun," the emotion engine analyzes the emotion "joy" from this text. The input is the user's text or voice data, and the output is the analyzed emotion data.

[1153] Step 4:

[1154] The server adjusts the learning of the NLP model based on the analysis result of the emotion engine. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion "sad". The input is the analyzed emotion data, and the output is the adjusted NLP model.

[1155] Step 5:

[1156] The server recommends optimal content based on the user's emotions and behavior. For example, if a user posts on social media that "I'm tired today," the emotion engine analyzes the emotion of "tiredness," and based on this information, the NLP model recommends relaxing music and refreshing videos. The input is the tuned NLP model and the analyzed emotion data, and the output is the recommended content.

[1157] Step 6:

[1158] Users can "allow / deny" API connections to their NLP models to various applications in the world. Users can control data usage for each application. The input is the user's permission settings, and the output is API connection control information.

[1159] In this way, a system is realized that provides personalized content based on the user's emotions and behavior.

[1160] Example 2

[1161] Next, a second embodiment of the second form will be described. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal".

[1162] Conventional natural language processing (NLP) systems have the ability to understand general language patterns, but they have difficulty generating personalized responses based on the characteristics and emotions of individual users. It is also a challenge to build individual NLP models while protecting the privacy of users. Furthermore, users lack the means to freely control which applications can use their data.

[1163] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1164] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile terminal in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes and analyzes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This enables highly personalized NLP based on the user's characteristics and emotions, and allows the user to freely control which applications can use their data by building individual NLP models while protecting the user's privacy.

[1165] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[1166] "Base training" is an initial learning process that is carried out using a large amount of data in advance for a specific task.

[1167] "Full security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[1168] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[1169] "Interaction" refers to communication activities such as emails, social media posts, and search history that users conduct via their mobile devices.

[1170] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[1171] "API connection" refers to a means of connection that allows different software to share data and functions through an application program interface.

[1172] "Application" means a software program that provides specific functions or services.

[1173] An "emotion engine" refers to technology that analyzes emotions from a user's text or voice input.

[1174] "Learning tuning" refers to the process of changing the parameters of a model based on analyzed data to improve performance.

[1175] "Privacy protection" refers to protective measures to prevent personal information from being used improperly.

[1176] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs the customer's interactions on a mobile device in a completely secure environment, and builds an individual NLP model for each customer based on that data.Furthermore, by combining an emotion engine that recognizes and analyzes the user's emotions and adjusting the learning of the NLP based on the analysis results, personalized NLP that adapts to the user's characteristics and emotions is realized.

[1177] Hardware and Software Used

[1178] server

[1179] The server prepares an NLP model that has been trained in advance. Specifically, it uses existing NLP models such as GPT-3 and BERT and deploys them on the cloud. The server analyzes the collected data and executes machine learning algorithms to build individual NLP models for each customer.

[1180] Terminal

[1181] The terminal collects the customer's mobile device interactions and transmits them to the server in a fully secure environment. The terminal collects data through a dedicated application, encrypts the data using SSL / TLS protocol, and transmits it securely to the server.

[1182] Emotion Engine

[1183] The emotion engine analyzes emotions from the user's text and voice input. The emotion engine sends the analysis results to the server and uses them to adjust the learning of the NLP model. For example, if a user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" and reflects it in the NLP model.

[1184] Examples

[1185] For example, if a user posts on social media, "I'm so tired today," the emotion engine will analyze the emotion of "fatigue." The server will then feed this emotion data into the NLP model and adjust the learning to generate appropriate responses when the user is tired.

[1186] Examples of prompt statements

[1187] "If a user posts on social media, 'I'm so tired today,' please explain what emotion the emotion engine analyzes and how the NLP model adjusts its learning."

[1188] The system enables highly personalized NLP based on user characteristics and emotions, building individual NLP models while protecting user privacy and giving users full control over which applications can use their data.

[1189] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1190] Step 1:

[1191] The server prepares a natural language processing (NLP) model that has been pre-trained as a base.

[1192] Input: A large amount of text data

[1193] Data processing: Training NLP models using text data

[1194] Output: An NLP model with basic language understanding capabilities.

[1195] How it works: The server uses existing NLP models such as GPT-3 and BERT and deploys them on the cloud.

[1196] Step 2:

[1197] The terminal inputs all of the customer's mobile interactions in a completely secure environment.

[1198] Input: Customer emails, social media posts, search history, etc.

[1199] Data processing: Collect and encrypt data

[1200] Output: Encrypted customer data

[1201] How it works: The device collects data through a dedicated application, encrypts the data using the SSL / TLS protocol, and securely transmits it to the server.

[1202] Step 3:

[1203] The server uses the collected data to build an individual NLP model for each customer.

[1204] Input: Encrypted customer data

[1205] Data processing: Analyze data and learn customer language and behavior patterns

[1206] Output: A separate NLP model for each customer

[1207] What happens: The server uses machine learning algorithms to train the model on specific phrases and expressions used by the customer.

[1208] Step 4:

[1209] The server combines the NLP system with an emotion engine that recognizes and analyzes the user's emotions.

[1210] Input: User text and voice input

[1211] Data processing: Emotion engine analyzes emotions

[1212] Output: Parsed emotion data

[1213] How it works: The server integrates an emotion engine into the NLP model to analyze emotions from the user's text and voice input.

[1214] Step 5:

[1215] The server adjusts the NLP learning based on the emotions analyzed by the emotion engine.

[1216] Input: Parsed emotion data

[1217] Data processing: Adjusting parameters of NLP models

[1218] Output: Emotion-adapted NLP model

[1219] Specific operation: The server uses the emotion data obtained from the emotion engine to adjust the parameters of the NLP model.

[1220] Step 6:

[1221] Users can "allow / deny" API connections to their NLP from various applications out there.

[1222] Input: User permission settings

[1223] Data processing: Control API connections based on permission settings

[1224] Output: API connections to authorized applications

[1225] How it works: Through a dedicated admin screen, users configure which applications can access their NLP models.

[1226] (Application example 2)

[1227] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server", and the headset type terminal 314 will be referred to as a "terminal".

[1228] Conventional content delivery services have the problem that they do not adequately recommend personalized content based on users' interests and emotions. In addition, there is no established method for effectively utilizing user data while protecting user privacy. Furthermore, there is a problem that it is difficult to gain users' trust because users lack the means to control the use of their data.

[1229] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1230] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their own NLP from various apps in the world, a means for analyzing the customer's emotions and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the customer's emotions and interests. This enables personalized content recommendations based on the user's emotions and interests, and allows the user's data to be effectively utilized while protecting the user's privacy. In addition, by providing a means for the user to control the use of their own data, the user's trust can be gained.

[1231] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[1232] "Full security" means that advanced protective measures are in place to prevent unauthorized access or leakage of data.

[1233] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[1234] "Personality-trained NLP for each customer" is a natural language processing model that learns the behavior and thought patterns of each individual customer and is specialized for that customer.

[1235] An "API connection" is an interface for exchanging data and functions between different software programs.

[1236] "Emotion analysis" is the process of identifying a user's current emotional state from their text or voice input.

[1237] "Tuning NLP learning" means changing the learning parameters of a natural language processing model based on analyzed sentiment data.

[1238] "Recommending content" means presenting information such as the most suitable movies, music, articles, etc. based on the user's interests and emotions.

[1239] The following system configuration will be described as an embodiment of the present invention.

[1240] System Programming

[1241] The system consists of the following main components:

[1242] 1. Natural Language Processing (NLP) Model: Prepare a pre-trained NLP model, which is trained from a large amount of text data.

[1243] 2. Data entry method: In a fully secure location, all of the customer's mobile phone interactions (emails, social media posts, search history, etc.) are entered.

[1244] 3. Individual NLP model construction method: Based on the input data, an NLP model trained on the personality of each customer is constructed.

[1245] 4. Mind-reading: The NLP model built can predict what customers will think.

[1246] 5. API connection control: Allows customers to "allow / deny" API connections to their NLP from various apps around the world.

[1247] 6. Sentiment analysis tools: Analyze emotions from customer text and voice input.

[1248] 7. Learning adjustment measures: Adjust the learning of the NLP model based on the analyzed sentiment data.

[1249] 8. Content Recommendation: Recommend the most relevant content based on customer sentiment and interests.

[1250] Processing Description

[1251] The server first prepares an NLP model that has already undergone base training. This NLP model is trained from a large amount of text data. Next, in a completely secure location, it inputs all of the customer's mobile phone interactions (emails, SNS posts, search history, etc.). Based on this data, it builds an NLP model trained on each customer's personality.

[1252] The NLP model that is built is tuned to be able to predict what the customer will think, and it also provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[1253] The emotion analysis means analyzes emotions from the customer's text input or voice input. For example, if a customer inputs the text "I had fun today," the emotion analysis means analyzes the emotion of "joy" from this text. Based on the analyzed emotion data, the learning of the NLP model is adjusted.

[1254] Finally, the system recommends the most suitable content based on the customer's emotions and interests. For example, if a customer sends an email saying "I had fun today" and searches for "latest movies," the system will analyze that the customer has the emotion of "joy" and recommend the latest comedy movies.

[1255] Examples of concrete examples and prompts

[1256] As a specific example, if a user sends an email saying "I had fun today" and searches for "latest movies," the system will analyze that the user has the emotion of "joy" and recommend the latest comedy movies.

[1257] An example of a prompt is as follows:

[1258] Develop an application that analyzes user sentiment based on user emails, SNS posts, and search history, and recommends the most suitable content. If a user sends an email saying "I had fun today" and searches for "latest movies," recommend the latest comedy movies.

[1259] The flow of the specific process in the application example 2 will be described with reference to FIG.

[1260] Step 1:

[1261] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data. The existing text data is used as input, and the pre-trained NLP model is obtained as output.

[1262] Step 2:

[1263] The server is a completely secure location and inputs all of the customer's mobile phone interactions (emails, SNS posts, search history, etc.) The input is the customer's communication data, and the output is that this data is safely stored on the server.

[1264] Step 3:

[1265] The server builds an NLP model trained on the personality of each customer based on the input data. The customer's communication data is used as input, and the NLP model for each customer is obtained as output.

[1266] Step 4:

[1267] The server adjusts the constructed NLP model so that it can predict the customer's thoughts. The NLP model for each customer is used as input, and the output is an NLP model with a thought-predicting function.

[1268] Step 5:

[1269] The server provides a way for customers to "allow / deny" API connections to their NLP from various apps out there. The input is the customer's permission information, and the output is the allow / deny setting for the API connection.

[1270] Step 6:

[1271] The server analyzes emotions from the customer's text and voice input, using the customer's text and voice data as input and obtaining analyzed emotion data as output.

[1272] Step 7:

[1273] The server adjusts the learning of the NLP model based on the analyzed emotion data: the emotion data and the NLP model are used as inputs, and the emotion-adapted NLP model is obtained as output.

[1274] Step 8:

[1275] The server recommends optimal content based on the customer's emotions and interests. As input, emotion data and customer interest data are used, and as output, the recommended content is obtained.

[1276] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input for the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1277] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by making a neural network perform deep learning. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating a voice, text data indicating a text, and image data indicating an image is input. The data generation model 58 performs inference on the input inference data according to the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1278] Another example of generative AI is Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) are mentioned.

[1279] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1280] [Fourth embodiment]

[1281] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1282] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1283] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a wide area network (WAN) and / or a local area network (LAN).

[1284] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. In addition, the microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1285] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs the voice according to instructions from the processor 46.

[1286] Camera 42 is a small digital camera equipped with an optical system including a lens, an aperture, and a shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (e.g., an imaging range defined by an angle of view equivalent to the width of the field of vision of an average healthy person).

[1287] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for transmitting and receiving various types of information between the processor 46 and the processor 28 via the network 54. The transmission and reception of various types of information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is performed in a secure state.

[1288] The control target 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, legs, etc. The posture and behavior of the robot 414 are controlled by controlling the motors of the arms, hands, legs, etc. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1289] Fig. 8 shows an example of main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1290] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32, and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1291] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1292] In the robot 414, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50, and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1293] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[1294] "Example 1"

[1295] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[1296] "Example 2"

[1297] As one embodiment of the present invention, a natural language processing (NLP) that has been pre-trained is prepared. This NLP is trained, for example, from a large amount of text data. Next, all of the customer's interactions on the mobile phone are input in a completely secure location. This input includes, for example, the customer's emails, SNS posts, search history, etc. Using this data, an NLP is built for each customer that is trained on that person's personality. This NLP has the ability to predict the customer's thoughts and actions. Finally, the customer is allowed to "allow / deny" API connections to their NLP from various apps in the world. This allows the customer to freely control which apps can use their data.

[1298] The process flow of each embodiment will be described below.

[1299] "Example 1"

[1300] Step 1: Prepare a pre-trained natural language processing (NLP) model. This NLP model is trained, for example, from a large amount of text data.

[1301] Step 2: In a fully secured location, enter all of the customer's mobile phone interactions, including, for example, the customer's emails, social media posts, search history, etc.

[1302] Step 3: Using this data, we build a personalized NLP that is trained on that person's personality and has the ability to predict what the customer will think and do.

[1303] Step 4: Allow customers to "allow / deny" API connections to their NLP from various apps around the world, giving customers the freedom to control which apps can use their data.

[1304] Example 1

[1305] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1306] Conventional natural language processing (NLP) systems have difficulty building personalized models that reflect the characteristics and behavior of individual customers. In addition, there was no established method for safely collecting customer data and using it while protecting privacy. Furthermore, there was a lack of means for customers to freely control which applications could use their data.

[1307] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1308] In this invention, the server includes a means for preparing a natural language processing (NLP) model that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure location and constructing an NLP model for each customer that has been trained with that person's personality, a means for enabling the NLP model to predict the customer's thoughts and actions, a means for collecting customer data, encrypting it and transmitting it to the server, a means for the server to fine-tune the NLP model using the customer data, and a means for the terminal to provide an interface for managing the customer's API connection settings. This makes it possible to construct a personalized NLP model that reflects the characteristics of the customer, safely collect and use customer data, and enable the customer to freely control the use of their own data.

[1309] A "natural language processing (NLP) model" is a machine learning model that is trained based on large amounts of text data and has the ability to understand and generate natural language.

[1310] "Complete security" means that advanced security measures are in place to prevent unauthorized access and information leaks during data collection, transmission, and storage.

[1311] "Communication equipment" refers to devices used by customers, such as mobile phones and tablets, through which data is entered and transmitted.

[1312] "Interactions" refers to all communication activities conducted by customers through communication devices, such as emails, social media posts, search history, etc.

[1313] A "personality-trained NLP model for each customer" is an NLP model trained to reflect the characteristics and behavioral patterns of each individual customer based on their communication activity data.

[1314] An "API connection" is a method of connecting different software to exchange data and functions through an application program interface.

[1315] "Allow / Deny" refers to the act of a customer choosing whether to allow or deny a particular application access to their NLP models.

[1316] "Data collection" is the process of obtaining interaction data from a customer's communications device.

[1317] "Encryption" is a security technology that converts data into a form that cannot be deciphered by third parties.

[1318] "Fine-tuning" is the process of fine-tuning an existing base model to a specific dataset to improve its performance.

[1319] An "interface" refers to the screen or operating means that allows a user to interact with the system, and in this case refers to the means used to set whether to allow or deny API connections.

[1320] This invention is a system that prepares a natural language processing (NLP) model that has been trained beforehand, inputs all of the customer's interactions on communication devices, and builds an NLP model for each customer that is trained based on that person's personality. Furthermore, it enables this NLP model to predict the customer's thoughts and actions, and enables the customer to "allow / reject" API connections to their NLP model from various applications in the world.

[1321] Hardware and Software Used

[1322] server

[1323] The server uses machine learning frameworks such as TensorFlow or PyTorch to train NLP models on large amounts of text data. The models use advanced natural language processing techniques such as BERT and GPT-3. The server saves the trained models and prepares them for later processing.

[1324] Terminal

[1325] The device installs a dedicated app that allows the user to input interactions on their communication device. With the user's permission, the app collects data such as emails, social media posts, and search history. The collected data is encrypted and sent to a server. The device provides an interface for the user to manage API connections to their NLP models.

[1326] User

[1327] The user uses a dedicated app to input interactions on their communication device. The user's communication device transmits this data to the server. Through the device interface, the user can allow or deny the API connection to their NLP model.

[1328] Data processing and calculation

[1329] The server uses the received customer data to build an NLP model specific to that customer. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data preprocessing, model training, and evaluation. For example, the server analyzes the content of a customer's emails to learn what topics the customer is interested in.

[1330] Examples

[1331] For example, if a user allows their NLP model to be connected when using a shopping app, the app can recommend products that the user might be interested in based on the user's past purchase history and search history. On the other hand, they can protect their privacy by refusing to connect to SNS apps.

[1332] Examples of prompt statements

[1333] Examples of prompt sentences to be input to a generative AI model include the following:

[1334] "Predict the next topic that a user will be interested in based on their past emails and social media posts."

[1335] Using this prompt, the generative AI model can predict the user's interests and provide appropriate information.

[1336] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1337] Step 1:

[1338] The server prepares a natural language processing (NLP) model that has been pre-trained. Specifically, the server uses machine learning frameworks such as TensorFlow and PyTorch to train an NLP model based on a large amount of text data. This model utilizes advanced natural language processing techniques such as BERT and GPT-3. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[1339] Step 2:

[1340] A user installs a dedicated app to input communications on their communication device. With the user's permission, this app collects data such as emails, SNS posts, and search history. In concrete terms, the app obtains this data from the user's communication device, encrypts it, and sends it to a server. The input required is the user's communication data, and the output is the encrypted data sent to the server.

[1341] Step 3:

[1342] The server uses the received customer data to build a customer-specific NLP model. Specifically, the server fine-tunes an existing base model to create a model specific to the customer's data. This process includes data pre-processing, model training, and evaluation. As input, encrypted customer data is required, and as output, a customer-specific NLP model is obtained.

[1343] Step 4:

[1344] The terminal provides an interface for the user to manage API connections to their NLP models. In concrete terms, the terminal allows the user to select which applications are allowed or denied to use the NLP model. The terminal requires the user's selection information as input, and obtains the permission / denial setting for API connections as output.

[1345] Step 5:

[1346] The server controls the access of each application based on the user's API connection settings. Specifically, the server receives the user's setting information and allows or denies API access based on that information. The input is the user's API connection settings, and the output is access control for each application.

[1347] Step 6:

[1348] A user uses his / her own NLP model through an authorized application. Specifically, the user receives personalized services using his / her own NLP model in a shopping application, information providing application, etc. The input is a request from the application, and the output is the prediction result and recommendation information by the NLP model.

[1349] (Application example 1)

[1350] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1351] Conventional online shopping sites did not adequately recommend products based on customers' interests and purchase history, making it difficult to suggest products that matched customers' needs in a timely manner. In addition, there was a lack of a system that could respond to individual needs while protecting customers' privacy.

[1352] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means. In this invention, the server includes a means for preparing natural language processing (NLP) that has been previously trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that is trained with the personality of that person, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to his / her NLP from various applications in the world, a means for using the customer's NLP to recommend products based on the customer's interests and purchase history, and a means for predicting the customer's behavioral patterns and making timely product suggestions. This makes it possible to timely suggest products that meet the customer's needs.

[1353] "Natural Language Processing (NLP)" is a technology that uses algorithms trained from large amounts of text data to understand and generate human language.

[1354] "Full security" means that advanced protective measures are in place to prevent data leakage and unauthorized access.

[1355] "Mobile device" refers to a portable electronic device such as a smartphone or tablet.

[1356] "Interaction" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile devices.

[1357] "Personality-trained NLP for each customer" is an NLP model that learns each individual customer's thinking and behavior patterns based on their communication activities.

[1358] An "API connection" is an interface for exchanging data and functions between different software programs.

[1359] "Allow / Deny" is the act of a Customer choosing to allow or deny an API connection to their NLP.

[1360] "Means of product recommendation" is a function that uses customer NLP to suggest products based on the customer's interests and purchasing history.

[1361] A "means for predicting behavioral patterns" is a function that predicts future behavior based on a customer's past behavioral data.

[1362] "Timely product proposals" refers to proposing products that meet customer needs at the right time.

[1363] The system for implementing this invention has the following configuration. First, the server prepares natural language processing (NLP) that has undergone base learning in advance. Next, in a completely secure location, all interactions on the customer's mobile device (emails, SNS posts, search history, etc.) are entered, and an NLP for each customer that is trained on that person's personality is constructed. This NLP is designed to be able to predict the customer's thoughts.

[1364] Customers can "allow / deny" API connections to their NLP to various applications in the world, giving them the freedom to control which applications can use their data.

[1365] In addition, the server uses NLP to recommend products based on the customer's interests and purchase history, making it possible to predict customer behavior patterns and provide timely product suggestions.

[1366] Hardware and Software Used

[1367] Hardware: Servers (for data processing), smartphones (for collecting customer data)

[1368] Software: Python, scikit-learn (machine learning library), TfidfVectorizer (vectorization of text data)

[1369] Data processing and calculation

[1370] 1. Data collection:

[1371] The server collects customer emails, social media posts, and search history, and stores this data in JSON format.

[1372] 2. Building the NLP model:

[1373] The server vectorizes the text data using TfidfVectorizer, and uses the vectorized data to build the client's NLP model.

[1374] 3. Product recommendation:

[1375] The server compares the customer's NLP model with the product list, calculates the cosine similarity, and recommends products with high similarity.

[1376] Examples

[1377] For example, if customer A mentions in an email that he / she "want a new smartphone," posts on social media that he / she is "interested in the latest gadgets," and his / her search history includes "smartphone reviews," the server will use this data to build an NLP model for customer A. The server will then use this NLP model to recommend the latest smartphones, smartwatches, gadget accessories, etc.

[1378] Examples of prompt statements

[1379] Collect customer emails, social media posts, and search history, and vectorize them using TfidfVectorizer. Then build an NLP model of the customer and compare it with the product list to calculate the cosine similarity. Recommend products with high similarity.

[1380] In this way, a personalized shopping assistant can be realized that utilizes individual NLP for each customer.

[1381] The flow of the specific process in the application example 1 will be described with reference to FIG.

[1382] Step 1:

[1383] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and has basic language understanding capabilities. The input is the pre-trained NLP model, and the output is a usable NLP model.

[1384] Step 2:

[1385] The server, located in a completely secure location, collects all of the customer's mobile device interactions (emails, social media posts, search history, etc.). This data is encrypted and stored to protect the customer's privacy. The input is the customer's communication data, and the output is encrypted customer data.

[1386] Step 3:

[1387] The server uses the collected customer data to build an NLP model for each customer that is trained on that person's personality. Specifically, it uses TfidfVectorizer to vectorize the text data and generate an NLP model for the customer. The input is the encrypted customer data, and the output is an NLP model for each customer.

[1388] Step 4:

[1389] The server uses the customer's NLP model to predict the customer's thoughts and actions. To do this, it applies an algorithm that predicts future behavior based on past data. The input is the customer's NLP model, and the output is the predicted customer behavior pattern.

[1390] Step 5:

[1391] Users can "allow / deny" API connections to their NLP models from various applications in the world. The server controls API connections based on user permissions. The input is the user's permission information, and the output is a list of allowed applications.

[1392] Step 6:

[1393] The server uses the customer's NLP model to recommend products based on the customer's interests and purchase history. Specifically, it compares the customer's NLP model with the product list and calculates the cosine similarity. The inputs are the customer's NLP model and the product list, and the output is the recommended product list.

[1394] Step 7:

[1395] The server predicts customer behavior patterns and makes timely product suggestions. This makes it possible to suggest products that meet customer needs at the right time. The inputs are the predicted customer behavior patterns and a product list, and the output is a timely suggested product list.

[1396] Example 2

[1397] Next, a second embodiment of the second embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1398] Conventional natural language processing (NLP) systems have difficulty accurately predicting the thoughts and behavior of individual users, and lack the means to freely control which applications can use user data. As a result, there is a high risk of violating user privacy and a poor user experience.

[1399] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1400] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's communication device in a completely secure place and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for collecting customer data and constructing an individual NLP model, and a means for the customer to manage API connections using a dedicated management application. This makes it possible to accurately predict the user's thoughts and actions and freely control which applications can use the user's data.

[1401] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[1402] "Base learning" is the process of training a basic model using a large amount of data in advance.

[1403] "Complete Security" refers to the highest level of protection to prevent unauthorized access or leakage of data.

[1404] A "communication device" is a device used by a user, such as a mobile phone or tablet.

[1405] "Interaction" refers to activities such as emails, social media posts, and search history that users conduct through communication devices.

[1406] "Personality-trained NLP for each customer" is an NLP model that learns the characteristics and behavioral patterns of each individual customer based on their data.

[1407] An "API connection" is an interface that allows an application to communicate with other software or services.

[1408] "Allow / deny" is the act of a user choosing whether to allow or deny a particular application access to their data.

[1409] "Management Application" means dedicated software that allows a User to manage API connections to their NLP models.

[1410] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on communication devices in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality. This system makes it possible to predict the customer's thoughts, and enables the customer to "allow / reject" API connections to their own NLP from various applications in the world.

[1411] Hardware and Software Used

[1412] server

[1413] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. This NLP model uses a generative AI model such as GPT-4. The server is responsible for collecting customer data and building individual NLP models.

[1414] Terminal

[1415] The terminal is a communication device (such as a mobile phone or tablet) used by the user. The terminal collects data such as the user's emails, SNS posts, search history, etc., and transmits it to the server.

[1416] User

[1417] Users enter their interactions on their communication devices and manage API connections to their NLP models using a dedicated management application, and users choose to allow or deny specific applications access to their data.

[1418] Data processing and calculation

[1419] The server analyzes the collected customer data and builds an individual NLP model. Specifically, the customer data is tokenized and input into the base NLP model. The model learns the customer's language patterns and behavioral patterns and generates an individual NLP model. This model has the ability to predict the customer's thoughts and behavior.

[1420] Examples

[1421] For example, if a user is using a social networking app, the app needs the user's permission to access the user's NLP model. If the user gives permission, the social networking app can provide content tailored to the user's interests based on the user's past posts and search history.

[1422] Examples of prompt statements

[1423] "Recommend new movies that you think might interest me based on my past social media posts and search history."

[1424] When this prompt is input into a generative AI model, the model analyzes the user's past data and recommends movies that might interest them.

[1425] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1426] Step 1:

[1427] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, it uses a generative AI model such as GPT-4 to analyze and tokenize the text data and train the model. The input of this process is a large amount of text data, and the output is a base NLP model.

[1428] Step 2:

[1429] A user inputs interactions on their communication device. Specifically, they open an email app, select past emails, and send them to the server. They also open a social networking app, select past posts and comments, and send them to the server. They also export their browser search history and send it to the server. The input to this process is the data on the user's communication device, and the output is customer data sent to the server.

[1430] Step 3:

[1431] The server analyzes the collected customer data and builds an individual NLP model. Specifically, it tokenizes the customer data and inputs it into the base NLP model. The model learns the customer's language and behavioral patterns and generates an individual NLP model. The input of this process is the customer data and the output is an individual NLP model.

[1432] Step 4:

[1433] Users manage API connections to their NLP models using a dedicated management application: they open the management application and choose whether to allow or deny a particular application access to their data. The input to this process is the user's choice, and the output is the setting to allow or deny the API connection.

[1434] Step 5:

[1435] The server controls API connections based on user settings, granting access to NLP models to authorized applications and denying access to denied applications. The input to this process is the user settings, and the output is the API connection control result.

[1436] Step 6:

[1437] A user inputs a prompt sentence into the generative AI model using a specific application, for example, "Please recommend new movies that might interest me based on my past SNS posts and search history." The input of this process is the prompt sentence, and the output is the recommendation result from the generative AI model.

[1438] Step 7:

[1439] The server provides the user with the recommendation results from the generative AI model. Specifically, it analyzes the user's past data and recommends movies that the user may be interested in. The input of this process is the recommendation results from the generative AI model, and the output is the recommendation information provided to the user.

[1440] (Application example 2)

[1441] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1442] Conventional online shopping sites did not adequately recommend products based on individual customer needs and preferences, making it difficult to improve customer satisfaction. It was also difficult to provide a personalized shopping experience while protecting customer privacy.

[1443] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1444] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for the customer to "allow / reject" API connections to their NLP from various apps in the world, a means for using the NLP for each customer to recommend products that are optimal for each customer, and a means for displaying the recommended products to the customer. This makes it possible to recommend products personalized based on individual needs and preferences while protecting the customer's privacy.

[1445] "Natural Language Processing (NLP)" is a technology that learns from large amounts of text data to understand and generate human language.

[1446] "Full Security" refers to advanced protection measures to prevent data leakage and unauthorized access.

[1447] "Customer mobile phone interactions" refers to communication activities such as emails, social media posts, and search history that customers conduct via their mobile phones.

[1448] "Customer-specific NLP" is a natural language processing model that is trained based on individual customer data and has the ability to predict what those customers will think and do.

[1449] An "API connection" is an interface for exchanging data and functions between different software systems.

[1450] "Allow / deny" is the act of a customer choosing which apps can use their data.

[1451] "Means of product recommendation" refers to a method that uses NLP for each customer to suggest the most suitable products to each individual customer.

[1452] A "means for displaying products" is a method for visually presenting the recommended products to a customer.

[1453] As an embodiment of the present invention, a smartphone application for a mail-order site will be described as an example. This application has a function of recommending the best products to each individual customer by using natural language processing (NLP) for each customer.

[1454] System Configuration

[1455] 1. Pre-trained Natural Language Processing (NLP):

[1456] The server prepares a base NLP model trained from a large amount of text data, which serves as the basis for building individual NLP models based on the customer's interactions on their mobile phone.

[1457] 2. Fully secure data entry:

[1458] Customers' mobile phone interactions (emails, SNS posts, search history, etc.) are entered in a completely secure environment, allowing necessary data to be collected while protecting customer privacy.

[1459] 3. Build customer-specific NLP:

[1460] The server uses the collected data to build an NLP model that reflects each customer's personality, and this NLP model has the ability to predict the customer's thoughts and behavior.

[1461] 4. Allow / Deny API Connection:

[1462] Customers can "allow / deny" API connections to their NLP to various apps around the world, giving them the freedom to control which apps can use their data.

[1463] 5. Generate product recommendations:

[1464] The server uses per-customer NLP to recommend the best products for each individual customer, based on the customer's past behavior and preferences.

[1465] 6. Display of recommended products:

[1466] The recommended products are displayed to customers through a smartphone application, making it easy for them to find the products that best suit them.

[1467] Program processing explanation

[1468] The server uses Python and the requests library to collect customer data and send it to the NLP API, which generates an NLP model for each customer and makes product recommendations based on that model. The recommended products are displayed to the customer through a smartphone application.

[1469] As a concrete example, if a customer has recently searched for "new smartphone," an NLP model can use that information to recommend "latest smartphones" and "smartphone accessories."

[1470] Example of a prompt to be input to a generative AI model:

[1471] Your customer's recent search history includes "new smartphone." Use this information to recommend the best products to them.

[1472] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[1473] The flow of the specific process in the application example 2 will be described with reference to FIG.

[1474] Step 1:

[1475] The server prepares a pre-trained natural language processing (NLP) model. This NLP model is trained from a large amount of text data and serves as the basis for building a customer-specific NLP model. The input is the pre-trained text data, and the output is the base NLP model.

[1476] Step 2:

[1477] The terminal collects the customer's mobile phone interactions (emails, SNS posts, search history, etc.) in a completely secure environment. The input is the customer's mobile phone interaction data, and the output is the collected customer data.

[1478] Step 3:

[1479] The server builds an NLP model that reflects the personality of each individual customer based on the collected customer data. The input is the collected customer data and the base NLP model, and the output is an NLP model for each customer. In concrete terms, the server sends customer data to the NLP API and generates an individual NLP model.

[1480] Step 4:

[1481] Users can "allow / deny" API connections to their NLP from various apps out there. The input is the user's choice of allow / deny, and the output is the allow / deny setting for API connections. In concrete terms, the user selects allow / deny on the application's settings screen.

[1482] Step 5:

[1483] The server uses an NLP model for each customer to recommend the best products to each individual customer. The input is the NLP model for each customer and the customer's past behavioral data, and the output is a list of recommended products. In concrete terms, the server uses the NLP model to analyze the customer's preferences and select the best products.

[1484] Step 6:

[1485] The terminal displays the recommended products to the customer. The input is a list of recommended products, and the output is product information that is displayed to the customer. In concrete terms, the terminal visually presents the product information through a smartphone application.

[1486] In this way, they can predict customer behavior and thoughts and provide a personalized shopping experience.

[1487] In addition, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1488] "Example 1"

[1489] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[1490] "Example 2"

[1491] As one embodiment of the present invention, a system combining an emotion engine is provided. As described in claim 4, this system includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with the person's personality, a means for enabling the NLP to predict the customer's thoughts, and a means for the customer to "allow / reject" API connections to the NLP from various apps in the world. Furthermore, this system combines an emotion engine that recognizes the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text input or voice input. For example, if a user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. In addition, the emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if a user frequently expresses the emotion "sad," the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad." This allows a personalized NLP that is more adapted to the user's emotions to be constructed.

[1492] The process flow of each embodiment will be described below.

[1493] "Example 1"

[1494] Step 1: The system prepares a pre-trained natural language processing (NLP) model.

[1495] Step 2: The system inputs all of the customer's mobile phone interactions in a fully secure location and builds an NLP model for each customer that is trained on that person's personality.

[1496] Step 3: The system uses the NLP model it built to predict what the customer will think.

[1497] Step 4: The system provides a way for customers to "allow / deny" API connections to their NLP from various apps out there.

[1498] Step 5: The system combines an emotion engine to recognize the user's emotions. Specifically, the emotion engine analyzes emotions from the user's text and voice inputs.

[1499] Step 6: The emotion engine adjusts the learning of the NLP based on the analyzed emotion. For example, if the user frequently expresses the emotion "sad", the emotion engine adjusts the learning of the NLP to correspond to the user's emotion "sad".

[1500] Example 1

[1501] Next, a description will be given of Example 1 of the embodiment 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1502] Conventional natural language processing (NLP) systems have difficulty in fully reflecting the individual personality and emotions of users, making it difficult to provide personalized experiences. They also lack the ability to safely manage user data and allow or deny the use of data for specific applications. Furthermore, they lack the ability to recognize user emotions and adjust NLP learning based on them.

[1503] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1504] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a mobile terminal of a customer in a completely secure place and constructing an NLP for each customer that has been trained with the personality of that person, a means for enabling the NLP to predict the thoughts of the customer, a means for the customer to "allow / reject" API connections to the NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, and a means for adjusting the learning of the NLP based on the emotions analyzed by the emotion engine. This provides a personalized experience that reflects the individual personality and emotions of the user, and enables safe management and use control of data.

[1505] "Natural Language Processing (NLP)" is a technology that enables computers to understand, interpret, and generate human language.

[1506] "Base training" is the process of training a model using a large amount of data in advance.

[1507] "Full Security" refers to advanced protection measures to prevent unauthorized access or leakage of data.

[1508] A "mobile terminal" is a portable electronic device such as a smartphone or tablet.

[1509] "Interactions" refers to user activity data such as emails, social media posts, and search history.

[1510] "Personality" refers to the characteristics and behavioral patterns of an individual user.

[1511] An "API connection" is a way of exchanging data between systems through an application programming interface.

[1512] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input.

[1513] "Learning and tuning" is the process of changing the parameters of a model based on analyzed data to improve performance.

[1514] A "personalized experience" refers to providing services and features that are customized based on a user's individual characteristics and preferences.

[1515] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[1516] This invention is a system that prepares natural language processing (NLP) that has been trained in advance, inputs all of the customer's interactions on a mobile device in a completely secure location, and builds an NLP for each customer that is trained based on that person's personality.Furthermore, by combining an emotion engine that recognizes the user's emotions and adjusting the learning of the NLP based on the analyzed emotions, a personalized experience that reflects the user's individual personality and emotions is provided.

[1517] Hardware and Software Used

[1518] The server prepares natural language processing (NLP) models that have been trained in advance from large amounts of text data. For example, advanced language models such as GPT-3 and BERT are used for these NLP models. The server deploys these models on the cloud and makes them accessible via API.

[1519] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and transmits it to the server using a secure communication protocol (e.g., HTTPS).

[1520] The server analyzes the received data and builds an NLP model for each customer. Specifically, the server preprocesses the data, tokenizing the text, removing stop words, stemming, etc. Then, the server uses the data to retrain the NLP model.

[1521] Users can select which applications they want to allow or deny API connections to their NLP models from a dedicated management screen. The device sends this setting to the server, which then stores it.

[1522] The server receives the user's text or voice input and analyzes the emotion using the emotion engine. For example, if the user inputs "Today was fun," the emotion engine analyzes the emotion of "joy" from this text.

[1523] The server adjusts the learning of the NLP model using the analyzed emotion data. For example, if the user frequently expresses the emotion "sad," the server adjusts the learning of the NLP model to correspond to the user's emotion "sad." This creates a personalized NLP that adapts to the user's emotions.

[1524] Examples

[1525] Example 1: Parsing an email

[1526] If a user sends an email saying "I'm very tired today," the device sends this text to the server, which uses an emotion engine to analyze the emotion of "tiredness" and incorporates it into the NLP model.

[1527] Example 2: Social media posts

[1528] If a user posts on social media, "I'm happy that a new cafe has opened," the device sends this text to the server. The server uses an emotion engine to analyze the emotion of "happiness" and reflects it in the NLP model.

[1529] Examples of prompt statements

[1530] "How can we analyze the sentiment of text posted by users on social media and adjust NLP learning based on that sentiment?"

[1531] The system allows users to safely manage their data and enjoy personalized experiences that respond to their emotions.

[1532] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1533] Step 1:

[1534] The server prepares a natural language processing (NLP) model that has been trained in advance from a large amount of text data. Specifically, the server uses advanced language models such as GPT-3 and BERT, and deploys these models on the cloud. A large amount of text data is required as input, and a trained NLP model is obtained as output.

[1535] Step 2:

[1536] A user inputs data such as emails, SNS posts, search history, etc. from his / her mobile device. The device encrypts this data and sends it to the server using a secure communication protocol (e.g., HTTPS). As input, the user's activity data is required, and as output, the encrypted data is sent to the server.

[1537] Step 3:

[1538] The server analyzes the received data and builds a per-customer NLP model. Specifically, the server pre-processes the data, tokenizing the text, removing stop words, stemming, etc. Then it retrains the NLP model with these data. As input, it requires encrypted user data, and as output, it gets a per-customer NLP model.

[1539] Step 4:

[1540] From a dedicated management screen, the user selects the applications for which they want to allow or deny API connections to their NLP models. The device sends this setting to the server, which then stores the setting. As input, the user's permission setting is required, and as output, the setting for allowing or denying API connections is stored on the server.

[1541] Step 5:

[1542] The server receives the user's text or voice input and analyzes the emotion using an emotion engine. For example, if the user inputs "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. The input requires the user's text or voice data, and the output is the analyzed emotion data.

[1543] Step 6:

[1544] The server uses the analyzed emotion data to adjust the learning of the NLP model. For example, if the user frequently expresses the emotion "sad", the server adjusts the learning of the NLP model to correspond to the user's emotion "sad". As input, the analyzed emotion data is required, and as output, an emotion-adapted NLP model is obtained.

[1545] (Application example 1)

[1546] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal".

[1547] Conventional content delivery services have had difficulty providing personalized content based on users' emotions and behavior. They also lacked technology to accurately analyze users' thoughts and emotions while protecting their privacy. Furthermore, there were limited ways for users to freely control which applications could use their data.

[1548] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1549] In this invention, the server includes a means for preparing a natural language processing (NLP) that has been pre-trained as a base, a means for inputting all interactions on a customer's mobile phone in a completely secure location and constructing an NLP for each customer that has been trained with that person's personality, a means for enabling the NLP to predict the customer's thoughts, a means for a customer to "allow / reject" API connections to their own NLP from various applications in the world, a means for combining an emotion engine that recognizes the user's emotions, analyzing emotions from the user's text input or voice input, and adjusting the learning of the NLP based on the analysis results, and a means for recommending optimal content based on the user's emotions and behavior. This makes it possible to provide personalized content based on the user's emotions and behavior.

[1550] "Natural Language Processing (NLP)" is a technology that understands and generates human language using models trained from large amounts of text data.

[1551] "Full security" refers to a state in which advanced security measures are in place to prevent unauthorized access or leakage of data and protect customer privacy.

[1552] "Customer-specific NLP" is an individual natural language processing model that learns about a customer's personality and behavioral patterns based on their interactions on their mobile phone.

[1553] An "API connection" is a connection that exchanges data and functions between different software through an application program interface.

[1554] An "emotion engine" is a technology that analyzes emotions from a user's text or voice input and adjusts the system's operation based on the results of that analysis.

[1555] "Content recommendation" is a function that selects and provides the most suitable content, such as videos, articles, and music, based on the user's emotions and behavior.

[1556] A system for implementing the present invention is configured as follows: First, a server prepares a natural language processing (NLP) model that has undergone base training in advance. This NLP model has been trained from a large amount of text data and has the ability to understand and generate the language of a user.

[1557] Next, the device (e.g., a smartphone or head-mounted display) inputs all of the user's mobile phone interactions in a fully secure environment. This input data includes the user's emails, social media posts, search history, etc. Based on this data, the server builds an NLP model for each user. This model learns the user's personality and behavioral patterns and has the ability to predict the user's thoughts.

[1558] Furthermore, the server combines an emotion engine. This emotion engine analyzes emotions from the user's text input or voice input, and adjusts the learning of the NLP model based on the analysis results. For example, if the user inputs the text "I had fun today," the emotion engine analyzes the emotion of "joy" from this text. Also, if the user frequently shows the emotion of "sad," the emotion engine adjusts the learning of the NLP model to correspond to the user's emotion of "sad."

[1559] The system also has the ability to recommend optimal content based on a user's emotions and behavior. For example, if a user posts on social media that they are "tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos.

[1560] Users can also "allow / deny" API connections to their NLP models to various applications out there, giving users the freedom to control which applications can use their data.

[1561] For example, if a user posts on social media that "I'm tired today," the emotion engine will analyze the emotion of "tiredness," and based on this information, the NLP model will recommend relaxing music and refreshing videos. In this way, it is possible to provide personalized content based on the user's emotions and behavior.

[1562] An example of a prompt sentence might be:

[1563] "If a user posts on social media, 'I'm tired today,' what emotion will the emotion engine analyze, and what content ...

Claims

1. A means for preparing a natural language processing (NLP) model that has been trained in advance; A means to collect customer mobile interactions in a fully secure location; and A means for retraining a customer-specific NLP model using the collected data; and A means for adjusting the learning of an NLP model to correspond to the emotion analyzed by an emotion engine that recognizes and analyzes the emotion of a user; A system including:

2. The system of claim 1 , further comprising: means for outputting content corresponding to the emotion based on the data collected by the collecting means, the emotion analyzed by the emotion engine, and the NLP model.

3. The system according to claim 1, wherein the means for adjusting the learning of the NLP model adjusts the learning of the NLP model using a prompt sentence and a generative AI that instructs the emotion engine to analyze emotions and adjust the learning of the NLP model according to the content of the data collected by the collecting means.

Images